Merge branch 'feature/ffmpeg' into feature/codec

.circleci/config.yml

@@ -5,7 +5,7 @@ jobs:
       - image: ossrs/srs:dev
     steps:
       - checkout
-      - run: cd trunk && ./configure && make
+      - run: cd trunk && ./configure --without-rtc && make && ./configure && make
   test:
     docker:
       - image: ossrs/srs:dev

trunk/3rdparty/ffmpeg-4.2-fit/.gitignore

@@ -0,0 +1,18 @@
+# Ignore files.
+*.d
+*.pc
+*.o
+config.asm
+config.h
+_release
+doc
+ffbuild/config.fate
+ffbuild/config.log
+ffbuild/config.mak
+ffbuild/config.sh
+ffbuild/.config
+libavutil/lib.version
+libavcodec/libavcodec.version
+libavutil/libavutil.version
+libswresample/libswresample.version
+libavutil/ffversion.h

trunk/3rdparty/ffmpeg-4.2-fit/Makefile

@@ -0,0 +1,169 @@
+MAIN_MAKEFILE=1
+include ffbuild/config.mak
+
+vpath %.c    $(SRC_PATH)
+vpath %.cpp  $(SRC_PATH)
+vpath %.h    $(SRC_PATH)
+vpath %.inc  $(SRC_PATH)
+vpath %.m    $(SRC_PATH)
+vpath %.S    $(SRC_PATH)
+vpath %.asm  $(SRC_PATH)
+vpath %.rc   $(SRC_PATH)
+vpath %.v    $(SRC_PATH)
+vpath %.texi $(SRC_PATH)
+vpath %.cu   $(SRC_PATH)
+vpath %.ptx  $(SRC_PATH)
+vpath %/fate_config.sh.template $(SRC_PATH)
+
+TESTTOOLS   = audiogen videogen rotozoom tiny_psnr tiny_ssim base64 audiomatch
+HOSTPROGS  := $(TESTTOOLS:%=tests/%) doc/print_options
+
+# $(FFLIBS-yes) needs to be in linking order
+FFLIBS-$(CONFIG_AVDEVICE)   += avdevice
+FFLIBS-$(CONFIG_AVFILTER)   += avfilter
+FFLIBS-$(CONFIG_AVFORMAT)   += avformat
+FFLIBS-$(CONFIG_AVCODEC)    += avcodec
+FFLIBS-$(CONFIG_AVRESAMPLE) += avresample
+FFLIBS-$(CONFIG_POSTPROC)   += postproc
+FFLIBS-$(CONFIG_SWRESAMPLE) += swresample
+FFLIBS-$(CONFIG_SWSCALE)    += swscale
+
+FFLIBS := avutil
+
+DATA_FILES := $(wildcard $(SRC_PATH)/presets/*.ffpreset) $(SRC_PATH)/doc/ffprobe.xsd
+
+SKIPHEADERS = compat/w32pthreads.h
+
+# first so "all" becomes default target
+all: all-yes
+
+include $(SRC_PATH)/ffbuild/common.mak
+
+FF_EXTRALIBS := $(FFEXTRALIBS)
+FF_DEP_LIBS  := $(DEP_LIBS)
+FF_STATIC_DEP_LIBS := $(STATIC_DEP_LIBS)
+
+$(TOOLS): %$(EXESUF): %.o
+	$(LD) $(LDFLAGS) $(LDEXEFLAGS) $(LD_O) $^ $(EXTRALIBS-$(*F)) $(EXTRALIBS) $(ELIBS)
+
+target_dec_%_fuzzer$(EXESUF): target_dec_%_fuzzer.o $(FF_DEP_LIBS)
+	$(LD) $(LDFLAGS) $(LDEXEFLAGS) $(LD_O) $^ $(ELIBS) $(FF_EXTRALIBS) $(LIBFUZZER_PATH)
+
+tools/target_dem_fuzzer$(EXESUF): tools/target_dem_fuzzer.o $(FF_DEP_LIBS)
+	$(LD) $(LDFLAGS) $(LDEXEFLAGS) $(LD_O) $^ $(ELIBS) $(FF_EXTRALIBS) $(LIBFUZZER_PATH)
+
+tools/sofa2wavs$(EXESUF): ELIBS = $(FF_EXTRALIBS)
+tools/uncoded_frame$(EXESUF): $(FF_DEP_LIBS)
+tools/uncoded_frame$(EXESUF): ELIBS = $(FF_EXTRALIBS)
+tools/target_dec_%_fuzzer$(EXESUF): $(FF_DEP_LIBS)
+
+CONFIGURABLE_COMPONENTS =                                           \
+    $(wildcard $(FFLIBS:%=$(SRC_PATH)/lib%/all*.c))                 \
+    $(SRC_PATH)/libavcodec/bitstream_filters.c                      \
+    $(SRC_PATH)/libavcodec/parsers.c                                \
+    $(SRC_PATH)/libavformat/protocols.c                             \
+
+config.h: ffbuild/.config
+ffbuild/.config: $(CONFIGURABLE_COMPONENTS)
+	@-tput bold 2>/dev/null
+	@-printf '\nWARNING: $(?) newer than config.h, rerun configure\n\n'
+	@-tput sgr0 2>/dev/null
+
+SUBDIR_VARS := CLEANFILES FFLIBS HOSTPROGS TESTPROGS TOOLS               \
+               HEADERS ARCH_HEADERS BUILT_HEADERS SKIPHEADERS            \
+               ARMV5TE-OBJS ARMV6-OBJS ARMV8-OBJS VFP-OBJS NEON-OBJS     \
+               ALTIVEC-OBJS VSX-OBJS MMX-OBJS X86ASM-OBJS                \
+               MIPSFPU-OBJS MIPSDSPR2-OBJS MIPSDSP-OBJS MSA-OBJS         \
+               MMI-OBJS OBJS SLIBOBJS HOSTOBJS TESTOBJS
+
+define RESET
+$(1) :=
+$(1)-yes :=
+endef
+
+define DOSUBDIR
+$(foreach V,$(SUBDIR_VARS),$(eval $(call RESET,$(V))))
+SUBDIR := $(1)/
+include $(SRC_PATH)/$(1)/Makefile
+-include $(SRC_PATH)/$(1)/$(ARCH)/Makefile
+-include $(SRC_PATH)/$(1)/$(INTRINSICS)/Makefile
+include $(SRC_PATH)/ffbuild/library.mak
+endef
+
+$(foreach D,$(FFLIBS),$(eval $(call DOSUBDIR,lib$(D))))
+
+libavcodec/utils.o libavformat/utils.o libavdevice/avdevice.o libavfilter/avfilter.o libavutil/utils.o libpostproc/postprocess.o libswresample/swresample.o libswscale/utils.o : libavutil/ffversion.h
+
+$(PROGS): %$(PROGSSUF)$(EXESUF): %$(PROGSSUF)_g$(EXESUF)
+ifeq ($(STRIPTYPE),direct)
+	$(STRIP) -o $@ $<
+else
+	$(CP) $< $@
+	$(STRIP) $@
+endif
+
+%$(PROGSSUF)_g$(EXESUF): $(FF_DEP_LIBS)
+	$(LD) $(LDFLAGS) $(LDEXEFLAGS) $(LD_O) $(OBJS-$*) $(FF_EXTRALIBS)
+
+VERSION_SH  = $(SRC_PATH)/ffbuild/version.sh
+GIT_LOG     = $(SRC_PATH)/.git/logs/HEAD
+
+.version: $(wildcard $(GIT_LOG)) $(VERSION_SH) ffbuild/config.mak
+.version: M=@
+
+libavutil/ffversion.h .version:
+	$(M)$(VERSION_SH) $(SRC_PATH) libavutil/ffversion.h $(EXTRA_VERSION)
+	$(Q)touch .version
+
+# force version.sh to run whenever version might have changed
+-include .version
+
+install: install-libs install-headers
+
+install-libs: install-libs-yes
+
+install-data: $(DATA_FILES)
+	$(Q)mkdir -p "$(DATADIR)"
+	$(INSTALL) -m 644 $(DATA_FILES) "$(DATADIR)"
+
+uninstall: uninstall-data uninstall-headers uninstall-libs uninstall-pkgconfig
+
+uninstall-data:
+	$(RM) -r "$(DATADIR)"
+
+clean::
+	$(RM) $(CLEANSUFFIXES)
+	$(RM) $(addprefix compat/,$(CLEANSUFFIXES)) $(addprefix compat/*/,$(CLEANSUFFIXES)) $(addprefix compat/*/*/,$(CLEANSUFFIXES))
+	$(RM) -r coverage-html
+	$(RM) -rf coverage.info coverage.info.in lcov
+
+distclean:: clean
+	$(RM) .version avversion.h config.asm config.h mapfile  \
+		ffbuild/.config ffbuild/config.* libavutil/avconfig.h \
+		version.h libavutil/ffversion.h libavcodec/codec_names.h \
+		libavcodec/bsf_list.c libavformat/protocol_list.c \
+		libavcodec/codec_list.c libavcodec/parser_list.c \
+		libavformat/muxer_list.c libavformat/demuxer_list.c
+ifeq ($(SRC_LINK),src)
+	$(RM) src
+endif
+
+config:
+	$(SRC_PATH)/configure $(value FFMPEG_CONFIGURATION)
+
+build: all alltools testprogs
+check: all alltools testprogs fate
+
+$(sort $(OUTDIRS)):
+	$(Q)mkdir -p $@
+
+# Dummy rule to stop make trying to rebuild removed or renamed headers
+%.h:
+	@:
+
+# Disable suffix rules.  Most of the builtin rules are suffix rules,
+# so this saves some time on slow systems.
+.SUFFIXES:
+
+.PHONY: all all-yes alltools build check config testprogs
+.PHONY: *clean install* uninstall*

trunk/3rdparty/ffmpeg-4.2-fit/compat/atomics/gcc/stdatomic.h

@@ -0,0 +1,173 @@
+/*
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+/*
+ * based on vlc_atomic.h from VLC
+ * Copyright (C) 2010 Rémi Denis-Courmont
+ */
+
+#ifndef COMPAT_ATOMICS_GCC_STDATOMIC_H
+#define COMPAT_ATOMICS_GCC_STDATOMIC_H
+
+#include <stddef.h>
+#include <stdint.h>
+
+#define ATOMIC_FLAG_INIT 0
+
+#define ATOMIC_VAR_INIT(value) (value)
+
+#define atomic_init(obj, value) \
+do {                            \
+    *(obj) = (value);           \
+} while(0)
+
+#define kill_dependency(y) ((void)0)
+
+#define atomic_thread_fence(order) \
+    __sync_synchronize()
+
+#define atomic_signal_fence(order) \
+    ((void)0)
+
+#define atomic_is_lock_free(obj) 0
+
+typedef         _Bool      atomic_flag;
+typedef         _Bool      atomic_bool;
+typedef          char      atomic_char;
+typedef   signed char      atomic_schar;
+typedef unsigned char      atomic_uchar;
+typedef          short     atomic_short;
+typedef unsigned short     atomic_ushort;
+typedef          int       atomic_int;
+typedef unsigned int       atomic_uint;
+typedef          long      atomic_long;
+typedef unsigned long      atomic_ulong;
+typedef          long long atomic_llong;
+typedef unsigned long long atomic_ullong;
+typedef          wchar_t   atomic_wchar_t;
+typedef       int_least8_t atomic_int_least8_t;
+typedef      uint_least8_t atomic_uint_least8_t;
+typedef      int_least16_t atomic_int_least16_t;
+typedef     uint_least16_t atomic_uint_least16_t;
+typedef      int_least32_t atomic_int_least32_t;
+typedef     uint_least32_t atomic_uint_least32_t;
+typedef      int_least64_t atomic_int_least64_t;
+typedef     uint_least64_t atomic_uint_least64_t;
+typedef       int_fast8_t atomic_int_fast8_t;
+typedef      uint_fast8_t atomic_uint_fast8_t;
+typedef      int_fast16_t atomic_int_fast16_t;
+typedef     uint_fast16_t atomic_uint_fast16_t;
+typedef      int_fast32_t atomic_int_fast32_t;
+typedef     uint_fast32_t atomic_uint_fast32_t;
+typedef      int_fast64_t atomic_int_fast64_t;
+typedef     uint_fast64_t atomic_uint_fast64_t;
+typedef          intptr_t atomic_intptr_t;
+typedef         uintptr_t atomic_uintptr_t;
+typedef            size_t atomic_size_t;
+typedef         ptrdiff_t atomic_ptrdiff_t;
+typedef          intmax_t atomic_intmax_t;
+typedef         uintmax_t atomic_uintmax_t;
+
+#define atomic_store(object, desired)   \
+do {                                    \
+    *(object) = (desired);              \
+    __sync_synchronize();               \
+} while (0)
+
+#define atomic_store_explicit(object, desired, order) \
+    atomic_store(object, desired)
+
+#define atomic_load(object) \
+    (__sync_synchronize(), *(object))
+
+#define atomic_load_explicit(object, order) \
+    atomic_load(object)
+
+#define atomic_exchange(object, desired)                            \
+({                                                                  \
+    __typeof__(object) _obj = (object);                             \
+    __typeof__(*object) _old;                                       \
+    do                                                              \
+        _old = atomic_load(_obj);                                   \
+    while (!__sync_bool_compare_and_swap(_obj, _old, (desired)));   \
+    _old;                                                           \
+})
+
+#define atomic_exchange_explicit(object, desired, order) \
+    atomic_exchange(object, desired)
+
+#define atomic_compare_exchange_strong(object, expected, desired)   \
+({                                                                  \
+    __typeof__(object) _exp = (expected);                           \
+    __typeof__(*object) _old = *_exp;                               \
+    *_exp = __sync_val_compare_and_swap((object), _old, (desired)); \
+    *_exp == _old;                                                  \
+})
+
+#define atomic_compare_exchange_strong_explicit(object, expected, desired, success, failure) \
+    atomic_compare_exchange_strong(object, expected, desired)
+
+#define atomic_compare_exchange_weak(object, expected, desired) \
+    atomic_compare_exchange_strong(object, expected, desired)
+
+#define atomic_compare_exchange_weak_explicit(object, expected, desired, success, failure) \
+    atomic_compare_exchange_weak(object, expected, desired)
+
+#define atomic_fetch_add(object, operand) \
+    __sync_fetch_and_add(object, operand)
+
+#define atomic_fetch_add_explicit(object, operand, order) \
+    atomic_fetch_add(object, operand)
+
+#define atomic_fetch_sub(object, operand) \
+    __sync_fetch_and_sub(object, operand)
+
+#define atomic_fetch_sub_explicit(object, operand, order) \
+    atomic_fetch_sub(object, operand)
+
+#define atomic_fetch_or(object, operand) \
+    __sync_fetch_and_or(object, operand)
+
+#define atomic_fetch_or_explicit(object, operand, order) \
+    atomic_fetch_or(object, operand)
+
+#define atomic_fetch_xor(object, operand) \
+    __sync_fetch_and_xor(object, operand)
+
+#define atomic_fetch_xor_explicit(object, operand, order) \
+    atomic_fetch_xor(object, operand)
+
+#define atomic_fetch_and(object, operand) \
+    __sync_fetch_and_and(object, operand)
+
+#define atomic_fetch_and_explicit(object, operand, order) \
+    atomic_fetch_and(object, operand)
+
+#define atomic_flag_test_and_set(object) \
+    atomic_exchange(object, 1)
+
+#define atomic_flag_test_and_set_explicit(object, order) \
+    atomic_flag_test_and_set(object)
+
+#define atomic_flag_clear(object) \
+    atomic_store(object, 0)
+
+#define atomic_flag_clear_explicit(object, order) \
+    atomic_flag_clear(object)
+
+#endif /* COMPAT_ATOMICS_GCC_STDATOMIC_H */

trunk/3rdparty/ffmpeg-4.2-fit/compat/va_copy.h

@@ -0,0 +1,34 @@
+/*
+ * MSVC Compatible va_copy macro
+ * Copyright (c) 2012 Derek Buitenhuis
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#ifndef COMPAT_VA_COPY_H
+#define COMPAT_VA_COPY_H
+
+#include <stdarg.h>
+
+#if !defined(va_copy) && defined(_MSC_VER)
+#define va_copy(dst, src) ((dst) = (src))
+#endif
+#if !defined(va_copy) && defined(__GNUC__) && __GNUC__ < 3
+#define va_copy(dst, src) __va_copy(dst, src)
+#endif
+
+#endif /* COMPAT_VA_COPY_H */

trunk/3rdparty/ffmpeg-4.2-fit/ffbuild/arch.mak

@@ -0,0 +1,17 @@
+OBJS-$(HAVE_ARMV5TE) += $(ARMV5TE-OBJS) $(ARMV5TE-OBJS-yes)
+OBJS-$(HAVE_ARMV6)   += $(ARMV6-OBJS)   $(ARMV6-OBJS-yes)
+OBJS-$(HAVE_ARMV8)   += $(ARMV8-OBJS)   $(ARMV8-OBJS-yes)
+OBJS-$(HAVE_VFP)     += $(VFP-OBJS)     $(VFP-OBJS-yes)
+OBJS-$(HAVE_NEON)    += $(NEON-OBJS)    $(NEON-OBJS-yes)
+
+OBJS-$(HAVE_MIPSFPU)   += $(MIPSFPU-OBJS)    $(MIPSFPU-OBJS-yes)
+OBJS-$(HAVE_MIPSDSP)   += $(MIPSDSP-OBJS)    $(MIPSDSP-OBJS-yes)
+OBJS-$(HAVE_MIPSDSPR2) += $(MIPSDSPR2-OBJS)  $(MIPSDSPR2-OBJS-yes)
+OBJS-$(HAVE_MSA)       += $(MSA-OBJS)        $(MSA-OBJS-yes)
+OBJS-$(HAVE_MMI)   += $(MMI-OBJS)   $(MMI-OBJS-yes)
+
+OBJS-$(HAVE_ALTIVEC) += $(ALTIVEC-OBJS) $(ALTIVEC-OBJS-yes)
+OBJS-$(HAVE_VSX)     += $(VSX-OBJS) $(VSX-OBJS-yes)
+
+OBJS-$(HAVE_MMX)     += $(MMX-OBJS)     $(MMX-OBJS-yes)
+OBJS-$(HAVE_X86ASM)  += $(X86ASM-OBJS)  $(X86ASM-OBJS-yes)

trunk/3rdparty/ffmpeg-4.2-fit/ffbuild/common.mak

@@ -0,0 +1,175 @@
+#
+# common bits used by all libraries
+#
+
+DEFAULT_X86ASMD=.dbg
+
+ifeq ($(DBG),1)
+X86ASMD=$(DEFAULT_X86ASMD)
+else
+X86ASMD=
+endif
+
+ifndef SUBDIR
+
+ifndef V
+Q      = @
+ECHO   = printf "$(1)\t%s\n" $(2)
+BRIEF  = CC CXX OBJCC HOSTCC HOSTLD AS X86ASM AR LD STRIP CP WINDRES NVCC
+SILENT = DEPCC DEPHOSTCC DEPAS DEPX86ASM RANLIB RM
+
+MSG    = $@
+M      = @$(call ECHO,$(TAG),$@);
+$(foreach VAR,$(BRIEF), \
+    $(eval override $(VAR) = @$$(call ECHO,$(VAR),$$(MSG)); $($(VAR))))
+$(foreach VAR,$(SILENT),$(eval override $(VAR) = @$($(VAR))))
+$(eval INSTALL = @$(call ECHO,INSTALL,$$(^:$(SRC_DIR)/%=%)); $(INSTALL))
+endif
+
+ALLFFLIBS = avcodec avdevice avfilter avformat avresample avutil postproc swscale swresample
+
+# NASM requires -I path terminated with /
+IFLAGS     := -I. -I$(SRC_LINK)/
+CPPFLAGS   := $(IFLAGS) $(CPPFLAGS)
+CFLAGS     += $(ECFLAGS)
+CCFLAGS     = $(CPPFLAGS) $(CFLAGS)
+OBJCFLAGS  += $(EOBJCFLAGS)
+OBJCCFLAGS  = $(CPPFLAGS) $(CFLAGS) $(OBJCFLAGS)
+ASFLAGS    := $(CPPFLAGS) $(ASFLAGS)
+CXXFLAGS   := $(CPPFLAGS) $(CFLAGS) $(CXXFLAGS)
+X86ASMFLAGS += $(IFLAGS:%=%/) -I$(<D)/ -Pconfig.asm
+
+HOSTCCFLAGS = $(IFLAGS) $(HOSTCPPFLAGS) $(HOSTCFLAGS)
+LDFLAGS    := $(ALLFFLIBS:%=$(LD_PATH)lib%) $(LDFLAGS)
+
+define COMPILE
+       $(call $(1)DEP,$(1))
+       $($(1)) $($(1)FLAGS) $($(1)_DEPFLAGS) $($(1)_C) $($(1)_O) $(patsubst $(SRC_PATH)/%,$(SRC_LINK)/%,$<)
+endef
+
+COMPILE_C = $(call COMPILE,CC)
+COMPILE_CXX = $(call COMPILE,CXX)
+COMPILE_S = $(call COMPILE,AS)
+COMPILE_M = $(call COMPILE,OBJCC)
+COMPILE_X86ASM = $(call COMPILE,X86ASM)
+COMPILE_HOSTC = $(call COMPILE,HOSTCC)
+COMPILE_NVCC = $(call COMPILE,NVCC)
+
+%.o: %.c
+	$(COMPILE_C)
+
+%.o: %.cpp
+	$(COMPILE_CXX)
+
+%.o: %.m
+	$(COMPILE_M)
+
+%.s: %.c
+	$(CC) $(CCFLAGS) -S -o $@ $<
+
+%.o: %.S
+	$(COMPILE_S)
+
+%_host.o: %.c
+	$(COMPILE_HOSTC)
+
+%$(DEFAULT_X86ASMD).asm: %.asm
+	$(DEPX86ASM) $(X86ASMFLAGS) -M -o $@ $< > $(@:.asm=.d)
+	$(X86ASM) $(X86ASMFLAGS) -e $< | sed '/^%/d;/^$$/d;' > $@
+
+%.o: %.asm
+	$(COMPILE_X86ASM)
+	-$(if $(ASMSTRIPFLAGS), $(STRIP) $(ASMSTRIPFLAGS) $@)
+
+%.o: %.rc
+	$(WINDRES) $(IFLAGS) --preprocessor "$(DEPWINDRES) -E -xc-header -DRC_INVOKED $(CC_DEPFLAGS)" -o $@ $<
+
+%.i: %.c
+	$(CC) $(CCFLAGS) $(CC_E) $<
+
+%.h.c:
+	$(Q)echo '#include "$*.h"' >$@
+
+%.ptx: %.cu $(SRC_PATH)/compat/cuda/cuda_runtime.h
+	$(COMPILE_NVCC)
+
+%.ptx.c: %.ptx
+	$(Q)sh $(SRC_PATH)/compat/cuda/ptx2c.sh $@ $(patsubst $(SRC_PATH)/%,$(SRC_LINK)/%,$<)
+
+%.c %.h %.pc %.ver %.version: TAG = GEN
+
+# Dummy rule to stop make trying to rebuild removed or renamed headers
+%.h:
+	@:
+
+# Disable suffix rules.  Most of the builtin rules are suffix rules,
+# so this saves some time on slow systems.
+.SUFFIXES:
+
+# Do not delete intermediate files from chains of implicit rules
+$(OBJS):
+endif
+
+include $(SRC_PATH)/ffbuild/arch.mak
+
+OBJS      += $(OBJS-yes)
+SLIBOBJS  += $(SLIBOBJS-yes)
+FFLIBS    := $($(NAME)_FFLIBS) $(FFLIBS-yes) $(FFLIBS)
+TESTPROGS += $(TESTPROGS-yes)
+
+LDLIBS       = $(FFLIBS:%=%$(BUILDSUF))
+FFEXTRALIBS := $(LDLIBS:%=$(LD_LIB)) $(foreach lib,EXTRALIBS-$(NAME) $(FFLIBS:%=EXTRALIBS-%),$($(lib))) $(EXTRALIBS)
+
+OBJS      := $(sort $(OBJS:%=$(SUBDIR)%))
+SLIBOBJS  := $(sort $(SLIBOBJS:%=$(SUBDIR)%))
+TESTOBJS  := $(TESTOBJS:%=$(SUBDIR)tests/%) $(TESTPROGS:%=$(SUBDIR)tests/%.o)
+TESTPROGS := $(TESTPROGS:%=$(SUBDIR)tests/%$(EXESUF))
+HOSTOBJS  := $(HOSTPROGS:%=$(SUBDIR)%.o)
+HOSTPROGS := $(HOSTPROGS:%=$(SUBDIR)%$(HOSTEXESUF))
+TOOLS     += $(TOOLS-yes)
+TOOLOBJS  := $(TOOLS:%=tools/%.o)
+TOOLS     := $(TOOLS:%=tools/%$(EXESUF))
+HEADERS   += $(HEADERS-yes)
+
+PATH_LIBNAME = $(foreach NAME,$(1),lib$(NAME)/$($(2)LIBNAME))
+DEP_LIBS := $(foreach lib,$(FFLIBS),$(call PATH_LIBNAME,$(lib),$(CONFIG_SHARED:yes=S)))
+STATIC_DEP_LIBS := $(foreach lib,$(FFLIBS),$(call PATH_LIBNAME,$(lib)))
+
+SRC_DIR    := $(SRC_PATH)/lib$(NAME)
+ALLHEADERS := $(subst $(SRC_DIR)/,$(SUBDIR),$(wildcard $(SRC_DIR)/*.h $(SRC_DIR)/$(ARCH)/*.h))
+SKIPHEADERS += $(ARCH_HEADERS:%=$(ARCH)/%) $(SKIPHEADERS-)
+SKIPHEADERS := $(SKIPHEADERS:%=$(SUBDIR)%)
+HOBJS        = $(filter-out $(SKIPHEADERS:.h=.h.o),$(ALLHEADERS:.h=.h.o))
+PTXOBJS      = $(filter %.ptx.o,$(OBJS))
+$(HOBJS):     CCFLAGS += $(CFLAGS_HEADERS)
+checkheaders: $(HOBJS)
+.SECONDARY:   $(HOBJS:.o=.c) $(PTXOBJS:.o=.c) $(PTXOBJS:.o=)
+
+alltools: $(TOOLS)
+
+$(HOSTOBJS): %.o: %.c
+	$(COMPILE_HOSTC)
+
+$(HOSTPROGS): %$(HOSTEXESUF): %.o
+	$(HOSTLD) $(HOSTLDFLAGS) $(HOSTLD_O) $^ $(HOSTEXTRALIBS)
+
+$(OBJS):     | $(sort $(dir $(OBJS)))
+$(HOBJS):    | $(sort $(dir $(HOBJS)))
+$(HOSTOBJS): | $(sort $(dir $(HOSTOBJS)))
+$(SLIBOBJS): | $(sort $(dir $(SLIBOBJS)))
+$(TESTOBJS): | $(sort $(dir $(TESTOBJS)))
+$(TOOLOBJS): | tools
+
+OUTDIRS := $(OUTDIRS) $(dir $(OBJS) $(HOBJS) $(HOSTOBJS) $(SLIBOBJS) $(TESTOBJS))
+
+CLEANSUFFIXES     = *.d *.gcda *.gcno *.h.c *.ho *.map *.o *.pc *.ptx *.ptx.c *.ver *.version *$(DEFAULT_X86ASMD).asm *~
+LIBSUFFIXES       = *.a *.lib *.so *.so.* *.dylib *.dll *.def *.dll.a
+
+define RULES
+clean::
+	$(RM) $(HOSTPROGS) $(TESTPROGS) $(TOOLS)
+endef
+
+$(eval $(RULES))
+
+-include $(wildcard $(OBJS:.o=.d) $(HOSTOBJS:.o=.d) $(TESTOBJS:.o=.d) $(HOBJS:.o=.d) $(SLIBOBJS:.o=.d)) $(OBJS:.o=$(DEFAULT_X86ASMD).d)

trunk/3rdparty/ffmpeg-4.2-fit/ffbuild/library.mak

@@ -0,0 +1,107 @@
+include $(SRC_PATH)/ffbuild/common.mak
+
+ifeq (,$(filter %clean config,$(MAKECMDGOALS)))
+-include $(SUBDIR)lib$(NAME).version
+endif
+
+LIBVERSION := $(lib$(NAME)_VERSION)
+LIBMAJOR   := $(lib$(NAME)_VERSION_MAJOR)
+LIBMINOR   := $(lib$(NAME)_VERSION_MINOR)
+INCINSTDIR := $(INCDIR)/lib$(NAME)
+
+INSTHEADERS := $(INSTHEADERS) $(HEADERS:%=$(SUBDIR)%)
+
+all-$(CONFIG_STATIC): $(SUBDIR)$(LIBNAME)  $(SUBDIR)lib$(FULLNAME).pc
+all-$(CONFIG_SHARED): $(SUBDIR)$(SLIBNAME) $(SUBDIR)lib$(FULLNAME).pc
+
+LIBOBJS := $(OBJS) $(SUBDIR)%.h.o $(TESTOBJS)
+$(LIBOBJS) $(LIBOBJS:.o=.s) $(LIBOBJS:.o=.i):   CPPFLAGS += -DHAVE_AV_CONFIG_H
+
+$(SUBDIR)$(LIBNAME): $(OBJS)
+	$(RM) $@
+	$(AR) $(ARFLAGS) $(AR_O) $^
+	$(RANLIB) $@
+
+install-headers: install-lib$(NAME)-headers install-lib$(NAME)-pkgconfig
+
+install-libs-$(CONFIG_STATIC): install-lib$(NAME)-static
+install-libs-$(CONFIG_SHARED): install-lib$(NAME)-shared
+
+define RULES
+$(TOOLS):     THISLIB = $(FULLNAME:%=$(LD_LIB))
+$(TESTPROGS): THISLIB = $(SUBDIR)$(LIBNAME)
+
+$(LIBOBJS): CPPFLAGS += -DBUILDING_$(NAME)
+
+$(TESTPROGS) $(TOOLS): %$(EXESUF): %.o
+	$$(LD) $(LDFLAGS) $(LDEXEFLAGS) $$(LD_O) $$(filter %.o,$$^) $$(THISLIB) $(FFEXTRALIBS) $$(EXTRALIBS-$$(*F)) $$(ELIBS)
+
+$(SUBDIR)lib$(NAME).version: $(SUBDIR)version.h | $(SUBDIR)
+	$$(M) $$(SRC_PATH)/ffbuild/libversion.sh $(NAME) $$< > $$@
+
+$(SUBDIR)lib$(FULLNAME).pc: $(SUBDIR)version.h ffbuild/config.sh | $(SUBDIR)
+	$$(M) $$(SRC_PATH)/ffbuild/pkgconfig_generate.sh $(NAME) "$(DESC)"
+
+$(SUBDIR)lib$(NAME).ver: $(SUBDIR)lib$(NAME).v $(OBJS)
+	$$(M)sed 's/MAJOR/$(lib$(NAME)_VERSION_MAJOR)/' $$< | $(VERSION_SCRIPT_POSTPROCESS_CMD) > $$@
+
+$(SUBDIR)$(SLIBNAME): $(SUBDIR)$(SLIBNAME_WITH_MAJOR)
+	$(Q)cd ./$(SUBDIR) && $(LN_S) $(SLIBNAME_WITH_MAJOR) $(SLIBNAME)
+
+$(SUBDIR)$(SLIBNAME_WITH_MAJOR): $(OBJS) $(SLIBOBJS) $(SUBDIR)lib$(NAME).ver
+	$(SLIB_CREATE_DEF_CMD)
+	$$(LD) $(SHFLAGS) $(LDFLAGS) $(LDSOFLAGS) $$(LD_O) $$(filter %.o,$$^) $(FFEXTRALIBS)
+	$(SLIB_EXTRA_CMD)
+
+ifdef SUBDIR
+$(SUBDIR)$(SLIBNAME_WITH_MAJOR): $(DEP_LIBS)
+endif
+
+clean::
+	$(RM) $(addprefix $(SUBDIR),$(CLEANFILES) $(CLEANSUFFIXES) $(LIBSUFFIXES)) \
+	    $(CLEANSUFFIXES:%=$(SUBDIR)$(ARCH)/%) $(CLEANSUFFIXES:%=$(SUBDIR)tests/%)
+
+install-lib$(NAME)-shared: $(SUBDIR)$(SLIBNAME)
+	$(Q)mkdir -p "$(SHLIBDIR)"
+	$$(INSTALL) -m 755 $$< "$(SHLIBDIR)/$(SLIB_INSTALL_NAME)"
+	$$(STRIP) "$(SHLIBDIR)/$(SLIB_INSTALL_NAME)"
+	$(Q)$(foreach F,$(SLIB_INSTALL_LINKS),(cd "$(SHLIBDIR)" && $(LN_S) $(SLIB_INSTALL_NAME) $(F));)
+	$(if $(SLIB_INSTALL_EXTRA_SHLIB),$$(INSTALL) -m 644 $(SLIB_INSTALL_EXTRA_SHLIB:%=$(SUBDIR)%) "$(SHLIBDIR)")
+	$(if $(SLIB_INSTALL_EXTRA_LIB),$(Q)mkdir -p "$(LIBDIR)")
+	$(if $(SLIB_INSTALL_EXTRA_LIB),$$(INSTALL) -m 644 $(SLIB_INSTALL_EXTRA_LIB:%=$(SUBDIR)%) "$(LIBDIR)")
+
+install-lib$(NAME)-static: $(SUBDIR)$(LIBNAME)
+	$(Q)mkdir -p "$(LIBDIR)"
+	$$(INSTALL) -m 644 $$< "$(LIBDIR)"
+	$(LIB_INSTALL_EXTRA_CMD)
+
+install-lib$(NAME)-headers: $(addprefix $(SUBDIR),$(HEADERS) $(BUILT_HEADERS))
+	$(Q)mkdir -p "$(INCINSTDIR)"
+	$$(INSTALL) -m 644 $$^ "$(INCINSTDIR)"
+
+install-lib$(NAME)-pkgconfig: $(SUBDIR)lib$(FULLNAME).pc
+	$(Q)mkdir -p "$(PKGCONFIGDIR)"
+	$$(INSTALL) -m 644 $$^ "$(PKGCONFIGDIR)"
+
+uninstall-libs::
+	-$(RM) "$(SHLIBDIR)/$(SLIBNAME_WITH_MAJOR)" \
+	       "$(SHLIBDIR)/$(SLIBNAME)"            \
+	       "$(SHLIBDIR)/$(SLIBNAME_WITH_VERSION)"
+	-$(RM)  $(SLIB_INSTALL_EXTRA_SHLIB:%="$(SHLIBDIR)/%")
+	-$(RM)  $(SLIB_INSTALL_EXTRA_LIB:%="$(LIBDIR)/%")
+	-$(RM) "$(LIBDIR)/$(LIBNAME)"
+
+uninstall-headers::
+	$(RM) $(addprefix "$(INCINSTDIR)/",$(HEADERS) $(BUILT_HEADERS))
+	-rmdir "$(INCINSTDIR)"
+
+uninstall-pkgconfig::
+	$(RM) "$(PKGCONFIGDIR)/lib$(FULLNAME).pc"
+endef
+
+$(eval $(RULES))
+
+$(TOOLS):     $(DEP_LIBS) $(SUBDIR)$($(CONFIG_SHARED:yes=S)LIBNAME)
+$(TESTPROGS): $(DEP_LIBS) $(SUBDIR)$(LIBNAME)
+
+testprogs: $(TESTPROGS)

trunk/3rdparty/ffmpeg-4.2-fit/ffbuild/libversion.sh

@@ -0,0 +1,13 @@
+toupper(){
+    echo "$@" | tr abcdefghijklmnopqrstuvwxyz ABCDEFGHIJKLMNOPQRSTUVWXYZ
+}
+
+name=lib$1
+ucname=$(toupper ${name})
+file=$2
+
+eval $(awk "/#define ${ucname}_VERSION_M/ { print \$2 \"=\" \$3 }" "$file")
+eval ${ucname}_VERSION=\$${ucname}_VERSION_MAJOR.\$${ucname}_VERSION_MINOR.\$${ucname}_VERSION_MICRO
+eval echo "${name}_VERSION=\$${ucname}_VERSION"
+eval echo "${name}_VERSION_MAJOR=\$${ucname}_VERSION_MAJOR"
+eval echo "${name}_VERSION_MINOR=\$${ucname}_VERSION_MINOR"

trunk/3rdparty/ffmpeg-4.2-fit/ffbuild/pkgconfig_generate.sh

@@ -0,0 +1,62 @@
+#!/bin/sh
+
+. ffbuild/config.sh
+
+if test "$shared" = "yes"; then
+    shared=true
+else
+    shared=false
+fi
+
+shortname=$1
+name=lib${shortname}
+fullname=${name}${build_suffix}
+comment=$2
+libs=$(eval echo \$extralibs_${shortname})
+deps=$(eval echo \$${shortname}_deps)
+
+for dep in $deps; do
+    depname=lib${dep}
+    fulldepname=${depname}${build_suffix}
+    . ${depname}/${depname}.version
+    depversion=$(eval echo \$${depname}_VERSION)
+    requires="$requires ${fulldepname} >= ${depversion}, "
+done
+requires=${requires%, }
+
+version=$(grep ${name}_VERSION= $name/${name}.version | cut -d= -f2)
+
+cat <<EOF > $name/$fullname.pc
+prefix=$prefix
+exec_prefix=\${prefix}
+libdir=$libdir
+includedir=$incdir
+
+Name: $fullname
+Description: $comment
+Version: $version
+Requires: $($shared || echo $requires)
+Requires.private: $($shared && echo $requires)
+Conflicts:
+Libs: -L\${libdir} $rpath -l${fullname#lib} $($shared || echo $libs)
+Libs.private: $($shared && echo $libs)
+Cflags: -I\${includedir}
+EOF
+
+mkdir -p doc/examples/pc-uninstalled
+includedir=${source_path}
+[ "$includedir" = . ] && includedir="\${pcfiledir}/../../.."
+    cat <<EOF > doc/examples/pc-uninstalled/${name}-uninstalled.pc
+prefix=
+exec_prefix=
+libdir=\${pcfiledir}/../../../$name
+includedir=${source_path}
+
+Name: $fullname
+Description: $comment
+Version: $version
+Requires: $requires
+Conflicts:
+Libs: -L\${libdir} -Wl,-rpath,\${libdir} -l${fullname#lib} $($shared || echo $libs)
+Cflags: -I\${includedir}
+EOF

trunk/3rdparty/ffmpeg-4.2-fit/ffbuild/version.sh

@@ -0,0 +1,63 @@
+#!/bin/sh
+
+# Usage: version.sh <ffmpeg-root-dir> <output-version.h> <extra-version>
+
+# check for git short hash
+if ! test "$revision"; then
+    if (cd "$1" && grep git RELEASE 2> /dev/null >/dev/null) ; then
+        revision=$(cd "$1" && git describe --tags --match N 2> /dev/null)
+    else
+        revision=$(cd "$1" && git describe --tags --always 2> /dev/null)
+    fi
+fi
+
+# Shallow Git clones (--depth) do not have the N tag:
+# use 'git-YYYY-MM-DD-hhhhhhh'.
+test "$revision" || revision=$(cd "$1" &&
+  git log -1 --pretty=format:"git-%cd-%h" --date=short 2> /dev/null)
+
+# Snapshots from gitweb are in a directory called ffmpeg-hhhhhhh or
+# ffmpeg-HEAD-hhhhhhh.
+if [ -z "$revision" ]; then
+  srcdir=$(cd "$1" && pwd)
+  case "$srcdir" in
+    */ffmpeg-[0-9a-f][0-9a-f][0-9a-f][0-9a-f][0-9a-f][0-9a-f][0-9a-f])
+      git_hash="${srcdir##*-}";;
+    */ffmpeg-HEAD-[0-9a-f][0-9a-f][0-9a-f][0-9a-f][0-9a-f][0-9a-f][0-9a-f])
+      git_hash="${srcdir##*-}";;
+  esac
+fi
+
+# no revision number found
+test "$revision" || revision=$(cd "$1" && cat RELEASE 2> /dev/null)
+
+# Append the Git hash if we have one
+test "$revision" && test "$git_hash" && revision="$revision-$git_hash"
+
+# releases extract the version number from the VERSION file
+version=$(cd "$1" && cat VERSION 2> /dev/null)
+test "$version" || version=$revision
+
+test -n "$3" && version=$version-$3
+
+if [ -z "$2" ]; then
+    echo "$version"
+    exit
+fi
+
+NEW_REVISION="#define FFMPEG_VERSION \"$version\""
+OLD_REVISION=$(cat "$2" 2> /dev/null | head -4 | tail -1)
+
+# String used for preprocessor guard
+GUARD=$(echo "$2" | sed 's/\//_/' | sed 's/\./_/' | tr '[:lower:]' '[:upper:]' | sed 's/LIB//')
+
+# Update version header only on revision changes to avoid spurious rebuilds
+if test "$NEW_REVISION" != "$OLD_REVISION"; then
+    cat << EOF > "$2"
+/* Automatically generated by version.sh, do not manually edit! */
+#ifndef $GUARD
+#define $GUARD
+$NEW_REVISION
+#endif /* $GUARD */
+EOF
+fi

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aac.h

@@ -0,0 +1,377 @@
+/*
+ * AAC definitions and structures
+ * Copyright (c) 2005-2006 Oded Shimon ( ods15 ods15 dyndns org )
+ * Copyright (c) 2006-2007 Maxim Gavrilov ( maxim.gavrilov gmail com )
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+/**
+ * @file
+ * AAC definitions and structures
+ * @author Oded Shimon  ( ods15 ods15 dyndns org )
+ * @author Maxim Gavrilov ( maxim.gavrilov gmail com )
+ */
+
+#ifndef AVCODEC_AAC_H
+#define AVCODEC_AAC_H
+
+
+#include "aac_defines.h"
+#include "libavutil/float_dsp.h"
+#include "libavutil/fixed_dsp.h"
+#include "avcodec.h"
+#if !USE_FIXED
+#include "mdct15.h"
+#endif
+#include "fft.h"
+#include "mpeg4audio.h"
+#include "sbr.h"
+
+#include <stdint.h>
+
+#define MAX_CHANNELS 64
+#define MAX_ELEM_ID 16
+
+#define TNS_MAX_ORDER 20
+#define MAX_LTP_LONG_SFB 40
+
+#define CLIP_AVOIDANCE_FACTOR 0.95f
+
+enum RawDataBlockType {
+    TYPE_SCE,
+    TYPE_CPE,
+    TYPE_CCE,
+    TYPE_LFE,
+    TYPE_DSE,
+    TYPE_PCE,
+    TYPE_FIL,
+    TYPE_END,
+};
+
+enum ExtensionPayloadID {
+    EXT_FILL,
+    EXT_FILL_DATA,
+    EXT_DATA_ELEMENT,
+    EXT_DYNAMIC_RANGE = 0xb,
+    EXT_SBR_DATA      = 0xd,
+    EXT_SBR_DATA_CRC  = 0xe,
+};
+
+enum WindowSequence {
+    ONLY_LONG_SEQUENCE,
+    LONG_START_SEQUENCE,
+    EIGHT_SHORT_SEQUENCE,
+    LONG_STOP_SEQUENCE,
+};
+
+enum BandType {
+    ZERO_BT        = 0,     ///< Scalefactors and spectral data are all zero.
+    FIRST_PAIR_BT  = 5,     ///< This and later band types encode two values (rather than four) with one code word.
+    ESC_BT         = 11,    ///< Spectral data are coded with an escape sequence.
+    RESERVED_BT    = 12,    ///< Band types following are encoded differently from others.
+    NOISE_BT       = 13,    ///< Spectral data are scaled white noise not coded in the bitstream.
+    INTENSITY_BT2  = 14,    ///< Scalefactor data are intensity stereo positions (out of phase).
+    INTENSITY_BT   = 15,    ///< Scalefactor data are intensity stereo positions (in phase).
+};
+
+#define IS_CODEBOOK_UNSIGNED(x) (((x) - 1) & 10)
+
+enum ChannelPosition {
+    AAC_CHANNEL_OFF   = 0,
+    AAC_CHANNEL_FRONT = 1,
+    AAC_CHANNEL_SIDE  = 2,
+    AAC_CHANNEL_BACK  = 3,
+    AAC_CHANNEL_LFE   = 4,
+    AAC_CHANNEL_CC    = 5,
+};
+
+/**
+ * The point during decoding at which channel coupling is applied.
+ */
+enum CouplingPoint {
+    BEFORE_TNS,
+    BETWEEN_TNS_AND_IMDCT,
+    AFTER_IMDCT = 3,
+};
+
+/**
+ * Output configuration status
+ */
+enum OCStatus {
+    OC_NONE,        ///< Output unconfigured
+    OC_TRIAL_PCE,   ///< Output configuration under trial specified by an inband PCE
+    OC_TRIAL_FRAME, ///< Output configuration under trial specified by a frame header
+    OC_GLOBAL_HDR,  ///< Output configuration set in a global header but not yet locked
+    OC_LOCKED,      ///< Output configuration locked in place
+};
+
+typedef struct OutputConfiguration {
+    MPEG4AudioConfig m4ac;
+    uint8_t layout_map[MAX_ELEM_ID*4][3];
+    int layout_map_tags;
+    int channels;
+    uint64_t channel_layout;
+    enum OCStatus status;
+} OutputConfiguration;
+
+/**
+ * Predictor State
+ */
+typedef struct PredictorState {
+    AAC_FLOAT cor0;
+    AAC_FLOAT cor1;
+    AAC_FLOAT var0;
+    AAC_FLOAT var1;
+    AAC_FLOAT r0;
+    AAC_FLOAT r1;
+    AAC_FLOAT k1;
+    AAC_FLOAT x_est;
+} PredictorState;
+
+#define MAX_PREDICTORS 672
+
+#define SCALE_DIV_512    36    ///< scalefactor difference that corresponds to scale difference in 512 times
+#define SCALE_ONE_POS   140    ///< scalefactor index that corresponds to scale=1.0
+#define SCALE_MAX_POS   255    ///< scalefactor index maximum value
+#define SCALE_MAX_DIFF   60    ///< maximum scalefactor difference allowed by standard
+#define SCALE_DIFF_ZERO  60    ///< codebook index corresponding to zero scalefactor indices difference
+
+#define POW_SF2_ZERO    200    ///< ff_aac_pow2sf_tab index corresponding to pow(2, 0);
+
+#define NOISE_PRE       256    ///< preamble for NOISE_BT, put in bitstream with the first noise band
+#define NOISE_PRE_BITS    9    ///< length of preamble
+#define NOISE_OFFSET     90    ///< subtracted from global gain, used as offset for the preamble
+
+/**
+ * Long Term Prediction
+ */
+typedef struct LongTermPrediction {
+    int8_t present;
+    int16_t lag;
+    int coef_idx;
+    INTFLOAT coef;
+    int8_t used[MAX_LTP_LONG_SFB];
+} LongTermPrediction;
+
+/**
+ * Individual Channel Stream
+ */
+typedef struct IndividualChannelStream {
+    uint8_t max_sfb;            ///< number of scalefactor bands per group
+    enum WindowSequence window_sequence[2];
+    uint8_t use_kb_window[2];   ///< If set, use Kaiser-Bessel window, otherwise use a sine window.
+    int num_window_groups;
+    uint8_t group_len[8];
+    LongTermPrediction ltp;
+    const uint16_t *swb_offset; ///< table of offsets to the lowest spectral coefficient of a scalefactor band, sfb, for a particular window
+    const uint8_t *swb_sizes;   ///< table of scalefactor band sizes for a particular window
+    int num_swb;                ///< number of scalefactor window bands
+    int num_windows;
+    int tns_max_bands;
+    int predictor_present;
+    int predictor_initialized;
+    int predictor_reset_group;
+    int predictor_reset_count[31];  ///< used by encoder to count prediction resets
+    uint8_t prediction_used[41];
+    uint8_t window_clipping[8]; ///< set if a certain window is near clipping
+    float clip_avoidance_factor; ///< set if any window is near clipping to the necessary atennuation factor to avoid it
+} IndividualChannelStream;
+
+/**
+ * Temporal Noise Shaping
+ */
+typedef struct TemporalNoiseShaping {
+    int present;
+    int n_filt[8];
+    int length[8][4];
+    int direction[8][4];
+    int order[8][4];
+    int coef_idx[8][4][TNS_MAX_ORDER];
+    INTFLOAT coef[8][4][TNS_MAX_ORDER];
+} TemporalNoiseShaping;
+
+/**
+ * Dynamic Range Control - decoded from the bitstream but not processed further.
+ */
+typedef struct DynamicRangeControl {
+    int pce_instance_tag;                           ///< Indicates with which program the DRC info is associated.
+    int dyn_rng_sgn[17];                            ///< DRC sign information; 0 - positive, 1 - negative
+    int dyn_rng_ctl[17];                            ///< DRC magnitude information
+    int exclude_mask[MAX_CHANNELS];                 ///< Channels to be excluded from DRC processing.
+    int band_incr;                                  ///< Number of DRC bands greater than 1 having DRC info.
+    int interpolation_scheme;                       ///< Indicates the interpolation scheme used in the SBR QMF domain.
+    int band_top[17];                               ///< Indicates the top of the i-th DRC band in units of 4 spectral lines.
+    int prog_ref_level;                             /**< A reference level for the long-term program audio level for all
+                                                     *   channels combined.
+                                                     */
+} DynamicRangeControl;
+
+typedef struct Pulse {
+    int num_pulse;
+    int start;
+    int pos[4];
+    int amp[4];
+} Pulse;
+
+/**
+ * coupling parameters
+ */
+typedef struct ChannelCoupling {
+    enum CouplingPoint coupling_point;  ///< The point during decoding at which coupling is applied.
+    int num_coupled;       ///< number of target elements
+    enum RawDataBlockType type[8];   ///< Type of channel element to be coupled - SCE or CPE.
+    int id_select[8];      ///< element id
+    int ch_select[8];      /**< [0] shared list of gains; [1] list of gains for right channel;
+                            *   [2] list of gains for left channel; [3] lists of gains for both channels
+                            */
+    INTFLOAT gain[16][120];
+} ChannelCoupling;
+
+/**
+ * Single Channel Element - used for both SCE and LFE elements.
+ */
+typedef struct SingleChannelElement {
+    IndividualChannelStream ics;
+    TemporalNoiseShaping tns;
+    Pulse pulse;
+    enum BandType band_type[128];                   ///< band types
+    enum BandType band_alt[128];                    ///< alternative band type (used by encoder)
+    int band_type_run_end[120];                     ///< band type run end points
+    INTFLOAT sf[120];                               ///< scalefactors
+    int sf_idx[128];                                ///< scalefactor indices (used by encoder)
+    uint8_t zeroes[128];                            ///< band is not coded (used by encoder)
+    uint8_t can_pns[128];                           ///< band is allowed to PNS (informative)
+    float  is_ener[128];                            ///< Intensity stereo pos (used by encoder)
+    float pns_ener[128];                            ///< Noise energy values (used by encoder)
+    DECLARE_ALIGNED(32, INTFLOAT, pcoeffs)[1024];   ///< coefficients for IMDCT, pristine
+    DECLARE_ALIGNED(32, INTFLOAT, coeffs)[1024];    ///< coefficients for IMDCT, maybe processed
+    DECLARE_ALIGNED(32, INTFLOAT, saved)[1536];     ///< overlap
+    DECLARE_ALIGNED(32, INTFLOAT, ret_buf)[2048];   ///< PCM output buffer
+    DECLARE_ALIGNED(16, INTFLOAT, ltp_state)[3072]; ///< time signal for LTP
+    DECLARE_ALIGNED(32, AAC_FLOAT, lcoeffs)[1024];  ///< MDCT of LTP coefficients (used by encoder)
+    DECLARE_ALIGNED(32, AAC_FLOAT, prcoeffs)[1024]; ///< Main prediction coefs (used by encoder)
+    PredictorState predictor_state[MAX_PREDICTORS];
+    INTFLOAT *ret;                                  ///< PCM output
+} SingleChannelElement;
+
+/**
+ * channel element - generic struct for SCE/CPE/CCE/LFE
+ */
+typedef struct ChannelElement {
+    int present;
+    // CPE specific
+    int common_window;        ///< Set if channels share a common 'IndividualChannelStream' in bitstream.
+    int     ms_mode;          ///< Signals mid/side stereo flags coding mode (used by encoder)
+    uint8_t is_mode;          ///< Set if any bands have been encoded using intensity stereo (used by encoder)
+    uint8_t ms_mask[128];     ///< Set if mid/side stereo is used for each scalefactor window band
+    uint8_t is_mask[128];     ///< Set if intensity stereo is used (used by encoder)
+    // shared
+    SingleChannelElement ch[2];
+    // CCE specific
+    ChannelCoupling coup;
+    SpectralBandReplication sbr;
+} ChannelElement;
+
+/**
+ * main AAC context
+ */
+struct AACContext {
+    AVClass        *class;
+    AVCodecContext *avctx;
+    AVFrame *frame;
+
+    int is_saved;                 ///< Set if elements have stored overlap from previous frame.
+    DynamicRangeControl che_drc;
+
+    /**
+     * @name Channel element related data
+     * @{
+     */
+    ChannelElement          *che[4][MAX_ELEM_ID];
+    ChannelElement  *tag_che_map[4][MAX_ELEM_ID];
+    int tags_mapped;
+    int warned_remapping_once;
+    /** @} */
+
+    /**
+     * @name temporary aligned temporary buffers
+     * (We do not want to have these on the stack.)
+     * @{
+     */
+    DECLARE_ALIGNED(32, INTFLOAT, buf_mdct)[1024];
+    /** @} */
+
+    /**
+     * @name Computed / set up during initialization
+     * @{
+     */
+    FFTContext mdct;
+    FFTContext mdct_small;
+    FFTContext mdct_ld;
+    FFTContext mdct_ltp;
+#if USE_FIXED
+    AVFixedDSPContext *fdsp;
+#else
+    MDCT15Context *mdct120;
+    MDCT15Context *mdct480;
+    MDCT15Context *mdct960;
+    AVFloatDSPContext *fdsp;
+#endif /* USE_FIXED */
+    int random_state;
+    /** @} */
+
+    /**
+     * @name Members used for output
+     * @{
+     */
+    SingleChannelElement *output_element[MAX_CHANNELS]; ///< Points to each SingleChannelElement
+    /** @} */
+
+
+    /**
+     * @name Japanese DTV specific extension
+     * @{
+     */
+    int force_dmono_mode;///< 0->not dmono, 1->use first channel, 2->use second channel
+    int dmono_mode;      ///< 0->not dmono, 1->use first channel, 2->use second channel
+    /** @} */
+
+    DECLARE_ALIGNED(32, INTFLOAT, temp)[128];
+
+    OutputConfiguration oc[2];
+    int warned_num_aac_frames;
+    int warned_960_sbr;
+
+    int warned_gain_control;
+
+    /* aacdec functions pointers */
+    void (*imdct_and_windowing)(AACContext *ac, SingleChannelElement *sce);
+    void (*apply_ltp)(AACContext *ac, SingleChannelElement *sce);
+    void (*apply_tns)(INTFLOAT coef[1024], TemporalNoiseShaping *tns,
+                      IndividualChannelStream *ics, int decode);
+    void (*windowing_and_mdct_ltp)(AACContext *ac, INTFLOAT *out,
+                                   INTFLOAT *in, IndividualChannelStream *ics);
+    void (*update_ltp)(AACContext *ac, SingleChannelElement *sce);
+    void (*vector_pow43)(int *coefs, int len);
+    void (*subband_scale)(int *dst, int *src, int scale, int offset, int len, void *log_context);
+
+};
+
+void ff_aacdec_init_mips(AACContext *c);
+
+#endif /* AVCODEC_AAC_H */

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aac_ac3_parser.c

@@ -0,0 +1,105 @@
+/*
+ * Common AAC and AC-3 parser
+ * Copyright (c) 2003 Fabrice Bellard
+ * Copyright (c) 2003 Michael Niedermayer
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include "libavutil/channel_layout.h"
+#include "libavutil/common.h"
+#include "parser.h"
+#include "aac_ac3_parser.h"
+
+int ff_aac_ac3_parse(AVCodecParserContext *s1,
+                     AVCodecContext *avctx,
+                     const uint8_t **poutbuf, int *poutbuf_size,
+                     const uint8_t *buf, int buf_size)
+{
+    AACAC3ParseContext *s = s1->priv_data;
+    ParseContext *pc = &s->pc;
+    int len, i;
+    int new_frame_start;
+    int got_frame = 0;
+
+get_next:
+    i=END_NOT_FOUND;
+    if(s->remaining_size <= buf_size){
+        if(s->remaining_size && !s->need_next_header){
+            i= s->remaining_size;
+            s->remaining_size = 0;
+        }else{ //we need a header first
+            len=0;
+            for(i=s->remaining_size; i<buf_size; i++){
+                s->state = (s->state<<8) + buf[i];
+                if((len=s->sync(s->state, s, &s->need_next_header, &new_frame_start)))
+                    break;
+            }
+            if(len<=0){
+                i=END_NOT_FOUND;
+            }else{
+                got_frame = 1;
+                s->state=0;
+                i-= s->header_size -1;
+                s->remaining_size = len;
+                if(!new_frame_start || pc->index+i<=0){
+                    s->remaining_size += i;
+                    goto get_next;
+                }
+                else if (i < 0) {
+                    s->remaining_size += i;
+                }
+            }
+        }
+    }
+
+    if(ff_combine_frame(pc, i, &buf, &buf_size)<0){
+        s->remaining_size -= FFMIN(s->remaining_size, buf_size);
+        *poutbuf = NULL;
+        *poutbuf_size = 0;
+        return buf_size;
+    }
+
+    *poutbuf = buf;
+    *poutbuf_size = buf_size;
+
+    /* update codec info */
+    if(s->codec_id)
+        avctx->codec_id = s->codec_id;
+
+    if (got_frame) {
+        /* Due to backwards compatible HE-AAC the sample rate, channel count,
+           and total number of samples found in an AAC ADTS header are not
+           reliable. Bit rate is still accurate because the total frame
+           duration in seconds is still correct (as is the number of bits in
+           the frame). */
+        if (avctx->codec_id != AV_CODEC_ID_AAC) {
+            avctx->sample_rate = s->sample_rate;
+            if (avctx->codec_id != AV_CODEC_ID_EAC3) {
+                avctx->channels = s->channels;
+                avctx->channel_layout = s->channel_layout;
+            }
+            s1->duration = s->samples;
+            avctx->audio_service_type = s->service_type;
+        }
+
+        if (avctx->codec_id != AV_CODEC_ID_EAC3)
+            avctx->bit_rate = s->bit_rate;
+    }
+
+    return i;
+}

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aac_ac3_parser.h

@@ -0,0 +1,66 @@
+/*
+ * Common AAC and AC-3 parser prototypes
+ * Copyright (c) 2003 Fabrice Bellard
+ * Copyright (c) 2003 Michael Niedermayer
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#ifndef AVCODEC_AAC_AC3_PARSER_H
+#define AVCODEC_AAC_AC3_PARSER_H
+
+#include <stdint.h>
+#include "avcodec.h"
+#include "parser.h"
+
+typedef enum {
+    AAC_AC3_PARSE_ERROR_SYNC        = -0x1030c0a,
+    AAC_AC3_PARSE_ERROR_BSID        = -0x2030c0a,
+    AAC_AC3_PARSE_ERROR_SAMPLE_RATE = -0x3030c0a,
+    AAC_AC3_PARSE_ERROR_FRAME_SIZE  = -0x4030c0a,
+    AAC_AC3_PARSE_ERROR_FRAME_TYPE  = -0x5030c0a,
+    AAC_AC3_PARSE_ERROR_CRC         = -0x6030c0a,
+    AAC_AC3_PARSE_ERROR_CHANNEL_CFG = -0x7030c0a,
+} AACAC3ParseError;
+
+typedef struct AACAC3ParseContext {
+    ParseContext pc;
+    int frame_size;
+    int header_size;
+    int (*sync)(uint64_t state, struct AACAC3ParseContext *hdr_info,
+            int *need_next_header, int *new_frame_start);
+
+    int channels;
+    int sample_rate;
+    int bit_rate;
+    int samples;
+    uint64_t channel_layout;
+    int service_type;
+
+    int remaining_size;
+    uint64_t state;
+
+    int need_next_header;
+    enum AVCodecID codec_id;
+} AACAC3ParseContext;
+
+int ff_aac_ac3_parse(AVCodecParserContext *s1,
+                     AVCodecContext *avctx,
+                     const uint8_t **poutbuf, int *poutbuf_size,
+                     const uint8_t *buf, int buf_size);
+
+#endif /* AVCODEC_AAC_AC3_PARSER_H */

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aac_adtstoasc_bsf.c

@@ -0,0 +1,158 @@
+/*
+ * MPEG-2/4 AAC ADTS to MPEG-4 Audio Specific Configuration bitstream filter
+ * Copyright (c) 2009 Alex Converse <alex.converse@gmail.com>
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include "adts_header.h"
+#include "adts_parser.h"
+#include "avcodec.h"
+#include "bsf.h"
+#include "put_bits.h"
+#include "get_bits.h"
+#include "mpeg4audio.h"
+#include "internal.h"
+
+typedef struct AACBSFContext {
+    int first_frame_done;
+} AACBSFContext;
+
+/**
+ * This filter creates an MPEG-4 AudioSpecificConfig from an MPEG-2/4
+ * ADTS header and removes the ADTS header.
+ */
+static int aac_adtstoasc_filter(AVBSFContext *bsfc, AVPacket *pkt)
+{
+    AACBSFContext *ctx = bsfc->priv_data;
+
+    GetBitContext gb;
+    PutBitContext pb;
+    AACADTSHeaderInfo hdr;
+    int ret;
+
+    ret = ff_bsf_get_packet_ref(bsfc, pkt);
+    if (ret < 0)
+        return ret;
+
+    if (bsfc->par_in->extradata && pkt->size >= 2 && (AV_RB16(pkt->data) >> 4) != 0xfff)
+        return 0;
+
+    if (pkt->size < AV_AAC_ADTS_HEADER_SIZE)
+        goto packet_too_small;
+
+    init_get_bits(&gb, pkt->data, AV_AAC_ADTS_HEADER_SIZE * 8);
+
+    if (ff_adts_header_parse(&gb, &hdr) < 0) {
+        av_log(bsfc, AV_LOG_ERROR, "Error parsing ADTS frame header!\n");
+        ret = AVERROR_INVALIDDATA;
+        goto fail;
+    }
+
+    if (!hdr.crc_absent && hdr.num_aac_frames > 1) {
+        avpriv_report_missing_feature(bsfc,
+                                      "Multiple RDBs per frame with CRC");
+        ret = AVERROR_PATCHWELCOME;
+        goto fail;
+    }
+
+    pkt->size -= AV_AAC_ADTS_HEADER_SIZE + 2 * !hdr.crc_absent;
+    if (pkt->size <= 0)
+        goto packet_too_small;
+    pkt->data += AV_AAC_ADTS_HEADER_SIZE + 2 * !hdr.crc_absent;
+
+    if (!ctx->first_frame_done) {
+        int            pce_size = 0;
+        uint8_t        pce_data[MAX_PCE_SIZE];
+        uint8_t       *extradata;
+
+        if (!hdr.chan_config) {
+            init_get_bits(&gb, pkt->data, pkt->size * 8);
+            if (get_bits(&gb, 3) != 5) {
+                avpriv_report_missing_feature(bsfc,
+                                              "PCE-based channel configuration "
+                                              "without PCE as first syntax "
+                                              "element");
+                ret = AVERROR_PATCHWELCOME;
+                goto fail;
+            }
+            init_put_bits(&pb, pce_data, MAX_PCE_SIZE);
+            pce_size = ff_copy_pce_data(&pb, &gb) / 8;
+            flush_put_bits(&pb);
+            pkt->size -= get_bits_count(&gb)/8;
+            pkt->data += get_bits_count(&gb)/8;
+        }
+
+        extradata = av_packet_new_side_data(pkt, AV_PKT_DATA_NEW_EXTRADATA,
+                                            2 + pce_size);
+        if (!extradata) {
+            ret = AVERROR(ENOMEM);
+            goto fail;
+        }
+
+        init_put_bits(&pb, extradata, 2 + pce_size);
+        put_bits(&pb, 5, hdr.object_type);
+        put_bits(&pb, 4, hdr.sampling_index);
+        put_bits(&pb, 4, hdr.chan_config);
+        put_bits(&pb, 1, 0); //frame length - 1024 samples
+        put_bits(&pb, 1, 0); //does not depend on core coder
+        put_bits(&pb, 1, 0); //is not extension
+        flush_put_bits(&pb);
+        if (pce_size) {
+            memcpy(extradata + 2, pce_data, pce_size);
+        }
+
+        ctx->first_frame_done = 1;
+    }
+
+    return 0;
+
+packet_too_small:
+    av_log(bsfc, AV_LOG_ERROR, "Input packet too small\n");
+    ret = AVERROR_INVALIDDATA;
+fail:
+    av_packet_unref(pkt);
+    return ret;
+}
+
+static int aac_adtstoasc_init(AVBSFContext *ctx)
+{
+    /* Validate the extradata if the stream is already MPEG-4 AudioSpecificConfig */
+    if (ctx->par_in->extradata) {
+        MPEG4AudioConfig mp4ac;
+        int ret = avpriv_mpeg4audio_get_config(&mp4ac, ctx->par_in->extradata,
+                                               ctx->par_in->extradata_size * 8, 1);
+        if (ret < 0) {
+            av_log(ctx, AV_LOG_ERROR, "Error parsing AudioSpecificConfig extradata!\n");
+            return ret;
+        }
+    }
+
+    return 0;
+}
+
+static const enum AVCodecID codec_ids[] = {
+    AV_CODEC_ID_AAC, AV_CODEC_ID_NONE,
+};
+
+const AVBitStreamFilter ff_aac_adtstoasc_bsf = {
+    .name           = "aac_adtstoasc",
+    .priv_data_size = sizeof(AACBSFContext),
+    .init           = aac_adtstoasc_init,
+    .filter         = aac_adtstoasc_filter,
+    .codec_ids      = codec_ids,
+};

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aac_defines.h

@@ -0,0 +1,116 @@
+/*
+ * AAC defines
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#ifndef AVCODEC_AAC_DEFINES_H
+#define AVCODEC_AAC_DEFINES_H
+
+#ifndef USE_FIXED
+#define USE_FIXED 0
+#endif
+
+#if USE_FIXED
+
+#include "libavutil/softfloat.h"
+
+#define FFT_FLOAT    0
+#define FFT_FIXED_32 1
+
+#define AAC_RENAME(x)       x ## _fixed
+#define AAC_RENAME_32(x)    x ## _fixed_32
+typedef int                 INTFLOAT;
+typedef unsigned            UINTFLOAT;  ///< Equivalent to INTFLOAT, Used as temporal cast to avoid undefined sign overflow operations.
+typedef int64_t             INT64FLOAT;
+typedef int16_t             SHORTFLOAT;
+typedef SoftFloat           AAC_FLOAT;
+typedef int                 AAC_SIGNE;
+#define FIXR(a)             ((int)((a) * 1 + 0.5))
+#define FIXR10(a)           ((int)((a) * 1024.0 + 0.5))
+#define Q23(a)              (int)((a) * 8388608.0 + 0.5)
+#define Q30(x)              (int)((x)*1073741824.0 + 0.5)
+#define Q31(x)              (int)((x)*2147483648.0 + 0.5)
+#define RANGE15(x)          x
+#define GET_GAIN(x, y)      (-(y) * (1 << (x))) + 1024
+#define AAC_MUL16(x, y)     (int)(((int64_t)(x) * (y) + 0x8000) >> 16)
+#define AAC_MUL26(x, y)     (int)(((int64_t)(x) * (y) + 0x2000000) >> 26)
+#define AAC_MUL30(x, y)     (int)(((int64_t)(x) * (y) + 0x20000000) >> 30)
+#define AAC_MUL31(x, y)     (int)(((int64_t)(x) * (y) + 0x40000000) >> 31)
+#define AAC_MADD28(x, y, a, b) (int)((((int64_t)(x) * (y)) + \
+                                      ((int64_t)(a) * (b)) + \
+                                        0x8000000) >> 28)
+#define AAC_MADD30(x, y, a, b) (int)((((int64_t)(x) * (y)) + \
+                                      ((int64_t)(a) * (b)) + \
+                                        0x20000000) >> 30)
+#define AAC_MADD30_V8(x, y, a, b, c, d, e, f) (int)((((int64_t)(x) * (y)) + \
+                                                     ((int64_t)(a) * (b)) + \
+                                                     ((int64_t)(c) * (d)) + \
+                                                     ((int64_t)(e) * (f)) + \
+                                                       0x20000000) >> 30)
+#define AAC_MSUB30(x, y, a, b) (int)((((int64_t)(x) * (y)) - \
+                                      ((int64_t)(a) * (b)) + \
+                                        0x20000000) >> 30)
+#define AAC_MSUB30_V8(x, y, a, b, c, d, e, f) (int)((((int64_t)(x) * (y)) + \
+                                                     ((int64_t)(a) * (b)) - \
+                                                     ((int64_t)(c) * (d)) - \
+                                                     ((int64_t)(e) * (f)) + \
+                                                       0x20000000) >> 30)
+#define AAC_MSUB31_V3(x, y, z)    (int)((((int64_t)(x) * (z)) - \
+                                      ((int64_t)(y) * (z)) + \
+                                        0x40000000) >> 31)
+#define AAC_HALF_SUM(x, y)  (((x) >> 1) + ((y) >> 1))
+#define AAC_SRA_R(x, y)     (int)(((x) + (1 << ((y) - 1))) >> (y))
+
+#else
+
+#define FFT_FLOAT    1
+#define FFT_FIXED_32 0
+
+#define AAC_RENAME(x)       x
+#define AAC_RENAME_32(x)    x
+typedef float               INTFLOAT;
+typedef float               UINTFLOAT;
+typedef float               INT64FLOAT;
+typedef float               SHORTFLOAT;
+typedef float               AAC_FLOAT;
+typedef unsigned            AAC_SIGNE;
+#define FIXR(x)             ((float)(x))
+#define FIXR10(x)           ((float)(x))
+#define Q23(x)              ((float)(x))
+#define Q30(x)              ((float)(x))
+#define Q31(x)              ((float)(x))
+#define RANGE15(x)          (32768.0 * (x))
+#define GET_GAIN(x, y)      powf((x), -(y))
+#define AAC_MUL16(x, y)     ((x) * (y))
+#define AAC_MUL26(x, y)     ((x) * (y))
+#define AAC_MUL30(x, y)     ((x) * (y))
+#define AAC_MUL31(x, y)     ((x) * (y))
+#define AAC_MADD28(x, y, a, b) ((x) * (y) + (a) * (b))
+#define AAC_MADD30(x, y, a, b) ((x) * (y) + (a) * (b))
+#define AAC_MADD30_V8(x, y, a, b, c, d, e, f) ((x) * (y) + (a) * (b) + \
+                                               (c) * (d) + (e) * (f))
+#define AAC_MSUB30(x, y, a, b) ((x) * (y) - (a) * (b))
+#define AAC_MSUB30_V8(x, y, a, b, c, d, e, f) ((x) * (y) + (a) * (b) - \
+                                               (c) * (d) - (e) * (f))
+#define AAC_MSUB31_V3(x, y, z)    ((x) - (y)) * (z)
+#define AAC_HALF_SUM(x, y)  ((x) + (y)) * 0.5f
+#define AAC_SRA_R(x, y)     (x)
+
+#endif /* USE_FIXED */
+
+#endif /* AVCODEC_AAC_DEFINES_H */

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aac_parser.c

@@ -0,0 +1,71 @@
+/*
+ * Audio and Video frame extraction
+ * Copyright (c) 2003 Fabrice Bellard
+ * Copyright (c) 2003 Michael Niedermayer
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include "parser.h"
+#include "aac_ac3_parser.h"
+#include "adts_header.h"
+#include "adts_parser.h"
+#include "get_bits.h"
+#include "mpeg4audio.h"
+
+static int aac_sync(uint64_t state, AACAC3ParseContext *hdr_info,
+        int *need_next_header, int *new_frame_start)
+{
+    GetBitContext bits;
+    AACADTSHeaderInfo hdr;
+    int size;
+    union {
+        uint64_t u64;
+        uint8_t  u8[8 + AV_INPUT_BUFFER_PADDING_SIZE];
+    } tmp;
+
+    tmp.u64 = av_be2ne64(state);
+    init_get_bits(&bits, tmp.u8 + 8 - AV_AAC_ADTS_HEADER_SIZE,
+                  AV_AAC_ADTS_HEADER_SIZE * 8);
+
+    if ((size = ff_adts_header_parse(&bits, &hdr)) < 0)
+        return 0;
+    *need_next_header = 0;
+    *new_frame_start  = 1;
+    hdr_info->sample_rate = hdr.sample_rate;
+    hdr_info->channels    = ff_mpeg4audio_channels[hdr.chan_config];
+    hdr_info->samples     = hdr.samples;
+    hdr_info->bit_rate    = hdr.bit_rate;
+    return size;
+}
+
+static av_cold int aac_parse_init(AVCodecParserContext *s1)
+{
+    AACAC3ParseContext *s = s1->priv_data;
+    s->header_size = AV_AAC_ADTS_HEADER_SIZE;
+    s->sync = aac_sync;
+    return 0;
+}
+
+
+AVCodecParser ff_aac_parser = {
+    .codec_ids      = { AV_CODEC_ID_AAC },
+    .priv_data_size = sizeof(AACAC3ParseContext),
+    .parser_init    = aac_parse_init,
+    .parser_parse   = ff_aac_ac3_parse,
+    .parser_close   = ff_parse_close,
+};

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aaccoder.c

@@ -0,0 +1,964 @@
+/*
+ * AAC coefficients encoder
+ * Copyright (C) 2008-2009 Konstantin Shishkov
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+/**
+ * @file
+ * AAC coefficients encoder
+ */
+
+/***********************************
+ *              TODOs:
+ * speedup quantizer selection
+ * add sane pulse detection
+ ***********************************/
+
+#include "libavutil/libm.h" // brought forward to work around cygwin header breakage
+
+#include <float.h>
+
+#include "libavutil/mathematics.h"
+#include "mathops.h"
+#include "avcodec.h"
+#include "put_bits.h"
+#include "aac.h"
+#include "aacenc.h"
+#include "aactab.h"
+#include "aacenctab.h"
+#include "aacenc_utils.h"
+#include "aacenc_quantization.h"
+
+#include "aacenc_is.h"
+#include "aacenc_tns.h"
+#include "aacenc_ltp.h"
+#include "aacenc_pred.h"
+
+#include "libavcodec/aaccoder_twoloop.h"
+
+/* Parameter of f(x) = a*(lambda/100), defines the maximum fourier spread
+ * beyond which no PNS is used (since the SFBs contain tone rather than noise) */
+#define NOISE_SPREAD_THRESHOLD 0.9f
+
+/* Parameter of f(x) = a*(100/lambda), defines how much PNS is allowed to
+ * replace low energy non zero bands */
+#define NOISE_LAMBDA_REPLACE 1.948f
+
+#include "libavcodec/aaccoder_trellis.h"
+
+/**
+ * structure used in optimal codebook search
+ */
+typedef struct BandCodingPath {
+    int prev_idx; ///< pointer to the previous path point
+    float cost;   ///< path cost
+    int run;
+} BandCodingPath;
+
+/**
+ * Encode band info for single window group bands.
+ */
+static void encode_window_bands_info(AACEncContext *s, SingleChannelElement *sce,
+                                     int win, int group_len, const float lambda)
+{
+    BandCodingPath path[120][CB_TOT_ALL];
+    int w, swb, cb, start, size;
+    int i, j;
+    const int max_sfb  = sce->ics.max_sfb;
+    const int run_bits = sce->ics.num_windows == 1 ? 5 : 3;
+    const int run_esc  = (1 << run_bits) - 1;
+    int idx, ppos, count;
+    int stackrun[120], stackcb[120], stack_len;
+    float next_minrd = INFINITY;
+    int next_mincb = 0;
+
+    s->abs_pow34(s->scoefs, sce->coeffs, 1024);
+    start = win*128;
+    for (cb = 0; cb < CB_TOT_ALL; cb++) {
+        path[0][cb].cost     = 0.0f;
+        path[0][cb].prev_idx = -1;
+        path[0][cb].run      = 0;
+    }
+    for (swb = 0; swb < max_sfb; swb++) {
+        size = sce->ics.swb_sizes[swb];
+        if (sce->zeroes[win*16 + swb]) {
+            for (cb = 0; cb < CB_TOT_ALL; cb++) {
+                path[swb+1][cb].prev_idx = cb;
+                path[swb+1][cb].cost     = path[swb][cb].cost;
+                path[swb+1][cb].run      = path[swb][cb].run + 1;
+            }
+        } else {
+            float minrd = next_minrd;
+            int mincb = next_mincb;
+            next_minrd = INFINITY;
+            next_mincb = 0;
+            for (cb = 0; cb < CB_TOT_ALL; cb++) {
+                float cost_stay_here, cost_get_here;
+                float rd = 0.0f;
+                if (cb >= 12 && sce->band_type[win*16+swb] < aac_cb_out_map[cb] ||
+                    cb  < aac_cb_in_map[sce->band_type[win*16+swb]] && sce->band_type[win*16+swb] > aac_cb_out_map[cb]) {
+                    path[swb+1][cb].prev_idx = -1;
+                    path[swb+1][cb].cost     = INFINITY;
+                    path[swb+1][cb].run      = path[swb][cb].run + 1;
+                    continue;
+                }
+                for (w = 0; w < group_len; w++) {
+                    FFPsyBand *band = &s->psy.ch[s->cur_channel].psy_bands[(win+w)*16+swb];
+                    rd += quantize_band_cost(s, &sce->coeffs[start + w*128],
+                                             &s->scoefs[start + w*128], size,
+                                             sce->sf_idx[(win+w)*16+swb], aac_cb_out_map[cb],
+                                             lambda / band->threshold, INFINITY, NULL, NULL, 0);
+                }
+                cost_stay_here = path[swb][cb].cost + rd;
+                cost_get_here  = minrd              + rd + run_bits + 4;
+                if (   run_value_bits[sce->ics.num_windows == 8][path[swb][cb].run]
+                    != run_value_bits[sce->ics.num_windows == 8][path[swb][cb].run+1])
+                    cost_stay_here += run_bits;
+                if (cost_get_here < cost_stay_here) {
+                    path[swb+1][cb].prev_idx = mincb;
+                    path[swb+1][cb].cost     = cost_get_here;
+                    path[swb+1][cb].run      = 1;
+                } else {
+                    path[swb+1][cb].prev_idx = cb;
+                    path[swb+1][cb].cost     = cost_stay_here;
+                    path[swb+1][cb].run      = path[swb][cb].run + 1;
+                }
+                if (path[swb+1][cb].cost < next_minrd) {
+                    next_minrd = path[swb+1][cb].cost;
+                    next_mincb = cb;
+                }
+            }
+        }
+        start += sce->ics.swb_sizes[swb];
+    }
+
+    //convert resulting path from backward-linked list
+    stack_len = 0;
+    idx       = 0;
+    for (cb = 1; cb < CB_TOT_ALL; cb++)
+        if (path[max_sfb][cb].cost < path[max_sfb][idx].cost)
+            idx = cb;
+    ppos = max_sfb;
+    while (ppos > 0) {
+        av_assert1(idx >= 0);
+        cb = idx;
+        stackrun[stack_len] = path[ppos][cb].run;
+        stackcb [stack_len] = cb;
+        idx = path[ppos-path[ppos][cb].run+1][cb].prev_idx;
+        ppos -= path[ppos][cb].run;
+        stack_len++;
+    }
+    //perform actual band info encoding
+    start = 0;
+    for (i = stack_len - 1; i >= 0; i--) {
+        cb = aac_cb_out_map[stackcb[i]];
+        put_bits(&s->pb, 4, cb);
+        count = stackrun[i];
+        memset(sce->zeroes + win*16 + start, !cb, count);
+        //XXX: memset when band_type is also uint8_t
+        for (j = 0; j < count; j++) {
+            sce->band_type[win*16 + start] = cb;
+            start++;
+        }
+        while (count >= run_esc) {
+            put_bits(&s->pb, run_bits, run_esc);
+            count -= run_esc;
+        }
+        put_bits(&s->pb, run_bits, count);
+    }
+}
+
+
+typedef struct TrellisPath {
+    float cost;
+    int prev;
+} TrellisPath;
+
+#define TRELLIS_STAGES 121
+#define TRELLIS_STATES (SCALE_MAX_DIFF+1)
+
+static void set_special_band_scalefactors(AACEncContext *s, SingleChannelElement *sce)
+{
+    int w, g;
+    int prevscaler_n = -255, prevscaler_i = 0;
+    int bands = 0;
+
+    for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) {
+        for (g = 0; g < sce->ics.num_swb; g++) {
+            if (sce->zeroes[w*16+g])
+                continue;
+            if (sce->band_type[w*16+g] == INTENSITY_BT || sce->band_type[w*16+g] == INTENSITY_BT2) {
+                sce->sf_idx[w*16+g] = av_clip(roundf(log2f(sce->is_ener[w*16+g])*2), -155, 100);
+                bands++;
+            } else if (sce->band_type[w*16+g] == NOISE_BT) {
+                sce->sf_idx[w*16+g] = av_clip(3+ceilf(log2f(sce->pns_ener[w*16+g])*2), -100, 155);
+                if (prevscaler_n == -255)
+                    prevscaler_n = sce->sf_idx[w*16+g];
+                bands++;
+            }
+        }
+    }
+
+    if (!bands)
+        return;
+
+    /* Clip the scalefactor indices */
+    for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) {
+        for (g = 0; g < sce->ics.num_swb; g++) {
+            if (sce->zeroes[w*16+g])
+                continue;
+            if (sce->band_type[w*16+g] == INTENSITY_BT || sce->band_type[w*16+g] == INTENSITY_BT2) {
+                sce->sf_idx[w*16+g] = prevscaler_i = av_clip(sce->sf_idx[w*16+g], prevscaler_i - SCALE_MAX_DIFF, prevscaler_i + SCALE_MAX_DIFF);
+            } else if (sce->band_type[w*16+g] == NOISE_BT) {
+                sce->sf_idx[w*16+g] = prevscaler_n = av_clip(sce->sf_idx[w*16+g], prevscaler_n - SCALE_MAX_DIFF, prevscaler_n + SCALE_MAX_DIFF);
+            }
+        }
+    }
+}
+
+static void search_for_quantizers_anmr(AVCodecContext *avctx, AACEncContext *s,
+                                       SingleChannelElement *sce,
+                                       const float lambda)
+{
+    int q, w, w2, g, start = 0;
+    int i, j;
+    int idx;
+    TrellisPath paths[TRELLIS_STAGES][TRELLIS_STATES];
+    int bandaddr[TRELLIS_STAGES];
+    int minq;
+    float mincost;
+    float q0f = FLT_MAX, q1f = 0.0f, qnrgf = 0.0f;
+    int q0, q1, qcnt = 0;
+
+    for (i = 0; i < 1024; i++) {
+        float t = fabsf(sce->coeffs[i]);
+        if (t > 0.0f) {
+            q0f = FFMIN(q0f, t);
+            q1f = FFMAX(q1f, t);
+            qnrgf += t*t;
+            qcnt++;
+        }
+    }
+
+    if (!qcnt) {
+        memset(sce->sf_idx, 0, sizeof(sce->sf_idx));
+        memset(sce->zeroes, 1, sizeof(sce->zeroes));
+        return;
+    }
+
+    //minimum scalefactor index is when minimum nonzero coefficient after quantizing is not clipped
+    q0 = av_clip(coef2minsf(q0f), 0, SCALE_MAX_POS-1);
+    //maximum scalefactor index is when maximum coefficient after quantizing is still not zero
+    q1 = av_clip(coef2maxsf(q1f), 1, SCALE_MAX_POS);
+    if (q1 - q0 > 60) {
+        int q0low  = q0;
+        int q1high = q1;
+        //minimum scalefactor index is when maximum nonzero coefficient after quantizing is not clipped
+        int qnrg = av_clip_uint8(log2f(sqrtf(qnrgf/qcnt))*4 - 31 + SCALE_ONE_POS - SCALE_DIV_512);
+        q1 = qnrg + 30;
+        q0 = qnrg - 30;
+        if (q0 < q0low) {
+            q1 += q0low - q0;
+            q0  = q0low;
+        } else if (q1 > q1high) {
+            q0 -= q1 - q1high;
+            q1  = q1high;
+        }
+    }
+    // q0 == q1 isn't really a legal situation
+    if (q0 == q1) {
+        // the following is indirect but guarantees q1 != q0 && q1 near q0
+        q1 = av_clip(q0+1, 1, SCALE_MAX_POS);
+        q0 = av_clip(q1-1, 0, SCALE_MAX_POS - 1);
+    }
+
+    for (i = 0; i < TRELLIS_STATES; i++) {
+        paths[0][i].cost    = 0.0f;
+        paths[0][i].prev    = -1;
+    }
+    for (j = 1; j < TRELLIS_STAGES; j++) {
+        for (i = 0; i < TRELLIS_STATES; i++) {
+            paths[j][i].cost    = INFINITY;
+            paths[j][i].prev    = -2;
+        }
+    }
+    idx = 1;
+    s->abs_pow34(s->scoefs, sce->coeffs, 1024);
+    for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) {
+        start = w*128;
+        for (g = 0; g < sce->ics.num_swb; g++) {
+            const float *coefs = &sce->coeffs[start];
+            float qmin, qmax;
+            int nz = 0;
+
+            bandaddr[idx] = w * 16 + g;
+            qmin = INT_MAX;
+            qmax = 0.0f;
+            for (w2 = 0; w2 < sce->ics.group_len[w]; w2++) {
+                FFPsyBand *band = &s->psy.ch[s->cur_channel].psy_bands[(w+w2)*16+g];
+                if (band->energy <= band->threshold || band->threshold == 0.0f) {
+                    sce->zeroes[(w+w2)*16+g] = 1;
+                    continue;
+                }
+                sce->zeroes[(w+w2)*16+g] = 0;
+                nz = 1;
+                for (i = 0; i < sce->ics.swb_sizes[g]; i++) {
+                    float t = fabsf(coefs[w2*128+i]);
+                    if (t > 0.0f)
+                        qmin = FFMIN(qmin, t);
+                    qmax = FFMAX(qmax, t);
+                }
+            }
+            if (nz) {
+                int minscale, maxscale;
+                float minrd = INFINITY;
+                float maxval;
+                //minimum scalefactor index is when minimum nonzero coefficient after quantizing is not clipped
+                minscale = coef2minsf(qmin);
+                //maximum scalefactor index is when maximum coefficient after quantizing is still not zero
+                maxscale = coef2maxsf(qmax);
+                minscale = av_clip(minscale - q0, 0, TRELLIS_STATES - 1);
+                maxscale = av_clip(maxscale - q0, 0, TRELLIS_STATES);
+                if (minscale == maxscale) {
+                    maxscale = av_clip(minscale+1, 1, TRELLIS_STATES);
+                    minscale = av_clip(maxscale-1, 0, TRELLIS_STATES - 1);
+                }
+                maxval = find_max_val(sce->ics.group_len[w], sce->ics.swb_sizes[g], s->scoefs+start);
+                for (q = minscale; q < maxscale; q++) {
+                    float dist = 0;
+                    int cb = find_min_book(maxval, sce->sf_idx[w*16+g]);
+                    for (w2 = 0; w2 < sce->ics.group_len[w]; w2++) {
+                        FFPsyBand *band = &s->psy.ch[s->cur_channel].psy_bands[(w+w2)*16+g];
+                        dist += quantize_band_cost(s, coefs + w2*128, s->scoefs + start + w2*128, sce->ics.swb_sizes[g],
+                                                   q + q0, cb, lambda / band->threshold, INFINITY, NULL, NULL, 0);
+                    }
+                    minrd = FFMIN(minrd, dist);
+
+                    for (i = 0; i < q1 - q0; i++) {
+                        float cost;
+                        cost = paths[idx - 1][i].cost + dist
+                               + ff_aac_scalefactor_bits[q - i + SCALE_DIFF_ZERO];
+                        if (cost < paths[idx][q].cost) {
+                            paths[idx][q].cost    = cost;
+                            paths[idx][q].prev    = i;
+                        }
+                    }
+                }
+            } else {
+                for (q = 0; q < q1 - q0; q++) {
+                    paths[idx][q].cost = paths[idx - 1][q].cost + 1;
+                    paths[idx][q].prev = q;
+                }
+            }
+            sce->zeroes[w*16+g] = !nz;
+            start += sce->ics.swb_sizes[g];
+            idx++;
+        }
+    }
+    idx--;
+    mincost = paths[idx][0].cost;
+    minq    = 0;
+    for (i = 1; i < TRELLIS_STATES; i++) {
+        if (paths[idx][i].cost < mincost) {
+            mincost = paths[idx][i].cost;
+            minq = i;
+        }
+    }
+    while (idx) {
+        sce->sf_idx[bandaddr[idx]] = minq + q0;
+        minq = FFMAX(paths[idx][minq].prev, 0);
+        idx--;
+    }
+    //set the same quantizers inside window groups
+    for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w])
+        for (g = 0; g < sce->ics.num_swb; g++)
+            for (w2 = 1; w2 < sce->ics.group_len[w]; w2++)
+                sce->sf_idx[(w+w2)*16+g] = sce->sf_idx[w*16+g];
+}
+
+static void search_for_quantizers_fast(AVCodecContext *avctx, AACEncContext *s,
+                                       SingleChannelElement *sce,
+                                       const float lambda)
+{
+    int start = 0, i, w, w2, g;
+    int destbits = avctx->bit_rate * 1024.0 / avctx->sample_rate / avctx->channels * (lambda / 120.f);
+    float dists[128] = { 0 }, uplims[128] = { 0 };
+    float maxvals[128];
+    int fflag, minscaler;
+    int its  = 0;
+    int allz = 0;
+    float minthr = INFINITY;
+
+    // for values above this the decoder might end up in an endless loop
+    // due to always having more bits than what can be encoded.
+    destbits = FFMIN(destbits, 5800);
+    //some heuristic to determine initial quantizers will reduce search time
+    //determine zero bands and upper limits
+    for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) {
+        start = 0;
+        for (g = 0; g < sce->ics.num_swb; g++) {
+            int nz = 0;
+            float uplim = 0.0f, energy = 0.0f;
+            for (w2 = 0; w2 < sce->ics.group_len[w]; w2++) {
+                FFPsyBand *band = &s->psy.ch[s->cur_channel].psy_bands[(w+w2)*16+g];
+                uplim += band->threshold;
+                energy += band->energy;
+                if (band->energy <= band->threshold || band->threshold == 0.0f) {
+                    sce->zeroes[(w+w2)*16+g] = 1;
+                    continue;
+                }
+                nz = 1;
+            }
+            uplims[w*16+g] = uplim *512;
+            sce->band_type[w*16+g] = 0;
+            sce->zeroes[w*16+g] = !nz;
+            if (nz)
+                minthr = FFMIN(minthr, uplim);
+            allz |= nz;
+            start += sce->ics.swb_sizes[g];
+        }
+    }
+    for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) {
+        for (g = 0; g < sce->ics.num_swb; g++) {
+            if (sce->zeroes[w*16+g]) {
+                sce->sf_idx[w*16+g] = SCALE_ONE_POS;
+                continue;
+            }
+            sce->sf_idx[w*16+g] = SCALE_ONE_POS + FFMIN(log2f(uplims[w*16+g]/minthr)*4,59);
+        }
+    }
+
+    if (!allz)
+        return;
+    s->abs_pow34(s->scoefs, sce->coeffs, 1024);
+    ff_quantize_band_cost_cache_init(s);
+
+    for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) {
+        start = w*128;
+        for (g = 0; g < sce->ics.num_swb; g++) {
+            const float *scaled = s->scoefs + start;
+            maxvals[w*16+g] = find_max_val(sce->ics.group_len[w], sce->ics.swb_sizes[g], scaled);
+            start += sce->ics.swb_sizes[g];
+        }
+    }
+
+    //perform two-loop search
+    //outer loop - improve quality
+    do {
+        int tbits, qstep;
+        minscaler = sce->sf_idx[0];
+        //inner loop - quantize spectrum to fit into given number of bits
+        qstep = its ? 1 : 32;
+        do {
+            int prev = -1;
+            tbits = 0;
+            for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) {
+                start = w*128;
+                for (g = 0; g < sce->ics.num_swb; g++) {
+                    const float *coefs = sce->coeffs + start;
+                    const float *scaled = s->scoefs + start;
+                    int bits = 0;
+                    int cb;
+                    float dist = 0.0f;
+
+                    if (sce->zeroes[w*16+g] || sce->sf_idx[w*16+g] >= 218) {
+                        start += sce->ics.swb_sizes[g];
+                        continue;
+                    }
+                    minscaler = FFMIN(minscaler, sce->sf_idx[w*16+g]);
+                    cb = find_min_book(maxvals[w*16+g], sce->sf_idx[w*16+g]);
+                    for (w2 = 0; w2 < sce->ics.group_len[w]; w2++) {
+                        int b;
+                        dist += quantize_band_cost_cached(s, w + w2, g,
+                                                          coefs + w2*128,
+                                                          scaled + w2*128,
+                                                          sce->ics.swb_sizes[g],
+                                                          sce->sf_idx[w*16+g],
+                                                          cb, 1.0f, INFINITY,
+                                                          &b, NULL, 0);
+                        bits += b;
+                    }
+                    dists[w*16+g] = dist - bits;
+                    if (prev != -1) {
+                        bits += ff_aac_scalefactor_bits[sce->sf_idx[w*16+g] - prev + SCALE_DIFF_ZERO];
+                    }
+                    tbits += bits;
+                    start += sce->ics.swb_sizes[g];
+                    prev = sce->sf_idx[w*16+g];
+                }
+            }
+            if (tbits > destbits) {
+                for (i = 0; i < 128; i++)
+                    if (sce->sf_idx[i] < 218 - qstep)
+                        sce->sf_idx[i] += qstep;
+            } else {
+                for (i = 0; i < 128; i++)
+                    if (sce->sf_idx[i] > 60 - qstep)
+                        sce->sf_idx[i] -= qstep;
+            }
+            qstep >>= 1;
+            if (!qstep && tbits > destbits*1.02 && sce->sf_idx[0] < 217)
+                qstep = 1;
+        } while (qstep);
+
+        fflag = 0;
+        minscaler = av_clip(minscaler, 60, 255 - SCALE_MAX_DIFF);
+
+        for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) {
+            for (g = 0; g < sce->ics.num_swb; g++) {
+                int prevsc = sce->sf_idx[w*16+g];
+                if (dists[w*16+g] > uplims[w*16+g] && sce->sf_idx[w*16+g] > 60) {
+                    if (find_min_book(maxvals[w*16+g], sce->sf_idx[w*16+g]-1))
+                        sce->sf_idx[w*16+g]--;
+                    else //Try to make sure there is some energy in every band
+                        sce->sf_idx[w*16+g]-=2;
+                }
+                sce->sf_idx[w*16+g] = av_clip(sce->sf_idx[w*16+g], minscaler, minscaler + SCALE_MAX_DIFF);
+                sce->sf_idx[w*16+g] = FFMIN(sce->sf_idx[w*16+g], 219);
+                if (sce->sf_idx[w*16+g] != prevsc)
+                    fflag = 1;
+                sce->band_type[w*16+g] = find_min_book(maxvals[w*16+g], sce->sf_idx[w*16+g]);
+            }
+        }
+        its++;
+    } while (fflag && its < 10);
+}
+
+static void search_for_pns(AACEncContext *s, AVCodecContext *avctx, SingleChannelElement *sce)
+{
+    FFPsyBand *band;
+    int w, g, w2, i;
+    int wlen = 1024 / sce->ics.num_windows;
+    int bandwidth, cutoff;
+    float *PNS = &s->scoefs[0*128], *PNS34 = &s->scoefs[1*128];
+    float *NOR34 = &s->scoefs[3*128];
+    uint8_t nextband[128];
+    const float lambda = s->lambda;
+    const float freq_mult = avctx->sample_rate*0.5f/wlen;
+    const float thr_mult = NOISE_LAMBDA_REPLACE*(100.0f/lambda);
+    const float spread_threshold = FFMIN(0.75f, NOISE_SPREAD_THRESHOLD*FFMAX(0.5f, lambda/100.f));
+    const float dist_bias = av_clipf(4.f * 120 / lambda, 0.25f, 4.0f);
+    const float pns_transient_energy_r = FFMIN(0.7f, lambda / 140.f);
+
+    int refbits = avctx->bit_rate * 1024.0 / avctx->sample_rate
+        / ((avctx->flags & AV_CODEC_FLAG_QSCALE) ? 2.0f : avctx->channels)
+        * (lambda / 120.f);
+
+    /** Keep this in sync with twoloop's cutoff selection */
+    float rate_bandwidth_multiplier = 1.5f;
+    int prev = -1000, prev_sf = -1;
+    int frame_bit_rate = (avctx->flags & AV_CODEC_FLAG_QSCALE)
+        ? (refbits * rate_bandwidth_multiplier * avctx->sample_rate / 1024)
+        : (avctx->bit_rate / avctx->channels);
+
+    frame_bit_rate *= 1.15f;
+
+    if (avctx->cutoff > 0) {
+        bandwidth = avctx->cutoff;
+    } else {
+        bandwidth = FFMAX(3000, AAC_CUTOFF_FROM_BITRATE(frame_bit_rate, 1, avctx->sample_rate));
+    }
+
+    cutoff = bandwidth * 2 * wlen / avctx->sample_rate;
+
+    memcpy(sce->band_alt, sce->band_type, sizeof(sce->band_type));
+    ff_init_nextband_map(sce, nextband);
+    for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) {
+        int wstart = w*128;
+        for (g = 0; g < sce->ics.num_swb; g++) {
+            int noise_sfi;
+            float dist1 = 0.0f, dist2 = 0.0f, noise_amp;
+            float pns_energy = 0.0f, pns_tgt_energy, energy_ratio, dist_thresh;
+            float sfb_energy = 0.0f, threshold = 0.0f, spread = 2.0f;
+            float min_energy = -1.0f, max_energy = 0.0f;
+            const int start = wstart+sce->ics.swb_offset[g];
+            const float freq = (start-wstart)*freq_mult;
+            const float freq_boost = FFMAX(0.88f*freq/NOISE_LOW_LIMIT, 1.0f);
+            if (freq < NOISE_LOW_LIMIT || (start-wstart) >= cutoff) {
+                if (!sce->zeroes[w*16+g])
+                    prev_sf = sce->sf_idx[w*16+g];
+                continue;
+            }
+            for (w2 = 0; w2 < sce->ics.group_len[w]; w2++) {
+                band = &s->psy.ch[s->cur_channel].psy_bands[(w+w2)*16+g];
+                sfb_energy += band->energy;
+                spread     = FFMIN(spread, band->spread);
+                threshold  += band->threshold;
+                if (!w2) {
+                    min_energy = max_energy = band->energy;
+                } else {
+                    min_energy = FFMIN(min_energy, band->energy);
+                    max_energy = FFMAX(max_energy, band->energy);
+                }
+            }
+
+            /* Ramps down at ~8000Hz and loosens the dist threshold */
+            dist_thresh = av_clipf(2.5f*NOISE_LOW_LIMIT/freq, 0.5f, 2.5f) * dist_bias;
+
+            /* PNS is acceptable when all of these are true:
+             * 1. high spread energy (noise-like band)
+             * 2. near-threshold energy (high PE means the random nature of PNS content will be noticed)
+             * 3. on short window groups, all windows have similar energy (variations in energy would be destroyed by PNS)
+             *
+             * At this stage, point 2 is relaxed for zeroed bands near the noise threshold (hole avoidance is more important)
+             */
+            if ((!sce->zeroes[w*16+g] && !ff_sfdelta_can_remove_band(sce, nextband, prev_sf, w*16+g)) ||
+                ((sce->zeroes[w*16+g] || !sce->band_alt[w*16+g]) && sfb_energy < threshold*sqrtf(1.0f/freq_boost)) || spread < spread_threshold ||
+                (!sce->zeroes[w*16+g] && sce->band_alt[w*16+g] && sfb_energy > threshold*thr_mult*freq_boost) ||
+                min_energy < pns_transient_energy_r * max_energy ) {
+                sce->pns_ener[w*16+g] = sfb_energy;
+                if (!sce->zeroes[w*16+g])
+                    prev_sf = sce->sf_idx[w*16+g];
+                continue;
+            }
+
+            pns_tgt_energy = sfb_energy*FFMIN(1.0f, spread*spread);
+            noise_sfi = av_clip(roundf(log2f(pns_tgt_energy)*2), -100, 155); /* Quantize */
+            noise_amp = -ff_aac_pow2sf_tab[noise_sfi + POW_SF2_ZERO];    /* Dequantize */
+            if (prev != -1000) {
+                int noise_sfdiff = noise_sfi - prev + SCALE_DIFF_ZERO;
+                if (noise_sfdiff < 0 || noise_sfdiff > 2*SCALE_MAX_DIFF) {
+                    if (!sce->zeroes[w*16+g])
+                        prev_sf = sce->sf_idx[w*16+g];
+                    continue;
+                }
+            }
+            for (w2 = 0; w2 < sce->ics.group_len[w]; w2++) {
+                float band_energy, scale, pns_senergy;
+                const int start_c = (w+w2)*128+sce->ics.swb_offset[g];
+                band = &s->psy.ch[s->cur_channel].psy_bands[(w+w2)*16+g];
+                for (i = 0; i < sce->ics.swb_sizes[g]; i++) {
+                    s->random_state  = lcg_random(s->random_state);
+                    PNS[i] = s->random_state;
+                }
+                band_energy = s->fdsp->scalarproduct_float(PNS, PNS, sce->ics.swb_sizes[g]);
+                scale = noise_amp/sqrtf(band_energy);
+                s->fdsp->vector_fmul_scalar(PNS, PNS, scale, sce->ics.swb_sizes[g]);
+                pns_senergy = s->fdsp->scalarproduct_float(PNS, PNS, sce->ics.swb_sizes[g]);
+                pns_energy += pns_senergy;
+                s->abs_pow34(NOR34, &sce->coeffs[start_c], sce->ics.swb_sizes[g]);
+                s->abs_pow34(PNS34, PNS, sce->ics.swb_sizes[g]);
+                dist1 += quantize_band_cost(s, &sce->coeffs[start_c],
+                                            NOR34,
+                                            sce->ics.swb_sizes[g],
+                                            sce->sf_idx[(w+w2)*16+g],
+                                            sce->band_alt[(w+w2)*16+g],
+                                            lambda/band->threshold, INFINITY, NULL, NULL, 0);
+                /* Estimate rd on average as 5 bits for SF, 4 for the CB, plus spread energy * lambda/thr */
+                dist2 += band->energy/(band->spread*band->spread)*lambda*dist_thresh/band->threshold;
+            }
+            if (g && sce->band_type[w*16+g-1] == NOISE_BT) {
+                dist2 += 5;
+            } else {
+                dist2 += 9;
+            }
+            energy_ratio = pns_tgt_energy/pns_energy; /* Compensates for quantization error */
+            sce->pns_ener[w*16+g] = energy_ratio*pns_tgt_energy;
+            if (sce->zeroes[w*16+g] || !sce->band_alt[w*16+g] || (energy_ratio > 0.85f && energy_ratio < 1.25f && dist2 < dist1)) {
+                sce->band_type[w*16+g] = NOISE_BT;
+                sce->zeroes[w*16+g] = 0;
+                prev = noise_sfi;
+            } else {
+                if (!sce->zeroes[w*16+g])
+                    prev_sf = sce->sf_idx[w*16+g];
+            }
+        }
+    }
+}
+
+static void mark_pns(AACEncContext *s, AVCodecContext *avctx, SingleChannelElement *sce)
+{
+    FFPsyBand *band;
+    int w, g, w2;
+    int wlen = 1024 / sce->ics.num_windows;
+    int bandwidth, cutoff;
+    const float lambda = s->lambda;
+    const float freq_mult = avctx->sample_rate*0.5f/wlen;
+    const float spread_threshold = FFMIN(0.75f, NOISE_SPREAD_THRESHOLD*FFMAX(0.5f, lambda/100.f));
+    const float pns_transient_energy_r = FFMIN(0.7f, lambda / 140.f);
+
+    int refbits = avctx->bit_rate * 1024.0 / avctx->sample_rate
+        / ((avctx->flags & AV_CODEC_FLAG_QSCALE) ? 2.0f : avctx->channels)
+        * (lambda / 120.f);
+
+    /** Keep this in sync with twoloop's cutoff selection */
+    float rate_bandwidth_multiplier = 1.5f;
+    int frame_bit_rate = (avctx->flags & AV_CODEC_FLAG_QSCALE)
+        ? (refbits * rate_bandwidth_multiplier * avctx->sample_rate / 1024)
+        : (avctx->bit_rate / avctx->channels);
+
+    frame_bit_rate *= 1.15f;
+
+    if (avctx->cutoff > 0) {
+        bandwidth = avctx->cutoff;
+    } else {
+        bandwidth = FFMAX(3000, AAC_CUTOFF_FROM_BITRATE(frame_bit_rate, 1, avctx->sample_rate));
+    }
+
+    cutoff = bandwidth * 2 * wlen / avctx->sample_rate;
+
+    memcpy(sce->band_alt, sce->band_type, sizeof(sce->band_type));
+    for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) {
+        for (g = 0; g < sce->ics.num_swb; g++) {
+            float sfb_energy = 0.0f, threshold = 0.0f, spread = 2.0f;
+            float min_energy = -1.0f, max_energy = 0.0f;
+            const int start = sce->ics.swb_offset[g];
+            const float freq = start*freq_mult;
+            const float freq_boost = FFMAX(0.88f*freq/NOISE_LOW_LIMIT, 1.0f);
+            if (freq < NOISE_LOW_LIMIT || start >= cutoff) {
+                sce->can_pns[w*16+g] = 0;
+                continue;
+            }
+            for (w2 = 0; w2 < sce->ics.group_len[w]; w2++) {
+                band = &s->psy.ch[s->cur_channel].psy_bands[(w+w2)*16+g];
+                sfb_energy += band->energy;
+                spread     = FFMIN(spread, band->spread);
+                threshold  += band->threshold;
+                if (!w2) {
+                    min_energy = max_energy = band->energy;
+                } else {
+                    min_energy = FFMIN(min_energy, band->energy);
+                    max_energy = FFMAX(max_energy, band->energy);
+                }
+            }
+
+            /* PNS is acceptable when all of these are true:
+             * 1. high spread energy (noise-like band)
+             * 2. near-threshold energy (high PE means the random nature of PNS content will be noticed)
+             * 3. on short window groups, all windows have similar energy (variations in energy would be destroyed by PNS)
+             */
+            sce->pns_ener[w*16+g] = sfb_energy;
+            if (sfb_energy < threshold*sqrtf(1.5f/freq_boost) || spread < spread_threshold || min_energy < pns_transient_energy_r * max_energy) {
+                sce->can_pns[w*16+g] = 0;
+            } else {
+                sce->can_pns[w*16+g] = 1;
+            }
+        }
+    }
+}
+
+static void search_for_ms(AACEncContext *s, ChannelElement *cpe)
+{
+    int start = 0, i, w, w2, g, sid_sf_boost, prev_mid, prev_side;
+    uint8_t nextband0[128], nextband1[128];
+    float *M   = s->scoefs + 128*0, *S   = s->scoefs + 128*1;
+    float *L34 = s->scoefs + 128*2, *R34 = s->scoefs + 128*3;
+    float *M34 = s->scoefs + 128*4, *S34 = s->scoefs + 128*5;
+    const float lambda = s->lambda;
+    const float mslambda = FFMIN(1.0f, lambda / 120.f);
+    SingleChannelElement *sce0 = &cpe->ch[0];
+    SingleChannelElement *sce1 = &cpe->ch[1];
+    if (!cpe->common_window)
+        return;
+
+    /** Scout out next nonzero bands */
+    ff_init_nextband_map(sce0, nextband0);
+    ff_init_nextband_map(sce1, nextband1);
+
+    prev_mid = sce0->sf_idx[0];
+    prev_side = sce1->sf_idx[0];
+    for (w = 0; w < sce0->ics.num_windows; w += sce0->ics.group_len[w]) {
+        start = 0;
+        for (g = 0; g < sce0->ics.num_swb; g++) {
+            float bmax = bval2bmax(g * 17.0f / sce0->ics.num_swb) / 0.0045f;
+            if (!cpe->is_mask[w*16+g])
+                cpe->ms_mask[w*16+g] = 0;
+            if (!sce0->zeroes[w*16+g] && !sce1->zeroes[w*16+g] && !cpe->is_mask[w*16+g]) {
+                float Mmax = 0.0f, Smax = 0.0f;
+
+                /* Must compute mid/side SF and book for the whole window group */
+                for (w2 = 0; w2 < sce0->ics.group_len[w]; w2++) {
+                    for (i = 0; i < sce0->ics.swb_sizes[g]; i++) {
+                        M[i] = (sce0->coeffs[start+(w+w2)*128+i]
+                              + sce1->coeffs[start+(w+w2)*128+i]) * 0.5;
+                        S[i] =  M[i]
+                              - sce1->coeffs[start+(w+w2)*128+i];
+                    }
+                    s->abs_pow34(M34, M, sce0->ics.swb_sizes[g]);
+                    s->abs_pow34(S34, S, sce0->ics.swb_sizes[g]);
+                    for (i = 0; i < sce0->ics.swb_sizes[g]; i++ ) {
+                        Mmax = FFMAX(Mmax, M34[i]);
+                        Smax = FFMAX(Smax, S34[i]);
+                    }
+                }
+
+                for (sid_sf_boost = 0; sid_sf_boost < 4; sid_sf_boost++) {
+                    float dist1 = 0.0f, dist2 = 0.0f;
+                    int B0 = 0, B1 = 0;
+                    int minidx;
+                    int mididx, sididx;
+                    int midcb, sidcb;
+
+                    minidx = FFMIN(sce0->sf_idx[w*16+g], sce1->sf_idx[w*16+g]);
+                    mididx = av_clip(minidx, 0, SCALE_MAX_POS - SCALE_DIV_512);
+                    sididx = av_clip(minidx - sid_sf_boost * 3, 0, SCALE_MAX_POS - SCALE_DIV_512);
+                    if (sce0->band_type[w*16+g] != NOISE_BT && sce1->band_type[w*16+g] != NOISE_BT
+                        && (   !ff_sfdelta_can_replace(sce0, nextband0, prev_mid, mididx, w*16+g)
+                            || !ff_sfdelta_can_replace(sce1, nextband1, prev_side, sididx, w*16+g))) {
+                        /* scalefactor range violation, bad stuff, will decrease quality unacceptably */
+                        continue;
+                    }
+
+                    midcb = find_min_book(Mmax, mididx);
+                    sidcb = find_min_book(Smax, sididx);
+
+                    /* No CB can be zero */
+                    midcb = FFMAX(1,midcb);
+                    sidcb = FFMAX(1,sidcb);
+
+                    for (w2 = 0; w2 < sce0->ics.group_len[w]; w2++) {
+                        FFPsyBand *band0 = &s->psy.ch[s->cur_channel+0].psy_bands[(w+w2)*16+g];
+                        FFPsyBand *band1 = &s->psy.ch[s->cur_channel+1].psy_bands[(w+w2)*16+g];
+                        float minthr = FFMIN(band0->threshold, band1->threshold);
+                        int b1,b2,b3,b4;
+                        for (i = 0; i < sce0->ics.swb_sizes[g]; i++) {
+                            M[i] = (sce0->coeffs[start+(w+w2)*128+i]
+                                  + sce1->coeffs[start+(w+w2)*128+i]) * 0.5;
+                            S[i] =  M[i]
+                                  - sce1->coeffs[start+(w+w2)*128+i];
+                        }
+
+                        s->abs_pow34(L34, sce0->coeffs+start+(w+w2)*128, sce0->ics.swb_sizes[g]);
+                        s->abs_pow34(R34, sce1->coeffs+start+(w+w2)*128, sce0->ics.swb_sizes[g]);
+                        s->abs_pow34(M34, M,                         sce0->ics.swb_sizes[g]);
+                        s->abs_pow34(S34, S,                         sce0->ics.swb_sizes[g]);
+                        dist1 += quantize_band_cost(s, &sce0->coeffs[start + (w+w2)*128],
+                                                    L34,
+                                                    sce0->ics.swb_sizes[g],
+                                                    sce0->sf_idx[w*16+g],
+                                                    sce0->band_type[w*16+g],
+                                                    lambda / band0->threshold, INFINITY, &b1, NULL, 0);
+                        dist1 += quantize_band_cost(s, &sce1->coeffs[start + (w+w2)*128],
+                                                    R34,
+                                                    sce1->ics.swb_sizes[g],
+                                                    sce1->sf_idx[w*16+g],
+                                                    sce1->band_type[w*16+g],
+                                                    lambda / band1->threshold, INFINITY, &b2, NULL, 0);
+                        dist2 += quantize_band_cost(s, M,
+                                                    M34,
+                                                    sce0->ics.swb_sizes[g],
+                                                    mididx,
+                                                    midcb,
+                                                    lambda / minthr, INFINITY, &b3, NULL, 0);
+                        dist2 += quantize_band_cost(s, S,
+                                                    S34,
+                                                    sce1->ics.swb_sizes[g],
+                                                    sididx,
+                                                    sidcb,
+                                                    mslambda / (minthr * bmax), INFINITY, &b4, NULL, 0);
+                        B0 += b1+b2;
+                        B1 += b3+b4;
+                        dist1 -= b1+b2;
+                        dist2 -= b3+b4;
+                    }
+                    cpe->ms_mask[w*16+g] = dist2 <= dist1 && B1 < B0;
+                    if (cpe->ms_mask[w*16+g]) {
+                        if (sce0->band_type[w*16+g] != NOISE_BT && sce1->band_type[w*16+g] != NOISE_BT) {
+                            sce0->sf_idx[w*16+g] = mididx;
+                            sce1->sf_idx[w*16+g] = sididx;
+                            sce0->band_type[w*16+g] = midcb;
+                            sce1->band_type[w*16+g] = sidcb;
+                        } else if ((sce0->band_type[w*16+g] != NOISE_BT) ^ (sce1->band_type[w*16+g] != NOISE_BT)) {
+                            /* ms_mask unneeded, and it confuses some decoders */
+                            cpe->ms_mask[w*16+g] = 0;
+                        }
+                        break;
+                    } else if (B1 > B0) {
+                        /* More boost won't fix this */
+                        break;
+                    }
+                }
+            }
+            if (!sce0->zeroes[w*16+g] && sce0->band_type[w*16+g] < RESERVED_BT)
+                prev_mid = sce0->sf_idx[w*16+g];
+            if (!sce1->zeroes[w*16+g] && !cpe->is_mask[w*16+g] && sce1->band_type[w*16+g] < RESERVED_BT)
+                prev_side = sce1->sf_idx[w*16+g];
+            start += sce0->ics.swb_sizes[g];
+        }
+    }
+}
+
+const AACCoefficientsEncoder ff_aac_coders[AAC_CODER_NB] = {
+    [AAC_CODER_ANMR] = {
+        search_for_quantizers_anmr,
+        encode_window_bands_info,
+        quantize_and_encode_band,
+        ff_aac_encode_tns_info,
+        ff_aac_encode_ltp_info,
+        ff_aac_encode_main_pred,
+        ff_aac_adjust_common_pred,
+        ff_aac_adjust_common_ltp,
+        ff_aac_apply_main_pred,
+        ff_aac_apply_tns,
+        ff_aac_update_ltp,
+        ff_aac_ltp_insert_new_frame,
+        set_special_band_scalefactors,
+        search_for_pns,
+        mark_pns,
+        ff_aac_search_for_tns,
+        ff_aac_search_for_ltp,
+        search_for_ms,
+        ff_aac_search_for_is,
+        ff_aac_search_for_pred,
+    },
+    [AAC_CODER_TWOLOOP] = {
+        search_for_quantizers_twoloop,
+        codebook_trellis_rate,
+        quantize_and_encode_band,
+        ff_aac_encode_tns_info,
+        ff_aac_encode_ltp_info,
+        ff_aac_encode_main_pred,
+        ff_aac_adjust_common_pred,
+        ff_aac_adjust_common_ltp,
+        ff_aac_apply_main_pred,
+        ff_aac_apply_tns,
+        ff_aac_update_ltp,
+        ff_aac_ltp_insert_new_frame,
+        set_special_band_scalefactors,
+        search_for_pns,
+        mark_pns,
+        ff_aac_search_for_tns,
+        ff_aac_search_for_ltp,
+        search_for_ms,
+        ff_aac_search_for_is,
+        ff_aac_search_for_pred,
+    },
+    [AAC_CODER_FAST] = {
+        search_for_quantizers_fast,
+        codebook_trellis_rate,
+        quantize_and_encode_band,
+        ff_aac_encode_tns_info,
+        ff_aac_encode_ltp_info,
+        ff_aac_encode_main_pred,
+        ff_aac_adjust_common_pred,
+        ff_aac_adjust_common_ltp,
+        ff_aac_apply_main_pred,
+        ff_aac_apply_tns,
+        ff_aac_update_ltp,
+        ff_aac_ltp_insert_new_frame,
+        set_special_band_scalefactors,
+        search_for_pns,
+        mark_pns,
+        ff_aac_search_for_tns,
+        ff_aac_search_for_ltp,
+        search_for_ms,
+        ff_aac_search_for_is,
+        ff_aac_search_for_pred,
+    },
+};

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aaccoder_trellis.h

@@ -0,0 +1,192 @@
+/*
+ * AAC encoder trellis codebook selector
+ * Copyright (C) 2008-2009 Konstantin Shishkov
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+/**
+ * @file
+ * AAC encoder trellis codebook selector
+ * @author Konstantin Shishkov
+ */
+
+/**
+ * This file contains a template for the codebook_trellis_rate selector function.
+ * It needs to be provided, externally, as an already included declaration,
+ * the following functions from aacenc_quantization/util.h. They're not included
+ * explicitly here to make it possible to provide alternative implementations:
+ *  - quantize_band_cost_bits
+ *  - abs_pow34_v
+ */
+
+#ifndef AVCODEC_AACCODER_TRELLIS_H
+#define AVCODEC_AACCODER_TRELLIS_H
+
+#include <float.h>
+#include "libavutil/mathematics.h"
+#include "avcodec.h"
+#include "put_bits.h"
+#include "aac.h"
+#include "aacenc.h"
+#include "aactab.h"
+#include "aacenctab.h"
+
+/**
+ * structure used in optimal codebook search
+ */
+typedef struct TrellisBandCodingPath {
+    int prev_idx; ///< pointer to the previous path point
+    float cost;   ///< path cost
+    int run;
+} TrellisBandCodingPath;
+
+
+static void codebook_trellis_rate(AACEncContext *s, SingleChannelElement *sce,
+                                  int win, int group_len, const float lambda)
+{
+    TrellisBandCodingPath path[120][CB_TOT_ALL];
+    int w, swb, cb, start, size;
+    int i, j;
+    const int max_sfb  = sce->ics.max_sfb;
+    const int run_bits = sce->ics.num_windows == 1 ? 5 : 3;
+    const int run_esc  = (1 << run_bits) - 1;
+    int idx, ppos, count;
+    int stackrun[120], stackcb[120], stack_len;
+    float next_minbits = INFINITY;
+    int next_mincb = 0;
+
+    s->abs_pow34(s->scoefs, sce->coeffs, 1024);
+    start = win*128;
+    for (cb = 0; cb < CB_TOT_ALL; cb++) {
+        path[0][cb].cost     = run_bits+4;
+        path[0][cb].prev_idx = -1;
+        path[0][cb].run      = 0;
+    }
+    for (swb = 0; swb < max_sfb; swb++) {
+        size = sce->ics.swb_sizes[swb];
+        if (sce->zeroes[win*16 + swb]) {
+            float cost_stay_here = path[swb][0].cost;
+            float cost_get_here  = next_minbits + run_bits + 4;
+            if (   run_value_bits[sce->ics.num_windows == 8][path[swb][0].run]
+                != run_value_bits[sce->ics.num_windows == 8][path[swb][0].run+1])
+                cost_stay_here += run_bits;
+            if (cost_get_here < cost_stay_here) {
+                path[swb+1][0].prev_idx = next_mincb;
+                path[swb+1][0].cost     = cost_get_here;
+                path[swb+1][0].run      = 1;
+            } else {
+                path[swb+1][0].prev_idx = 0;
+                path[swb+1][0].cost     = cost_stay_here;
+                path[swb+1][0].run      = path[swb][0].run + 1;
+            }
+            next_minbits = path[swb+1][0].cost;
+            next_mincb = 0;
+            for (cb = 1; cb < CB_TOT_ALL; cb++) {
+                path[swb+1][cb].cost = 61450;
+                path[swb+1][cb].prev_idx = -1;
+                path[swb+1][cb].run = 0;
+            }
+        } else {
+            float minbits = next_minbits;
+            int mincb = next_mincb;
+            int startcb = sce->band_type[win*16+swb];
+            startcb = aac_cb_in_map[startcb];
+            next_minbits = INFINITY;
+            next_mincb = 0;
+            for (cb = 0; cb < startcb; cb++) {
+                path[swb+1][cb].cost = 61450;
+                path[swb+1][cb].prev_idx = -1;
+                path[swb+1][cb].run = 0;
+            }
+            for (cb = startcb; cb < CB_TOT_ALL; cb++) {
+                float cost_stay_here, cost_get_here;
+                float bits = 0.0f;
+                if (cb >= 12 && sce->band_type[win*16+swb] != aac_cb_out_map[cb]) {
+                    path[swb+1][cb].cost = 61450;
+                    path[swb+1][cb].prev_idx = -1;
+                    path[swb+1][cb].run = 0;
+                    continue;
+                }
+                for (w = 0; w < group_len; w++) {
+                    bits += quantize_band_cost_bits(s, &sce->coeffs[start + w*128],
+                                               &s->scoefs[start + w*128], size,
+                                               sce->sf_idx[win*16+swb],
+                                               aac_cb_out_map[cb],
+                                               0, INFINITY, NULL, NULL, 0);
+                }
+                cost_stay_here = path[swb][cb].cost + bits;
+                cost_get_here  = minbits            + bits + run_bits + 4;
+                if (   run_value_bits[sce->ics.num_windows == 8][path[swb][cb].run]
+                    != run_value_bits[sce->ics.num_windows == 8][path[swb][cb].run+1])
+                    cost_stay_here += run_bits;
+                if (cost_get_here < cost_stay_here) {
+                    path[swb+1][cb].prev_idx = mincb;
+                    path[swb+1][cb].cost     = cost_get_here;
+                    path[swb+1][cb].run      = 1;
+                } else {
+                    path[swb+1][cb].prev_idx = cb;
+                    path[swb+1][cb].cost     = cost_stay_here;
+                    path[swb+1][cb].run      = path[swb][cb].run + 1;
+                }
+                if (path[swb+1][cb].cost < next_minbits) {
+                    next_minbits = path[swb+1][cb].cost;
+                    next_mincb = cb;
+                }
+            }
+        }
+        start += sce->ics.swb_sizes[swb];
+    }
+
+    //convert resulting path from backward-linked list
+    stack_len = 0;
+    idx       = 0;
+    for (cb = 1; cb < CB_TOT_ALL; cb++)
+        if (path[max_sfb][cb].cost < path[max_sfb][idx].cost)
+            idx = cb;
+    ppos = max_sfb;
+    while (ppos > 0) {
+        av_assert1(idx >= 0);
+        cb = idx;
+        stackrun[stack_len] = path[ppos][cb].run;
+        stackcb [stack_len] = cb;
+        idx = path[ppos-path[ppos][cb].run+1][cb].prev_idx;
+        ppos -= path[ppos][cb].run;
+        stack_len++;
+    }
+    //perform actual band info encoding
+    start = 0;
+    for (i = stack_len - 1; i >= 0; i--) {
+        cb = aac_cb_out_map[stackcb[i]];
+        put_bits(&s->pb, 4, cb);
+        count = stackrun[i];
+        memset(sce->zeroes + win*16 + start, !cb, count);
+        //XXX: memset when band_type is also uint8_t
+        for (j = 0; j < count; j++) {
+            sce->band_type[win*16 + start] = cb;
+            start++;
+        }
+        while (count >= run_esc) {
+            put_bits(&s->pb, run_bits, run_esc);
+            count -= run_esc;
+        }
+        put_bits(&s->pb, run_bits, count);
+    }
+}
+
+
+#endif /* AVCODEC_AACCODER_TRELLIS_H */

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aaccoder_twoloop.h

@@ -0,0 +1,763 @@
+/*
+ * AAC encoder twoloop coder
+ * Copyright (C) 2008-2009 Konstantin Shishkov
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+/**
+ * @file
+ * AAC encoder twoloop coder
+ * @author Konstantin Shishkov, Claudio Freire
+ */
+
+/**
+ * This file contains a template for the twoloop coder function.
+ * It needs to be provided, externally, as an already included declaration,
+ * the following functions from aacenc_quantization/util.h. They're not included
+ * explicitly here to make it possible to provide alternative implementations:
+ *  - quantize_band_cost
+ *  - abs_pow34_v
+ *  - find_max_val
+ *  - find_min_book
+ *  - find_form_factor
+ */
+
+#ifndef AVCODEC_AACCODER_TWOLOOP_H
+#define AVCODEC_AACCODER_TWOLOOP_H
+
+#include <float.h>
+#include "libavutil/mathematics.h"
+#include "mathops.h"
+#include "avcodec.h"
+#include "put_bits.h"
+#include "aac.h"
+#include "aacenc.h"
+#include "aactab.h"
+#include "aacenctab.h"
+
+/** Frequency in Hz for lower limit of noise substitution **/
+#define NOISE_LOW_LIMIT 4000
+
+#define sclip(x) av_clip(x,60,218)
+
+/* Reflects the cost to change codebooks */
+static inline int ff_pns_bits(SingleChannelElement *sce, int w, int g)
+{
+    return (!g || !sce->zeroes[w*16+g-1] || !sce->can_pns[w*16+g-1]) ? 9 : 5;
+}
+
+/**
+ * two-loop quantizers search taken from ISO 13818-7 Appendix C
+ */
+static void search_for_quantizers_twoloop(AVCodecContext *avctx,
+                                          AACEncContext *s,
+                                          SingleChannelElement *sce,
+                                          const float lambda)
+{
+    int start = 0, i, w, w2, g, recomprd;
+    int destbits = avctx->bit_rate * 1024.0 / avctx->sample_rate
+        / ((avctx->flags & AV_CODEC_FLAG_QSCALE) ? 2.0f : avctx->channels)
+        * (lambda / 120.f);
+    int refbits = destbits;
+    int toomanybits, toofewbits;
+    char nzs[128];
+    uint8_t nextband[128];
+    int maxsf[128], minsf[128];
+    float dists[128] = { 0 }, qenergies[128] = { 0 }, uplims[128], euplims[128], energies[128];
+    float maxvals[128], spread_thr_r[128];
+    float min_spread_thr_r, max_spread_thr_r;
+
+    /**
+     * rdlambda controls the maximum tolerated distortion. Twoloop
+     * will keep iterating until it fails to lower it or it reaches
+     * ulimit * rdlambda. Keeping it low increases quality on difficult
+     * signals, but lower it too much, and bits will be taken from weak
+     * signals, creating "holes". A balance is necessary.
+     * rdmax and rdmin specify the relative deviation from rdlambda
+     * allowed for tonality compensation
+     */
+    float rdlambda = av_clipf(2.0f * 120.f / lambda, 0.0625f, 16.0f);
+    const float nzslope = 1.5f;
+    float rdmin = 0.03125f;
+    float rdmax = 1.0f;
+
+    /**
+     * sfoffs controls an offset of optmium allocation that will be
+     * applied based on lambda. Keep it real and modest, the loop
+     * will take care of the rest, this just accelerates convergence
+     */
+    float sfoffs = av_clipf(log2f(120.0f / lambda) * 4.0f, -5, 10);
+
+    int fflag, minscaler, maxscaler, nminscaler;
+    int its  = 0;
+    int maxits = 30;
+    int allz = 0;
+    int tbits;
+    int cutoff = 1024;
+    int pns_start_pos;
+    int prev;
+
+    /**
+     * zeroscale controls a multiplier of the threshold, if band energy
+     * is below this, a zero is forced. Keep it lower than 1, unless
+     * low lambda is used, because energy < threshold doesn't mean there's
+     * no audible signal outright, it's just energy. Also make it rise
+     * slower than rdlambda, as rdscale has due compensation with
+     * noisy band depriorization below, whereas zeroing logic is rather dumb
+     */
+    float zeroscale;
+    if (lambda > 120.f) {
+        zeroscale = av_clipf(powf(120.f / lambda, 0.25f), 0.0625f, 1.0f);
+    } else {
+        zeroscale = 1.f;
+    }
+
+    if (s->psy.bitres.alloc >= 0) {
+        /**
+         * Psy granted us extra bits to use, from the reservoire
+         * adjust for lambda except what psy already did
+         */
+        destbits = s->psy.bitres.alloc
+            * (lambda / (avctx->global_quality ? avctx->global_quality : 120));
+    }
+
+    if (avctx->flags & AV_CODEC_FLAG_QSCALE) {
+        /**
+         * Constant Q-scale doesn't compensate MS coding on its own
+         * No need to be overly precise, this only controls RD
+         * adjustment CB limits when going overboard
+         */
+        if (s->options.mid_side && s->cur_type == TYPE_CPE)
+            destbits *= 2;
+
+        /**
+         * When using a constant Q-scale, don't adjust bits, just use RD
+         * Don't let it go overboard, though... 8x psy target is enough
+         */
+        toomanybits = 5800;
+        toofewbits = destbits / 16;
+
+        /** Don't offset scalers, just RD */
+        sfoffs = sce->ics.num_windows - 1;
+        rdlambda = sqrtf(rdlambda);
+
+        /** search further */
+        maxits *= 2;
+    } else {
+        /* When using ABR, be strict, but a reasonable leeway is
+         * critical to allow RC to smoothly track desired bitrate
+         * without sudden quality drops that cause audible artifacts.
+         * Symmetry is also desirable, to avoid systematic bias.
+         */
+        toomanybits = destbits + destbits/8;
+        toofewbits = destbits - destbits/8;
+
+        sfoffs = 0;
+        rdlambda = sqrtf(rdlambda);
+    }
+
+    /** and zero out above cutoff frequency */
+    {
+        int wlen = 1024 / sce->ics.num_windows;
+        int bandwidth;
+
+        /**
+         * Scale, psy gives us constant quality, this LP only scales
+         * bitrate by lambda, so we save bits on subjectively unimportant HF
+         * rather than increase quantization noise. Adjust nominal bitrate
+         * to effective bitrate according to encoding parameters,
+         * AAC_CUTOFF_FROM_BITRATE is calibrated for effective bitrate.
+         */
+        float rate_bandwidth_multiplier = 1.5f;
+        int frame_bit_rate = (avctx->flags & AV_CODEC_FLAG_QSCALE)
+            ? (refbits * rate_bandwidth_multiplier * avctx->sample_rate / 1024)
+            : (avctx->bit_rate / avctx->channels);
+
+        /** Compensate for extensions that increase efficiency */
+        if (s->options.pns || s->options.intensity_stereo)
+            frame_bit_rate *= 1.15f;
+
+        if (avctx->cutoff > 0) {
+            bandwidth = avctx->cutoff;
+        } else {
+            bandwidth = FFMAX(3000, AAC_CUTOFF_FROM_BITRATE(frame_bit_rate, 1, avctx->sample_rate));
+            s->psy.cutoff = bandwidth;
+        }
+
+        cutoff = bandwidth * 2 * wlen / avctx->sample_rate;
+        pns_start_pos = NOISE_LOW_LIMIT * 2 * wlen / avctx->sample_rate;
+    }
+
+    /**
+     * for values above this the decoder might end up in an endless loop
+     * due to always having more bits than what can be encoded.
+     */
+    destbits = FFMIN(destbits, 5800);
+    toomanybits = FFMIN(toomanybits, 5800);
+    toofewbits = FFMIN(toofewbits, 5800);
+    /**
+     * XXX: some heuristic to determine initial quantizers will reduce search time
+     * determine zero bands and upper distortion limits
+     */
+    min_spread_thr_r = -1;
+    max_spread_thr_r = -1;
+    for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) {
+        for (g = start = 0;  g < sce->ics.num_swb; start += sce->ics.swb_sizes[g++]) {
+            int nz = 0;
+            float uplim = 0.0f, energy = 0.0f, spread = 0.0f;
+            for (w2 = 0; w2 < sce->ics.group_len[w]; w2++) {
+                FFPsyBand *band = &s->psy.ch[s->cur_channel].psy_bands[(w+w2)*16+g];
+                if (start >= cutoff || band->energy <= (band->threshold * zeroscale) || band->threshold == 0.0f) {
+                    sce->zeroes[(w+w2)*16+g] = 1;
+                    continue;
+                }
+                nz = 1;
+            }
+            if (!nz) {
+                uplim = 0.0f;
+            } else {
+                nz = 0;
+                for (w2 = 0; w2 < sce->ics.group_len[w]; w2++) {
+                    FFPsyBand *band = &s->psy.ch[s->cur_channel].psy_bands[(w+w2)*16+g];
+                    if (band->energy <= (band->threshold * zeroscale) || band->threshold == 0.0f)
+                        continue;
+                    uplim += band->threshold;
+                    energy += band->energy;
+                    spread += band->spread;
+                    nz++;
+                }
+            }
+            uplims[w*16+g] = uplim;
+            energies[w*16+g] = energy;
+            nzs[w*16+g] = nz;
+            sce->zeroes[w*16+g] = !nz;
+            allz |= nz;
+            if (nz && sce->can_pns[w*16+g]) {
+                spread_thr_r[w*16+g] = energy * nz / (uplim * spread);
+                if (min_spread_thr_r < 0) {
+                    min_spread_thr_r = max_spread_thr_r = spread_thr_r[w*16+g];
+                } else {
+                    min_spread_thr_r = FFMIN(min_spread_thr_r, spread_thr_r[w*16+g]);
+                    max_spread_thr_r = FFMAX(max_spread_thr_r, spread_thr_r[w*16+g]);
+                }
+            }
+        }
+    }
+
+    /** Compute initial scalers */
+    minscaler = 65535;
+    for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) {
+        for (g = 0;  g < sce->ics.num_swb; g++) {
+            if (sce->zeroes[w*16+g]) {
+                sce->sf_idx[w*16+g] = SCALE_ONE_POS;
+                continue;
+            }
+            /**
+             * log2f-to-distortion ratio is, technically, 2 (1.5db = 4, but it's power vs level so it's 2).
+             * But, as offsets are applied, low-frequency signals are too sensitive to the induced distortion,
+             * so we make scaling more conservative by choosing a lower log2f-to-distortion ratio, and thus
+             * more robust.
+             */
+            sce->sf_idx[w*16+g] = av_clip(
+                SCALE_ONE_POS
+                    + 1.75*log2f(FFMAX(0.00125f,uplims[w*16+g]) / sce->ics.swb_sizes[g])
+                    + sfoffs,
+                60, SCALE_MAX_POS);
+            minscaler = FFMIN(minscaler, sce->sf_idx[w*16+g]);
+        }
+    }
+
+    /** Clip */
+    minscaler = av_clip(minscaler, SCALE_ONE_POS - SCALE_DIV_512, SCALE_MAX_POS - SCALE_DIV_512);
+    for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w])
+        for (g = 0;  g < sce->ics.num_swb; g++)
+            if (!sce->zeroes[w*16+g])
+                sce->sf_idx[w*16+g] = av_clip(sce->sf_idx[w*16+g], minscaler, minscaler + SCALE_MAX_DIFF - 1);
+
+    if (!allz)
+        return;
+    s->abs_pow34(s->scoefs, sce->coeffs, 1024);
+    ff_quantize_band_cost_cache_init(s);
+
+    for (i = 0; i < sizeof(minsf) / sizeof(minsf[0]); ++i)
+        minsf[i] = 0;
+    for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) {
+        start = w*128;
+        for (g = 0;  g < sce->ics.num_swb; g++) {
+            const float *scaled = s->scoefs + start;
+            int minsfidx;
+            maxvals[w*16+g] = find_max_val(sce->ics.group_len[w], sce->ics.swb_sizes[g], scaled);
+            if (maxvals[w*16+g] > 0) {
+                minsfidx = coef2minsf(maxvals[w*16+g]);
+                for (w2 = 0; w2 < sce->ics.group_len[w]; w2++)
+                    minsf[(w+w2)*16+g] = minsfidx;
+            }
+            start += sce->ics.swb_sizes[g];
+        }
+    }
+
+    /**
+     * Scale uplims to match rate distortion to quality
+     * bu applying noisy band depriorization and tonal band priorization.
+     * Maxval-energy ratio gives us an idea of how noisy/tonal the band is.
+     * If maxval^2 ~ energy, then that band is mostly noise, and we can relax
+     * rate distortion requirements.
+     */
+    memcpy(euplims, uplims, sizeof(euplims));
+    for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) {
+        /** psy already priorizes transients to some extent */
+        float de_psy_factor = (sce->ics.num_windows > 1) ? 8.0f / sce->ics.group_len[w] : 1.0f;
+        start = w*128;
+        for (g = 0;  g < sce->ics.num_swb; g++) {
+            if (nzs[g] > 0) {
+                float cleanup_factor = ff_sqrf(av_clipf(start / (cutoff * 0.75f), 1.0f, 2.0f));
+                float energy2uplim = find_form_factor(
+                    sce->ics.group_len[w], sce->ics.swb_sizes[g],
+                    uplims[w*16+g] / (nzs[g] * sce->ics.swb_sizes[w]),
+                    sce->coeffs + start,
+                    nzslope * cleanup_factor);
+                energy2uplim *= de_psy_factor;
+                if (!(avctx->flags & AV_CODEC_FLAG_QSCALE)) {
+                    /** In ABR, we need to priorize less and let rate control do its thing */
+                    energy2uplim = sqrtf(energy2uplim);
+                }
+                energy2uplim = FFMAX(0.015625f, FFMIN(1.0f, energy2uplim));
+                uplims[w*16+g] *= av_clipf(rdlambda * energy2uplim, rdmin, rdmax)
+                                  * sce->ics.group_len[w];
+
+                energy2uplim = find_form_factor(
+                    sce->ics.group_len[w], sce->ics.swb_sizes[g],
+                    uplims[w*16+g] / (nzs[g] * sce->ics.swb_sizes[w]),
+                    sce->coeffs + start,
+                    2.0f);
+                energy2uplim *= de_psy_factor;
+                if (!(avctx->flags & AV_CODEC_FLAG_QSCALE)) {
+                    /** In ABR, we need to priorize less and let rate control do its thing */
+                    energy2uplim = sqrtf(energy2uplim);
+                }
+                energy2uplim = FFMAX(0.015625f, FFMIN(1.0f, energy2uplim));
+                euplims[w*16+g] *= av_clipf(rdlambda * energy2uplim * sce->ics.group_len[w],
+                    0.5f, 1.0f);
+            }
+            start += sce->ics.swb_sizes[g];
+        }
+    }
+
+    for (i = 0; i < sizeof(maxsf) / sizeof(maxsf[0]); ++i)
+        maxsf[i] = SCALE_MAX_POS;
+
+    //perform two-loop search
+    //outer loop - improve quality
+    do {
+        //inner loop - quantize spectrum to fit into given number of bits
+        int overdist;
+        int qstep = its ? 1 : 32;
+        do {
+            int changed = 0;
+            prev = -1;
+            recomprd = 0;
+            tbits = 0;
+            for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) {
+                start = w*128;
+                for (g = 0;  g < sce->ics.num_swb; g++) {
+                    const float *coefs = &sce->coeffs[start];
+                    const float *scaled = &s->scoefs[start];
+                    int bits = 0;
+                    int cb;
+                    float dist = 0.0f;
+                    float qenergy = 0.0f;
+
+                    if (sce->zeroes[w*16+g] || sce->sf_idx[w*16+g] >= 218) {
+                        start += sce->ics.swb_sizes[g];
+                        if (sce->can_pns[w*16+g]) {
+                            /** PNS isn't free */
+                            tbits += ff_pns_bits(sce, w, g);
+                        }
+                        continue;
+                    }
+                    cb = find_min_book(maxvals[w*16+g], sce->sf_idx[w*16+g]);
+                    for (w2 = 0; w2 < sce->ics.group_len[w]; w2++) {
+                        int b;
+                        float sqenergy;
+                        dist += quantize_band_cost_cached(s, w + w2, g, coefs + w2*128,
+                                                   scaled + w2*128,
+                                                   sce->ics.swb_sizes[g],
+                                                   sce->sf_idx[w*16+g],
+                                                   cb,
+                                                   1.0f,
+                                                   INFINITY,
+                                                   &b, &sqenergy,
+                                                   0);
+                        bits += b;
+                        qenergy += sqenergy;
+                    }
+                    dists[w*16+g] = dist - bits;
+                    qenergies[w*16+g] = qenergy;
+                    if (prev != -1) {
+                        int sfdiff = av_clip(sce->sf_idx[w*16+g] - prev + SCALE_DIFF_ZERO, 0, 2*SCALE_MAX_DIFF);
+                        bits += ff_aac_scalefactor_bits[sfdiff];
+                    }
+                    tbits += bits;
+                    start += sce->ics.swb_sizes[g];
+                    prev = sce->sf_idx[w*16+g];
+                }
+            }
+            if (tbits > toomanybits) {
+                recomprd = 1;
+                for (i = 0; i < 128; i++) {
+                    if (sce->sf_idx[i] < (SCALE_MAX_POS - SCALE_DIV_512)) {
+                        int maxsf_i = (tbits > 5800) ? SCALE_MAX_POS : maxsf[i];
+                        int new_sf = FFMIN(maxsf_i, sce->sf_idx[i] + qstep);
+                        if (new_sf != sce->sf_idx[i]) {
+                            sce->sf_idx[i] = new_sf;
+                            changed = 1;
+                        }
+                    }
+                }
+            } else if (tbits < toofewbits) {
+                recomprd = 1;
+                for (i = 0; i < 128; i++) {
+                    if (sce->sf_idx[i] > SCALE_ONE_POS) {
+                        int new_sf = FFMAX3(minsf[i], SCALE_ONE_POS, sce->sf_idx[i] - qstep);
+                        if (new_sf != sce->sf_idx[i]) {
+                            sce->sf_idx[i] = new_sf;
+                            changed = 1;
+                        }
+                    }
+                }
+            }
+            qstep >>= 1;
+            if (!qstep && tbits > toomanybits && sce->sf_idx[0] < 217 && changed)
+                qstep = 1;
+        } while (qstep);
+
+        overdist = 1;
+        fflag = tbits < toofewbits;
+        for (i = 0; i < 2 && (overdist || recomprd); ++i) {
+            if (recomprd) {
+                /** Must recompute distortion */
+                prev = -1;
+                tbits = 0;
+                for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) {
+                    start = w*128;
+                    for (g = 0;  g < sce->ics.num_swb; g++) {
+                        const float *coefs = sce->coeffs + start;
+                        const float *scaled = s->scoefs + start;
+                        int bits = 0;
+                        int cb;
+                        float dist = 0.0f;
+                        float qenergy = 0.0f;
+
+                        if (sce->zeroes[w*16+g] || sce->sf_idx[w*16+g] >= 218) {
+                            start += sce->ics.swb_sizes[g];
+                            if (sce->can_pns[w*16+g]) {
+                                /** PNS isn't free */
+                                tbits += ff_pns_bits(sce, w, g);
+                            }
+                            continue;
+                        }
+                        cb = find_min_book(maxvals[w*16+g], sce->sf_idx[w*16+g]);
+                        for (w2 = 0; w2 < sce->ics.group_len[w]; w2++) {
+                            int b;
+                            float sqenergy;
+                            dist += quantize_band_cost_cached(s, w + w2, g, coefs + w2*128,
+                                                    scaled + w2*128,
+                                                    sce->ics.swb_sizes[g],
+                                                    sce->sf_idx[w*16+g],
+                                                    cb,
+                                                    1.0f,
+                                                    INFINITY,
+                                                    &b, &sqenergy,
+                                                    0);
+                            bits += b;
+                            qenergy += sqenergy;
+                        }
+                        dists[w*16+g] = dist - bits;
+                        qenergies[w*16+g] = qenergy;
+                        if (prev != -1) {
+                            int sfdiff = av_clip(sce->sf_idx[w*16+g] - prev + SCALE_DIFF_ZERO, 0, 2*SCALE_MAX_DIFF);
+                            bits += ff_aac_scalefactor_bits[sfdiff];
+                        }
+                        tbits += bits;
+                        start += sce->ics.swb_sizes[g];
+                        prev = sce->sf_idx[w*16+g];
+                    }
+                }
+            }
+            if (!i && s->options.pns && its > maxits/2 && tbits > toofewbits) {
+                float maxoverdist = 0.0f;
+                float ovrfactor = 1.f+(maxits-its)*16.f/maxits;
+                overdist = recomprd = 0;
+                for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) {
+                    for (g = start = 0;  g < sce->ics.num_swb; start += sce->ics.swb_sizes[g++]) {
+                        if (!sce->zeroes[w*16+g] && sce->sf_idx[w*16+g] > SCALE_ONE_POS && dists[w*16+g] > uplims[w*16+g]*ovrfactor) {
+                            float ovrdist = dists[w*16+g] / FFMAX(uplims[w*16+g],euplims[w*16+g]);
+                            maxoverdist = FFMAX(maxoverdist, ovrdist);
+                            overdist++;
+                        }
+                    }
+                }
+                if (overdist) {
+                    /* We have overdistorted bands, trade for zeroes (that can be noise)
+                     * Zero the bands in the lowest 1.25% spread-energy-threshold ranking
+                     */
+                    float minspread = max_spread_thr_r;
+                    float maxspread = min_spread_thr_r;
+                    float zspread;
+                    int zeroable = 0;
+                    int zeroed = 0;
+                    int maxzeroed, zloop;
+                    for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) {
+                        for (g = start = 0;  g < sce->ics.num_swb; start += sce->ics.swb_sizes[g++]) {
+                            if (start >= pns_start_pos && !sce->zeroes[w*16+g] && sce->can_pns[w*16+g]) {
+                                minspread = FFMIN(minspread, spread_thr_r[w*16+g]);
+                                maxspread = FFMAX(maxspread, spread_thr_r[w*16+g]);
+                                zeroable++;
+                            }
+                        }
+                    }
+                    zspread = (maxspread-minspread) * 0.0125f + minspread;
+                    /* Don't PNS everything even if allowed. It suppresses bit starvation signals from RC,
+                     * and forced the hand of the later search_for_pns step.
+                     * Instead, PNS a fraction of the spread_thr_r range depending on how starved for bits we are,
+                     * and leave further PNSing to search_for_pns if worthwhile.
+                     */
+                    zspread = FFMIN3(min_spread_thr_r * 8.f, zspread,
+                        ((toomanybits - tbits) * min_spread_thr_r + (tbits - toofewbits) * max_spread_thr_r) / (toomanybits - toofewbits + 1));
+                    maxzeroed = FFMIN(zeroable, FFMAX(1, (zeroable * its + maxits - 1) / (2 * maxits)));
+                    for (zloop = 0; zloop < 2; zloop++) {
+                        /* Two passes: first distorted stuff - two birds in one shot and all that,
+                         * then anything viable. Viable means not zero, but either CB=zero-able
+                         * (too high SF), not SF <= 1 (that means we'd be operating at very high
+                         * quality, we don't want PNS when doing VHQ), PNS allowed, and within
+                         * the lowest ranking percentile.
+                         */
+                        float loopovrfactor = (zloop) ? 1.0f : ovrfactor;
+                        int loopminsf = (zloop) ? (SCALE_ONE_POS - SCALE_DIV_512) : SCALE_ONE_POS;
+                        int mcb;
+                        for (g = sce->ics.num_swb-1; g > 0 && zeroed < maxzeroed; g--) {
+                            if (sce->ics.swb_offset[g] < pns_start_pos)
+                                continue;
+                            for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) {
+                                if (!sce->zeroes[w*16+g] && sce->can_pns[w*16+g] && spread_thr_r[w*16+g] <= zspread
+                                    && sce->sf_idx[w*16+g] > loopminsf
+                                    && (dists[w*16+g] > loopovrfactor*uplims[w*16+g] || !(mcb = find_min_book(maxvals[w*16+g], sce->sf_idx[w*16+g]))
+                                        || (mcb <= 1 && dists[w*16+g] > FFMIN(uplims[w*16+g], euplims[w*16+g]))) ) {
+                                    sce->zeroes[w*16+g] = 1;
+                                    sce->band_type[w*16+g] = 0;
+                                    zeroed++;
+                                }
+                            }
+                        }
+                    }
+                    if (zeroed)
+                        recomprd = fflag = 1;
+                } else {
+                    overdist = 0;
+                }
+            }
+        }
+
+        minscaler = SCALE_MAX_POS;
+        maxscaler = 0;
+        for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) {
+            for (g = 0;  g < sce->ics.num_swb; g++) {
+                if (!sce->zeroes[w*16+g]) {
+                    minscaler = FFMIN(minscaler, sce->sf_idx[w*16+g]);
+                    maxscaler = FFMAX(maxscaler, sce->sf_idx[w*16+g]);
+                }
+            }
+        }
+
+        minscaler = nminscaler = av_clip(minscaler, SCALE_ONE_POS - SCALE_DIV_512, SCALE_MAX_POS - SCALE_DIV_512);
+        prev = -1;
+        for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) {
+            /** Start with big steps, end up fine-tunning */
+            int depth = (its > maxits/2) ? ((its > maxits*2/3) ? 1 : 3) : 10;
+            int edepth = depth+2;
+            float uplmax = its / (maxits*0.25f) + 1.0f;
+            uplmax *= (tbits > destbits) ? FFMIN(2.0f, tbits / (float)FFMAX(1,destbits)) : 1.0f;
+            start = w * 128;
+            for (g = 0; g < sce->ics.num_swb; g++) {
+                int prevsc = sce->sf_idx[w*16+g];
+                if (prev < 0 && !sce->zeroes[w*16+g])
+                    prev = sce->sf_idx[0];
+                if (!sce->zeroes[w*16+g]) {
+                    const float *coefs = sce->coeffs + start;
+                    const float *scaled = s->scoefs + start;
+                    int cmb = find_min_book(maxvals[w*16+g], sce->sf_idx[w*16+g]);
+                    int mindeltasf = FFMAX(0, prev - SCALE_MAX_DIFF);
+                    int maxdeltasf = FFMIN(SCALE_MAX_POS - SCALE_DIV_512, prev + SCALE_MAX_DIFF);
+                    if ((!cmb || dists[w*16+g] > uplims[w*16+g]) && sce->sf_idx[w*16+g] > FFMAX(mindeltasf, minsf[w*16+g])) {
+                        /* Try to make sure there is some energy in every nonzero band
+                         * NOTE: This algorithm must be forcibly imbalanced, pushing harder
+                         *  on holes or more distorted bands at first, otherwise there's
+                         *  no net gain (since the next iteration will offset all bands
+                         *  on the opposite direction to compensate for extra bits)
+                         */
+                        for (i = 0; i < edepth && sce->sf_idx[w*16+g] > mindeltasf; ++i) {
+                            int cb, bits;
+                            float dist, qenergy;
+                            int mb = find_min_book(maxvals[w*16+g], sce->sf_idx[w*16+g]-1);
+                            cb = find_min_book(maxvals[w*16+g], sce->sf_idx[w*16+g]);
+                            dist = qenergy = 0.f;
+                            bits = 0;
+                            if (!cb) {
+                                maxsf[w*16+g] = FFMIN(sce->sf_idx[w*16+g]-1, maxsf[w*16+g]);
+                            } else if (i >= depth && dists[w*16+g] < euplims[w*16+g]) {
+                                break;
+                            }
+                            /* !g is the DC band, it's important, since quantization error here
+                             * applies to less than a cycle, it creates horrible intermodulation
+                             * distortion if it doesn't stick to what psy requests
+                             */
+                            if (!g && sce->ics.num_windows > 1 && dists[w*16+g] >= euplims[w*16+g])
+                                maxsf[w*16+g] = FFMIN(sce->sf_idx[w*16+g], maxsf[w*16+g]);
+                            for (w2 = 0; w2 < sce->ics.group_len[w]; w2++) {
+                                int b;
+                                float sqenergy;
+                                dist += quantize_band_cost_cached(s, w + w2, g, coefs + w2*128,
+                                                        scaled + w2*128,
+                                                        sce->ics.swb_sizes[g],
+                                                        sce->sf_idx[w*16+g]-1,
+                                                        cb,
+                                                        1.0f,
+                                                        INFINITY,
+                                                        &b, &sqenergy,
+                                                        0);
+                                bits += b;
+                                qenergy += sqenergy;
+                            }
+                            sce->sf_idx[w*16+g]--;
+                            dists[w*16+g] = dist - bits;
+                            qenergies[w*16+g] = qenergy;
+                            if (mb && (sce->sf_idx[w*16+g] < mindeltasf || (
+                                    (dists[w*16+g] < FFMIN(uplmax*uplims[w*16+g], euplims[w*16+g]))
+                                    && (fabsf(qenergies[w*16+g]-energies[w*16+g]) < euplims[w*16+g])
+                                ) )) {
+                                break;
+                            }
+                        }
+                    } else if (tbits > toofewbits && sce->sf_idx[w*16+g] < FFMIN(maxdeltasf, maxsf[w*16+g])
+                            && (dists[w*16+g] < FFMIN(euplims[w*16+g], uplims[w*16+g]))
+                            && (fabsf(qenergies[w*16+g]-energies[w*16+g]) < euplims[w*16+g])
+                        ) {
+                        /** Um... over target. Save bits for more important stuff. */
+                        for (i = 0; i < depth && sce->sf_idx[w*16+g] < maxdeltasf; ++i) {
+                            int cb, bits;
+                            float dist, qenergy;
+                            cb = find_min_book(maxvals[w*16+g], sce->sf_idx[w*16+g]+1);
+                            if (cb > 0) {
+                                dist = qenergy = 0.f;
+                                bits = 0;
+                                for (w2 = 0; w2 < sce->ics.group_len[w]; w2++) {
+                                    int b;
+                                    float sqenergy;
+                                    dist += quantize_band_cost_cached(s, w + w2, g, coefs + w2*128,
+                                                            scaled + w2*128,
+                                                            sce->ics.swb_sizes[g],
+                                                            sce->sf_idx[w*16+g]+1,
+                                                            cb,
+                                                            1.0f,
+                                                            INFINITY,
+                                                            &b, &sqenergy,
+                                                            0);
+                                    bits += b;
+                                    qenergy += sqenergy;
+                                }
+                                dist -= bits;
+                                if (dist < FFMIN(euplims[w*16+g], uplims[w*16+g])) {
+                                    sce->sf_idx[w*16+g]++;
+                                    dists[w*16+g] = dist;
+                                    qenergies[w*16+g] = qenergy;
+                                } else {
+                                    break;
+                                }
+                            } else {
+                                maxsf[w*16+g] = FFMIN(sce->sf_idx[w*16+g], maxsf[w*16+g]);
+                                break;
+                            }
+                        }
+                    }
+                    prev = sce->sf_idx[w*16+g] = av_clip(sce->sf_idx[w*16+g], mindeltasf, maxdeltasf);
+                    if (sce->sf_idx[w*16+g] != prevsc)
+                        fflag = 1;
+                    nminscaler = FFMIN(nminscaler, sce->sf_idx[w*16+g]);
+                    sce->band_type[w*16+g] = find_min_book(maxvals[w*16+g], sce->sf_idx[w*16+g]);
+                }
+                start += sce->ics.swb_sizes[g];
+            }
+        }
+
+        /** SF difference limit violation risk. Must re-clamp. */
+        prev = -1;
+        for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) {
+            for (g = 0; g < sce->ics.num_swb; g++) {
+                if (!sce->zeroes[w*16+g]) {
+                    int prevsf = sce->sf_idx[w*16+g];
+                    if (prev < 0)
+                        prev = prevsf;
+                    sce->sf_idx[w*16+g] = av_clip(sce->sf_idx[w*16+g], prev - SCALE_MAX_DIFF, prev + SCALE_MAX_DIFF);
+                    sce->band_type[w*16+g] = find_min_book(maxvals[w*16+g], sce->sf_idx[w*16+g]);
+                    prev = sce->sf_idx[w*16+g];
+                    if (!fflag && prevsf != sce->sf_idx[w*16+g])
+                        fflag = 1;
+                }
+            }
+        }
+
+        its++;
+    } while (fflag && its < maxits);
+
+    /** Scout out next nonzero bands */
+    ff_init_nextband_map(sce, nextband);
+
+    prev = -1;
+    for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) {
+        /** Make sure proper codebooks are set */
+        for (g = 0; g < sce->ics.num_swb; g++) {
+            if (!sce->zeroes[w*16+g]) {
+                sce->band_type[w*16+g] = find_min_book(maxvals[w*16+g], sce->sf_idx[w*16+g]);
+                if (sce->band_type[w*16+g] <= 0) {
+                    if (!ff_sfdelta_can_remove_band(sce, nextband, prev, w*16+g)) {
+                        /** Cannot zero out, make sure it's not attempted */
+                        sce->band_type[w*16+g] = 1;
+                    } else {
+                        sce->zeroes[w*16+g] = 1;
+                        sce->band_type[w*16+g] = 0;
+                    }
+                }
+            } else {
+                sce->band_type[w*16+g] = 0;
+            }
+            /** Check that there's no SF delta range violations */
+            if (!sce->zeroes[w*16+g]) {
+                if (prev != -1) {
+                    av_unused int sfdiff = sce->sf_idx[w*16+g] - prev + SCALE_DIFF_ZERO;
+                    av_assert1(sfdiff >= 0 && sfdiff <= 2*SCALE_MAX_DIFF);
+                } else if (sce->zeroes[0]) {
+                    /** Set global gain to something useful */
+                    sce->sf_idx[0] = sce->sf_idx[w*16+g];
+                }
+                prev = sce->sf_idx[w*16+g];
+            }
+        }
+    }
+}
+
+#endif /* AVCODEC_AACCODER_TWOLOOP_H */

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aacdec.c

@@ -0,0 +1,591 @@
+/*
+ * AAC decoder
+ * Copyright (c) 2005-2006 Oded Shimon ( ods15 ods15 dyndns org )
+ * Copyright (c) 2006-2007 Maxim Gavrilov ( maxim.gavrilov gmail com )
+ * Copyright (c) 2008-2013 Alex Converse <alex.converse@gmail.com>
+ *
+ * AAC LATM decoder
+ * Copyright (c) 2008-2010 Paul Kendall <paul@kcbbs.gen.nz>
+ * Copyright (c) 2010      Janne Grunau <janne-libav@jannau.net>
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+/**
+ * @file
+ * AAC decoder
+ * @author Oded Shimon  ( ods15 ods15 dyndns org )
+ * @author Maxim Gavrilov ( maxim.gavrilov gmail com )
+ */
+
+#define FFT_FLOAT 1
+#define FFT_FIXED_32 0
+#define USE_FIXED 0
+
+#include "libavutil/float_dsp.h"
+#include "libavutil/opt.h"
+#include "avcodec.h"
+#include "internal.h"
+#include "get_bits.h"
+#include "fft.h"
+#include "mdct15.h"
+#include "lpc.h"
+#include "kbdwin.h"
+#include "sinewin.h"
+
+#include "aac.h"
+#include "aactab.h"
+#include "aacdectab.h"
+#include "adts_header.h"
+#include "cbrt_data.h"
+#include "sbr.h"
+#include "aacsbr.h"
+#include "mpeg4audio.h"
+#include "profiles.h"
+#include "libavutil/intfloat.h"
+
+#include <errno.h>
+#include <math.h>
+#include <stdint.h>
+#include <string.h>
+
+#if ARCH_ARM
+#   include "arm/aac.h"
+#elif ARCH_MIPS
+#   include "mips/aacdec_mips.h"
+#endif
+
+static av_always_inline void reset_predict_state(PredictorState *ps)
+{
+    ps->r0   = 0.0f;
+    ps->r1   = 0.0f;
+    ps->cor0 = 0.0f;
+    ps->cor1 = 0.0f;
+    ps->var0 = 1.0f;
+    ps->var1 = 1.0f;
+}
+
+#ifndef VMUL2
+static inline float *VMUL2(float *dst, const float *v, unsigned idx,
+                           const float *scale)
+{
+    float s = *scale;
+    *dst++ = v[idx    & 15] * s;
+    *dst++ = v[idx>>4 & 15] * s;
+    return dst;
+}
+#endif
+
+#ifndef VMUL4
+static inline float *VMUL4(float *dst, const float *v, unsigned idx,
+                           const float *scale)
+{
+    float s = *scale;
+    *dst++ = v[idx    & 3] * s;
+    *dst++ = v[idx>>2 & 3] * s;
+    *dst++ = v[idx>>4 & 3] * s;
+    *dst++ = v[idx>>6 & 3] * s;
+    return dst;
+}
+#endif
+
+#ifndef VMUL2S
+static inline float *VMUL2S(float *dst, const float *v, unsigned idx,
+                            unsigned sign, const float *scale)
+{
+    union av_intfloat32 s0, s1;
+
+    s0.f = s1.f = *scale;
+    s0.i ^= sign >> 1 << 31;
+    s1.i ^= sign      << 31;
+
+    *dst++ = v[idx    & 15] * s0.f;
+    *dst++ = v[idx>>4 & 15] * s1.f;
+
+    return dst;
+}
+#endif
+
+#ifndef VMUL4S
+static inline float *VMUL4S(float *dst, const float *v, unsigned idx,
+                            unsigned sign, const float *scale)
+{
+    unsigned nz = idx >> 12;
+    union av_intfloat32 s = { .f = *scale };
+    union av_intfloat32 t;
+
+    t.i = s.i ^ (sign & 1U<<31);
+    *dst++ = v[idx    & 3] * t.f;
+
+    sign <<= nz & 1; nz >>= 1;
+    t.i = s.i ^ (sign & 1U<<31);
+    *dst++ = v[idx>>2 & 3] * t.f;
+
+    sign <<= nz & 1; nz >>= 1;
+    t.i = s.i ^ (sign & 1U<<31);
+    *dst++ = v[idx>>4 & 3] * t.f;
+
+    sign <<= nz & 1;
+    t.i = s.i ^ (sign & 1U<<31);
+    *dst++ = v[idx>>6 & 3] * t.f;
+
+    return dst;
+}
+#endif
+
+static av_always_inline float flt16_round(float pf)
+{
+    union av_intfloat32 tmp;
+    tmp.f = pf;
+    tmp.i = (tmp.i + 0x00008000U) & 0xFFFF0000U;
+    return tmp.f;
+}
+
+static av_always_inline float flt16_even(float pf)
+{
+    union av_intfloat32 tmp;
+    tmp.f = pf;
+    tmp.i = (tmp.i + 0x00007FFFU + (tmp.i & 0x00010000U >> 16)) & 0xFFFF0000U;
+    return tmp.f;
+}
+
+static av_always_inline float flt16_trunc(float pf)
+{
+    union av_intfloat32 pun;
+    pun.f = pf;
+    pun.i &= 0xFFFF0000U;
+    return pun.f;
+}
+
+static av_always_inline void predict(PredictorState *ps, float *coef,
+                                     int output_enable)
+{
+    const float a     = 0.953125; // 61.0 / 64
+    const float alpha = 0.90625;  // 29.0 / 32
+    float e0, e1;
+    float pv;
+    float k1, k2;
+    float   r0 = ps->r0,     r1 = ps->r1;
+    float cor0 = ps->cor0, cor1 = ps->cor1;
+    float var0 = ps->var0, var1 = ps->var1;
+
+    k1 = var0 > 1 ? cor0 * flt16_even(a / var0) : 0;
+    k2 = var1 > 1 ? cor1 * flt16_even(a / var1) : 0;
+
+    pv = flt16_round(k1 * r0 + k2 * r1);
+    if (output_enable)
+        *coef += pv;
+
+    e0 = *coef;
+    e1 = e0 - k1 * r0;
+
+    ps->cor1 = flt16_trunc(alpha * cor1 + r1 * e1);
+    ps->var1 = flt16_trunc(alpha * var1 + 0.5f * (r1 * r1 + e1 * e1));
+    ps->cor0 = flt16_trunc(alpha * cor0 + r0 * e0);
+    ps->var0 = flt16_trunc(alpha * var0 + 0.5f * (r0 * r0 + e0 * e0));
+
+    ps->r1 = flt16_trunc(a * (r0 - k1 * e0));
+    ps->r0 = flt16_trunc(a * e0);
+}
+
+/**
+ * Apply dependent channel coupling (applied before IMDCT).
+ *
+ * @param   index   index into coupling gain array
+ */
+static void apply_dependent_coupling(AACContext *ac,
+                                     SingleChannelElement *target,
+                                     ChannelElement *cce, int index)
+{
+    IndividualChannelStream *ics = &cce->ch[0].ics;
+    const uint16_t *offsets = ics->swb_offset;
+    float *dest = target->coeffs;
+    const float *src = cce->ch[0].coeffs;
+    int g, i, group, k, idx = 0;
+    if (ac->oc[1].m4ac.object_type == AOT_AAC_LTP) {
+        av_log(ac->avctx, AV_LOG_ERROR,
+               "Dependent coupling is not supported together with LTP\n");
+        return;
+    }
+    for (g = 0; g < ics->num_window_groups; g++) {
+        for (i = 0; i < ics->max_sfb; i++, idx++) {
+            if (cce->ch[0].band_type[idx] != ZERO_BT) {
+                const float gain = cce->coup.gain[index][idx];
+                for (group = 0; group < ics->group_len[g]; group++) {
+                    for (k = offsets[i]; k < offsets[i + 1]; k++) {
+                        // FIXME: SIMDify
+                        dest[group * 128 + k] += gain * src[group * 128 + k];
+                    }
+                }
+            }
+        }
+        dest += ics->group_len[g] * 128;
+        src  += ics->group_len[g] * 128;
+    }
+}
+
+/**
+ * Apply independent channel coupling (applied after IMDCT).
+ *
+ * @param   index   index into coupling gain array
+ */
+static void apply_independent_coupling(AACContext *ac,
+                                       SingleChannelElement *target,
+                                       ChannelElement *cce, int index)
+{
+    const float gain = cce->coup.gain[index][0];
+    const float *src = cce->ch[0].ret;
+    float *dest = target->ret;
+    const int len = 1024 << (ac->oc[1].m4ac.sbr == 1);
+
+    ac->fdsp->vector_fmac_scalar(dest, src, gain, len);
+}
+
+#include "aacdec_template.c"
+
+#define LOAS_SYNC_WORD   0x2b7       ///< 11 bits LOAS sync word
+
+struct LATMContext {
+    AACContext aac_ctx;     ///< containing AACContext
+    int initialized;        ///< initialized after a valid extradata was seen
+
+    // parser data
+    int audio_mux_version_A; ///< LATM syntax version
+    int frame_length_type;   ///< 0/1 variable/fixed frame length
+    int frame_length;        ///< frame length for fixed frame length
+};
+
+static inline uint32_t latm_get_value(GetBitContext *b)
+{
+    int length = get_bits(b, 2);
+
+    return get_bits_long(b, (length+1)*8);
+}
+
+static int latm_decode_audio_specific_config(struct LATMContext *latmctx,
+                                             GetBitContext *gb, int asclen)
+{
+    AACContext *ac        = &latmctx->aac_ctx;
+    AVCodecContext *avctx = ac->avctx;
+    MPEG4AudioConfig m4ac = { 0 };
+    GetBitContext gbc;
+    int config_start_bit  = get_bits_count(gb);
+    int sync_extension    = 0;
+    int bits_consumed, esize, i;
+
+    if (asclen > 0) {
+        sync_extension = 1;
+        asclen         = FFMIN(asclen, get_bits_left(gb));
+        init_get_bits(&gbc, gb->buffer, config_start_bit + asclen);
+        skip_bits_long(&gbc, config_start_bit);
+    } else if (asclen == 0) {
+        gbc = *gb;
+    } else {
+        return AVERROR_INVALIDDATA;
+    }
+
+    if (get_bits_left(gb) <= 0)
+        return AVERROR_INVALIDDATA;
+
+    bits_consumed = decode_audio_specific_config_gb(NULL, avctx, &m4ac,
+                                                    &gbc, config_start_bit,
+                                                    sync_extension);
+
+    if (bits_consumed < config_start_bit)
+        return AVERROR_INVALIDDATA;
+    bits_consumed -= config_start_bit;
+
+    if (asclen == 0)
+      asclen = bits_consumed;
+
+    if (!latmctx->initialized ||
+        ac->oc[1].m4ac.sample_rate != m4ac.sample_rate ||
+        ac->oc[1].m4ac.chan_config != m4ac.chan_config) {
+
+        if (latmctx->initialized) {
+            av_log(avctx, AV_LOG_INFO, "audio config changed (sample_rate=%d, chan_config=%d)\n", m4ac.sample_rate, m4ac.chan_config);
+        } else {
+            av_log(avctx, AV_LOG_DEBUG, "initializing latmctx\n");
+        }
+        latmctx->initialized = 0;
+
+        esize = (asclen + 7) / 8;
+
+        if (avctx->extradata_size < esize) {
+            av_free(avctx->extradata);
+            avctx->extradata = av_malloc(esize + AV_INPUT_BUFFER_PADDING_SIZE);
+            if (!avctx->extradata)
+                return AVERROR(ENOMEM);
+        }
+
+        avctx->extradata_size = esize;
+        gbc = *gb;
+        for (i = 0; i < esize; i++) {
+          avctx->extradata[i] = get_bits(&gbc, 8);
+        }
+        memset(avctx->extradata+esize, 0, AV_INPUT_BUFFER_PADDING_SIZE);
+    }
+    skip_bits_long(gb, asclen);
+
+    return 0;
+}
+
+static int read_stream_mux_config(struct LATMContext *latmctx,
+                                  GetBitContext *gb)
+{
+    int ret, audio_mux_version = get_bits(gb, 1);
+
+    latmctx->audio_mux_version_A = 0;
+    if (audio_mux_version)
+        latmctx->audio_mux_version_A = get_bits(gb, 1);
+
+    if (!latmctx->audio_mux_version_A) {
+
+        if (audio_mux_version)
+            latm_get_value(gb);                 // taraFullness
+
+        skip_bits(gb, 1);                       // allStreamSameTimeFraming
+        skip_bits(gb, 6);                       // numSubFrames
+        // numPrograms
+        if (get_bits(gb, 4)) {                  // numPrograms
+            avpriv_request_sample(latmctx->aac_ctx.avctx, "Multiple programs");
+            return AVERROR_PATCHWELCOME;
+        }
+
+        // for each program (which there is only one in DVB)
+
+        // for each layer (which there is only one in DVB)
+        if (get_bits(gb, 3)) {                   // numLayer
+            avpriv_request_sample(latmctx->aac_ctx.avctx, "Multiple layers");
+            return AVERROR_PATCHWELCOME;
+        }
+
+        // for all but first stream: use_same_config = get_bits(gb, 1);
+        if (!audio_mux_version) {
+            if ((ret = latm_decode_audio_specific_config(latmctx, gb, 0)) < 0)
+                return ret;
+        } else {
+            int ascLen = latm_get_value(gb);
+            if ((ret = latm_decode_audio_specific_config(latmctx, gb, ascLen)) < 0)
+                return ret;
+        }
+
+        latmctx->frame_length_type = get_bits(gb, 3);
+        switch (latmctx->frame_length_type) {
+        case 0:
+            skip_bits(gb, 8);       // latmBufferFullness
+            break;
+        case 1:
+            latmctx->frame_length = get_bits(gb, 9);
+            break;
+        case 3:
+        case 4:
+        case 5:
+            skip_bits(gb, 6);       // CELP frame length table index
+            break;
+        case 6:
+        case 7:
+            skip_bits(gb, 1);       // HVXC frame length table index
+            break;
+        }
+
+        if (get_bits(gb, 1)) {                  // other data
+            if (audio_mux_version) {
+                latm_get_value(gb);             // other_data_bits
+            } else {
+                int esc;
+                do {
+                    esc = get_bits(gb, 1);
+                    skip_bits(gb, 8);
+                } while (esc);
+            }
+        }
+
+        if (get_bits(gb, 1))                     // crc present
+            skip_bits(gb, 8);                    // config_crc
+    }
+
+    return 0;
+}
+
+static int read_payload_length_info(struct LATMContext *ctx, GetBitContext *gb)
+{
+    uint8_t tmp;
+
+    if (ctx->frame_length_type == 0) {
+        int mux_slot_length = 0;
+        do {
+            if (get_bits_left(gb) < 8)
+                return AVERROR_INVALIDDATA;
+            tmp = get_bits(gb, 8);
+            mux_slot_length += tmp;
+        } while (tmp == 255);
+        return mux_slot_length;
+    } else if (ctx->frame_length_type == 1) {
+        return ctx->frame_length;
+    } else if (ctx->frame_length_type == 3 ||
+               ctx->frame_length_type == 5 ||
+               ctx->frame_length_type == 7) {
+        skip_bits(gb, 2);          // mux_slot_length_coded
+    }
+    return 0;
+}
+
+static int read_audio_mux_element(struct LATMContext *latmctx,
+                                  GetBitContext *gb)
+{
+    int err;
+    uint8_t use_same_mux = get_bits(gb, 1);
+    if (!use_same_mux) {
+        if ((err = read_stream_mux_config(latmctx, gb)) < 0)
+            return err;
+    } else if (!latmctx->aac_ctx.avctx->extradata) {
+        av_log(latmctx->aac_ctx.avctx, AV_LOG_DEBUG,
+               "no decoder config found\n");
+        return 1;
+    }
+    if (latmctx->audio_mux_version_A == 0) {
+        int mux_slot_length_bytes = read_payload_length_info(latmctx, gb);
+        if (mux_slot_length_bytes < 0 || mux_slot_length_bytes * 8LL > get_bits_left(gb)) {
+            av_log(latmctx->aac_ctx.avctx, AV_LOG_ERROR, "incomplete frame\n");
+            return AVERROR_INVALIDDATA;
+        } else if (mux_slot_length_bytes * 8 + 256 < get_bits_left(gb)) {
+            av_log(latmctx->aac_ctx.avctx, AV_LOG_ERROR,
+                   "frame length mismatch %d << %d\n",
+                   mux_slot_length_bytes * 8, get_bits_left(gb));
+            return AVERROR_INVALIDDATA;
+        }
+    }
+    return 0;
+}
+
+
+static int latm_decode_frame(AVCodecContext *avctx, void *out,
+                             int *got_frame_ptr, AVPacket *avpkt)
+{
+    struct LATMContext *latmctx = avctx->priv_data;
+    int                 muxlength, err;
+    GetBitContext       gb;
+
+    if ((err = init_get_bits8(&gb, avpkt->data, avpkt->size)) < 0)
+        return err;
+
+    // check for LOAS sync word
+    if (get_bits(&gb, 11) != LOAS_SYNC_WORD)
+        return AVERROR_INVALIDDATA;
+
+    muxlength = get_bits(&gb, 13) + 3;
+    // not enough data, the parser should have sorted this out
+    if (muxlength > avpkt->size)
+        return AVERROR_INVALIDDATA;
+
+    if ((err = read_audio_mux_element(latmctx, &gb)))
+        return (err < 0) ? err : avpkt->size;
+
+    if (!latmctx->initialized) {
+        if (!avctx->extradata) {
+            *got_frame_ptr = 0;
+            return avpkt->size;
+        } else {
+            push_output_configuration(&latmctx->aac_ctx);
+            if ((err = decode_audio_specific_config(
+                    &latmctx->aac_ctx, avctx, &latmctx->aac_ctx.oc[1].m4ac,
+                    avctx->extradata, avctx->extradata_size*8LL, 1)) < 0) {
+                pop_output_configuration(&latmctx->aac_ctx);
+                return err;
+            }
+            latmctx->initialized = 1;
+        }
+    }
+
+    if (show_bits(&gb, 12) == 0xfff) {
+        av_log(latmctx->aac_ctx.avctx, AV_LOG_ERROR,
+               "ADTS header detected, probably as result of configuration "
+               "misparsing\n");
+        return AVERROR_INVALIDDATA;
+    }
+
+    switch (latmctx->aac_ctx.oc[1].m4ac.object_type) {
+    case AOT_ER_AAC_LC:
+    case AOT_ER_AAC_LTP:
+    case AOT_ER_AAC_LD:
+    case AOT_ER_AAC_ELD:
+        err = aac_decode_er_frame(avctx, out, got_frame_ptr, &gb);
+        break;
+    default:
+        err = aac_decode_frame_int(avctx, out, got_frame_ptr, &gb, avpkt);
+    }
+    if (err < 0)
+        return err;
+
+    return muxlength;
+}
+
+static av_cold int latm_decode_init(AVCodecContext *avctx)
+{
+    struct LATMContext *latmctx = avctx->priv_data;
+    int ret = aac_decode_init(avctx);
+
+    if (avctx->extradata_size > 0)
+        latmctx->initialized = !ret;
+
+    return ret;
+}
+
+AVCodec ff_aac_decoder = {
+    .name            = "aac",
+    .long_name       = NULL_IF_CONFIG_SMALL("AAC (Advanced Audio Coding)"),
+    .type            = AVMEDIA_TYPE_AUDIO,
+    .id              = AV_CODEC_ID_AAC,
+    .priv_data_size  = sizeof(AACContext),
+    .init            = aac_decode_init,
+    .close           = aac_decode_close,
+    .decode          = aac_decode_frame,
+    .sample_fmts     = (const enum AVSampleFormat[]) {
+        AV_SAMPLE_FMT_FLTP, AV_SAMPLE_FMT_NONE
+    },
+    .capabilities    = AV_CODEC_CAP_CHANNEL_CONF | AV_CODEC_CAP_DR1,
+    .caps_internal   = FF_CODEC_CAP_INIT_THREADSAFE,
+    .channel_layouts = aac_channel_layout,
+    .flush = flush,
+    .priv_class      = &aac_decoder_class,
+    .profiles        = NULL_IF_CONFIG_SMALL(ff_aac_profiles),
+};
+
+/*
+    Note: This decoder filter is intended to decode LATM streams transferred
+    in MPEG transport streams which only contain one program.
+    To do a more complex LATM demuxing a separate LATM demuxer should be used.
+*/
+AVCodec ff_aac_latm_decoder = {
+    .name            = "aac_latm",
+    .long_name       = NULL_IF_CONFIG_SMALL("AAC LATM (Advanced Audio Coding LATM syntax)"),
+    .type            = AVMEDIA_TYPE_AUDIO,
+    .id              = AV_CODEC_ID_AAC_LATM,
+    .priv_data_size  = sizeof(struct LATMContext),
+    .init            = latm_decode_init,
+    .close           = aac_decode_close,
+    .decode          = latm_decode_frame,
+    .sample_fmts     = (const enum AVSampleFormat[]) {
+        AV_SAMPLE_FMT_FLTP, AV_SAMPLE_FMT_NONE
+    },
+    .capabilities    = AV_CODEC_CAP_CHANNEL_CONF | AV_CODEC_CAP_DR1,
+    .caps_internal   = FF_CODEC_CAP_INIT_THREADSAFE,
+    .channel_layouts = aac_channel_layout,
+    .flush = flush,
+    .profiles        = NULL_IF_CONFIG_SMALL(ff_aac_profiles),
+};

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aacdec_fixed.c

@@ -0,0 +1,468 @@
+/*
+ * Copyright (c) 2013
+ *      MIPS Technologies, Inc., California.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ * 3. Neither the name of the MIPS Technologies, Inc., nor the names of its
+ *    contributors may be used to endorse or promote products derived from
+ *    this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE MIPS TECHNOLOGIES, INC. ``AS IS'' AND
+ * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE MIPS TECHNOLOGIES, INC. BE LIABLE
+ * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
+ * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
+ * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
+ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
+ * SUCH DAMAGE.
+ *
+ * AAC decoder fixed-point implementation
+ *
+ * Copyright (c) 2005-2006 Oded Shimon ( ods15 ods15 dyndns org )
+ * Copyright (c) 2006-2007 Maxim Gavrilov ( maxim.gavrilov gmail com )
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+/**
+ * @file
+ * AAC decoder
+ * @author Oded Shimon  ( ods15 ods15 dyndns org )
+ * @author Maxim Gavrilov ( maxim.gavrilov gmail com )
+ *
+ * Fixed point implementation
+ * @author Stanislav Ocovaj ( stanislav.ocovaj imgtec com )
+ */
+
+#define FFT_FLOAT 0
+#define FFT_FIXED_32 1
+#define USE_FIXED 1
+
+#include "libavutil/fixed_dsp.h"
+#include "libavutil/opt.h"
+#include "avcodec.h"
+#include "internal.h"
+#include "get_bits.h"
+#include "fft.h"
+#include "lpc.h"
+#include "kbdwin.h"
+#include "sinewin.h"
+
+#include "aac.h"
+#include "aactab.h"
+#include "aacdectab.h"
+#include "adts_header.h"
+#include "cbrt_data.h"
+#include "sbr.h"
+#include "aacsbr.h"
+#include "mpeg4audio.h"
+#include "profiles.h"
+#include "libavutil/intfloat.h"
+
+#include <math.h>
+#include <string.h>
+
+static av_always_inline void reset_predict_state(PredictorState *ps)
+{
+    ps->r0.mant   = 0;
+    ps->r0.exp   = 0;
+    ps->r1.mant   = 0;
+    ps->r1.exp   = 0;
+    ps->cor0.mant = 0;
+    ps->cor0.exp = 0;
+    ps->cor1.mant = 0;
+    ps->cor1.exp = 0;
+    ps->var0.mant = 0x20000000;
+    ps->var0.exp = 1;
+    ps->var1.mant = 0x20000000;
+    ps->var1.exp = 1;
+}
+
+static const int exp2tab[4] = { Q31(1.0000000000/2), Q31(1.1892071150/2), Q31(1.4142135624/2), Q31(1.6817928305/2) };  // 2^0, 2^0.25, 2^0.5, 2^0.75
+
+static inline int *DEC_SPAIR(int *dst, unsigned idx)
+{
+    dst[0] = (idx & 15) - 4;
+    dst[1] = (idx >> 4 & 15) - 4;
+
+    return dst + 2;
+}
+
+static inline int *DEC_SQUAD(int *dst, unsigned idx)
+{
+    dst[0] = (idx & 3) - 1;
+    dst[1] = (idx >> 2 & 3) - 1;
+    dst[2] = (idx >> 4 & 3) - 1;
+    dst[3] = (idx >> 6 & 3) - 1;
+
+    return dst + 4;
+}
+
+static inline int *DEC_UPAIR(int *dst, unsigned idx, unsigned sign)
+{
+    dst[0] = (idx & 15) * (1 - (sign & 0xFFFFFFFE));
+    dst[1] = (idx >> 4 & 15) * (1 - ((sign & 1) * 2));
+
+    return dst + 2;
+}
+
+static inline int *DEC_UQUAD(int *dst, unsigned idx, unsigned sign)
+{
+    unsigned nz = idx >> 12;
+
+    dst[0] = (idx & 3) * (1 + (((int)sign >> 31) * 2));
+    sign <<= nz & 1;
+    nz >>= 1;
+    dst[1] = (idx >> 2 & 3) * (1 + (((int)sign >> 31) * 2));
+    sign <<= nz & 1;
+    nz >>= 1;
+    dst[2] = (idx >> 4 & 3) * (1 + (((int)sign >> 31) * 2));
+    sign <<= nz & 1;
+    nz >>= 1;
+    dst[3] = (idx >> 6 & 3) * (1 + (((int)sign >> 31) * 2));
+
+    return dst + 4;
+}
+
+static void vector_pow43(int *coefs, int len)
+{
+    int i, coef;
+
+    for (i=0; i<len; i++) {
+        coef = coefs[i];
+        if (coef < 0)
+            coef = -(int)ff_cbrt_tab_fixed[-coef];
+        else
+            coef = (int)ff_cbrt_tab_fixed[coef];
+        coefs[i] = coef;
+    }
+}
+
+static void subband_scale(int *dst, int *src, int scale, int offset, int len, void *log_context)
+{
+    int ssign = scale < 0 ? -1 : 1;
+    int s = FFABS(scale);
+    unsigned int round;
+    int i, out, c = exp2tab[s & 3];
+
+    s = offset - (s >> 2);
+
+    if (s > 31) {
+        for (i=0; i<len; i++) {
+            dst[i] = 0;
+        }
+    } else if (s > 0) {
+        round = 1 << (s-1);
+        for (i=0; i<len; i++) {
+            out = (int)(((int64_t)src[i] * c) >> 32);
+            dst[i] = ((int)(out+round) >> s) * ssign;
+        }
+    } else if (s > -32) {
+        s = s + 32;
+        round = 1U << (s-1);
+        for (i=0; i<len; i++) {
+            out = (int)((int64_t)((int64_t)src[i] * c + round) >> s);
+            dst[i] = out * (unsigned)ssign;
+        }
+    } else {
+        av_log(log_context, AV_LOG_ERROR, "Overflow in subband_scale()\n");
+    }
+}
+
+static void noise_scale(int *coefs, int scale, int band_energy, int len)
+{
+    int s = -scale;
+    unsigned int round;
+    int i, out, c = exp2tab[s & 3];
+    int nlz = 0;
+
+    av_assert0(s >= 0);
+    while (band_energy > 0x7fff) {
+        band_energy >>= 1;
+        nlz++;
+    }
+    c /= band_energy;
+    s = 21 + nlz - (s >> 2);
+
+    if (s > 31) {
+        for (i=0; i<len; i++) {
+            coefs[i] = 0;
+        }
+    } else if (s >= 0) {
+        round = s ? 1 << (s-1) : 0;
+        for (i=0; i<len; i++) {
+            out = (int)(((int64_t)coefs[i] * c) >> 32);
+            coefs[i] = -((int)(out+round) >> s);
+        }
+    }
+    else {
+        s = s + 32;
+        if (s > 0) {
+            round = 1 << (s-1);
+            for (i=0; i<len; i++) {
+                out = (int)((int64_t)((int64_t)coefs[i] * c + round) >> s);
+                coefs[i] = -out;
+            }
+        } else {
+            for (i=0; i<len; i++)
+                coefs[i] = -(int64_t)coefs[i] * c * (1 << -s);
+        }
+    }
+}
+
+static av_always_inline SoftFloat flt16_round(SoftFloat pf)
+{
+    SoftFloat tmp;
+    int s;
+
+    tmp.exp = pf.exp;
+    s = pf.mant >> 31;
+    tmp.mant = (pf.mant ^ s) - s;
+    tmp.mant = (tmp.mant + 0x00200000U) & 0xFFC00000U;
+    tmp.mant = (tmp.mant ^ s) - s;
+
+    return tmp;
+}
+
+static av_always_inline SoftFloat flt16_even(SoftFloat pf)
+{
+    SoftFloat tmp;
+    int s;
+
+    tmp.exp = pf.exp;
+    s = pf.mant >> 31;
+    tmp.mant = (pf.mant ^ s) - s;
+    tmp.mant = (tmp.mant + 0x001FFFFFU + (tmp.mant & 0x00400000U >> 16)) & 0xFFC00000U;
+    tmp.mant = (tmp.mant ^ s) - s;
+
+    return tmp;
+}
+
+static av_always_inline SoftFloat flt16_trunc(SoftFloat pf)
+{
+    SoftFloat pun;
+    int s;
+
+    pun.exp = pf.exp;
+    s = pf.mant >> 31;
+    pun.mant = (pf.mant ^ s) - s;
+    pun.mant = pun.mant & 0xFFC00000U;
+    pun.mant = (pun.mant ^ s) - s;
+
+    return pun;
+}
+
+static av_always_inline void predict(PredictorState *ps, int *coef,
+                                     int output_enable)
+{
+    const SoftFloat a     = { 1023410176, 0 };  // 61.0 / 64
+    const SoftFloat alpha = {  973078528, 0 };  // 29.0 / 32
+    SoftFloat e0, e1;
+    SoftFloat pv;
+    SoftFloat k1, k2;
+    SoftFloat   r0 = ps->r0,     r1 = ps->r1;
+    SoftFloat cor0 = ps->cor0, cor1 = ps->cor1;
+    SoftFloat var0 = ps->var0, var1 = ps->var1;
+    SoftFloat tmp;
+
+    if (var0.exp > 1 || (var0.exp == 1 && var0.mant > 0x20000000)) {
+        k1 = av_mul_sf(cor0, flt16_even(av_div_sf(a, var0)));
+    }
+    else {
+        k1.mant = 0;
+        k1.exp = 0;
+    }
+
+    if (var1.exp > 1 || (var1.exp == 1 && var1.mant > 0x20000000)) {
+        k2 = av_mul_sf(cor1, flt16_even(av_div_sf(a, var1)));
+    }
+    else {
+        k2.mant = 0;
+        k2.exp = 0;
+    }
+
+    tmp = av_mul_sf(k1, r0);
+    pv = flt16_round(av_add_sf(tmp, av_mul_sf(k2, r1)));
+    if (output_enable) {
+        int shift = 28 - pv.exp;
+
+        if (shift < 31) {
+            if (shift > 0) {
+                *coef += (unsigned)((pv.mant + (1 << (shift - 1))) >> shift);
+            } else
+                *coef += (unsigned)pv.mant << -shift;
+        }
+    }
+
+    e0 = av_int2sf(*coef, 2);
+    e1 = av_sub_sf(e0, tmp);
+
+    ps->cor1 = flt16_trunc(av_add_sf(av_mul_sf(alpha, cor1), av_mul_sf(r1, e1)));
+    tmp = av_add_sf(av_mul_sf(r1, r1), av_mul_sf(e1, e1));
+    tmp.exp--;
+    ps->var1 = flt16_trunc(av_add_sf(av_mul_sf(alpha, var1), tmp));
+    ps->cor0 = flt16_trunc(av_add_sf(av_mul_sf(alpha, cor0), av_mul_sf(r0, e0)));
+    tmp = av_add_sf(av_mul_sf(r0, r0), av_mul_sf(e0, e0));
+    tmp.exp--;
+    ps->var0 = flt16_trunc(av_add_sf(av_mul_sf(alpha, var0), tmp));
+
+    ps->r1 = flt16_trunc(av_mul_sf(a, av_sub_sf(r0, av_mul_sf(k1, e0))));
+    ps->r0 = flt16_trunc(av_mul_sf(a, e0));
+}
+
+
+static const int cce_scale_fixed[8] = {
+    Q30(1.0),          //2^(0/8)
+    Q30(1.0905077327), //2^(1/8)
+    Q30(1.1892071150), //2^(2/8)
+    Q30(1.2968395547), //2^(3/8)
+    Q30(1.4142135624), //2^(4/8)
+    Q30(1.5422108254), //2^(5/8)
+    Q30(1.6817928305), //2^(6/8)
+    Q30(1.8340080864), //2^(7/8)
+};
+
+/**
+ * Apply dependent channel coupling (applied before IMDCT).
+ *
+ * @param   index   index into coupling gain array
+ */
+static void apply_dependent_coupling_fixed(AACContext *ac,
+                                     SingleChannelElement *target,
+                                     ChannelElement *cce, int index)
+{
+    IndividualChannelStream *ics = &cce->ch[0].ics;
+    const uint16_t *offsets = ics->swb_offset;
+    int *dest = target->coeffs;
+    const int *src = cce->ch[0].coeffs;
+    int g, i, group, k, idx = 0;
+    if (ac->oc[1].m4ac.object_type == AOT_AAC_LTP) {
+        av_log(ac->avctx, AV_LOG_ERROR,
+               "Dependent coupling is not supported together with LTP\n");
+        return;
+    }
+    for (g = 0; g < ics->num_window_groups; g++) {
+        for (i = 0; i < ics->max_sfb; i++, idx++) {
+            if (cce->ch[0].band_type[idx] != ZERO_BT) {
+                const int gain = cce->coup.gain[index][idx];
+                int shift, round, c, tmp;
+
+                if (gain < 0) {
+                    c = -cce_scale_fixed[-gain & 7];
+                    shift = (-gain-1024) >> 3;
+                }
+                else {
+                    c = cce_scale_fixed[gain & 7];
+                    shift = (gain-1024) >> 3;
+                }
+
+                if (shift < -31) {
+                    // Nothing to do
+                } else if (shift < 0) {
+                    shift = -shift;
+                    round = 1 << (shift - 1);
+
+                    for (group = 0; group < ics->group_len[g]; group++) {
+                        for (k = offsets[i]; k < offsets[i + 1]; k++) {
+                            tmp = (int)(((int64_t)src[group * 128 + k] * c + \
+                                       (int64_t)0x1000000000) >> 37);
+                            dest[group * 128 + k] += (tmp + (int64_t)round) >> shift;
+                        }
+                    }
+                }
+                else {
+                    for (group = 0; group < ics->group_len[g]; group++) {
+                        for (k = offsets[i]; k < offsets[i + 1]; k++) {
+                            tmp = (int)(((int64_t)src[group * 128 + k] * c + \
+                                        (int64_t)0x1000000000) >> 37);
+                            dest[group * 128 + k] += tmp * (1U << shift);
+                        }
+                    }
+                }
+            }
+        }
+        dest += ics->group_len[g] * 128;
+        src  += ics->group_len[g] * 128;
+    }
+}
+
+/**
+ * Apply independent channel coupling (applied after IMDCT).
+ *
+ * @param   index   index into coupling gain array
+ */
+static void apply_independent_coupling_fixed(AACContext *ac,
+                                       SingleChannelElement *target,
+                                       ChannelElement *cce, int index)
+{
+    int i, c, shift, round, tmp;
+    const int gain = cce->coup.gain[index][0];
+    const int *src = cce->ch[0].ret;
+    unsigned int *dest = target->ret;
+    const int len = 1024 << (ac->oc[1].m4ac.sbr == 1);
+
+    c = cce_scale_fixed[gain & 7];
+    shift = (gain-1024) >> 3;
+    if (shift < -31) {
+        return;
+    } else if (shift < 0) {
+        shift = -shift;
+        round = 1 << (shift - 1);
+
+        for (i = 0; i < len; i++) {
+            tmp = (int)(((int64_t)src[i] * c + (int64_t)0x1000000000) >> 37);
+            dest[i] += (tmp + round) >> shift;
+        }
+    }
+    else {
+      for (i = 0; i < len; i++) {
+          tmp = (int)(((int64_t)src[i] * c + (int64_t)0x1000000000) >> 37);
+          dest[i] += tmp * (1U << shift);
+      }
+    }
+}
+
+#include "aacdec_template.c"
+
+AVCodec ff_aac_fixed_decoder = {
+    .name            = "aac_fixed",
+    .long_name       = NULL_IF_CONFIG_SMALL("AAC (Advanced Audio Coding)"),
+    .type            = AVMEDIA_TYPE_AUDIO,
+    .id              = AV_CODEC_ID_AAC,
+    .priv_data_size  = sizeof(AACContext),
+    .init            = aac_decode_init,
+    .close           = aac_decode_close,
+    .decode          = aac_decode_frame,
+    .sample_fmts     = (const enum AVSampleFormat[]) {
+        AV_SAMPLE_FMT_S32P, AV_SAMPLE_FMT_NONE
+    },
+    .capabilities    = AV_CODEC_CAP_CHANNEL_CONF | AV_CODEC_CAP_DR1,
+    .caps_internal   = FF_CODEC_CAP_INIT_THREADSAFE,
+    .channel_layouts = aac_channel_layout,
+    .profiles        = NULL_IF_CONFIG_SMALL(ff_aac_profiles),
+    .flush = flush,
+};

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aacdectab.h

@@ -0,0 +1,74 @@
+/*
+ * AAC decoder data
+ * Copyright (c) 2005-2006 Oded Shimon ( ods15 ods15 dyndns org )
+ * Copyright (c) 2006-2007 Maxim Gavrilov ( maxim.gavrilov gmail com )
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+/**
+ * @file
+ * AAC decoder data
+ * @author Oded Shimon  ( ods15 ods15 dyndns org )
+ * @author Maxim Gavrilov ( maxim.gavrilov gmail com )
+ */
+
+#ifndef AVCODEC_AACDECTAB_H
+#define AVCODEC_AACDECTAB_H
+
+#include "libavutil/channel_layout.h"
+#include "aac.h"
+
+#include <stdint.h>
+
+static const int8_t tags_per_config[16] = { 0, 1, 1, 2, 3, 3, 4, 5, 0, 0, 0, 4, 5, 0, 5, 0 };
+
+static const uint8_t aac_channel_layout_map[16][5][3] = {
+    { { TYPE_SCE, 0, AAC_CHANNEL_FRONT }, },
+    { { TYPE_CPE, 0, AAC_CHANNEL_FRONT }, },
+    { { TYPE_SCE, 0, AAC_CHANNEL_FRONT }, { TYPE_CPE, 0, AAC_CHANNEL_FRONT }, },
+    { { TYPE_SCE, 0, AAC_CHANNEL_FRONT }, { TYPE_CPE, 0, AAC_CHANNEL_FRONT }, { TYPE_SCE, 1, AAC_CHANNEL_BACK }, },
+    { { TYPE_SCE, 0, AAC_CHANNEL_FRONT }, { TYPE_CPE, 0, AAC_CHANNEL_FRONT }, { TYPE_CPE, 1, AAC_CHANNEL_BACK }, },
+    { { TYPE_SCE, 0, AAC_CHANNEL_FRONT }, { TYPE_CPE, 0, AAC_CHANNEL_FRONT }, { TYPE_CPE, 1, AAC_CHANNEL_BACK }, { TYPE_LFE, 0, AAC_CHANNEL_LFE  }, },
+    { { TYPE_SCE, 0, AAC_CHANNEL_FRONT }, { TYPE_CPE, 0, AAC_CHANNEL_FRONT }, { TYPE_CPE, 1, AAC_CHANNEL_FRONT }, { TYPE_CPE, 2, AAC_CHANNEL_BACK }, { TYPE_LFE, 0, AAC_CHANNEL_LFE  }, },
+    { { 0, } },
+    { { 0, } },
+    { { 0, } },
+    { { TYPE_SCE, 0, AAC_CHANNEL_FRONT }, { TYPE_CPE, 0, AAC_CHANNEL_FRONT }, { TYPE_CPE, 1, AAC_CHANNEL_BACK }, { TYPE_SCE, 1, AAC_CHANNEL_BACK }, { TYPE_LFE, 0, AAC_CHANNEL_LFE  }, },
+    { { TYPE_SCE, 0, AAC_CHANNEL_FRONT }, { TYPE_CPE, 0, AAC_CHANNEL_FRONT }, { TYPE_CPE, 1, AAC_CHANNEL_SIDE }, { TYPE_CPE, 2, AAC_CHANNEL_BACK }, { TYPE_LFE, 0, AAC_CHANNEL_LFE  }, },
+    { { 0, } },
+    /* TODO: Add 7+1 TOP configuration */
+};
+
+static const uint64_t aac_channel_layout[16] = {
+    AV_CH_LAYOUT_MONO,
+    AV_CH_LAYOUT_STEREO,
+    AV_CH_LAYOUT_SURROUND,
+    AV_CH_LAYOUT_4POINT0,
+    AV_CH_LAYOUT_5POINT0_BACK,
+    AV_CH_LAYOUT_5POINT1_BACK,
+    AV_CH_LAYOUT_7POINT1_WIDE_BACK,
+    0,
+    0,
+    0,
+    AV_CH_LAYOUT_6POINT1,
+    AV_CH_LAYOUT_7POINT1,
+    0,
+    /* AV_CH_LAYOUT_7POINT1_TOP, */
+};
+
+#endif /* AVCODEC_AACDECTAB_H */

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aacenc.h

@@ -0,0 +1,428 @@
+/*
+ * AAC encoder
+ * Copyright (C) 2008 Konstantin Shishkov
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#ifndef AVCODEC_AACENC_H
+#define AVCODEC_AACENC_H
+
+#include "libavutil/float_dsp.h"
+#include "avcodec.h"
+#include "put_bits.h"
+
+#include "aac.h"
+#include "audio_frame_queue.h"
+#include "psymodel.h"
+
+#include "lpc.h"
+
+typedef enum AACCoder {
+    AAC_CODER_ANMR = 0,
+    AAC_CODER_TWOLOOP,
+    AAC_CODER_FAST,
+
+    AAC_CODER_NB,
+}AACCoder;
+
+typedef struct AACEncOptions {
+    int coder;
+    int pns;
+    int tns;
+    int ltp;
+    int pce;
+    int pred;
+    int mid_side;
+    int intensity_stereo;
+} AACEncOptions;
+
+struct AACEncContext;
+
+typedef struct AACCoefficientsEncoder {
+    void (*search_for_quantizers)(AVCodecContext *avctx, struct AACEncContext *s,
+                                  SingleChannelElement *sce, const float lambda);
+    void (*encode_window_bands_info)(struct AACEncContext *s, SingleChannelElement *sce,
+                                     int win, int group_len, const float lambda);
+    void (*quantize_and_encode_band)(struct AACEncContext *s, PutBitContext *pb, const float *in, float *out, int size,
+                                     int scale_idx, int cb, const float lambda, int rtz);
+    void (*encode_tns_info)(struct AACEncContext *s, SingleChannelElement *sce);
+    void (*encode_ltp_info)(struct AACEncContext *s, SingleChannelElement *sce, int common_window);
+    void (*encode_main_pred)(struct AACEncContext *s, SingleChannelElement *sce);
+    void (*adjust_common_pred)(struct AACEncContext *s, ChannelElement *cpe);
+    void (*adjust_common_ltp)(struct AACEncContext *s, ChannelElement *cpe);
+    void (*apply_main_pred)(struct AACEncContext *s, SingleChannelElement *sce);
+    void (*apply_tns_filt)(struct AACEncContext *s, SingleChannelElement *sce);
+    void (*update_ltp)(struct AACEncContext *s, SingleChannelElement *sce);
+    void (*ltp_insert_new_frame)(struct AACEncContext *s);
+    void (*set_special_band_scalefactors)(struct AACEncContext *s, SingleChannelElement *sce);
+    void (*search_for_pns)(struct AACEncContext *s, AVCodecContext *avctx, SingleChannelElement *sce);
+    void (*mark_pns)(struct AACEncContext *s, AVCodecContext *avctx, SingleChannelElement *sce);
+    void (*search_for_tns)(struct AACEncContext *s, SingleChannelElement *sce);
+    void (*search_for_ltp)(struct AACEncContext *s, SingleChannelElement *sce, int common_window);
+    void (*search_for_ms)(struct AACEncContext *s, ChannelElement *cpe);
+    void (*search_for_is)(struct AACEncContext *s, AVCodecContext *avctx, ChannelElement *cpe);
+    void (*search_for_pred)(struct AACEncContext *s, SingleChannelElement *sce);
+} AACCoefficientsEncoder;
+
+extern const AACCoefficientsEncoder ff_aac_coders[];
+
+typedef struct AACQuantizeBandCostCacheEntry {
+    float rd;
+    float energy;
+    int bits;
+    char cb;
+    char rtz;
+    uint16_t generation;
+} AACQuantizeBandCostCacheEntry;
+
+typedef struct AACPCEInfo {
+    int64_t layout;
+    int num_ele[4];                              ///< front, side, back, lfe
+    int pairing[3][8];                           ///< front, side, back
+    int index[4][8];                             ///< front, side, back, lfe
+    uint8_t config_map[16];                      ///< configs the encoder's channel specific settings
+    uint8_t reorder_map[16];                     ///< maps channels from lavc to aac order
+} AACPCEInfo;
+
+/**
+ * List of PCE (Program Configuration Element) for the channel layouts listed
+ * in channel_layout.h
+ *
+ * For those wishing in the future to add other layouts:
+ *
+ * - num_ele: number of elements in each group of front, side, back, lfe channels
+ *            (an element is of type SCE (single channel), CPE (channel pair) for
+ *            the first 3 groups; and is LFE for LFE group).
+ *
+ * - pairing: 0 for an SCE element or 1 for a CPE; does not apply to LFE group
+ *
+ * - index: there are three independent indices for SCE, CPE and LFE;
+ *     they are incremented irrespective of the group to which the element belongs;
+ *     they are not reset when going from one group to another
+ *
+ *     Example: for 7.0 channel layout,
+ *        .pairing = { { 1, 0 }, { 1 }, { 1 }, }, (3 CPE and 1 SCE in front group)
+ *        .index = { { 0, 0 }, { 1 }, { 2 }, },
+ *               (index is 0 for the single SCE but goes from 0 to 2 for the CPEs)
+ *
+ *     The index order impacts the channel ordering. But is otherwise arbitrary
+ *     (the sequence could have been 2, 0, 1 instead of 0, 1, 2).
+ *
+ *     Spec allows for discontinuous indices, e.g. if one has a total of two SCE,
+ *     SCE.0 SCE.15 is OK per spec; BUT it won't be decoded by our AAC decoder
+ *     which at this time requires that indices fully cover some range starting
+ *     from 0 (SCE.1 SCE.0 is OK but not SCE.0 SCE.15).
+ *
+ * - config_map: total number of elements and their types. Beware, the way the
+ *               types are ordered impacts the final channel ordering.
+ *
+ * - reorder_map: reorders the channels.
+ *
+ */
+static const AACPCEInfo aac_pce_configs[] = {
+    {
+        .layout = AV_CH_LAYOUT_MONO,
+        .num_ele = { 1, 0, 0, 0 },
+        .pairing = { { 0 }, },
+        .index = { { 0 }, },
+        .config_map = { 1, TYPE_SCE, },
+        .reorder_map = { 0 },
+    },
+    {
+        .layout = AV_CH_LAYOUT_STEREO,
+        .num_ele = { 1, 0, 0, 0 },
+        .pairing = { { 1 }, },
+        .index = { { 0 }, },
+        .config_map = { 1, TYPE_CPE, },
+        .reorder_map = { 0, 1 },
+    },
+    {
+        .layout = AV_CH_LAYOUT_2POINT1,
+        .num_ele = { 1, 0, 0, 1 },
+        .pairing = { { 1 }, },
+        .index = { { 0 },{ 0 },{ 0 },{ 0 } },
+        .config_map = { 2, TYPE_CPE, TYPE_LFE },
+        .reorder_map = { 0, 1, 2 },
+    },
+    {
+        .layout = AV_CH_LAYOUT_2_1,
+        .num_ele = { 1, 0, 1, 0 },
+        .pairing = { { 1 },{ 0 },{ 0 } },
+        .index = { { 0 },{ 0 },{ 0 }, },
+        .config_map = { 2, TYPE_CPE, TYPE_SCE },
+        .reorder_map = { 0, 1, 2 },
+    },
+    {
+        .layout = AV_CH_LAYOUT_SURROUND,
+        .num_ele = { 2, 0, 0, 0 },
+        .pairing = { { 1, 0 }, },
+        .index = { { 0, 0 }, },
+        .config_map = { 2, TYPE_CPE, TYPE_SCE, },
+        .reorder_map = { 0, 1, 2 },
+    },
+    {
+        .layout = AV_CH_LAYOUT_3POINT1,
+        .num_ele = { 2, 0, 0, 1 },
+        .pairing = { { 1, 0 }, },
+        .index = { { 0, 0 }, { 0 }, { 0 }, { 0 }, },
+        .config_map = { 3, TYPE_CPE, TYPE_SCE, TYPE_LFE },
+        .reorder_map = { 0, 1, 2, 3 },
+    },
+    {
+        .layout = AV_CH_LAYOUT_4POINT0,
+        .num_ele = { 2, 0, 1, 0 },
+        .pairing = { { 1, 0 }, { 0 }, { 0 }, },
+        .index = { { 0, 0 }, { 0 }, { 1 } },
+        .config_map = { 3, TYPE_CPE, TYPE_SCE, TYPE_SCE },
+        .reorder_map = {  0, 1, 2, 3 },
+    },
+    {
+        .layout = AV_CH_LAYOUT_4POINT1,
+        .num_ele = { 2, 1, 1, 0 },
+        .pairing = { { 1, 0 }, { 0 }, { 0 }, },
+        .index = { { 0, 0 }, { 1 }, { 2 }, { 0 } },
+        .config_map = { 4, TYPE_CPE, TYPE_SCE, TYPE_SCE, TYPE_SCE },
+        .reorder_map = { 0, 1, 2, 3, 4 },
+    },
+    {
+        .layout = AV_CH_LAYOUT_2_2,
+        .num_ele = { 1, 1, 0, 0 },
+        .pairing = { { 1 }, { 1 }, },
+        .index = { { 0 }, { 1 }, },
+        .config_map = { 2, TYPE_CPE, TYPE_CPE },
+        .reorder_map = { 0, 1, 2, 3 },
+    },
+    {
+        .layout = AV_CH_LAYOUT_QUAD,
+        .num_ele = { 1, 0, 1, 0 },
+        .pairing = { { 1 }, { 0 }, { 1 }, },
+        .index = { { 0 }, { 0 }, { 1 } },
+        .config_map = { 2, TYPE_CPE, TYPE_CPE },
+        .reorder_map = { 0, 1, 2, 3 },
+    },
+    {
+        .layout = AV_CH_LAYOUT_5POINT0,
+        .num_ele = { 2, 1, 0, 0 },
+        .pairing = { { 1, 0 }, { 1 }, },
+        .index = { { 0, 0 }, { 1 } },
+        .config_map = { 3, TYPE_CPE, TYPE_SCE, TYPE_CPE },
+        .reorder_map = { 0, 1, 2, 3, 4 },
+    },
+    {
+        .layout = AV_CH_LAYOUT_5POINT1,
+        .num_ele = { 2, 1, 1, 0 },
+        .pairing = { { 1, 0 }, { 0 }, { 1 }, },
+        .index = { { 0, 0 }, { 1 }, { 1 } },
+        .config_map = { 4, TYPE_CPE, TYPE_SCE, TYPE_SCE, TYPE_CPE },
+        .reorder_map = { 0, 1, 2, 3, 4, 5 },
+    },
+    {
+        .layout = AV_CH_LAYOUT_5POINT0_BACK,
+        .num_ele = { 2, 0, 1, 0 },
+        .pairing = { { 1, 0 }, { 0 }, { 1 } },
+        .index = { { 0, 0 }, { 0 }, { 1 } },
+        .config_map = { 3, TYPE_CPE, TYPE_SCE, TYPE_CPE },
+        .reorder_map = { 0, 1, 2, 3, 4 },
+    },
+    {
+        .layout = AV_CH_LAYOUT_5POINT1_BACK,
+        .num_ele = { 2, 1, 1, 0 },
+        .pairing = { { 1, 0 }, { 0 }, { 1 }, },
+        .index = { { 0, 0 }, { 1 }, { 1 } },
+        .config_map = { 4, TYPE_CPE, TYPE_SCE, TYPE_SCE, TYPE_CPE },
+        .reorder_map = { 0, 1, 2, 3, 4, 5 },
+    },
+    {
+        .layout = AV_CH_LAYOUT_6POINT0,
+        .num_ele = { 2, 1, 1, 0 },
+        .pairing = { { 1, 0 }, { 1 }, { 0 }, },
+        .index = { { 0, 0 }, { 1 }, { 1 } },
+        .config_map = { 4, TYPE_CPE, TYPE_SCE, TYPE_CPE, TYPE_SCE },
+        .reorder_map = { 0, 1, 2, 3, 4, 5 },
+    },
+    {
+        .layout = AV_CH_LAYOUT_6POINT0_FRONT,
+        .num_ele = { 2, 1, 0, 0 },
+        .pairing = { { 1, 1 }, { 1 } },
+        .index = { { 1, 0 }, { 2 }, },
+        .config_map = { 3, TYPE_CPE, TYPE_CPE, TYPE_CPE, },
+        .reorder_map = { 0, 1, 2, 3, 4, 5 },
+    },
+    {
+        .layout = AV_CH_LAYOUT_HEXAGONAL,
+        .num_ele = { 2, 0, 2, 0 },
+        .pairing = { { 1, 0 },{ 0 },{ 1, 0 }, },
+        .index = { { 0, 0 },{ 0 },{ 1, 1 } },
+        .config_map = { 4, TYPE_CPE, TYPE_SCE, TYPE_CPE, TYPE_SCE, },
+        .reorder_map = { 0, 1, 2, 3, 4, 5 },
+    },
+    {
+        .layout = AV_CH_LAYOUT_6POINT1,
+        .num_ele = { 2, 1, 2, 0 },
+        .pairing = { { 1, 0 },{ 0 },{ 1, 0 }, },
+        .index = { { 0, 0 },{ 1 },{ 1, 2 } },
+        .config_map = { 5, TYPE_CPE, TYPE_SCE, TYPE_SCE, TYPE_CPE, TYPE_SCE },
+        .reorder_map = { 0, 1, 2, 3, 4, 5, 6 },
+    },
+    {
+        .layout = AV_CH_LAYOUT_6POINT1_BACK,
+        .num_ele = { 2, 1, 2, 0 },
+        .pairing = { { 1, 0 }, { 0 }, { 1, 0 }, },
+        .index = { { 0, 0 }, { 1 }, { 1, 2 } },
+        .config_map = { 5, TYPE_CPE, TYPE_SCE, TYPE_SCE, TYPE_CPE, TYPE_SCE },
+        .reorder_map = { 0, 1, 2, 3, 4, 5, 6 },
+    },
+    {
+        .layout = AV_CH_LAYOUT_6POINT1_FRONT,
+        .num_ele = { 2, 1, 2, 0 },
+        .pairing = { { 1, 0 }, { 0 }, { 1, 0 }, },
+        .index = { { 0, 0 }, { 1 }, { 1, 2 } },
+        .config_map = { 5, TYPE_CPE, TYPE_SCE, TYPE_SCE, TYPE_CPE, TYPE_SCE },
+        .reorder_map = { 0, 1, 2, 3, 4, 5, 6 },
+    },
+    {
+        .layout = AV_CH_LAYOUT_7POINT0,
+        .num_ele = { 2, 1, 1, 0 },
+        .pairing = { { 1, 0 }, { 1 }, { 1 }, },
+        .index = { { 0, 0 }, { 1 }, { 2 }, },
+        .config_map = { 4, TYPE_CPE, TYPE_SCE, TYPE_CPE, TYPE_CPE },
+        .reorder_map = { 0, 1, 2, 3, 4, 5, 6 },
+    },
+    {
+        .layout = AV_CH_LAYOUT_7POINT0_FRONT,
+        .num_ele = { 2, 1, 1, 0 },
+        .pairing = { { 1, 0 }, { 1 }, { 1 }, },
+        .index = { { 0, 0 }, { 1 }, { 2 }, },
+        .config_map = { 4, TYPE_CPE, TYPE_SCE, TYPE_CPE, TYPE_CPE },
+        .reorder_map = { 0, 1, 2, 3, 4, 5, 6 },
+    },
+    {
+        .layout = AV_CH_LAYOUT_7POINT1,
+        .num_ele = { 2, 1, 2, 0 },
+        .pairing = { { 1, 0 }, { 0 }, { 1, 1 }, },
+        .index = { { 0, 0 }, { 1 }, { 1, 2 }, { 0 } },
+        .config_map = { 5, TYPE_CPE, TYPE_SCE,  TYPE_SCE, TYPE_CPE, TYPE_CPE },
+        .reorder_map = { 0, 1, 2, 3, 4, 5, 6, 7 },
+    },
+    {
+        .layout = AV_CH_LAYOUT_7POINT1_WIDE,
+        .num_ele = { 2, 1, 2, 0 },
+        .pairing = { { 1, 0 }, { 0 },{  1, 1 }, },
+        .index = { { 0, 0 }, { 1 }, { 1, 2 }, { 0 } },
+        .config_map = { 5, TYPE_CPE, TYPE_SCE, TYPE_SCE, TYPE_CPE, TYPE_CPE },
+        .reorder_map = { 0, 1, 2, 3, 4, 5, 6, 7 },
+    },
+    {
+        .layout = AV_CH_LAYOUT_7POINT1_WIDE_BACK,
+        .num_ele = { 2, 1, 2, 0 },
+        .pairing = { { 1, 0 }, { 0 }, { 1, 1 }, },
+        .index = { { 0, 0 }, { 1 }, { 1, 2 }, { 0 } },
+        .config_map = { 5, TYPE_CPE, TYPE_SCE, TYPE_SCE, TYPE_CPE, TYPE_CPE },
+        .reorder_map = { 0, 1, 2, 3, 4, 5, 6, 7 },
+    },
+    {
+        .layout = AV_CH_LAYOUT_OCTAGONAL,
+        .num_ele = { 2, 1, 2, 0 },
+        .pairing = { { 1, 0 }, { 1 }, { 1, 0 }, },
+        .index = { { 0, 0 }, { 1 }, { 2, 1 } },
+        .config_map = { 5, TYPE_CPE, TYPE_SCE, TYPE_CPE, TYPE_CPE, TYPE_SCE },
+        .reorder_map = { 0, 1, 2, 3, 4, 5, 6, 7 },
+    },
+    {   /* Meant for order 2/mixed ambisonics */
+        .layout = AV_CH_LAYOUT_OCTAGONAL | AV_CH_TOP_CENTER,
+        .num_ele = { 2, 2, 2, 0 },
+        .pairing = { { 1, 0 }, { 1, 0 }, { 1, 0 }, },
+        .index = { { 0, 0 }, { 1, 1 }, { 2, 2 } },
+        .config_map = { 6, TYPE_CPE, TYPE_SCE, TYPE_CPE, TYPE_SCE, TYPE_CPE, TYPE_SCE },
+        .reorder_map = { 0, 1, 2, 3, 4, 5, 6, 7, 8 },
+    },
+    {   /* Meant for order 2/mixed ambisonics */
+        .layout = AV_CH_LAYOUT_6POINT0_FRONT | AV_CH_BACK_CENTER |
+                  AV_CH_BACK_LEFT | AV_CH_BACK_RIGHT | AV_CH_TOP_CENTER,
+        .num_ele = { 2, 2, 2, 0 },
+        .pairing = { { 1, 1 }, { 1, 0 }, { 1, 0 }, },
+        .index = { { 0, 1 }, { 2, 0 }, { 3, 1 } },
+        .config_map = { 6, TYPE_CPE, TYPE_CPE, TYPE_CPE, TYPE_SCE, TYPE_CPE, TYPE_SCE },
+        .reorder_map = { 0, 1, 2, 3, 4, 5, 6, 7, 8, 9 },
+    },
+    {
+        .layout = AV_CH_LAYOUT_HEXADECAGONAL,
+        .num_ele = { 4, 2, 4, 0 },
+        .pairing = { { 1, 0, 1, 0 }, { 1, 1 }, { 1, 0, 1, 0 }, },
+        .index = { { 0, 0, 1, 1 }, { 2, 3 }, { 4, 2, 5, 3 } },
+        .config_map = { 10, TYPE_CPE, TYPE_SCE, TYPE_CPE, TYPE_SCE, TYPE_CPE, TYPE_CPE, TYPE_CPE, TYPE_SCE, TYPE_CPE, TYPE_SCE },
+        .reorder_map = { 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15 },
+    },
+};
+
+/**
+ * AAC encoder context
+ */
+typedef struct AACEncContext {
+    AVClass *av_class;
+    AACEncOptions options;                       ///< encoding options
+    PutBitContext pb;
+    FFTContext mdct1024;                         ///< long (1024 samples) frame transform context
+    FFTContext mdct128;                          ///< short (128 samples) frame transform context
+    AVFloatDSPContext *fdsp;
+    AACPCEInfo pce;                              ///< PCE data, if needed
+    float *planar_samples[16];                   ///< saved preprocessed input
+
+    int profile;                                 ///< copied from avctx
+    int needs_pce;                               ///< flag for non-standard layout
+    LPCContext lpc;                              ///< used by TNS
+    int samplerate_index;                        ///< MPEG-4 samplerate index
+    int channels;                                ///< channel count
+    const uint8_t *reorder_map;                  ///< lavc to aac reorder map
+    const uint8_t *chan_map;                     ///< channel configuration map
+
+    ChannelElement *cpe;                         ///< channel elements
+    FFPsyContext psy;
+    struct FFPsyPreprocessContext* psypp;
+    const AACCoefficientsEncoder *coder;
+    int cur_channel;                             ///< current channel for coder context
+    int random_state;
+    float lambda;
+    int last_frame_pb_count;                     ///< number of bits for the previous frame
+    float lambda_sum;                            ///< sum(lambda), for Qvg reporting
+    int lambda_count;                            ///< count(lambda), for Qvg reporting
+    enum RawDataBlockType cur_type;              ///< channel group type cur_channel belongs to
+
+    AudioFrameQueue afq;
+    DECLARE_ALIGNED(16, int,   qcoefs)[96];      ///< quantized coefficients
+    DECLARE_ALIGNED(32, float, scoefs)[1024];    ///< scaled coefficients
+
+    uint16_t quantize_band_cost_cache_generation;
+    AACQuantizeBandCostCacheEntry quantize_band_cost_cache[256][128]; ///< memoization area for quantize_band_cost
+
+    void (*abs_pow34)(float *out, const float *in, const int size);
+    void (*quant_bands)(int *out, const float *in, const float *scaled,
+                        int size, int is_signed, int maxval, const float Q34,
+                        const float rounding);
+
+    struct {
+        float *samples;
+    } buffer;
+} AACEncContext;
+
+void ff_aac_dsp_init_x86(AACEncContext *s);
+void ff_aac_coder_init_mips(AACEncContext *c);
+void ff_quantize_band_cost_cache_init(struct AACEncContext *s);
+
+
+#endif /* AVCODEC_AACENC_H */

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aacenc_is.c

@@ -0,0 +1,158 @@
+/*
+ * AAC encoder intensity stereo
+ * Copyright (C) 2015 Rostislav Pehlivanov
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+/**
+ * @file
+ * AAC encoder Intensity Stereo
+ * @author Rostislav Pehlivanov ( atomnuker gmail com )
+ */
+
+#include "aacenc.h"
+#include "aacenc_utils.h"
+#include "aacenc_is.h"
+#include "aacenc_quantization.h"
+
+struct AACISError ff_aac_is_encoding_err(AACEncContext *s, ChannelElement *cpe,
+                                         int start, int w, int g, float ener0,
+                                         float ener1, float ener01,
+                                         int use_pcoeffs, int phase)
+{
+    int i, w2;
+    SingleChannelElement *sce0 = &cpe->ch[0];
+    SingleChannelElement *sce1 = &cpe->ch[1];
+    float *L = use_pcoeffs ? sce0->pcoeffs : sce0->coeffs;
+    float *R = use_pcoeffs ? sce1->pcoeffs : sce1->coeffs;
+    float *L34 = &s->scoefs[256*0], *R34 = &s->scoefs[256*1];
+    float *IS  = &s->scoefs[256*2], *I34 = &s->scoefs[256*3];
+    float dist1 = 0.0f, dist2 = 0.0f;
+    struct AACISError is_error = {0};
+
+    if (ener01 <= 0 || ener0 <= 0) {
+        is_error.pass = 0;
+        return is_error;
+    }
+
+    for (w2 = 0; w2 < sce0->ics.group_len[w]; w2++) {
+        FFPsyBand *band0 = &s->psy.ch[s->cur_channel+0].psy_bands[(w+w2)*16+g];
+        FFPsyBand *band1 = &s->psy.ch[s->cur_channel+1].psy_bands[(w+w2)*16+g];
+        int is_band_type, is_sf_idx = FFMAX(1, sce0->sf_idx[w*16+g]-4);
+        float e01_34 = phase*pos_pow34(ener1/ener0);
+        float maxval, dist_spec_err = 0.0f;
+        float minthr = FFMIN(band0->threshold, band1->threshold);
+        for (i = 0; i < sce0->ics.swb_sizes[g]; i++)
+            IS[i] = (L[start+(w+w2)*128+i] + phase*R[start+(w+w2)*128+i])*sqrt(ener0/ener01);
+        s->abs_pow34(L34, &L[start+(w+w2)*128], sce0->ics.swb_sizes[g]);
+        s->abs_pow34(R34, &R[start+(w+w2)*128], sce0->ics.swb_sizes[g]);
+        s->abs_pow34(I34, IS,                   sce0->ics.swb_sizes[g]);
+        maxval = find_max_val(1, sce0->ics.swb_sizes[g], I34);
+        is_band_type = find_min_book(maxval, is_sf_idx);
+        dist1 += quantize_band_cost(s, &L[start + (w+w2)*128], L34,
+                                    sce0->ics.swb_sizes[g],
+                                    sce0->sf_idx[w*16+g],
+                                    sce0->band_type[w*16+g],
+                                    s->lambda / band0->threshold, INFINITY, NULL, NULL, 0);
+        dist1 += quantize_band_cost(s, &R[start + (w+w2)*128], R34,
+                                    sce1->ics.swb_sizes[g],
+                                    sce1->sf_idx[w*16+g],
+                                    sce1->band_type[w*16+g],
+                                    s->lambda / band1->threshold, INFINITY, NULL, NULL, 0);
+        dist2 += quantize_band_cost(s, IS, I34, sce0->ics.swb_sizes[g],
+                                    is_sf_idx, is_band_type,
+                                    s->lambda / minthr, INFINITY, NULL, NULL, 0);
+        for (i = 0; i < sce0->ics.swb_sizes[g]; i++) {
+            dist_spec_err += (L34[i] - I34[i])*(L34[i] - I34[i]);
+            dist_spec_err += (R34[i] - I34[i]*e01_34)*(R34[i] - I34[i]*e01_34);
+        }
+        dist_spec_err *= s->lambda / minthr;
+        dist2 += dist_spec_err;
+    }
+
+    is_error.pass = dist2 <= dist1;
+    is_error.phase = phase;
+    is_error.error = dist2 - dist1;
+    is_error.dist1 = dist1;
+    is_error.dist2 = dist2;
+    is_error.ener01 = ener01;
+
+    return is_error;
+}
+
+void ff_aac_search_for_is(AACEncContext *s, AVCodecContext *avctx, ChannelElement *cpe)
+{
+    SingleChannelElement *sce0 = &cpe->ch[0];
+    SingleChannelElement *sce1 = &cpe->ch[1];
+    int start = 0, count = 0, w, w2, g, i, prev_sf1 = -1, prev_bt = -1, prev_is = 0;
+    const float freq_mult = avctx->sample_rate/(1024.0f/sce0->ics.num_windows)/2.0f;
+    uint8_t nextband1[128];
+
+    if (!cpe->common_window)
+        return;
+
+    /** Scout out next nonzero bands */
+    ff_init_nextband_map(sce1, nextband1);
+
+    for (w = 0; w < sce0->ics.num_windows; w += sce0->ics.group_len[w]) {
+        start = 0;
+        for (g = 0;  g < sce0->ics.num_swb; g++) {
+            if (start*freq_mult > INT_STEREO_LOW_LIMIT*(s->lambda/170.0f) &&
+                cpe->ch[0].band_type[w*16+g] != NOISE_BT && !cpe->ch[0].zeroes[w*16+g] &&
+                cpe->ch[1].band_type[w*16+g] != NOISE_BT && !cpe->ch[1].zeroes[w*16+g] &&
+                ff_sfdelta_can_remove_band(sce1, nextband1, prev_sf1, w*16+g)) {
+                float ener0 = 0.0f, ener1 = 0.0f, ener01 = 0.0f, ener01p = 0.0f;
+                struct AACISError ph_err1, ph_err2, *best;
+                for (w2 = 0; w2 < sce0->ics.group_len[w]; w2++) {
+                    for (i = 0; i < sce0->ics.swb_sizes[g]; i++) {
+                        float coef0 = sce0->coeffs[start+(w+w2)*128+i];
+                        float coef1 = sce1->coeffs[start+(w+w2)*128+i];
+                        ener0  += coef0*coef0;
+                        ener1  += coef1*coef1;
+                        ener01 += (coef0 + coef1)*(coef0 + coef1);
+                        ener01p += (coef0 - coef1)*(coef0 - coef1);
+                    }
+                }
+                ph_err1 = ff_aac_is_encoding_err(s, cpe, start, w, g,
+                                                 ener0, ener1, ener01p, 0, -1);
+                ph_err2 = ff_aac_is_encoding_err(s, cpe, start, w, g,
+                                                 ener0, ener1, ener01, 0, +1);
+                best = (ph_err1.pass && ph_err1.error < ph_err2.error) ? &ph_err1 : &ph_err2;
+                if (best->pass) {
+                    cpe->is_mask[w*16+g] = 1;
+                    cpe->ms_mask[w*16+g] = 0;
+                    cpe->ch[0].is_ener[w*16+g] = sqrt(ener0 / best->ener01);
+                    cpe->ch[1].is_ener[w*16+g] = ener0/ener1;
+                    cpe->ch[1].band_type[w*16+g] = (best->phase > 0) ? INTENSITY_BT : INTENSITY_BT2;
+                    if (prev_is && prev_bt != cpe->ch[1].band_type[w*16+g]) {
+                        /** Flip M/S mask and pick the other CB, since it encodes more efficiently */
+                        cpe->ms_mask[w*16+g] = 1;
+                        cpe->ch[1].band_type[w*16+g] = (best->phase > 0) ? INTENSITY_BT2 : INTENSITY_BT;
+                    }
+                    prev_bt = cpe->ch[1].band_type[w*16+g];
+                    count++;
+                }
+            }
+            if (!sce1->zeroes[w*16+g] && sce1->band_type[w*16+g] < RESERVED_BT)
+                prev_sf1 = sce1->sf_idx[w*16+g];
+            prev_is = cpe->is_mask[w*16+g];
+            start += sce0->ics.swb_sizes[g];
+        }
+    }
+    cpe->is_mode = !!count;
+}

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aacenc_is.h

@@ -0,0 +1,51 @@
+/*
+ * AAC encoder intensity stereo
+ * Copyright (C) 2015 Rostislav Pehlivanov
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+/**
+ * @file
+ * AAC encoder Intensity Stereo
+ * @author Rostislav Pehlivanov ( atomnuker gmail com )
+ */
+
+#ifndef AVCODEC_AACENC_IS_H
+#define AVCODEC_AACENC_IS_H
+
+#include "aacenc.h"
+
+/** Frequency in Hz for lower limit of intensity stereo **/
+#define INT_STEREO_LOW_LIMIT 6100
+
+struct AACISError {
+    int pass;    /* 1 if dist2 <= dist1  */
+    int phase;   /* -1 or +1             */
+    float error; /* fabs(dist1 - dist2)  */
+    float dist1; /* From original coeffs */
+    float dist2; /* From IS'd coeffs     */
+    float ener01;
+};
+
+struct AACISError ff_aac_is_encoding_err(AACEncContext *s, ChannelElement *cpe,
+                                         int start, int w, int g, float ener0,
+                                         float ener1, float ener01,
+                                         int use_pcoeffs, int phase);
+void ff_aac_search_for_is(AACEncContext *s, AVCodecContext *avctx, ChannelElement *cpe);
+
+#endif /* AVCODEC_AACENC_IS_H */

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aacenc_ltp.c

@@ -0,0 +1,236 @@
+/*
+ * AAC encoder long term prediction extension
+ * Copyright (C) 2015 Rostislav Pehlivanov
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+/**
+ * @file
+ * AAC encoder long term prediction extension
+ * @author Rostislav Pehlivanov ( atomnuker gmail com )
+ */
+
+#include "aacenc_ltp.h"
+#include "aacenc_quantization.h"
+#include "aacenc_utils.h"
+
+/**
+ * Encode LTP data.
+ */
+void ff_aac_encode_ltp_info(AACEncContext *s, SingleChannelElement *sce,
+                            int common_window)
+{
+    int i;
+    IndividualChannelStream *ics = &sce->ics;
+    if (s->profile != FF_PROFILE_AAC_LTP || !ics->predictor_present)
+        return;
+    if (common_window)
+        put_bits(&s->pb, 1, 0);
+    put_bits(&s->pb, 1, ics->ltp.present);
+    if (!ics->ltp.present)
+        return;
+    put_bits(&s->pb, 11, ics->ltp.lag);
+    put_bits(&s->pb, 3,  ics->ltp.coef_idx);
+    for (i = 0; i < FFMIN(ics->max_sfb, MAX_LTP_LONG_SFB); i++)
+        put_bits(&s->pb, 1, ics->ltp.used[i]);
+}
+
+void ff_aac_ltp_insert_new_frame(AACEncContext *s)
+{
+    int i, ch, tag, chans, cur_channel, start_ch = 0;
+    ChannelElement *cpe;
+    SingleChannelElement *sce;
+    for (i = 0; i < s->chan_map[0]; i++) {
+        cpe = &s->cpe[i];
+        tag      = s->chan_map[i+1];
+        chans    = tag == TYPE_CPE ? 2 : 1;
+        for (ch = 0; ch < chans; ch++) {
+            sce = &cpe->ch[ch];
+            cur_channel = start_ch + ch;
+            /* New sample + overlap */
+            memcpy(&sce->ltp_state[0],    &sce->ltp_state[1024], 1024*sizeof(sce->ltp_state[0]));
+            memcpy(&sce->ltp_state[1024], &s->planar_samples[cur_channel][2048], 1024*sizeof(sce->ltp_state[0]));
+            memcpy(&sce->ltp_state[2048], &sce->ret_buf[0], 1024*sizeof(sce->ltp_state[0]));
+            sce->ics.ltp.lag = 0;
+        }
+        start_ch += chans;
+    }
+}
+
+static void get_lag(float *buf, const float *new, LongTermPrediction *ltp)
+{
+    int i, j, lag = 0, max_corr = 0;
+    float max_ratio = 0.0f;
+    for (i = 0; i < 2048; i++) {
+        float corr, s0 = 0.0f, s1 = 0.0f;
+        const int start = FFMAX(0, i - 1024);
+        for (j = start; j < 2048; j++) {
+            const int idx = j - i + 1024;
+            s0 += new[j]*buf[idx];
+            s1 += buf[idx]*buf[idx];
+        }
+        corr = s1 > 0.0f ? s0/sqrt(s1) : 0.0f;
+        if (corr > max_corr) {
+            max_corr = corr;
+            lag = i;
+            max_ratio = corr/(2048-start);
+        }
+    }
+    ltp->lag = FFMAX(av_clip_uintp2(lag, 11), 0);
+    ltp->coef_idx = quant_array_idx(max_ratio, ltp_coef, 8);
+    ltp->coef = ltp_coef[ltp->coef_idx];
+}
+
+static void generate_samples(float *buf, LongTermPrediction *ltp)
+{
+    int i, samples_num = 2048;
+    if (!ltp->lag) {
+        ltp->present = 0;
+        return;
+    } else if (ltp->lag < 1024) {
+        samples_num = ltp->lag + 1024;
+    }
+    for (i = 0; i < samples_num; i++)
+        buf[i] = ltp->coef*buf[i + 2048 - ltp->lag];
+    memset(&buf[i], 0, (2048 - i)*sizeof(float));
+}
+
+/**
+ * Process LTP parameters
+ * @see Patent WO2006070265A1
+ */
+void ff_aac_update_ltp(AACEncContext *s, SingleChannelElement *sce)
+{
+    float *pred_signal = &sce->ltp_state[0];
+    const float *samples = &s->planar_samples[s->cur_channel][1024];
+
+    if (s->profile != FF_PROFILE_AAC_LTP)
+        return;
+
+    /* Calculate lag */
+    get_lag(pred_signal, samples, &sce->ics.ltp);
+    generate_samples(pred_signal, &sce->ics.ltp);
+}
+
+void ff_aac_adjust_common_ltp(AACEncContext *s, ChannelElement *cpe)
+{
+    int sfb, count = 0;
+    SingleChannelElement *sce0 = &cpe->ch[0];
+    SingleChannelElement *sce1 = &cpe->ch[1];
+
+    if (!cpe->common_window ||
+        sce0->ics.window_sequence[0] == EIGHT_SHORT_SEQUENCE ||
+        sce1->ics.window_sequence[0] == EIGHT_SHORT_SEQUENCE) {
+        sce0->ics.ltp.present = 0;
+        return;
+    }
+
+    for (sfb = 0; sfb < FFMIN(sce0->ics.max_sfb, MAX_LTP_LONG_SFB); sfb++) {
+        int sum = sce0->ics.ltp.used[sfb] + sce1->ics.ltp.used[sfb];
+        if (sum != 2) {
+            sce0->ics.ltp.used[sfb] = 0;
+        } else {
+            count++;
+        }
+    }
+
+    sce0->ics.ltp.present = !!count;
+    sce0->ics.predictor_present = !!count;
+}
+
+/**
+ * Mark LTP sfb's
+ */
+void ff_aac_search_for_ltp(AACEncContext *s, SingleChannelElement *sce,
+                           int common_window)
+{
+    int w, g, w2, i, start = 0, count = 0;
+    int saved_bits = -(15 + FFMIN(sce->ics.max_sfb, MAX_LTP_LONG_SFB));
+    float *C34 = &s->scoefs[128*0], *PCD = &s->scoefs[128*1];
+    float *PCD34 = &s->scoefs[128*2];
+    const int max_ltp = FFMIN(sce->ics.max_sfb, MAX_LTP_LONG_SFB);
+
+    if (sce->ics.window_sequence[0] == EIGHT_SHORT_SEQUENCE) {
+        if (sce->ics.ltp.lag) {
+            memset(&sce->ltp_state[0], 0, 3072*sizeof(sce->ltp_state[0]));
+            memset(&sce->ics.ltp, 0, sizeof(LongTermPrediction));
+        }
+        return;
+    }
+
+    if (!sce->ics.ltp.lag || s->lambda > 120.0f)
+        return;
+
+    for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) {
+        start = 0;
+        for (g = 0;  g < sce->ics.num_swb; g++) {
+            int bits1 = 0, bits2 = 0;
+            float dist1 = 0.0f, dist2 = 0.0f;
+            if (w*16+g > max_ltp) {
+                start += sce->ics.swb_sizes[g];
+                continue;
+            }
+            for (w2 = 0; w2 < sce->ics.group_len[w]; w2++) {
+                int bits_tmp1, bits_tmp2;
+                FFPsyBand *band = &s->psy.ch[s->cur_channel].psy_bands[(w+w2)*16+g];
+                for (i = 0; i < sce->ics.swb_sizes[g]; i++)
+                    PCD[i] = sce->coeffs[start+(w+w2)*128+i] - sce->lcoeffs[start+(w+w2)*128+i];
+                s->abs_pow34(C34,  &sce->coeffs[start+(w+w2)*128],  sce->ics.swb_sizes[g]);
+                s->abs_pow34(PCD34, PCD, sce->ics.swb_sizes[g]);
+                dist1 += quantize_band_cost(s, &sce->coeffs[start+(w+w2)*128], C34, sce->ics.swb_sizes[g],
+                                            sce->sf_idx[(w+w2)*16+g], sce->band_type[(w+w2)*16+g],
+                                            s->lambda/band->threshold, INFINITY, &bits_tmp1, NULL, 0);
+                dist2 += quantize_band_cost(s, PCD, PCD34, sce->ics.swb_sizes[g],
+                                            sce->sf_idx[(w+w2)*16+g],
+                                            sce->band_type[(w+w2)*16+g],
+                                            s->lambda/band->threshold, INFINITY, &bits_tmp2, NULL, 0);
+                bits1 += bits_tmp1;
+                bits2 += bits_tmp2;
+            }
+            if (dist2 < dist1 && bits2 < bits1) {
+                for (w2 = 0; w2 < sce->ics.group_len[w]; w2++)
+                    for (i = 0; i < sce->ics.swb_sizes[g]; i++)
+                        sce->coeffs[start+(w+w2)*128+i] -= sce->lcoeffs[start+(w+w2)*128+i];
+                sce->ics.ltp.used[w*16+g] = 1;
+                saved_bits += bits1 - bits2;
+                count++;
+            }
+            start += sce->ics.swb_sizes[g];
+        }
+    }
+
+    sce->ics.ltp.present = !!count && (saved_bits >= 0);
+    sce->ics.predictor_present = !!sce->ics.ltp.present;
+
+    /* Reset any marked sfbs */
+    if (!sce->ics.ltp.present && !!count) {
+        for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) {
+            start = 0;
+            for (g = 0;  g < sce->ics.num_swb; g++) {
+                if (sce->ics.ltp.used[w*16+g]) {
+                    for (w2 = 0; w2 < sce->ics.group_len[w]; w2++) {
+                        for (i = 0; i < sce->ics.swb_sizes[g]; i++) {
+                            sce->coeffs[start+(w+w2)*128+i] += sce->lcoeffs[start+(w+w2)*128+i];
+                        }
+                    }
+                }
+                start += sce->ics.swb_sizes[g];
+            }
+        }
+    }
+}

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aacenc_ltp.h

@@ -0,0 +1,41 @@
+/*
+ * AAC encoder long term prediction extension
+ * Copyright (C) 2015 Rostislav Pehlivanov
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+/**
+ * @file
+ * AAC encoder long term prediction extension
+ * @author Rostislav Pehlivanov ( atomnuker gmail com )
+ */
+
+#ifndef AVCODEC_AACENC_LTP_H
+#define AVCODEC_AACENC_LTP_H
+
+#include "aacenc.h"
+
+void ff_aac_encode_ltp_info(AACEncContext *s, SingleChannelElement *sce,
+                            int common_window);
+void ff_aac_update_ltp(AACEncContext *s, SingleChannelElement *sce);
+void ff_aac_adjust_common_ltp(AACEncContext *s, ChannelElement *cpe);
+void ff_aac_ltp_insert_new_frame(AACEncContext *s);
+void ff_aac_search_for_ltp(AACEncContext *s, SingleChannelElement *sce,
+                           int common_window);
+
+#endif /* AVCODEC_AACENC_LTP_H */

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aacenc_pred.c

@@ -0,0 +1,347 @@
+/*
+ * AAC encoder main-type prediction
+ * Copyright (C) 2015 Rostislav Pehlivanov
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+/**
+ * @file
+ * AAC encoder main-type prediction
+ * @author Rostislav Pehlivanov ( atomnuker gmail com )
+ */
+
+#include "aactab.h"
+#include "aacenc_pred.h"
+#include "aacenc_utils.h"
+#include "aacenc_is.h"            /* <- Needed for common window distortions */
+#include "aacenc_quantization.h"
+
+#define RESTORE_PRED(sce, sfb) \
+        if (sce->ics.prediction_used[sfb]) {\
+            sce->ics.prediction_used[sfb] = 0;\
+            sce->band_type[sfb] = sce->band_alt[sfb];\
+        }
+
+static inline float flt16_round(float pf)
+{
+    union av_intfloat32 tmp;
+    tmp.f = pf;
+    tmp.i = (tmp.i + 0x00008000U) & 0xFFFF0000U;
+    return tmp.f;
+}
+
+static inline float flt16_even(float pf)
+{
+    union av_intfloat32 tmp;
+    tmp.f = pf;
+    tmp.i = (tmp.i + 0x00007FFFU + (tmp.i & 0x00010000U >> 16)) & 0xFFFF0000U;
+    return tmp.f;
+}
+
+static inline float flt16_trunc(float pf)
+{
+    union av_intfloat32 pun;
+    pun.f = pf;
+    pun.i &= 0xFFFF0000U;
+    return pun.f;
+}
+
+static inline void predict(PredictorState *ps, float *coef, float *rcoef, int set)
+{
+    float k2;
+    const float a     = 0.953125; // 61.0 / 64
+    const float alpha = 0.90625;  // 29.0 / 32
+    const float   k1 = ps->k1;
+    const float   r0 = ps->r0,     r1 = ps->r1;
+    const float cor0 = ps->cor0, cor1 = ps->cor1;
+    const float var0 = ps->var0, var1 = ps->var1;
+    const float e0 = *coef - ps->x_est;
+    const float e1 = e0 - k1 * r0;
+
+    if (set)
+        *coef = e0;
+
+    ps->cor1 = flt16_trunc(alpha * cor1 + r1 * e1);
+    ps->var1 = flt16_trunc(alpha * var1 + 0.5f * (r1 * r1 + e1 * e1));
+    ps->cor0 = flt16_trunc(alpha * cor0 + r0 * e0);
+    ps->var0 = flt16_trunc(alpha * var0 + 0.5f * (r0 * r0 + e0 * e0));
+    ps->r1   = flt16_trunc(a * (r0 - k1 * e0));
+    ps->r0   = flt16_trunc(a * e0);
+
+    /* Prediction for next frame */
+    ps->k1   = ps->var0 > 1 ? ps->cor0 * flt16_even(a / ps->var0) : 0;
+    k2       = ps->var1 > 1 ? ps->cor1 * flt16_even(a / ps->var1) : 0;
+    *rcoef   = ps->x_est = flt16_round(ps->k1*ps->r0 + k2*ps->r1);
+}
+
+static inline void reset_predict_state(PredictorState *ps)
+{
+    ps->r0    = 0.0f;
+    ps->r1    = 0.0f;
+    ps->k1    = 0.0f;
+    ps->cor0  = 0.0f;
+    ps->cor1  = 0.0f;
+    ps->var0  = 1.0f;
+    ps->var1  = 1.0f;
+    ps->x_est = 0.0f;
+}
+
+static inline void reset_all_predictors(PredictorState *ps)
+{
+    int i;
+    for (i = 0; i < MAX_PREDICTORS; i++)
+        reset_predict_state(&ps[i]);
+}
+
+static inline void reset_predictor_group(SingleChannelElement *sce, int group_num)
+{
+    int i;
+    PredictorState *ps = sce->predictor_state;
+    for (i = group_num - 1; i < MAX_PREDICTORS; i += 30)
+        reset_predict_state(&ps[i]);
+}
+
+void ff_aac_apply_main_pred(AACEncContext *s, SingleChannelElement *sce)
+{
+    int sfb, k;
+    const int pmax = FFMIN(sce->ics.max_sfb, ff_aac_pred_sfb_max[s->samplerate_index]);
+
+    if (sce->ics.window_sequence[0] != EIGHT_SHORT_SEQUENCE) {
+        for (sfb = 0; sfb < pmax; sfb++) {
+            for (k = sce->ics.swb_offset[sfb]; k < sce->ics.swb_offset[sfb + 1]; k++) {
+                predict(&sce->predictor_state[k], &sce->coeffs[k], &sce->prcoeffs[k],
+                        sce->ics.predictor_present && sce->ics.prediction_used[sfb]);
+            }
+        }
+        if (sce->ics.predictor_reset_group) {
+            reset_predictor_group(sce, sce->ics.predictor_reset_group);
+        }
+    } else {
+        reset_all_predictors(sce->predictor_state);
+    }
+}
+
+/* If inc = 0 you can check if this returns 0 to see if you can reset freely */
+static inline int update_counters(IndividualChannelStream *ics, int inc)
+{
+    int i;
+    for (i = 1; i < 31; i++) {
+        ics->predictor_reset_count[i] += inc;
+        if (ics->predictor_reset_count[i] > PRED_RESET_FRAME_MIN)
+            return i; /* Reset this immediately */
+    }
+    return 0;
+}
+
+void ff_aac_adjust_common_pred(AACEncContext *s, ChannelElement *cpe)
+{
+    int start, w, w2, g, i, count = 0;
+    SingleChannelElement *sce0 = &cpe->ch[0];
+    SingleChannelElement *sce1 = &cpe->ch[1];
+    const int pmax0 = FFMIN(sce0->ics.max_sfb, ff_aac_pred_sfb_max[s->samplerate_index]);
+    const int pmax1 = FFMIN(sce1->ics.max_sfb, ff_aac_pred_sfb_max[s->samplerate_index]);
+    const int pmax  = FFMIN(pmax0, pmax1);
+
+    if (!cpe->common_window ||
+        sce0->ics.window_sequence[0] == EIGHT_SHORT_SEQUENCE ||
+        sce1->ics.window_sequence[0] == EIGHT_SHORT_SEQUENCE)
+        return;
+
+    for (w = 0; w < sce0->ics.num_windows; w += sce0->ics.group_len[w]) {
+        start = 0;
+        for (g = 0; g < sce0->ics.num_swb; g++) {
+            int sfb = w*16+g;
+            int sum = sce0->ics.prediction_used[sfb] + sce1->ics.prediction_used[sfb];
+            float ener0 = 0.0f, ener1 = 0.0f, ener01 = 0.0f;
+            struct AACISError ph_err1, ph_err2, *erf;
+            if (sfb < PRED_SFB_START || sfb > pmax || sum != 2) {
+                RESTORE_PRED(sce0, sfb);
+                RESTORE_PRED(sce1, sfb);
+                start += sce0->ics.swb_sizes[g];
+                continue;
+            }
+            for (w2 = 0; w2 < sce0->ics.group_len[w]; w2++) {
+                for (i = 0; i < sce0->ics.swb_sizes[g]; i++) {
+                    float coef0 = sce0->pcoeffs[start+(w+w2)*128+i];
+                    float coef1 = sce1->pcoeffs[start+(w+w2)*128+i];
+                    ener0  += coef0*coef0;
+                    ener1  += coef1*coef1;
+                    ener01 += (coef0 + coef1)*(coef0 + coef1);
+                }
+            }
+            ph_err1 = ff_aac_is_encoding_err(s, cpe, start, w, g,
+                                             ener0, ener1, ener01, 1, -1);
+            ph_err2 = ff_aac_is_encoding_err(s, cpe, start, w, g,
+                                             ener0, ener1, ener01, 1, +1);
+            erf = ph_err1.error < ph_err2.error ? &ph_err1 : &ph_err2;
+            if (erf->pass) {
+                sce0->ics.prediction_used[sfb] = 1;
+                sce1->ics.prediction_used[sfb] = 1;
+                count++;
+            } else {
+                RESTORE_PRED(sce0, sfb);
+                RESTORE_PRED(sce1, sfb);
+            }
+            start += sce0->ics.swb_sizes[g];
+        }
+    }
+
+    sce1->ics.predictor_present = sce0->ics.predictor_present = !!count;
+}
+
+static void update_pred_resets(SingleChannelElement *sce)
+{
+    int i, max_group_id_c, max_frame = 0;
+    float avg_frame = 0.0f;
+    IndividualChannelStream *ics = &sce->ics;
+
+    /* Update the counters and immediately update any frame behind schedule */
+    if ((ics->predictor_reset_group = update_counters(&sce->ics, 1)))
+        return;
+
+    for (i = 1; i < 31; i++) {
+        /* Count-based */
+        if (ics->predictor_reset_count[i] > max_frame) {
+            max_group_id_c = i;
+            max_frame = ics->predictor_reset_count[i];
+        }
+        avg_frame = (ics->predictor_reset_count[i] + avg_frame)/2;
+    }
+
+    if (max_frame > PRED_RESET_MIN) {
+        ics->predictor_reset_group = max_group_id_c;
+    } else {
+        ics->predictor_reset_group = 0;
+    }
+}
+
+void ff_aac_search_for_pred(AACEncContext *s, SingleChannelElement *sce)
+{
+    int sfb, i, count = 0, cost_coeffs = 0, cost_pred = 0;
+    const int pmax = FFMIN(sce->ics.max_sfb, ff_aac_pred_sfb_max[s->samplerate_index]);
+    float *O34  = &s->scoefs[128*0], *P34 = &s->scoefs[128*1];
+    float *SENT = &s->scoefs[128*2], *S34 = &s->scoefs[128*3];
+    float *QERR = &s->scoefs[128*4];
+
+    if (sce->ics.window_sequence[0] == EIGHT_SHORT_SEQUENCE) {
+        sce->ics.predictor_present = 0;
+        return;
+    }
+
+    if (!sce->ics.predictor_initialized) {
+        reset_all_predictors(sce->predictor_state);
+        sce->ics.predictor_initialized = 1;
+        memcpy(sce->prcoeffs, sce->coeffs, 1024*sizeof(float));
+        for (i = 1; i < 31; i++)
+            sce->ics.predictor_reset_count[i] = i;
+    }
+
+    update_pred_resets(sce);
+    memcpy(sce->band_alt, sce->band_type, sizeof(sce->band_type));
+
+    for (sfb = PRED_SFB_START; sfb < pmax; sfb++) {
+        int cost1, cost2, cb_p;
+        float dist1, dist2, dist_spec_err = 0.0f;
+        const int cb_n = sce->zeroes[sfb] ? 0 : sce->band_type[sfb];
+        const int cb_min = sce->zeroes[sfb] ? 0 : 1;
+        const int cb_max = sce->zeroes[sfb] ? 0 : RESERVED_BT;
+        const int start_coef = sce->ics.swb_offset[sfb];
+        const int num_coeffs = sce->ics.swb_offset[sfb + 1] - start_coef;
+        const FFPsyBand *band = &s->psy.ch[s->cur_channel].psy_bands[sfb];
+
+        if (start_coef + num_coeffs > MAX_PREDICTORS ||
+            (s->cur_channel && sce->band_type[sfb] >= INTENSITY_BT2) ||
+            sce->band_type[sfb] == NOISE_BT)
+            continue;
+
+        /* Normal coefficients */
+        s->abs_pow34(O34, &sce->coeffs[start_coef], num_coeffs);
+        dist1 = quantize_and_encode_band_cost(s, NULL, &sce->coeffs[start_coef], NULL,
+                                              O34, num_coeffs, sce->sf_idx[sfb],
+                                              cb_n, s->lambda / band->threshold, INFINITY, &cost1, NULL, 0);
+        cost_coeffs += cost1;
+
+        /* Encoded coefficients - needed for #bits, band type and quant. error */
+        for (i = 0; i < num_coeffs; i++)
+            SENT[i] = sce->coeffs[start_coef + i] - sce->prcoeffs[start_coef + i];
+        s->abs_pow34(S34, SENT, num_coeffs);
+        if (cb_n < RESERVED_BT)
+            cb_p = av_clip(find_min_book(find_max_val(1, num_coeffs, S34), sce->sf_idx[sfb]), cb_min, cb_max);
+        else
+            cb_p = cb_n;
+        quantize_and_encode_band_cost(s, NULL, SENT, QERR, S34, num_coeffs,
+                                      sce->sf_idx[sfb], cb_p, s->lambda / band->threshold, INFINITY,
+                                      &cost2, NULL, 0);
+
+        /* Reconstructed coefficients - needed for distortion measurements */
+        for (i = 0; i < num_coeffs; i++)
+            sce->prcoeffs[start_coef + i] += QERR[i] != 0.0f ? (sce->prcoeffs[start_coef + i] - QERR[i]) : 0.0f;
+        s->abs_pow34(P34, &sce->prcoeffs[start_coef], num_coeffs);
+        if (cb_n < RESERVED_BT)
+            cb_p = av_clip(find_min_book(find_max_val(1, num_coeffs, P34), sce->sf_idx[sfb]), cb_min, cb_max);
+        else
+            cb_p = cb_n;
+        dist2 = quantize_and_encode_band_cost(s, NULL, &sce->prcoeffs[start_coef], NULL,
+                                              P34, num_coeffs, sce->sf_idx[sfb],
+                                              cb_p, s->lambda / band->threshold, INFINITY, NULL, NULL, 0);
+        for (i = 0; i < num_coeffs; i++)
+            dist_spec_err += (O34[i] - P34[i])*(O34[i] - P34[i]);
+        dist_spec_err *= s->lambda / band->threshold;
+        dist2 += dist_spec_err;
+
+        if (dist2 <= dist1 && cb_p <= cb_n) {
+            cost_pred += cost2;
+            sce->ics.prediction_used[sfb] = 1;
+            sce->band_alt[sfb]  = cb_n;
+            sce->band_type[sfb] = cb_p;
+            count++;
+        } else {
+            cost_pred += cost1;
+            sce->band_alt[sfb] = cb_p;
+        }
+    }
+
+    if (count && cost_coeffs < cost_pred) {
+        count = 0;
+        for (sfb = PRED_SFB_START; sfb < pmax; sfb++)
+            RESTORE_PRED(sce, sfb);
+        memset(&sce->ics.prediction_used, 0, sizeof(sce->ics.prediction_used));
+    }
+
+    sce->ics.predictor_present = !!count;
+}
+
+/**
+ * Encoder predictors data.
+ */
+void ff_aac_encode_main_pred(AACEncContext *s, SingleChannelElement *sce)
+{
+    int sfb;
+    IndividualChannelStream *ics = &sce->ics;
+    const int pmax = FFMIN(ics->max_sfb, ff_aac_pred_sfb_max[s->samplerate_index]);
+
+    if (s->profile != FF_PROFILE_AAC_MAIN ||
+        !ics->predictor_present)
+        return;
+
+    put_bits(&s->pb, 1, !!ics->predictor_reset_group);
+    if (ics->predictor_reset_group)
+        put_bits(&s->pb, 5, ics->predictor_reset_group);
+    for (sfb = 0; sfb < pmax; sfb++)
+        put_bits(&s->pb, 1, ics->prediction_used[sfb]);
+}

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aacenc_pred.h

@@ -0,0 +1,47 @@
+/*
+ * AAC encoder main-type prediction
+ * Copyright (C) 2015 Rostislav Pehlivanov
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+/**
+ * @file
+ * AAC encoder main-type prediction
+ * @author Rostislav Pehlivanov ( atomnuker gmail com )
+ */
+
+#ifndef AVCODEC_AACENC_PRED_H
+#define AVCODEC_AACENC_PRED_H
+
+#include "aacenc.h"
+
+/* Every predictor group needs to get reset at least once in this many frames */
+#define PRED_RESET_FRAME_MIN 240
+
+/* Any frame with less than this amount of frames since last reset is ok */
+#define PRED_RESET_MIN 64
+
+/* Raise to filter any low frequency artifacts due to prediction */
+#define PRED_SFB_START 10
+
+void ff_aac_apply_main_pred(AACEncContext *s, SingleChannelElement *sce);
+void ff_aac_adjust_common_pred(AACEncContext *s, ChannelElement *cpe);
+void ff_aac_search_for_pred(AACEncContext *s, SingleChannelElement *sce);
+void ff_aac_encode_main_pred(AACEncContext *s, SingleChannelElement *sce);
+
+#endif /* AVCODEC_AACENC_PRED_H */

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aacenc_quantization.h

@@ -0,0 +1,283 @@
+/*
+ * AAC encoder quantizer
+ * Copyright (C) 2015 Rostislav Pehlivanov
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+/**
+ * @file
+ * AAC encoder quantizer
+ * @author Rostislav Pehlivanov ( atomnuker gmail com )
+ */
+
+#ifndef AVCODEC_AACENC_QUANTIZATION_H
+#define AVCODEC_AACENC_QUANTIZATION_H
+
+#include "aactab.h"
+#include "aacenc.h"
+#include "aacenctab.h"
+#include "aacenc_utils.h"
+
+/**
+ * Calculate rate distortion cost for quantizing with given codebook
+ *
+ * @return quantization distortion
+ */
+static av_always_inline float quantize_and_encode_band_cost_template(
+                                struct AACEncContext *s,
+                                PutBitContext *pb, const float *in, float *out,
+                                const float *scaled, int size, int scale_idx,
+                                int cb, const float lambda, const float uplim,
+                                int *bits, float *energy, int BT_ZERO, int BT_UNSIGNED,
+                                int BT_PAIR, int BT_ESC, int BT_NOISE, int BT_STEREO,
+                                const float ROUNDING)
+{
+    const int q_idx = POW_SF2_ZERO - scale_idx + SCALE_ONE_POS - SCALE_DIV_512;
+    const float Q   = ff_aac_pow2sf_tab [q_idx];
+    const float Q34 = ff_aac_pow34sf_tab[q_idx];
+    const float IQ  = ff_aac_pow2sf_tab [POW_SF2_ZERO + scale_idx - SCALE_ONE_POS + SCALE_DIV_512];
+    const float CLIPPED_ESCAPE = 165140.0f*IQ;
+    int i, j;
+    float cost = 0;
+    float qenergy = 0;
+    const int dim = BT_PAIR ? 2 : 4;
+    int resbits = 0;
+    int off;
+
+    if (BT_ZERO || BT_NOISE || BT_STEREO) {
+        for (i = 0; i < size; i++)
+            cost += in[i]*in[i];
+        if (bits)
+            *bits = 0;
+        if (energy)
+            *energy = qenergy;
+        if (out) {
+            for (i = 0; i < size; i += dim)
+                for (j = 0; j < dim; j++)
+                    out[i+j] = 0.0f;
+        }
+        return cost * lambda;
+    }
+    if (!scaled) {
+        s->abs_pow34(s->scoefs, in, size);
+        scaled = s->scoefs;
+    }
+    s->quant_bands(s->qcoefs, in, scaled, size, !BT_UNSIGNED, aac_cb_maxval[cb], Q34, ROUNDING);
+    if (BT_UNSIGNED) {
+        off = 0;
+    } else {
+        off = aac_cb_maxval[cb];
+    }
+    for (i = 0; i < size; i += dim) {
+        const float *vec;
+        int *quants = s->qcoefs + i;
+        int curidx = 0;
+        int curbits;
+        float quantized, rd = 0.0f;
+        for (j = 0; j < dim; j++) {
+            curidx *= aac_cb_range[cb];
+            curidx += quants[j] + off;
+        }
+        curbits =  ff_aac_spectral_bits[cb-1][curidx];
+        vec     = &ff_aac_codebook_vectors[cb-1][curidx*dim];
+        if (BT_UNSIGNED) {
+            for (j = 0; j < dim; j++) {
+                float t = fabsf(in[i+j]);
+                float di;
+                if (BT_ESC && vec[j] == 64.0f) { //FIXME: slow
+                    if (t >= CLIPPED_ESCAPE) {
+                        quantized = CLIPPED_ESCAPE;
+                        curbits += 21;
+                    } else {
+                        int c = av_clip_uintp2(quant(t, Q, ROUNDING), 13);
+                        quantized = c*cbrtf(c)*IQ;
+                        curbits += av_log2(c)*2 - 4 + 1;
+                    }
+                } else {
+                    quantized = vec[j]*IQ;
+                }
+                di = t - quantized;
+                if (out)
+                    out[i+j] = in[i+j] >= 0 ? quantized : -quantized;
+                if (vec[j] != 0.0f)
+                    curbits++;
+                qenergy += quantized*quantized;
+                rd += di*di;
+            }
+        } else {
+            for (j = 0; j < dim; j++) {
+                quantized = vec[j]*IQ;
+                qenergy += quantized*quantized;
+                if (out)
+                    out[i+j] = quantized;
+                rd += (in[i+j] - quantized)*(in[i+j] - quantized);
+            }
+        }
+        cost    += rd * lambda + curbits;
+        resbits += curbits;
+        if (cost >= uplim)
+            return uplim;
+        if (pb) {
+            put_bits(pb, ff_aac_spectral_bits[cb-1][curidx], ff_aac_spectral_codes[cb-1][curidx]);
+            if (BT_UNSIGNED)
+                for (j = 0; j < dim; j++)
+                    if (ff_aac_codebook_vectors[cb-1][curidx*dim+j] != 0.0f)
+                        put_bits(pb, 1, in[i+j] < 0.0f);
+            if (BT_ESC) {
+                for (j = 0; j < 2; j++) {
+                    if (ff_aac_codebook_vectors[cb-1][curidx*2+j] == 64.0f) {
+                        int coef = av_clip_uintp2(quant(fabsf(in[i+j]), Q, ROUNDING), 13);
+                        int len = av_log2(coef);
+
+                        put_bits(pb, len - 4 + 1, (1 << (len - 4 + 1)) - 2);
+                        put_sbits(pb, len, coef);
+                    }
+                }
+            }
+        }
+    }
+
+    if (bits)
+        *bits = resbits;
+    if (energy)
+        *energy = qenergy;
+    return cost;
+}
+
+static inline float quantize_and_encode_band_cost_NONE(struct AACEncContext *s, PutBitContext *pb,
+                                                const float *in, float *quant, const float *scaled,
+                                                int size, int scale_idx, int cb,
+                                                const float lambda, const float uplim,
+                                                int *bits, float *energy) {
+    av_assert0(0);
+    return 0.0f;
+}
+
+#define QUANTIZE_AND_ENCODE_BAND_COST_FUNC(NAME, BT_ZERO, BT_UNSIGNED, BT_PAIR, BT_ESC, BT_NOISE, BT_STEREO, ROUNDING) \
+static float quantize_and_encode_band_cost_ ## NAME(                                         \
+                                struct AACEncContext *s,                                     \
+                                PutBitContext *pb, const float *in, float *quant,            \
+                                const float *scaled, int size, int scale_idx,                \
+                                int cb, const float lambda, const float uplim,               \
+                                int *bits, float *energy) {                                  \
+    return quantize_and_encode_band_cost_template(                                           \
+                                s, pb, in, quant, scaled, size, scale_idx,                   \
+                                BT_ESC ? ESC_BT : cb, lambda, uplim, bits, energy,           \
+                                BT_ZERO, BT_UNSIGNED, BT_PAIR, BT_ESC, BT_NOISE, BT_STEREO,  \
+                                ROUNDING);                                                   \
+}
+
+QUANTIZE_AND_ENCODE_BAND_COST_FUNC(ZERO,  1, 0, 0, 0, 0, 0, ROUND_STANDARD)
+QUANTIZE_AND_ENCODE_BAND_COST_FUNC(SQUAD, 0, 0, 0, 0, 0, 0, ROUND_STANDARD)
+QUANTIZE_AND_ENCODE_BAND_COST_FUNC(UQUAD, 0, 1, 0, 0, 0, 0, ROUND_STANDARD)
+QUANTIZE_AND_ENCODE_BAND_COST_FUNC(SPAIR, 0, 0, 1, 0, 0, 0, ROUND_STANDARD)
+QUANTIZE_AND_ENCODE_BAND_COST_FUNC(UPAIR, 0, 1, 1, 0, 0, 0, ROUND_STANDARD)
+QUANTIZE_AND_ENCODE_BAND_COST_FUNC(ESC,   0, 1, 1, 1, 0, 0, ROUND_STANDARD)
+QUANTIZE_AND_ENCODE_BAND_COST_FUNC(ESC_RTZ, 0, 1, 1, 1, 0, 0, ROUND_TO_ZERO)
+QUANTIZE_AND_ENCODE_BAND_COST_FUNC(NOISE, 0, 0, 0, 0, 1, 0, ROUND_STANDARD)
+QUANTIZE_AND_ENCODE_BAND_COST_FUNC(STEREO,0, 0, 0, 0, 0, 1, ROUND_STANDARD)
+
+static float (*const quantize_and_encode_band_cost_arr[])(
+                                struct AACEncContext *s,
+                                PutBitContext *pb, const float *in, float *quant,
+                                const float *scaled, int size, int scale_idx,
+                                int cb, const float lambda, const float uplim,
+                                int *bits, float *energy) = {
+    quantize_and_encode_band_cost_ZERO,
+    quantize_and_encode_band_cost_SQUAD,
+    quantize_and_encode_band_cost_SQUAD,
+    quantize_and_encode_band_cost_UQUAD,
+    quantize_and_encode_band_cost_UQUAD,
+    quantize_and_encode_band_cost_SPAIR,
+    quantize_and_encode_band_cost_SPAIR,
+    quantize_and_encode_band_cost_UPAIR,
+    quantize_and_encode_band_cost_UPAIR,
+    quantize_and_encode_band_cost_UPAIR,
+    quantize_and_encode_band_cost_UPAIR,
+    quantize_and_encode_band_cost_ESC,
+    quantize_and_encode_band_cost_NONE,     /* CB 12 doesn't exist */
+    quantize_and_encode_band_cost_NOISE,
+    quantize_and_encode_band_cost_STEREO,
+    quantize_and_encode_band_cost_STEREO,
+};
+
+static float (*const quantize_and_encode_band_cost_rtz_arr[])(
+                                struct AACEncContext *s,
+                                PutBitContext *pb, const float *in, float *quant,
+                                const float *scaled, int size, int scale_idx,
+                                int cb, const float lambda, const float uplim,
+                                int *bits, float *energy) = {
+    quantize_and_encode_band_cost_ZERO,
+    quantize_and_encode_band_cost_SQUAD,
+    quantize_and_encode_band_cost_SQUAD,
+    quantize_and_encode_band_cost_UQUAD,
+    quantize_and_encode_band_cost_UQUAD,
+    quantize_and_encode_band_cost_SPAIR,
+    quantize_and_encode_band_cost_SPAIR,
+    quantize_and_encode_band_cost_UPAIR,
+    quantize_and_encode_band_cost_UPAIR,
+    quantize_and_encode_band_cost_UPAIR,
+    quantize_and_encode_band_cost_UPAIR,
+    quantize_and_encode_band_cost_ESC_RTZ,
+    quantize_and_encode_band_cost_NONE,     /* CB 12 doesn't exist */
+    quantize_and_encode_band_cost_NOISE,
+    quantize_and_encode_band_cost_STEREO,
+    quantize_and_encode_band_cost_STEREO,
+};
+
+#define quantize_and_encode_band_cost(                                  \
+                                s, pb, in, quant, scaled, size, scale_idx, cb, \
+                                lambda, uplim, bits, energy, rtz)               \
+    ((rtz) ? quantize_and_encode_band_cost_rtz_arr : quantize_and_encode_band_cost_arr)[cb]( \
+                                s, pb, in, quant, scaled, size, scale_idx, cb, \
+                                lambda, uplim, bits, energy)
+
+static inline float quantize_band_cost(struct AACEncContext *s, const float *in,
+                                const float *scaled, int size, int scale_idx,
+                                int cb, const float lambda, const float uplim,
+                                int *bits, float *energy, int rtz)
+{
+    return quantize_and_encode_band_cost(s, NULL, in, NULL, scaled, size, scale_idx,
+                                         cb, lambda, uplim, bits, energy, rtz);
+}
+
+static inline int quantize_band_cost_bits(struct AACEncContext *s, const float *in,
+                                const float *scaled, int size, int scale_idx,
+                                int cb, const float lambda, const float uplim,
+                                int *bits, float *energy, int rtz)
+{
+    int auxbits;
+    quantize_and_encode_band_cost(s, NULL, in, NULL, scaled, size, scale_idx,
+                                         cb, 0.0f, uplim, &auxbits, energy, rtz);
+    if (bits) {
+        *bits = auxbits;
+    }
+    return auxbits;
+}
+
+static inline void quantize_and_encode_band(struct AACEncContext *s, PutBitContext *pb,
+                                            const float *in, float *out, int size, int scale_idx,
+                                            int cb, const float lambda, int rtz)
+{
+    quantize_and_encode_band_cost(s, pb, in, out, NULL, size, scale_idx, cb, lambda,
+                                  INFINITY, NULL, NULL, rtz);
+}
+
+#include "aacenc_quantization_misc.h"
+
+#endif /* AVCODEC_AACENC_QUANTIZATION_H */

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aacenc_quantization_misc.h

@@ -0,0 +1,53 @@
+/*
+ * AAC encoder quantization
+ * Copyright (C) 2015 Claudio Freire
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+/**
+ * @file
+ * AAC encoder quantization misc reusable function templates
+ * @author Claudio Freire ( klaussfreire gmail com )
+ */
+
+#ifndef AVCODEC_AACENC_QUANTIZATION_MISC_H
+#define AVCODEC_AACENC_QUANTIZATION_MISC_H
+
+static inline float quantize_band_cost_cached(struct AACEncContext *s, int w, int g, const float *in,
+                                const float *scaled, int size, int scale_idx,
+                                int cb, const float lambda, const float uplim,
+                                int *bits, float *energy, int rtz)
+{
+    AACQuantizeBandCostCacheEntry *entry;
+    av_assert1(scale_idx >= 0 && scale_idx < 256);
+    entry = &s->quantize_band_cost_cache[scale_idx][w*16+g];
+    if (entry->generation != s->quantize_band_cost_cache_generation || entry->cb != cb || entry->rtz != rtz) {
+        entry->rd = quantize_band_cost(s, in, scaled, size, scale_idx,
+                                       cb, lambda, uplim, &entry->bits, &entry->energy, rtz);
+        entry->cb = cb;
+        entry->rtz = rtz;
+        entry->generation = s->quantize_band_cost_cache_generation;
+    }
+    if (bits)
+        *bits = entry->bits;
+    if (energy)
+        *energy = entry->energy;
+    return entry->rd;
+}
+
+#endif /* AVCODEC_AACENC_QUANTIZATION_MISC_H */

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aacenc_tns.c

@@ -0,0 +1,215 @@
+/*
+ * AAC encoder TNS
+ * Copyright (C) 2015 Rostislav Pehlivanov
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+/**
+ * @file
+ * AAC encoder temporal noise shaping
+ * @author Rostislav Pehlivanov ( atomnuker gmail com )
+ */
+
+#include "libavutil/libm.h"
+#include "aacenc.h"
+#include "aacenc_tns.h"
+#include "aactab.h"
+#include "aacenc_utils.h"
+#include "aacenc_quantization.h"
+
+/* Could be set to 3 to save an additional bit at the cost of little quality */
+#define TNS_Q_BITS 4
+
+/* Coefficient resolution in short windows */
+#define TNS_Q_BITS_IS8 4
+
+/* We really need the bits we save here elsewhere */
+#define TNS_ENABLE_COEF_COMPRESSION
+
+/* TNS will only be used if the LPC gain is within these margins */
+#define TNS_GAIN_THRESHOLD_LOW      1.4f
+#define TNS_GAIN_THRESHOLD_HIGH     1.16f*TNS_GAIN_THRESHOLD_LOW
+
+static inline int compress_coeffs(int *coef, int order, int c_bits)
+{
+    int i;
+    const int low_idx   = c_bits ?  4 : 2;
+    const int shift_val = c_bits ?  8 : 4;
+    const int high_idx  = c_bits ? 11 : 5;
+#ifndef TNS_ENABLE_COEF_COMPRESSION
+    return 0;
+#endif /* TNS_ENABLE_COEF_COMPRESSION */
+    for (i = 0; i < order; i++)
+        if (coef[i] >= low_idx && coef[i] <= high_idx)
+            return 0;
+    for (i = 0; i < order; i++)
+        coef[i] -= (coef[i] > high_idx) ? shift_val : 0;
+    return 1;
+}
+
+/**
+ * Encode TNS data.
+ * Coefficient compression is simply not lossless as it should be
+ * on any decoder tested and as such is not active.
+ */
+void ff_aac_encode_tns_info(AACEncContext *s, SingleChannelElement *sce)
+{
+    TemporalNoiseShaping *tns = &sce->tns;
+    int i, w, filt, coef_compress = 0, coef_len;
+    const int is8 = sce->ics.window_sequence[0] == EIGHT_SHORT_SEQUENCE;
+    const int c_bits = is8 ? TNS_Q_BITS_IS8 == 4 : TNS_Q_BITS == 4;
+
+    if (!sce->tns.present)
+        return;
+
+    for (i = 0; i < sce->ics.num_windows; i++) {
+        put_bits(&s->pb, 2 - is8, sce->tns.n_filt[i]);
+        if (!tns->n_filt[i])
+            continue;
+        put_bits(&s->pb, 1, c_bits);
+        for (filt = 0; filt < tns->n_filt[i]; filt++) {
+            put_bits(&s->pb, 6 - 2 * is8, tns->length[i][filt]);
+            put_bits(&s->pb, 5 - 2 * is8, tns->order[i][filt]);
+            if (!tns->order[i][filt])
+                continue;
+            put_bits(&s->pb, 1, tns->direction[i][filt]);
+            coef_compress = compress_coeffs(tns->coef_idx[i][filt],
+                                            tns->order[i][filt], c_bits);
+            put_bits(&s->pb, 1, coef_compress);
+            coef_len = c_bits + 3 - coef_compress;
+            for (w = 0; w < tns->order[i][filt]; w++)
+                put_bits(&s->pb, coef_len, tns->coef_idx[i][filt][w]);
+        }
+    }
+}
+
+/* Apply TNS filter */
+void ff_aac_apply_tns(AACEncContext *s, SingleChannelElement *sce)
+{
+    TemporalNoiseShaping *tns = &sce->tns;
+    IndividualChannelStream *ics = &sce->ics;
+    int w, filt, m, i, top, order, bottom, start, end, size, inc;
+    const int mmm = FFMIN(ics->tns_max_bands, ics->max_sfb);
+    float lpc[TNS_MAX_ORDER];
+
+    for (w = 0; w < ics->num_windows; w++) {
+        bottom = ics->num_swb;
+        for (filt = 0; filt < tns->n_filt[w]; filt++) {
+            top    = bottom;
+            bottom = FFMAX(0, top - tns->length[w][filt]);
+            order  = tns->order[w][filt];
+            if (order == 0)
+                continue;
+
+            // tns_decode_coef
+            compute_lpc_coefs(tns->coef[w][filt], order, lpc, 0, 0, 0);
+
+            start = ics->swb_offset[FFMIN(bottom, mmm)];
+            end   = ics->swb_offset[FFMIN(   top, mmm)];
+            if ((size = end - start) <= 0)
+                continue;
+            if (tns->direction[w][filt]) {
+                inc = -1;
+                start = end - 1;
+            } else {
+                inc = 1;
+            }
+            start += w * 128;
+
+            /* AR filter */
+            for (m = 0; m < size; m++, start += inc) {
+                for (i = 1; i <= FFMIN(m, order); i++) {
+                    sce->coeffs[start] += lpc[i-1]*sce->pcoeffs[start - i*inc];
+                }
+            }
+        }
+    }
+}
+
+/*
+ * c_bits - 1 if 4 bit coefficients, 0 if 3 bit coefficients
+ */
+static inline void quantize_coefs(double *coef, int *idx, float *lpc, int order,
+                                  int c_bits)
+{
+    int i;
+    const float *quant_arr = tns_tmp2_map[c_bits];
+    for (i = 0; i < order; i++) {
+        idx[i] = quant_array_idx(coef[i], quant_arr, c_bits ? 16 : 8);
+        lpc[i] = quant_arr[idx[i]];
+    }
+}
+
+/*
+ * 3 bits per coefficient with 8 short windows
+ */
+void ff_aac_search_for_tns(AACEncContext *s, SingleChannelElement *sce)
+{
+    TemporalNoiseShaping *tns = &sce->tns;
+    int w, g, count = 0;
+    double gain, coefs[MAX_LPC_ORDER];
+    const int mmm = FFMIN(sce->ics.tns_max_bands, sce->ics.max_sfb);
+    const int is8 = sce->ics.window_sequence[0] == EIGHT_SHORT_SEQUENCE;
+    const int c_bits = is8 ? TNS_Q_BITS_IS8 == 4 : TNS_Q_BITS == 4;
+    const int sfb_start = av_clip(tns_min_sfb[is8][s->samplerate_index], 0, mmm);
+    const int sfb_end   = av_clip(sce->ics.num_swb, 0, mmm);
+    const int order = is8 ? 7 : s->profile == FF_PROFILE_AAC_LOW ? 12 : TNS_MAX_ORDER;
+    const int slant = sce->ics.window_sequence[0] == LONG_STOP_SEQUENCE  ? 1 :
+                      sce->ics.window_sequence[0] == LONG_START_SEQUENCE ? 0 : 2;
+    const int sfb_len = sfb_end - sfb_start;
+    const int coef_len = sce->ics.swb_offset[sfb_end] - sce->ics.swb_offset[sfb_start];
+
+    if (coef_len <= 0 || sfb_len <= 0) {
+        sce->tns.present = 0;
+        return;
+    }
+
+    for (w = 0; w < sce->ics.num_windows; w++) {
+        float en[2] = {0.0f, 0.0f};
+        int oc_start = 0, os_start = 0;
+        int coef_start = sce->ics.swb_offset[sfb_start];
+
+        for (g = sfb_start; g < sce->ics.num_swb && g <= sfb_end; g++) {
+            FFPsyBand *band = &s->psy.ch[s->cur_channel].psy_bands[w*16+g];
+            if (g > sfb_start + (sfb_len/2))
+                en[1] += band->energy;
+            else
+                en[0] += band->energy;
+        }
+
+        /* LPC */
+        gain = ff_lpc_calc_ref_coefs_f(&s->lpc, &sce->coeffs[w*128 + coef_start],
+                                       coef_len, order, coefs);
+
+        if (!order || !isfinite(gain) || gain < TNS_GAIN_THRESHOLD_LOW || gain > TNS_GAIN_THRESHOLD_HIGH)
+            continue;
+
+        tns->n_filt[w] = is8 ? 1 : order != TNS_MAX_ORDER ? 2 : 3;
+        for (g = 0; g < tns->n_filt[w]; g++) {
+            tns->direction[w][g] = slant != 2 ? slant : en[g] < en[!g];
+            tns->order[w][g] = g < tns->n_filt[w] ? order/tns->n_filt[w] : order - oc_start;
+            tns->length[w][g] = g < tns->n_filt[w] ? sfb_len/tns->n_filt[w] : sfb_len - os_start;
+            quantize_coefs(&coefs[oc_start], tns->coef_idx[w][g], tns->coef[w][g],
+                            tns->order[w][g], c_bits);
+            oc_start += tns->order[w][g];
+            os_start += tns->length[w][g];
+        }
+        count++;
+    }
+    sce->tns.present = !!count;
+}

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aacenc_tns.h

@@ -0,0 +1,37 @@
+/*
+ * AAC encoder TNS
+ * Copyright (C) 2015 Rostislav Pehlivanov
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+/**
+ * @file
+ * AAC encoder temporal noise shaping
+ * @author Rostislav Pehlivanov ( atomnuker gmail com )
+ */
+
+#ifndef AVCODEC_AACENC_TNS_H
+#define AVCODEC_AACENC_TNS_H
+
+#include "aacenc.h"
+
+void ff_aac_encode_tns_info(AACEncContext *s, SingleChannelElement *sce);
+void ff_aac_apply_tns(AACEncContext *s, SingleChannelElement *sce);
+void ff_aac_search_for_tns(AACEncContext *s, SingleChannelElement *sce);
+
+#endif /* AVCODEC_AACENC_TNS_H */

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aacenc_utils.h

@@ -0,0 +1,279 @@
+/*
+ * AAC encoder utilities
+ * Copyright (C) 2015 Rostislav Pehlivanov
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+/**
+ * @file
+ * AAC encoder utilities
+ * @author Rostislav Pehlivanov ( atomnuker gmail com )
+ */
+
+#ifndef AVCODEC_AACENC_UTILS_H
+#define AVCODEC_AACENC_UTILS_H
+
+#include "libavutil/ffmath.h"
+#include "aac.h"
+#include "aacenctab.h"
+#include "aactab.h"
+
+#define ROUND_STANDARD 0.4054f
+#define ROUND_TO_ZERO 0.1054f
+#define C_QUANT 0.4054f
+
+static inline void abs_pow34_v(float *out, const float *in, const int size)
+{
+    int i;
+    for (i = 0; i < size; i++) {
+        float a = fabsf(in[i]);
+        out[i] = sqrtf(a * sqrtf(a));
+    }
+}
+
+static inline float pos_pow34(float a)
+{
+    return sqrtf(a * sqrtf(a));
+}
+
+/**
+ * Quantize one coefficient.
+ * @return absolute value of the quantized coefficient
+ * @see 3GPP TS26.403 5.6.2 "Scalefactor determination"
+ */
+static inline int quant(float coef, const float Q, const float rounding)
+{
+    float a = coef * Q;
+    return sqrtf(a * sqrtf(a)) + rounding;
+}
+
+static inline void quantize_bands(int *out, const float *in, const float *scaled,
+                                  int size, int is_signed, int maxval, const float Q34,
+                                  const float rounding)
+{
+    int i;
+    for (i = 0; i < size; i++) {
+        float qc = scaled[i] * Q34;
+        int tmp = (int)FFMIN(qc + rounding, (float)maxval);
+        if (is_signed && in[i] < 0.0f) {
+            tmp = -tmp;
+        }
+        out[i] = tmp;
+    }
+}
+
+static inline float find_max_val(int group_len, int swb_size, const float *scaled)
+{
+    float maxval = 0.0f;
+    int w2, i;
+    for (w2 = 0; w2 < group_len; w2++) {
+        for (i = 0; i < swb_size; i++) {
+            maxval = FFMAX(maxval, scaled[w2*128+i]);
+        }
+    }
+    return maxval;
+}
+
+static inline int find_min_book(float maxval, int sf)
+{
+    float Q34 = ff_aac_pow34sf_tab[POW_SF2_ZERO - sf + SCALE_ONE_POS - SCALE_DIV_512];
+    int qmaxval, cb;
+    qmaxval = maxval * Q34 + C_QUANT;
+    if (qmaxval >= (FF_ARRAY_ELEMS(aac_maxval_cb)))
+        cb = 11;
+    else
+        cb = aac_maxval_cb[qmaxval];
+    return cb;
+}
+
+static inline float find_form_factor(int group_len, int swb_size, float thresh,
+                                     const float *scaled, float nzslope) {
+    const float iswb_size = 1.0f / swb_size;
+    const float iswb_sizem1 = 1.0f / (swb_size - 1);
+    const float ethresh = thresh;
+    float form = 0.0f, weight = 0.0f;
+    int w2, i;
+    for (w2 = 0; w2 < group_len; w2++) {
+        float e = 0.0f, e2 = 0.0f, var = 0.0f, maxval = 0.0f;
+        float nzl = 0;
+        for (i = 0; i < swb_size; i++) {
+            float s = fabsf(scaled[w2*128+i]);
+            maxval = FFMAX(maxval, s);
+            e += s;
+            e2 += s *= s;
+            /* We really don't want a hard non-zero-line count, since
+             * even below-threshold lines do add up towards band spectral power.
+             * So, fall steeply towards zero, but smoothly
+             */
+            if (s >= ethresh) {
+                nzl += 1.0f;
+            } else {
+                if (nzslope == 2.f)
+                    nzl += (s / ethresh) * (s / ethresh);
+                else
+                    nzl += ff_fast_powf(s / ethresh, nzslope);
+            }
+        }
+        if (e2 > thresh) {
+            float frm;
+            e *= iswb_size;
+
+            /** compute variance */
+            for (i = 0; i < swb_size; i++) {
+                float d = fabsf(scaled[w2*128+i]) - e;
+                var += d*d;
+            }
+            var = sqrtf(var * iswb_sizem1);
+
+            e2 *= iswb_size;
+            frm = e / FFMIN(e+4*var,maxval);
+            form += e2 * sqrtf(frm) / FFMAX(0.5f,nzl);
+            weight += e2;
+        }
+    }
+    if (weight > 0) {
+        return form / weight;
+    } else {
+        return 1.0f;
+    }
+}
+
+/** Return the minimum scalefactor where the quantized coef does not clip. */
+static inline uint8_t coef2minsf(float coef)
+{
+    return av_clip_uint8(log2f(coef)*4 - 69 + SCALE_ONE_POS - SCALE_DIV_512);
+}
+
+/** Return the maximum scalefactor where the quantized coef is not zero. */
+static inline uint8_t coef2maxsf(float coef)
+{
+    return av_clip_uint8(log2f(coef)*4 +  6 + SCALE_ONE_POS - SCALE_DIV_512);
+}
+
+/*
+ * Returns the closest possible index to an array of float values, given a value.
+ */
+static inline int quant_array_idx(const float val, const float *arr, const int num)
+{
+    int i, index = 0;
+    float quant_min_err = INFINITY;
+    for (i = 0; i < num; i++) {
+        float error = (val - arr[i])*(val - arr[i]);
+        if (error < quant_min_err) {
+            quant_min_err = error;
+            index = i;
+        }
+    }
+    return index;
+}
+
+/**
+ * approximates exp10f(-3.0f*(0.5f + 0.5f * cosf(FFMIN(b,15.5f) / 15.5f)))
+ */
+static av_always_inline float bval2bmax(float b)
+{
+    return 0.001f + 0.0035f * (b*b*b) / (15.5f*15.5f*15.5f);
+}
+
+/*
+ * Compute a nextband map to be used with SF delta constraint utilities.
+ * The nextband array should contain 128 elements, and positions that don't
+ * map to valid, nonzero bands of the form w*16+g (with w being the initial
+ * window of the window group, only) are left indetermined.
+ */
+static inline void ff_init_nextband_map(const SingleChannelElement *sce, uint8_t *nextband)
+{
+    unsigned char prevband = 0;
+    int w, g;
+    /** Just a safe default */
+    for (g = 0; g < 128; g++)
+        nextband[g] = g;
+
+    /** Now really navigate the nonzero band chain */
+    for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) {
+        for (g = 0; g < sce->ics.num_swb; g++) {
+            if (!sce->zeroes[w*16+g] && sce->band_type[w*16+g] < RESERVED_BT)
+                prevband = nextband[prevband] = w*16+g;
+        }
+    }
+    nextband[prevband] = prevband; /* terminate */
+}
+
+/*
+ * Updates nextband to reflect a removed band (equivalent to
+ * calling ff_init_nextband_map after marking a band as zero)
+ */
+static inline void ff_nextband_remove(uint8_t *nextband, int prevband, int band)
+{
+    nextband[prevband] = nextband[band];
+}
+
+/*
+ * Checks whether the specified band could be removed without inducing
+ * scalefactor delta that violates SF delta encoding constraints.
+ * prev_sf has to be the scalefactor of the previous nonzero, nonspecial
+ * band, in encoding order, or negative if there was no such band.
+ */
+static inline int ff_sfdelta_can_remove_band(const SingleChannelElement *sce,
+    const uint8_t *nextband, int prev_sf, int band)
+{
+    return prev_sf >= 0
+        && sce->sf_idx[nextband[band]] >= (prev_sf - SCALE_MAX_DIFF)
+        && sce->sf_idx[nextband[band]] <= (prev_sf + SCALE_MAX_DIFF);
+}
+
+/*
+ * Checks whether the specified band's scalefactor could be replaced
+ * with another one without violating SF delta encoding constraints.
+ * prev_sf has to be the scalefactor of the previous nonzero, nonsepcial
+ * band, in encoding order, or negative if there was no such band.
+ */
+static inline int ff_sfdelta_can_replace(const SingleChannelElement *sce,
+    const uint8_t *nextband, int prev_sf, int new_sf, int band)
+{
+    return new_sf >= (prev_sf - SCALE_MAX_DIFF)
+        && new_sf <= (prev_sf + SCALE_MAX_DIFF)
+        && sce->sf_idx[nextband[band]] >= (new_sf - SCALE_MAX_DIFF)
+        && sce->sf_idx[nextband[band]] <= (new_sf + SCALE_MAX_DIFF);
+}
+
+/**
+ * linear congruential pseudorandom number generator
+ *
+ * @param   previous_val    pointer to the current state of the generator
+ *
+ * @return  Returns a 32-bit pseudorandom integer
+ */
+static av_always_inline int lcg_random(unsigned previous_val)
+{
+    union { unsigned u; int s; } v = { previous_val * 1664525u + 1013904223 };
+    return v.s;
+}
+
+#define ERROR_IF(cond, ...) \
+    if (cond) { \
+        av_log(avctx, AV_LOG_ERROR, __VA_ARGS__); \
+        return AVERROR(EINVAL); \
+    }
+
+#define WARN_IF(cond, ...) \
+    if (cond) { \
+        av_log(avctx, AV_LOG_WARNING, __VA_ARGS__); \
+    }
+
+#endif /* AVCODEC_AACENC_UTILS_H */

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aacencdsp.asm

@@ -0,0 +1,86 @@
+;******************************************************************************
+;* SIMD optimized AAC encoder DSP functions
+;*
+;* Copyright (C) 2016 Rostislav Pehlivanov <atomnuker@gmail.com>
+;*
+;* This file is part of FFmpeg.
+;*
+;* FFmpeg is free software; you can redistribute it and/or
+;* modify it under the terms of the GNU Lesser General Public
+;* License as published by the Free Software Foundation; either
+;* version 2.1 of the License, or (at your option) any later version.
+;*
+;* FFmpeg is distributed in the hope that it will be useful,
+;* but WITHOUT ANY WARRANTY; without even the implied warranty of
+;* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+;* Lesser General Public License for more details.
+;*
+;* You should have received a copy of the GNU Lesser General Public
+;* License along with FFmpeg; if not, write to the Free Software
+;* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+;******************************************************************************
+
+%include "libavutil/x86/x86util.asm"
+
+SECTION_RODATA
+
+float_abs_mask: times 4 dd 0x7fffffff
+
+SECTION .text
+
+;*******************************************************************
+;void ff_abs_pow34(float *out, const float *in, const int size);
+;*******************************************************************
+INIT_XMM sse
+cglobal abs_pow34, 3, 3, 3, out, in, size
+    mova   m2, [float_abs_mask]
+    shl    sizeq, 2
+    add    inq, sizeq
+    add    outq, sizeq
+    neg    sizeq
+.loop:
+    andps  m0, m2, [inq+sizeq]
+    sqrtps m1, m0
+    mulps  m0, m1
+    sqrtps m0, m0
+    mova   [outq+sizeq], m0
+    add    sizeq, mmsize
+    jl    .loop
+    RET
+
+;*******************************************************************
+;void ff_aac_quantize_bands(int *out, const float *in, const float *scaled,
+;                           int size, int is_signed, int maxval, const float Q34,
+;                           const float rounding)
+;*******************************************************************
+INIT_XMM sse2
+cglobal aac_quantize_bands, 5, 5, 6, out, in, scaled, size, is_signed, maxval, Q34, rounding
+%if UNIX64 == 0
+    movss     m0, Q34m
+    movss     m1, roundingm
+    cvtsi2ss  m3, dword maxvalm
+%else
+    cvtsi2ss  m3, maxvald
+%endif
+    shufps    m0, m0, 0
+    shufps    m1, m1, 0
+    shufps    m3, m3, 0
+    shl       is_signedd, 31
+    movd      m4, is_signedd
+    shufps    m4, m4, 0
+    shl       sized,   2
+    add       inq, sizeq
+    add       outq, sizeq
+    add       scaledq, sizeq
+    neg       sizeq
+.loop:
+    mulps     m2, m0, [scaledq+sizeq]
+    addps     m2, m1
+    minps     m2, m3
+    andps     m5, m4, [inq+sizeq]
+    orps      m2, m5
+    cvttps2dq m2, m2
+    mova      [outq+sizeq], m2
+    add       sizeq, mmsize
+    jl       .loop
+    RET

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aacencdsp_init.c

@@ -0,0 +1,43 @@
+/*
+ * AAC encoder assembly optimizations
+ * Copyright (C) 2016 Rostislav Pehlivanov <atomnuker@gmail.com>
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include "config.h"
+
+#include "libavutil/float_dsp.h"
+#include "libavutil/x86/cpu.h"
+#include "libavcodec/aacenc.h"
+
+void ff_abs_pow34_sse(float *out, const float *in, const int size);
+
+void ff_aac_quantize_bands_sse2(int *out, const float *in, const float *scaled,
+                                int size, int is_signed, int maxval, const float Q34,
+                                const float rounding);
+
+av_cold void ff_aac_dsp_init_x86(AACEncContext *s)
+{
+    int cpu_flags = av_get_cpu_flags();
+
+    if (EXTERNAL_SSE(cpu_flags))
+        s->abs_pow34   = ff_abs_pow34_sse;
+
+    if (EXTERNAL_SSE2(cpu_flags))
+        s->quant_bands = ff_aac_quantize_bands_sse2;
+}

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aacenctab.c

@@ -0,0 +1,108 @@
+/*
+ * AAC encoder data
+ * Copyright (c) 2015 Rostislav Pehlivanov ( atomnuker gmail com )
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include "aacenctab.h"
+
+static const uint8_t swb_size_128_96[] = {
+    4, 4, 4, 4, 4, 4, 8, 8, 8, 16, 28, 36
+};
+
+static const uint8_t swb_size_128_64[] = {
+    4, 4, 4, 4, 4, 4, 8, 8, 8, 16, 28, 36
+};
+
+static const uint8_t swb_size_128_48[] = {
+    4, 4, 4, 4, 4, 8, 8, 8, 12, 12, 12, 16, 16, 16
+};
+
+static const uint8_t swb_size_128_24[] = {
+    4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 12, 12, 16, 16, 20
+};
+
+static const uint8_t swb_size_128_16[] = {
+    4, 4, 4, 4, 4, 4, 4, 4, 8, 8, 12, 12, 16, 20, 20
+};
+
+static const uint8_t swb_size_128_8[] = {
+    4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 8, 12, 16, 20, 20
+};
+
+static const uint8_t swb_size_1024_96[] = {
+    4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 8, 8,
+    12, 12, 12, 12, 12, 16, 16, 24, 28, 36, 44,
+    64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64
+};
+
+static const uint8_t swb_size_1024_64[] = {
+    4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 8,
+    12, 12, 12, 16, 16, 16, 20, 24, 24, 28, 36,
+    40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40
+};
+
+static const uint8_t swb_size_1024_48[] = {
+    4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 8, 8, 8, 8,
+    12, 12, 12, 12, 16, 16, 20, 20, 24, 24, 28, 28,
+    32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32,
+    96
+};
+
+static const uint8_t swb_size_1024_32[] = {
+    4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 8, 8, 8, 8,
+    12, 12, 12, 12, 16, 16, 20, 20, 24, 24, 28, 28,
+    32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32
+};
+
+static const uint8_t swb_size_1024_24[] = {
+    4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
+    12, 12, 12, 12, 16, 16, 16, 20, 20, 24, 24, 28, 28,
+    32, 36, 36, 40, 44, 48, 52, 52, 64, 64, 64, 64, 64
+};
+
+static const uint8_t swb_size_1024_16[] = {
+    8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
+    12, 12, 12, 12, 12, 12, 12, 12, 12, 16, 16, 16, 16, 20, 20, 20, 24, 24, 28, 28,
+    32, 36, 40, 40, 44, 48, 52, 56, 60, 64, 64, 64
+};
+
+static const uint8_t swb_size_1024_8[] = {
+    12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+    16, 16, 16, 16, 16, 16, 16, 20, 20, 20, 20, 24, 24, 24, 28, 28,
+    32, 36, 36, 40, 44, 48, 52, 56, 60, 64, 80
+};
+
+const uint8_t *ff_aac_swb_size_128[] = {
+    swb_size_128_96, swb_size_128_96, swb_size_128_64,
+    swb_size_128_48, swb_size_128_48, swb_size_128_48,
+    swb_size_128_24, swb_size_128_24, swb_size_128_16,
+    swb_size_128_16, swb_size_128_16, swb_size_128_8,
+    swb_size_128_8
+};
+
+const uint8_t *ff_aac_swb_size_1024[] = {
+    swb_size_1024_96, swb_size_1024_96, swb_size_1024_64,
+    swb_size_1024_48, swb_size_1024_48, swb_size_1024_32,
+    swb_size_1024_24, swb_size_1024_24, swb_size_1024_16,
+    swb_size_1024_16, swb_size_1024_16, swb_size_1024_8,
+    swb_size_1024_8
+};
+
+const int ff_aac_swb_size_128_len  = FF_ARRAY_ELEMS(ff_aac_swb_size_128);
+const int ff_aac_swb_size_1024_len = FF_ARRAY_ELEMS(ff_aac_swb_size_1024);

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aacenctab.h

@@ -0,0 +1,139 @@
+/*
+ * AAC encoder data
+ * Copyright (c) 2015 Rostislav Pehlivanov ( atomnuker gmail com )
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+/**
+ * @file
+ * AAC encoder data
+ * @author Rostislav Pehlivanov ( atomnuker gmail com )
+ */
+
+#ifndef AVCODEC_AACENCTAB_H
+#define AVCODEC_AACENCTAB_H
+
+#include "aac.h"
+
+/** Total number of usable codebooks **/
+#define CB_TOT 12
+
+/** Total number of codebooks, including special ones **/
+#define CB_TOT_ALL 15
+
+#define AAC_MAX_CHANNELS 16
+
+extern const uint8_t *ff_aac_swb_size_1024[];
+extern const int      ff_aac_swb_size_1024_len;
+extern const uint8_t *ff_aac_swb_size_128[];
+extern const int      ff_aac_swb_size_128_len;
+
+/* Supported layouts without using a PCE */
+static const int64_t aac_normal_chan_layouts[7] = {
+    AV_CH_LAYOUT_MONO,
+    AV_CH_LAYOUT_STEREO,
+    AV_CH_LAYOUT_SURROUND,
+    AV_CH_LAYOUT_4POINT0,
+    AV_CH_LAYOUT_5POINT0_BACK,
+    AV_CH_LAYOUT_5POINT1_BACK,
+    AV_CH_LAYOUT_7POINT1,
+};
+
+/** default channel configurations */
+static const uint8_t aac_chan_configs[AAC_MAX_CHANNELS][6] = {
+    {1, TYPE_SCE},                                         // 1 channel  - single channel element
+    {1, TYPE_CPE},                                         // 2 channels - channel pair
+    {2, TYPE_SCE, TYPE_CPE},                               // 3 channels - center + stereo
+    {3, TYPE_SCE, TYPE_CPE, TYPE_SCE},                     // 4 channels - front center + stereo + back center
+    {3, TYPE_SCE, TYPE_CPE, TYPE_CPE},                     // 5 channels - front center + stereo + back stereo
+    {4, TYPE_SCE, TYPE_CPE, TYPE_CPE, TYPE_LFE},           // 6 channels - front center + stereo + back stereo + LFE
+    {0},                                                   // 7 channels - invalid without PCE
+    {5, TYPE_SCE, TYPE_CPE, TYPE_CPE, TYPE_CPE, TYPE_LFE}, // 8 channels - front center + front stereo + side stereo + back stereo + LFE
+};
+
+/**
+ * Table to remap channels from libavcodec's default order to AAC order.
+ */
+static const uint8_t aac_chan_maps[AAC_MAX_CHANNELS][AAC_MAX_CHANNELS] = {
+    { 0 },
+    { 0, 1 },
+    { 2, 0, 1 },
+    { 2, 0, 1, 3 },
+    { 2, 0, 1, 3, 4 },
+    { 2, 0, 1, 4, 5, 3 },
+    { 0 },
+    { 2, 0, 1, 6, 7, 4, 5, 3 },
+};
+
+/* duplicated from avpriv_mpeg4audio_sample_rates to avoid shared build
+ * failures */
+static const int mpeg4audio_sample_rates[16] = {
+    96000, 88200, 64000, 48000, 44100, 32000,
+    24000, 22050, 16000, 12000, 11025, 8000, 7350
+};
+
+/** bits needed to code codebook run value for long windows */
+static const uint8_t run_value_bits_long[64] = {
+     5,  5,  5,  5,  5,  5,  5,  5,  5,  5,  5,  5,  5,  5,  5,  5,
+     5,  5,  5,  5,  5,  5,  5,  5,  5,  5,  5,  5,  5,  5,  5, 10,
+    10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+    10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 15
+};
+
+/** bits needed to code codebook run value for short windows */
+static const uint8_t run_value_bits_short[16] = {
+    3, 3, 3, 3, 3, 3, 3, 6, 6, 6, 6, 6, 6, 6, 6, 9
+};
+
+/* TNS starting SFBs for long and short windows */
+static const uint8_t tns_min_sfb_short[16] = {
+    2, 2, 2, 3, 3, 4, 6, 6, 8, 10, 10, 12, 12, 12, 12, 12
+};
+
+static const uint8_t tns_min_sfb_long[16] = {
+    12, 13, 15, 16, 17, 20, 25, 26, 24, 28, 30, 31, 31, 31, 31, 31
+};
+
+static const uint8_t * const tns_min_sfb[2] = {
+    tns_min_sfb_long, tns_min_sfb_short
+};
+
+static const uint8_t * const run_value_bits[2] = {
+    run_value_bits_long, run_value_bits_short
+};
+
+/** Map to convert values from BandCodingPath index to a codebook index **/
+static const uint8_t aac_cb_out_map[CB_TOT_ALL]  = {0,1,2,3,4,5,6,7,8,9,10,11,13,14,15};
+/** Inverse map to convert from codebooks to BandCodingPath indices **/
+static const uint8_t aac_cb_in_map[CB_TOT_ALL+1] = {0,1,2,3,4,5,6,7,8,9,10,11,0,12,13,14};
+
+static const uint8_t aac_cb_range [12] = {0, 3, 3, 3, 3, 9, 9, 8, 8, 13, 13, 17};
+static const uint8_t aac_cb_maxval[12] = {0, 1, 1, 2, 2, 4, 4, 7, 7, 12, 12, 16};
+
+static const unsigned char aac_maxval_cb[] = {
+    0, 1, 3, 5, 5, 7, 7, 7, 9, 9, 9, 9, 9, 11
+};
+
+static const int aacenc_profiles[] = {
+    FF_PROFILE_AAC_MAIN,
+    FF_PROFILE_AAC_LOW,
+    FF_PROFILE_AAC_LTP,
+    FF_PROFILE_MPEG2_AAC_LOW,
+};
+
+#endif /* AVCODEC_AACENCTAB_H */

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aacps.h

@@ -0,0 +1,86 @@
+/*
+ * MPEG-4 Parametric Stereo definitions and declarations
+ * Copyright (c) 2010 Alex Converse <alex.converse@gmail.com>
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#ifndef AVCODEC_AACPS_H
+#define AVCODEC_AACPS_H
+
+#include <stdint.h>
+
+#include "aacpsdsp.h"
+#include "avcodec.h"
+#include "get_bits.h"
+
+#define PS_MAX_NUM_ENV 5
+#define PS_MAX_NR_IIDICC 34
+#define PS_MAX_NR_IPDOPD 17
+#define PS_MAX_SSB 91
+#define PS_MAX_AP_BANDS 50
+#define PS_QMF_TIME_SLOTS 32
+#define PS_MAX_DELAY 14
+#define PS_AP_LINKS 3
+#define PS_MAX_AP_DELAY 5
+
+typedef struct PSContext {
+    int    start;
+    int    enable_iid;
+    int    iid_quant;
+    int    nr_iid_par;
+    int    nr_ipdopd_par;
+    int    enable_icc;
+    int    icc_mode;
+    int    nr_icc_par;
+    int    enable_ext;
+    int    frame_class;
+    int    num_env_old;
+    int    num_env;
+    int    enable_ipdopd;
+    int    border_position[PS_MAX_NUM_ENV+1];
+    int8_t iid_par[PS_MAX_NUM_ENV][PS_MAX_NR_IIDICC]; ///< Inter-channel Intensity Difference Parameters
+    int8_t icc_par[PS_MAX_NUM_ENV][PS_MAX_NR_IIDICC]; ///< Inter-Channel Coherence Parameters
+    /* ipd/opd is iid/icc sized so that the same functions can handle both */
+    int8_t ipd_par[PS_MAX_NUM_ENV][PS_MAX_NR_IIDICC]; ///< Inter-channel Phase Difference Parameters
+    int8_t opd_par[PS_MAX_NUM_ENV][PS_MAX_NR_IIDICC]; ///< Overall Phase Difference Parameters
+    int    is34bands;
+    int    is34bands_old;
+
+    DECLARE_ALIGNED(16, INTFLOAT, in_buf)[5][44][2];
+    DECLARE_ALIGNED(16, INTFLOAT, delay)[PS_MAX_SSB][PS_QMF_TIME_SLOTS + PS_MAX_DELAY][2];
+    DECLARE_ALIGNED(16, INTFLOAT, ap_delay)[PS_MAX_AP_BANDS][PS_AP_LINKS][PS_QMF_TIME_SLOTS + PS_MAX_AP_DELAY][2];
+    DECLARE_ALIGNED(16, INTFLOAT, peak_decay_nrg)[34];
+    DECLARE_ALIGNED(16, INTFLOAT, power_smooth)[34];
+    DECLARE_ALIGNED(16, INTFLOAT, peak_decay_diff_smooth)[34];
+    DECLARE_ALIGNED(16, INTFLOAT, H11)[2][PS_MAX_NUM_ENV+1][PS_MAX_NR_IIDICC];
+    DECLARE_ALIGNED(16, INTFLOAT, H12)[2][PS_MAX_NUM_ENV+1][PS_MAX_NR_IIDICC];
+    DECLARE_ALIGNED(16, INTFLOAT, H21)[2][PS_MAX_NUM_ENV+1][PS_MAX_NR_IIDICC];
+    DECLARE_ALIGNED(16, INTFLOAT, H22)[2][PS_MAX_NUM_ENV+1][PS_MAX_NR_IIDICC];
+    DECLARE_ALIGNED(16, INTFLOAT, Lbuf)[91][32][2];
+    DECLARE_ALIGNED(16, INTFLOAT, Rbuf)[91][32][2];
+    int8_t opd_hist[PS_MAX_NR_IIDICC];
+    int8_t ipd_hist[PS_MAX_NR_IIDICC];
+    PSDSPContext dsp;
+} PSContext;
+
+void AAC_RENAME(ff_ps_init)(void);
+void AAC_RENAME(ff_ps_ctx_init)(PSContext *ps);
+int AAC_RENAME(ff_ps_read_data)(AVCodecContext *avctx, GetBitContext *gb, PSContext *ps, int bits_left);
+int AAC_RENAME(ff_ps_apply)(AVCodecContext *avctx, PSContext *ps, INTFLOAT L[2][38][64], INTFLOAT R[2][38][64], int top);
+
+#endif /* AVCODEC_AACPS_H */

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aacps_fixed.c

@@ -0,0 +1,24 @@
+/*
+ * MPEG-4 Parametric Stereo decoding functions
+ * Copyright (c) 2010 Alex Converse <alex.converse@gmail.com>
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#define USE_FIXED 1
+
+#include "aacps.c"

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aacps_fixed_tablegen.c

@@ -0,0 +1,24 @@
+/*
+ * Generate a header file for hardcoded Parametric Stereo tables
+ *
+ * Copyright (c) 2010 Alex Converse <alex.converse@gmail.com>
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#define USE_FIXED 1
+#include "aacps_tablegen_template.c"

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aacps_fixed_tablegen.h

@@ -0,0 +1,403 @@
+/*
+ * Header file for hardcoded Parametric Stereo tables
+ *
+ * Copyright (c) 2010 Alex Converse <alex.converse@gmail.com>
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ *
+ * Note: Rounding-to-nearest used unless otherwise stated
+ *
+ */
+
+#ifndef AVCODEC_AACPS_FIXED_TABLEGEN_H
+#define AVCODEC_AACPS_FIXED_TABLEGEN_H
+
+#include <math.h>
+#include <stdint.h>
+
+#if CONFIG_HARDCODED_TABLES
+#define ps_tableinit()
+#define TABLE_CONST const
+#include "libavcodec/aacps_fixed_tables.h"
+#else
+#include "libavutil/common.h"
+#include "libavutil/mathematics.h"
+#include "libavutil/mem.h"
+
+#include "aac_defines.h"
+#include "libavutil/softfloat.h"
+#define NR_ALLPASS_BANDS20 30
+#define NR_ALLPASS_BANDS34 50
+#define PS_AP_LINKS 3
+#define TABLE_CONST
+static int pd_re_smooth[8*8*8];
+static int pd_im_smooth[8*8*8];
+static int HA[46][8][4];
+static int HB[46][8][4];
+static DECLARE_ALIGNED(16, int, f20_0_8) [ 8][8][2];
+static DECLARE_ALIGNED(16, int, f34_0_12)[12][8][2];
+static DECLARE_ALIGNED(16, int, f34_1_8) [ 8][8][2];
+static DECLARE_ALIGNED(16, int, f34_2_4) [ 4][8][2];
+static TABLE_CONST DECLARE_ALIGNED(16, int, Q_fract_allpass)[2][50][3][2];
+static DECLARE_ALIGNED(16, int, phi_fract)[2][50][2];
+
+static const int g0_Q8[] = {
+    Q31(0.00746082949812f), Q31(0.02270420949825f), Q31(0.04546865930473f), Q31(0.07266113929591f),
+    Q31(0.09885108575264f), Q31(0.11793710567217f), Q31(0.125f)
+};
+
+static const int g0_Q12[] = {
+    Q31(0.04081179924692f), Q31(0.03812810994926f), Q31(0.05144908135699f), Q31(0.06399831151592f),
+    Q31(0.07428313801106f), Q31(0.08100347892914f), Q31(0.08333333333333f)
+};
+
+static const int g1_Q8[] = {
+    Q31(0.01565675600122f), Q31(0.03752716391991f), Q31(0.05417891378782f), Q31(0.08417044116767f),
+    Q31(0.10307344158036f), Q31(0.12222452249753f), Q31(0.125f)
+};
+
+static const int g2_Q4[] = {
+    Q31(-0.05908211155639f), Q31(-0.04871498374946f), Q31(0.0f),   Q31(0.07778723915851f),
+    Q31( 0.16486303567403f), Q31( 0.23279856662996f), Q31(0.25f)
+};
+
+static const int sintbl_4[4]   = {           0,  1073741824,           0, -1073741824 };
+static const int costbl_4[4]   = {  1073741824,           0, -1073741824,           0 };
+static const int sintbl_8[8]   = {           0,   759250125,  1073741824,   759250125,
+                                             0,  -759250125, -1073741824,  -759250125 };
+static const int costbl_8[8]   = {  1073741824,   759250125,           0,  -759250125,
+                                   -1073741824,  -759250125,           0,   759250125 };
+static const int sintbl_12[12] = {           0,   536870912,   929887697,  1073741824,
+                                     929887697,   536870912,           0,  -536870912,
+                                    -929887697, -1073741824,  -929887697,  -536870912 };
+static const int costbl_12[12] = {  1073741824,   929887697,   536870912,           0,
+                                    -536870912,  -929887697, -1073741824,  -929887697,
+                                    -536870912,           0,   536870912,   929887697 };
+
+static void make_filters_from_proto(int (*filter)[8][2], const int *proto, int bands)
+{
+
+    const int *sinptr, *cosptr;
+    int s, c, sinhalf, coshalf;
+    int q, n;
+
+    if (bands == 4) {
+        sinptr = sintbl_4;
+        cosptr = costbl_4;
+        sinhalf = 759250125;
+        coshalf = 759250125;
+    } else if (bands == 8) {
+        sinptr = sintbl_8;
+        cosptr = costbl_8;
+        sinhalf = 410903207;
+        coshalf = 992008094;
+    } else {
+        sinptr = sintbl_12;
+        cosptr = costbl_12;
+        sinhalf = 277904834;
+        coshalf = 1037154959;
+    }
+
+    for (q = 0; q < bands; q++) {
+        for (n = 0; n < 7; n++) {
+            int theta = (q*(n-6) + (n>>1) - 3) % bands;
+
+            if (theta < 0)
+                theta += bands;
+            s = sinptr[theta];
+            c = cosptr[theta];
+
+            if (n & 1) {
+                theta = (int)(((int64_t)c * coshalf - (int64_t)s * sinhalf + 0x20000000) >> 30);
+                s = (int)(((int64_t)s * coshalf + (int64_t)c * sinhalf + 0x20000000) >> 30);
+                c = theta;
+            }
+            filter[q][n][0] = (int)(((int64_t)proto[n] * c + 0x20000000) >> 30);
+            filter[q][n][1] = -(int)(((int64_t)proto[n] * s + 0x20000000) >> 30);
+        }
+    }
+}
+
+static void ps_tableinit(void)
+{
+    static const int ipdopd_sin[] = { Q30(0), Q30(M_SQRT1_2), Q30(1), Q30( M_SQRT1_2), Q30( 0), Q30(-M_SQRT1_2), Q30(-1), Q30(-M_SQRT1_2) };
+    static const int ipdopd_cos[] = { Q30(1), Q30(M_SQRT1_2), Q30(0), Q30(-M_SQRT1_2), Q30(-1), Q30(-M_SQRT1_2), Q30( 0), Q30( M_SQRT1_2) };
+    int pd0, pd1, pd2;
+    int idx;
+
+    static const int alpha_tab[] =
+    {
+      Q30(1.5146213770f/M_PI), Q30(1.5181334019f/M_PI), Q30(1.5234849453f/M_PI), Q30(1.5369486809f/M_PI), Q30(1.5500687361f/M_PI), Q30(1.5679757595f/M_PI),
+      Q30(1.4455626011f/M_PI), Q30(1.4531552792f/M_PI), Q30(1.4648091793f/M_PI), Q30(1.4945238829f/M_PI), Q30(1.5239057541f/M_PI), Q30(1.5644006729f/M_PI),
+      Q30(1.3738563061f/M_PI), Q30(1.3851221800f/M_PI), Q30(1.4026404619f/M_PI), Q30(1.4484288692f/M_PI), Q30(1.4949874878f/M_PI), Q30(1.5604078770f/M_PI),
+      Q30(1.2645189762f/M_PI), Q30(1.2796478271f/M_PI), Q30(1.3038636446f/M_PI), Q30(1.3710125685f/M_PI), Q30(1.4443849325f/M_PI), Q30(1.5532352924f/M_PI),
+      Q30(1.1507037878f/M_PI), Q30(1.1669205427f/M_PI), Q30(1.1938756704f/M_PI), Q30(1.2754167318f/M_PI), Q30(1.3761177063f/M_PI), Q30(1.5429240465f/M_PI),
+      Q30(1.0079245567f/M_PI), Q30(1.0208238363f/M_PI), Q30(1.0433073044f/M_PI), Q30(1.1208510399f/M_PI), Q30(1.2424604893f/M_PI), Q30(1.5185726881f/M_PI),
+      Q30(0.8995233774f/M_PI), Q30(0.9069069624f/M_PI), Q30(0.9201194048f/M_PI), Q30(0.9698365927f/M_PI), Q30(1.0671583414f/M_PI), Q30(1.4647934437f/M_PI),
+      Q30(0.7853981853f/M_PI), Q30(0.7853981853f/M_PI), Q30(0.7853981853f/M_PI), Q30(0.7853981853f/M_PI), Q30(0.7853981853f/M_PI), Q30(0.7853981853f/M_PI),
+      Q30(0.6712729335f/M_PI), Q30(0.6638893485f/M_PI), Q30(0.6506769061f/M_PI), Q30(0.6009597182f/M_PI), Q30(0.5036380291f/M_PI), Q30(0.1060028747f/M_PI),
+      Q30(0.5628717542f/M_PI), Q30(0.5499725342f/M_PI), Q30(0.5274890065f/M_PI), Q30(0.4499453008f/M_PI), Q30(0.3283358216f/M_PI), Q30(0.0522236861f/M_PI),
+      Q30(0.4200925827f/M_PI), Q30(0.4038758278f/M_PI), Q30(0.3769206405f/M_PI), Q30(0.2953795493f/M_PI), Q30(0.1946786791f/M_PI), Q30(0.0278722942f/M_PI),
+      Q30(0.3062773645f/M_PI), Q30(0.2911485136f/M_PI), Q30(0.2669326365f/M_PI), Q30(0.1997837722f/M_PI), Q30(0.1264114529f/M_PI), Q30(0.0175609849f/M_PI),
+      Q30(0.1969399750f/M_PI), Q30(0.1856741160f/M_PI), Q30(0.1681558639f/M_PI), Q30(0.1223674342f/M_PI), Q30(0.0758088827f/M_PI), Q30(0.0103884479f/M_PI),
+      Q30(0.1252337098f/M_PI), Q30(0.1176410317f/M_PI), Q30(0.1059871912f/M_PI), Q30(0.0762724727f/M_PI), Q30(0.0468905345f/M_PI), Q30(0.0063956482f/M_PI),
+      Q30(0.0561749674f/M_PI), Q30(0.0526629239f/M_PI), Q30(0.0473113805f/M_PI), Q30(0.0338476151f/M_PI), Q30(0.0207276177f/M_PI), Q30(0.0028205961f/M_PI),
+      Q30(1.5676341057f/M_PI), Q30(1.5678333044f/M_PI), Q30(1.5681363344f/M_PI), Q30(1.5688960552f/M_PI), Q30(1.5696337223f/M_PI), Q30(1.5706381798f/M_PI),
+      Q30(1.5651730299f/M_PI), Q30(1.5655272007f/M_PI), Q30(1.5660660267f/M_PI), Q30(1.5674170256f/M_PI), Q30(1.5687289238f/M_PI), Q30(1.5705151558f/M_PI),
+      Q30(1.5607966185f/M_PI), Q30(1.5614265203f/M_PI), Q30(1.5623844862f/M_PI), Q30(1.5647867918f/M_PI), Q30(1.5671195984f/M_PI), Q30(1.5702962875f/M_PI),
+      Q30(1.5530153513f/M_PI), Q30(1.5541347265f/M_PI), Q30(1.5558375120f/M_PI), Q30(1.5601085424f/M_PI), Q30(1.5642569065f/M_PI), Q30(1.5699069500f/M_PI),
+      Q30(1.5391840935f/M_PI), Q30(1.5411708355f/M_PI), Q30(1.5441943407f/M_PI), Q30(1.5517836809f/M_PI), Q30(1.5591609478f/M_PI), Q30(1.5692136288f/M_PI),
+      Q30(1.5146213770f/M_PI), Q30(1.5181334019f/M_PI), Q30(1.5234849453f/M_PI), Q30(1.5369486809f/M_PI), Q30(1.5500687361f/M_PI), Q30(1.5679757595f/M_PI),
+      Q30(1.4915299416f/M_PI), Q30(1.4964480400f/M_PI), Q30(1.5039558411f/M_PI), Q30(1.5229074955f/M_PI), Q30(1.5414420366f/M_PI), Q30(1.5667995214f/M_PI),
+      Q30(1.4590617418f/M_PI), Q30(1.4658898115f/M_PI), Q30(1.4763505459f/M_PI), Q30(1.5029321909f/M_PI), Q30(1.5291173458f/M_PI), Q30(1.5651149750f/M_PI),
+      Q30(1.4136143923f/M_PI), Q30(1.4229322672f/M_PI), Q30(1.4373078346f/M_PI), Q30(1.4743183851f/M_PI), Q30(1.5113102198f/M_PI), Q30(1.5626684427f/M_PI),
+      Q30(1.3505556583f/M_PI), Q30(1.3628427982f/M_PI), Q30(1.3820509911f/M_PI), Q30(1.4327841997f/M_PI), Q30(1.4850014448f/M_PI), Q30(1.5590143204f/M_PI),
+      Q30(1.2645189762f/M_PI), Q30(1.2796478271f/M_PI), Q30(1.3038636446f/M_PI), Q30(1.3710125685f/M_PI), Q30(1.4443849325f/M_PI), Q30(1.5532352924f/M_PI),
+      Q30(1.1919227839f/M_PI), Q30(1.2081253529f/M_PI), Q30(1.2346779108f/M_PI), Q30(1.3123005629f/M_PI), Q30(1.4034168720f/M_PI), Q30(1.5471596718f/M_PI),
+      Q30(1.1061993837f/M_PI), Q30(1.1219338179f/M_PI), Q30(1.1484941244f/M_PI), Q30(1.2320860624f/M_PI), Q30(1.3421301842f/M_PI), Q30(1.5373806953f/M_PI),
+      Q30(1.0079245567f/M_PI), Q30(1.0208238363f/M_PI), Q30(1.0433073044f/M_PI), Q30(1.1208510399f/M_PI), Q30(1.2424604893f/M_PI), Q30(1.5185726881f/M_PI),
+      Q30(0.8995233774f/M_PI), Q30(0.9069069624f/M_PI), Q30(0.9201194048f/M_PI), Q30(0.9698365927f/M_PI), Q30(1.0671583414f/M_PI), Q30(1.4647934437f/M_PI),
+      Q30(0.7853981853f/M_PI), Q30(0.7853981853f/M_PI), Q30(0.7853981853f/M_PI), Q30(0.7853981853f/M_PI), Q30(0.7853981853f/M_PI), Q30(0.7853981853f/M_PI),
+      Q30(0.6712729335f/M_PI), Q30(0.6638893485f/M_PI), Q30(0.6506769061f/M_PI), Q30(0.6009597182f/M_PI), Q30(0.5036380291f/M_PI), Q30(0.1060028747f/M_PI),
+      Q30(0.5628717542f/M_PI), Q30(0.5499725342f/M_PI), Q30(0.5274890065f/M_PI), Q30(0.4499453008f/M_PI), Q30(0.3283358216f/M_PI), Q30(0.0522236861f/M_PI),
+      Q30(0.4645969570f/M_PI), Q30(0.4488625824f/M_PI), Q30(0.4223022461f/M_PI), Q30(0.3387103081f/M_PI), Q30(0.2286661267f/M_PI), Q30(0.0334156826f/M_PI),
+      Q30(0.3788735867f/M_PI), Q30(0.3626709878f/M_PI), Q30(0.3361184299f/M_PI), Q30(0.2584958076f/M_PI), Q30(0.1673794836f/M_PI), Q30(0.0236366931f/M_PI),
+      Q30(0.3062773645f/M_PI), Q30(0.2911485136f/M_PI), Q30(0.2669326365f/M_PI), Q30(0.1997837722f/M_PI), Q30(0.1264114529f/M_PI), Q30(0.0175609849f/M_PI),
+      Q30(0.2202406377f/M_PI), Q30(0.2079535723f/M_PI), Q30(0.1887452900f/M_PI), Q30(0.1380121708f/M_PI), Q30(0.0857949182f/M_PI), Q30(0.0117820343f/M_PI),
+      Q30(0.1571819335f/M_PI), Q30(0.1478640437f/M_PI), Q30(0.1334884763f/M_PI), Q30(0.0964778885f/M_PI), Q30(0.0594860613f/M_PI), Q30(0.0081279324f/M_PI),
+      Q30(0.1117345318f/M_PI), Q30(0.1049065739f/M_PI), Q30(0.0944457650f/M_PI), Q30(0.0678641573f/M_PI), Q30(0.0416790098f/M_PI), Q30(0.0056813755f/M_PI),
+      Q30(0.0792663917f/M_PI), Q30(0.0743482932f/M_PI), Q30(0.0668405443f/M_PI), Q30(0.0478888862f/M_PI), Q30(0.0293543357f/M_PI), Q30(0.0039967746f/M_PI),
+      Q30(0.0561749674f/M_PI), Q30(0.0526629239f/M_PI), Q30(0.0473113805f/M_PI), Q30(0.0338476151f/M_PI), Q30(0.0207276177f/M_PI), Q30(0.0028205961f/M_PI),
+      Q30(0.0316122435f/M_PI), Q30(0.0296254847f/M_PI), Q30(0.0266019460f/M_PI), Q30(0.0190126132f/M_PI), Q30(0.0116353342f/M_PI), Q30(0.0015827164f/M_PI),
+      Q30(0.0177809205f/M_PI), Q30(0.0166615788f/M_PI), Q30(0.0149587989f/M_PI), Q30(0.0106877899f/M_PI), Q30(0.0065393616f/M_PI), Q30(0.0008894200f/M_PI),
+      Q30(0.0099996664f/M_PI), Q30(0.0093698399f/M_PI), Q30(0.0084118480f/M_PI), Q30(0.0060095116f/M_PI), Q30(0.0036767013f/M_PI), Q30(0.0005000498f/M_PI),
+      Q30(0.0056233541f/M_PI), Q30(0.0052691097f/M_PI), Q30(0.0047303112f/M_PI), Q30(0.0033792770f/M_PI), Q30(0.0020674451f/M_PI), Q30(0.0002811795f/M_PI),
+      Q30(0.0031622672f/M_PI), Q30(0.0029630491f/M_PI), Q30(0.0026600463f/M_PI), Q30(0.0019002859f/M_PI), Q30(0.0011625893f/M_PI), Q30(0.0001581155f/M_PI)
+    };
+
+    static const int gamma_tab[] =
+    {
+      Q30(0.0000000000f/M_PI), Q30(0.0195873566f/M_PI), Q30(0.0303316917f/M_PI), Q30(0.0448668823f/M_PI), Q30(0.0522258915f/M_PI), Q30(0.0561044961f/M_PI),
+      Q30(0.0000000000f/M_PI), Q30(0.0433459543f/M_PI), Q30(0.0672172382f/M_PI), Q30(0.0997167900f/M_PI), Q30(0.1162951663f/M_PI), Q30(0.1250736862f/M_PI),
+      Q30(0.0000000000f/M_PI), Q30(0.0672341362f/M_PI), Q30(0.1045235619f/M_PI), Q30(0.1558904350f/M_PI), Q30(0.1824723780f/M_PI), Q30(0.1966800541f/M_PI),
+      Q30(0.0000000000f/M_PI), Q30(0.1011129096f/M_PI), Q30(0.1580764502f/M_PI), Q30(0.2387557179f/M_PI), Q30(0.2820728719f/M_PI), Q30(0.3058380187f/M_PI),
+      Q30(0.0000000000f/M_PI), Q30(0.1315985769f/M_PI), Q30(0.2072522491f/M_PI), Q30(0.3188187480f/M_PI), Q30(0.3825501204f/M_PI), Q30(0.4193951190f/M_PI),
+      Q30(0.0000000000f/M_PI), Q30(0.1603866369f/M_PI), Q30(0.2549437582f/M_PI), Q30(0.4029446840f/M_PI), Q30(0.4980689585f/M_PI), Q30(0.5615641475f/M_PI),
+      Q30(0.0000000000f/M_PI), Q30(0.1736015975f/M_PI), Q30(0.2773745656f/M_PI), Q30(0.4461984038f/M_PI), Q30(0.5666890144f/M_PI), Q30(0.6686112881f/M_PI),
+      Q30(0.0000000000f/M_PI), Q30(0.1784276664f/M_PI), Q30(0.2856673002f/M_PI), Q30(0.4630723596f/M_PI), Q30(0.5971632004f/M_PI), Q30(0.7603877187f/M_PI),
+      Q30(0.0000000000f/M_PI), Q30(0.1736015975f/M_PI), Q30(0.2773745656f/M_PI), Q30(0.4461984038f/M_PI), Q30(0.5666890144f/M_PI), Q30(0.6686112881f/M_PI),
+      Q30(0.0000000000f/M_PI), Q30(0.1603866369f/M_PI), Q30(0.2549437582f/M_PI), Q30(0.4029446840f/M_PI), Q30(0.4980689585f/M_PI), Q30(0.5615641475f/M_PI),
+      Q30(0.0000000000f/M_PI), Q30(0.1315985769f/M_PI), Q30(0.2072522491f/M_PI), Q30(0.3188187480f/M_PI), Q30(0.3825501204f/M_PI), Q30(0.4193951190f/M_PI),
+      Q30(0.0000000000f/M_PI), Q30(0.1011129096f/M_PI), Q30(0.1580764502f/M_PI), Q30(0.2387557179f/M_PI), Q30(0.2820728719f/M_PI), Q30(0.3058380187f/M_PI),
+      Q30(0.0000000000f/M_PI), Q30(0.0672341362f/M_PI), Q30(0.1045235619f/M_PI), Q30(0.1558904350f/M_PI), Q30(0.1824723780f/M_PI), Q30(0.1966800541f/M_PI),
+      Q30(0.0000000000f/M_PI), Q30(0.0433459543f/M_PI), Q30(0.0672172382f/M_PI), Q30(0.0997167900f/M_PI), Q30(0.1162951663f/M_PI), Q30(0.1250736862f/M_PI),
+      Q30(0.0000000000f/M_PI), Q30(0.0195873566f/M_PI), Q30(0.0303316917f/M_PI), Q30(0.0448668823f/M_PI), Q30(0.0522258915f/M_PI), Q30(0.0561044961f/M_PI),
+      Q30(0.0000000000f/M_PI), Q30(0.0011053939f/M_PI), Q30(0.0017089852f/M_PI), Q30(0.0025254129f/M_PI), Q30(0.0029398468f/M_PI), Q30(0.0031597170f/M_PI),
+      Q30(0.0000000000f/M_PI), Q30(0.0019607407f/M_PI), Q30(0.0030395309f/M_PI), Q30(0.0044951206f/M_PI), Q30(0.0052305623f/M_PI), Q30(0.0056152637f/M_PI),
+      Q30(0.0000000000f/M_PI), Q30(0.0034913034f/M_PI), Q30(0.0054070661f/M_PI), Q30(0.0079917293f/M_PI), Q30(0.0092999367f/M_PI), Q30(0.0099875759f/M_PI),
+      Q30(0.0000000000f/M_PI), Q30(0.0062100487f/M_PI), Q30(0.0096135242f/M_PI), Q30(0.0142110568f/M_PI), Q30(0.0165348612f/M_PI), Q30(0.0177587029f/M_PI),
+      Q30(0.0000000000f/M_PI), Q30(0.0110366223f/M_PI), Q30(0.0170863140f/M_PI), Q30(0.0252620988f/M_PI), Q30(0.0293955617f/M_PI), Q30(0.0315726399f/M_PI),
+      Q30(0.0000000000f/M_PI), Q30(0.0195873566f/M_PI), Q30(0.0303316917f/M_PI), Q30(0.0448668823f/M_PI), Q30(0.0522258915f/M_PI), Q30(0.0561044961f/M_PI),
+      Q30(0.0000000000f/M_PI), Q30(0.0275881495f/M_PI), Q30(0.0427365713f/M_PI), Q30(0.0632618815f/M_PI), Q30(0.0736731067f/M_PI), Q30(0.0791663304f/M_PI),
+      Q30(0.0000000000f/M_PI), Q30(0.0387469754f/M_PI), Q30(0.0600636788f/M_PI), Q30(0.0890387669f/M_PI), Q30(0.1037906483f/M_PI), Q30(0.1115923747f/M_PI),
+      Q30(0.0000000000f/M_PI), Q30(0.0541138873f/M_PI), Q30(0.0839984417f/M_PI), Q30(0.1248718798f/M_PI), Q30(0.1458375156f/M_PI), Q30(0.1569785923f/M_PI),
+      Q30(0.0000000000f/M_PI), Q30(0.0747506917f/M_PI), Q30(0.1163287833f/M_PI), Q30(0.1738867164f/M_PI), Q30(0.2038587779f/M_PI), Q30(0.2199459076f/M_PI),
+      Q30(0.0000000000f/M_PI), Q30(0.1011129096f/M_PI), Q30(0.1580764502f/M_PI), Q30(0.2387557179f/M_PI), Q30(0.2820728719f/M_PI), Q30(0.3058380187f/M_PI),
+      Q30(0.0000000000f/M_PI), Q30(0.1212290376f/M_PI), Q30(0.1903949380f/M_PI), Q30(0.2907958031f/M_PI), Q30(0.3466993868f/M_PI), Q30(0.3782821596f/M_PI),
+      Q30(0.0000000000f/M_PI), Q30(0.1418247074f/M_PI), Q30(0.2240308374f/M_PI), Q30(0.3474813402f/M_PI), Q30(0.4202919006f/M_PI), Q30(0.4637607038f/M_PI),
+      Q30(0.0000000000f/M_PI), Q30(0.1603866369f/M_PI), Q30(0.2549437582f/M_PI), Q30(0.4029446840f/M_PI), Q30(0.4980689585f/M_PI), Q30(0.5615641475f/M_PI),
+      Q30(0.0000000000f/M_PI), Q30(0.1736015975f/M_PI), Q30(0.2773745656f/M_PI), Q30(0.4461984038f/M_PI), Q30(0.5666890144f/M_PI), Q30(0.6686112881f/M_PI),
+      Q30(0.0000000000f/M_PI), Q30(0.1784276664f/M_PI), Q30(0.2856673002f/M_PI), Q30(0.4630723596f/M_PI), Q30(0.5971632004f/M_PI), Q30(0.7603877187f/M_PI),
+      Q30(0.0000000000f/M_PI), Q30(0.1736015975f/M_PI), Q30(0.2773745656f/M_PI), Q30(0.4461984038f/M_PI), Q30(0.5666890144f/M_PI), Q30(0.6686112881f/M_PI),
+      Q30(0.0000000000f/M_PI), Q30(0.1603866369f/M_PI), Q30(0.2549437582f/M_PI), Q30(0.4029446840f/M_PI), Q30(0.4980689585f/M_PI), Q30(0.5615641475f/M_PI),
+      Q30(0.0000000000f/M_PI), Q30(0.1418247074f/M_PI), Q30(0.2240308374f/M_PI), Q30(0.3474813402f/M_PI), Q30(0.4202919006f/M_PI), Q30(0.4637607038f/M_PI),
+      Q30(0.0000000000f/M_PI), Q30(0.1212290376f/M_PI), Q30(0.1903949380f/M_PI), Q30(0.2907958031f/M_PI), Q30(0.3466993868f/M_PI), Q30(0.3782821596f/M_PI),
+      Q30(0.0000000000f/M_PI), Q30(0.1011129096f/M_PI), Q30(0.1580764502f/M_PI), Q30(0.2387557179f/M_PI), Q30(0.2820728719f/M_PI), Q30(0.3058380187f/M_PI),
+      Q30(0.0000000000f/M_PI), Q30(0.0747506917f/M_PI), Q30(0.1163287833f/M_PI), Q30(0.1738867164f/M_PI), Q30(0.2038587779f/M_PI), Q30(0.2199459076f/M_PI),
+      Q30(0.0000000000f/M_PI), Q30(0.0541138873f/M_PI), Q30(0.0839984417f/M_PI), Q30(0.1248718798f/M_PI), Q30(0.1458375156f/M_PI), Q30(0.1569785923f/M_PI),
+      Q30(0.0000000000f/M_PI), Q30(0.0387469754f/M_PI), Q30(0.0600636788f/M_PI), Q30(0.0890387669f/M_PI), Q30(0.1037906483f/M_PI), Q30(0.1115923747f/M_PI),
+      Q30(0.0000000000f/M_PI), Q30(0.0275881495f/M_PI), Q30(0.0427365713f/M_PI), Q30(0.0632618815f/M_PI), Q30(0.0736731067f/M_PI), Q30(0.0791663304f/M_PI),
+      Q30(0.0000000000f/M_PI), Q30(0.0195873566f/M_PI), Q30(0.0303316917f/M_PI), Q30(0.0448668823f/M_PI), Q30(0.0522258915f/M_PI), Q30(0.0561044961f/M_PI),
+      Q30(0.0000000000f/M_PI), Q30(0.0110366223f/M_PI), Q30(0.0170863140f/M_PI), Q30(0.0252620988f/M_PI), Q30(0.0293955617f/M_PI), Q30(0.0315726399f/M_PI),
+      Q30(0.0000000000f/M_PI), Q30(0.0062100487f/M_PI), Q30(0.0096135242f/M_PI), Q30(0.0142110568f/M_PI), Q30(0.0165348612f/M_PI), Q30(0.0177587029f/M_PI),
+      Q30(0.0000000000f/M_PI), Q30(0.0034913034f/M_PI), Q30(0.0054070661f/M_PI), Q30(0.0079917293f/M_PI), Q30(0.0092999367f/M_PI), Q30(0.0099875759f/M_PI),
+      Q30(0.0000000000f/M_PI), Q30(0.0019607407f/M_PI), Q30(0.0030395309f/M_PI), Q30(0.0044951206f/M_PI), Q30(0.0052305623f/M_PI), Q30(0.0056152637f/M_PI),
+      Q30(0.0000000000f/M_PI), Q30(0.0011053939f/M_PI), Q30(0.0017089852f/M_PI), Q30(0.0025254129f/M_PI), Q30(0.0029398468f/M_PI), Q30(0.0031597170f/M_PI)
+    };
+
+    static const int iid_par_dequant_c1[] = {
+        //iid_par_dequant_default
+        Q30(1.41198278375959f), Q30(1.40313815268360f), Q30(1.38687670404960f), Q30(1.34839972492648f),
+        Q30(1.29124937110028f), Q30(1.19603741667993f), Q30(1.10737240362323f), Q30(1),
+        Q30(0.87961716655242f), Q30(0.75464859232732f), Q30(0.57677990744575f), Q30(0.42640143271122f),
+        Q30(0.27671828230984f), Q30(0.17664462766713f), Q30(0.07940162697653f),
+        //iid_par_dequant_fine
+        Q30(1.41420649135832f), Q30(1.41419120222364f), Q30(1.41414285699784f), Q30(1.41399000859438f),
+        Q30(1.41350698548044f), Q30(1.41198278375959f), Q30(1.40977302262355f), Q30(1.40539479488545f),
+        Q30(1.39677960498402f), Q30(1.38005309967827f), Q30(1.34839972492648f), Q30(1.31392017367631f),
+        Q30(1.26431008149654f), Q30(1.19603741667993f), Q30(1.10737240362323f), Q30(1),
+        Q30(0.87961716655242f), Q30(0.75464859232732f), Q30(0.63365607219232f), Q30(0.52308104267543f),
+        Q30(0.42640143271122f), Q30(0.30895540465965f), Q30(0.22137464873077f), Q30(0.15768788954414f),
+        Q30(0.11198225164225f), Q30(0.07940162697653f), Q30(0.04469901562677f), Q30(0.02514469318284f),
+        Q30(0.01414142856998f), Q30(0.00795258154731f), Q30(0.00447211359449f),
+    };
+
+    static const int acos_icc_invq[] = {
+        Q31(0), Q31(0.178427635f/M_PI), Q31(0.28566733f/M_PI), Q31(0.46307236f/M_PI), Q31(0.59716315f/M_PI), Q31(0.78539816f/M_PI), Q31(1.10030855f/M_PI), Q31(1.57079633f/M_PI)
+    };
+    int iid, icc;
+
+    int k, m;
+    static const int8_t f_center_20[] = {
+        -3, -1, 1, 3, 5, 7, 10, 14, 18, 22,
+    };
+    static const int32_t f_center_34[] = {
+      Q31(  2/768.0),Q31(  6/768.0),Q31(10/768.0),Q31(14/768.0),Q31( 18/768.0),Q31( 22/768.0),Q31( 26/768.0),Q31(30/768.0),
+      Q31( 34/768.0),Q31(-10/768.0),Q31(-6/768.0),Q31(-2/768.0),Q31( 51/768.0),Q31( 57/768.0),Q31( 15/768.0),Q31(21/768.0),
+      Q31( 27/768.0),Q31( 33/768.0),Q31(39/768.0),Q31(45/768.0),Q31( 54/768.0),Q31( 66/768.0),Q31( 78/768.0),Q31(42/768.0),
+      Q31(102/768.0),Q31( 66/768.0),Q31(78/768.0),Q31(90/768.0),Q31(102/768.0),Q31(114/768.0),Q31(126/768.0),Q31(90/768.0)
+    };
+    static const int fractional_delay_links[] = { Q31(0.43f), Q31(0.75f), Q31(0.347f) };
+    const int fractional_delay_gain = Q31(0.39f);
+
+    for (pd0 = 0; pd0 < 8; pd0++) {
+        int pd0_re = (ipdopd_cos[pd0]+2)>>2;
+        int pd0_im = (ipdopd_sin[pd0]+2)>>2;
+        for (pd1 = 0; pd1 < 8; pd1++) {
+            int pd1_re = ipdopd_cos[pd1] >> 1;
+            int pd1_im = ipdopd_sin[pd1] >> 1;
+            for (pd2 = 0; pd2 < 8; pd2++) {
+                int shift, round;
+                int pd2_re = ipdopd_cos[pd2];
+                int pd2_im = ipdopd_sin[pd2];
+                int re_smooth = pd0_re + pd1_re + pd2_re;
+                int im_smooth = pd0_im + pd1_im + pd2_im;
+
+                SoftFloat pd_mag = av_int2sf(((ipdopd_cos[(pd0-pd1)&7]+8)>>4) + ((ipdopd_cos[(pd0-pd2)&7]+4)>>3) +
+                                               ((ipdopd_cos[(pd1-pd2)&7]+2)>>2) + 0x15000000, 28);
+                pd_mag = av_div_sf(FLOAT_1, av_sqrt_sf(pd_mag));
+                shift = 30 - pd_mag.exp;
+                round = 1 << (shift-1);
+                pd_re_smooth[pd0*64+pd1*8+pd2] = (int)(((int64_t)re_smooth * pd_mag.mant + round) >> shift);
+                pd_im_smooth[pd0*64+pd1*8+pd2] = (int)(((int64_t)im_smooth * pd_mag.mant + round) >> shift);
+            }
+        }
+    }
+
+    idx = 0;
+    for (iid = 0; iid < 46; iid++) {
+        int c1, c2;
+
+        c1 = iid_par_dequant_c1[iid];
+        if (iid < 15)
+          c2 = iid_par_dequant_c1[14-iid];
+        else
+          c2 = iid_par_dequant_c1[60-iid];
+
+        for (icc = 0; icc < 8; icc++) {
+            /*if (PS_BASELINE || ps->icc_mode < 3)*/{
+                int alpha, beta;
+                int ca, sa, cb, sb;
+
+                alpha = acos_icc_invq[icc];
+                beta = (int)(((int64_t)alpha * 1518500250 + 0x40000000) >> 31);
+                alpha >>= 1;
+                beta = (int)(((int64_t)beta * (c1 - c2) + 0x40000000) >> 31);
+                av_sincos_sf(beta + alpha, &sa, &ca);
+                av_sincos_sf(beta - alpha, &sb, &cb);
+
+                HA[iid][icc][0] = (int)(((int64_t)c2 * ca + 0x20000000) >> 30);
+                HA[iid][icc][1] = (int)(((int64_t)c1 * cb + 0x20000000) >> 30);
+                HA[iid][icc][2] = (int)(((int64_t)c2 * sa + 0x20000000) >> 30);
+                HA[iid][icc][3] = (int)(((int64_t)c1 * sb + 0x20000000) >> 30);
+            } /* else */ {
+                int alpha_int, gamma_int;
+                int alpha_c_int, alpha_s_int, gamma_c_int, gamma_s_int;
+
+                alpha_int = alpha_tab[idx];
+                gamma_int = gamma_tab[idx];
+
+                av_sincos_sf(alpha_int, &alpha_s_int, &alpha_c_int);
+                av_sincos_sf(gamma_int, &gamma_s_int, &gamma_c_int);
+
+                alpha_c_int = (int)(((int64_t)alpha_c_int * 1518500250 + 0x20000000) >> 30);
+                alpha_s_int = (int)(((int64_t)alpha_s_int * 1518500250 + 0x20000000) >> 30);
+
+                HB[iid][icc][0] = (int)(((int64_t)alpha_c_int * gamma_c_int + 0x20000000) >> 30);
+                HB[iid][icc][1] = (int)(((int64_t)alpha_s_int * gamma_c_int + 0x20000000) >> 30);
+                HB[iid][icc][2] = -(int)(((int64_t)alpha_s_int * gamma_s_int + 0x20000000) >> 30);
+                HB[iid][icc][3] = (int)(((int64_t)alpha_c_int * gamma_s_int + 0x20000000) >> 30);
+            }
+
+            if (icc < 5 || icc > 6)
+              idx++;
+        }
+    }
+
+    for (k = 0; k < NR_ALLPASS_BANDS20; k++) {
+        int theta;
+        int64_t f_center;
+        int c, s;
+
+        if (k < FF_ARRAY_ELEMS(f_center_20))
+          f_center = f_center_20[k];
+        else
+          f_center = (k << 3) - 52;
+
+        for (m = 0; m < PS_AP_LINKS; m++) {
+            theta = (int)(((int64_t)fractional_delay_links[m] * f_center + 8) >> 4);
+            av_sincos_sf(-theta, &s, &c);
+            Q_fract_allpass[0][k][m][0] = c;
+            Q_fract_allpass[0][k][m][1] = s;
+        }
+
+        theta = (int)(((int64_t)fractional_delay_gain * f_center + 8) >> 4);
+        av_sincos_sf(-theta, &s, &c);
+        phi_fract[0][k][0] = c;
+        phi_fract[0][k][1] = s;
+    }
+
+    for (k = 0; k < NR_ALLPASS_BANDS34; k++) {
+        int theta, f_center;
+        int c, s;
+
+        if (k < FF_ARRAY_ELEMS(f_center_34))
+            f_center = f_center_34[k];
+        else
+            f_center = ((int64_t)k << 26) - (53 << 25);
+
+        for (m = 0; m < PS_AP_LINKS; m++) {
+            theta = (int)(((int64_t)fractional_delay_links[m] * f_center + 0x10000000) >> 27);
+            av_sincos_sf(-theta, &s, &c);
+            Q_fract_allpass[1][k][m][0] = c;
+            Q_fract_allpass[1][k][m][1] = s;
+        }
+
+        theta = (int)(((int64_t)fractional_delay_gain * f_center + 0x10000000) >> 27);
+        av_sincos_sf(-theta, &s, &c);
+        phi_fract[1][k][0] = c;
+        phi_fract[1][k][1] = s;
+    }
+
+    make_filters_from_proto(f20_0_8,  g0_Q8,   8);
+    make_filters_from_proto(f34_0_12, g0_Q12, 12);
+    make_filters_from_proto(f34_1_8,  g1_Q8,   8);
+    make_filters_from_proto(f34_2_4,  g2_Q4,   4);
+}
+#endif /* CONFIG_HARDCODED_TABLES */
+
+#endif /* AVCODEC_AACPS_FIXED_TABLEGEN_H */

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aacps_float.c

@@ -0,0 +1,24 @@
+/*
+ * MPEG-4 Parametric Stereo decoding functions
+ * Copyright (c) 2010 Alex Converse <alex.converse@gmail.com>
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#define USE_FIXED 0
+
+#include "aacps.c"

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aacps_tablegen.c

@@ -0,0 +1,24 @@
+/*
+ * Generate a header file for hardcoded Parametric Stereo tables
+ *
+ * Copyright (c) 2010 Alex Converse <alex.converse@gmail.com>
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#define USE_FIXED 0
+#include "aacps_tablegen_template.c"

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aacps_tablegen.h

@@ -0,0 +1,217 @@
+/*
+ * Header file for hardcoded Parametric Stereo tables
+ *
+ * Copyright (c) 2010 Alex Converse <alex.converse@gmail.com>
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#ifndef AVCODEC_AACPS_TABLEGEN_H
+#define AVCODEC_AACPS_TABLEGEN_H
+
+#include <math.h>
+#include <stdint.h>
+
+#if CONFIG_HARDCODED_TABLES
+#define ps_tableinit()
+#define TABLE_CONST const
+#include "libavcodec/aacps_tables.h"
+#else
+#include "libavutil/common.h"
+#include "libavutil/libm.h"
+#include "libavutil/mathematics.h"
+#include "libavutil/mem.h"
+#define NR_ALLPASS_BANDS20 30
+#define NR_ALLPASS_BANDS34 50
+#define PS_AP_LINKS 3
+#define TABLE_CONST
+static float pd_re_smooth[8*8*8];
+static float pd_im_smooth[8*8*8];
+static float HA[46][8][4];
+static float HB[46][8][4];
+static DECLARE_ALIGNED(16, float, f20_0_8) [ 8][8][2];
+static DECLARE_ALIGNED(16, float, f34_0_12)[12][8][2];
+static DECLARE_ALIGNED(16, float, f34_1_8) [ 8][8][2];
+static DECLARE_ALIGNED(16, float, f34_2_4) [ 4][8][2];
+static TABLE_CONST DECLARE_ALIGNED(16, float, Q_fract_allpass)[2][50][3][2];
+static DECLARE_ALIGNED(16, float, phi_fract)[2][50][2];
+
+static const float g0_Q8[] = {
+    0.00746082949812f, 0.02270420949825f, 0.04546865930473f, 0.07266113929591f,
+    0.09885108575264f, 0.11793710567217f, 0.125f
+};
+
+static const float g0_Q12[] = {
+    0.04081179924692f, 0.03812810994926f, 0.05144908135699f, 0.06399831151592f,
+    0.07428313801106f, 0.08100347892914f, 0.08333333333333f
+};
+
+static const float g1_Q8[] = {
+    0.01565675600122f, 0.03752716391991f, 0.05417891378782f, 0.08417044116767f,
+    0.10307344158036f, 0.12222452249753f, 0.125f
+};
+
+static const float g2_Q4[] = {
+    -0.05908211155639f, -0.04871498374946f, 0.0f,   0.07778723915851f,
+     0.16486303567403f,  0.23279856662996f, 0.25f
+};
+
+static av_cold void make_filters_from_proto(float (*filter)[8][2], const float *proto, int bands)
+{
+    int q, n;
+    for (q = 0; q < bands; q++) {
+        for (n = 0; n < 7; n++) {
+            double theta = 2 * M_PI * (q + 0.5) * (n - 6) / bands;
+            filter[q][n][0] = proto[n] *  cos(theta);
+            filter[q][n][1] = proto[n] * -sin(theta);
+        }
+    }
+}
+
+static av_cold void ps_tableinit(void)
+{
+    static const float ipdopd_sin[] = { 0, M_SQRT1_2, 1,  M_SQRT1_2,  0, -M_SQRT1_2, -1, -M_SQRT1_2 };
+    static const float ipdopd_cos[] = { 1, M_SQRT1_2, 0, -M_SQRT1_2, -1, -M_SQRT1_2,  0,  M_SQRT1_2 };
+    int pd0, pd1, pd2;
+
+    static const float iid_par_dequant[] = {
+        //iid_par_dequant_default
+        0.05623413251903, 0.12589254117942, 0.19952623149689, 0.31622776601684,
+        0.44668359215096, 0.63095734448019, 0.79432823472428, 1,
+        1.25892541179417, 1.58489319246111, 2.23872113856834, 3.16227766016838,
+        5.01187233627272, 7.94328234724282, 17.7827941003892,
+        //iid_par_dequant_fine
+        0.00316227766017, 0.00562341325190, 0.01,             0.01778279410039,
+        0.03162277660168, 0.05623413251903, 0.07943282347243, 0.11220184543020,
+        0.15848931924611, 0.22387211385683, 0.31622776601684, 0.39810717055350,
+        0.50118723362727, 0.63095734448019, 0.79432823472428, 1,
+        1.25892541179417, 1.58489319246111, 1.99526231496888, 2.51188643150958,
+        3.16227766016838, 4.46683592150963, 6.30957344480193, 8.91250938133745,
+        12.5892541179417, 17.7827941003892, 31.6227766016838, 56.2341325190349,
+        100,              177.827941003892, 316.227766016837,
+    };
+    static const float icc_invq[] = {
+        1, 0.937,      0.84118,    0.60092,    0.36764,   0,      -0.589,    -1
+    };
+    static const float acos_icc_invq[] = {
+        0, 0.35685527, 0.57133466, 0.92614472, 1.1943263, M_PI/2, 2.2006171, M_PI
+    };
+    int iid, icc;
+
+    int k, m;
+    static const int8_t f_center_20[] = {
+        -3, -1, 1, 3, 5, 7, 10, 14, 18, 22,
+    };
+    static const int8_t f_center_34[] = {
+         2,  6, 10, 14, 18, 22, 26, 30,
+        34,-10, -6, -2, 51, 57, 15, 21,
+        27, 33, 39, 45, 54, 66, 78, 42,
+       102, 66, 78, 90,102,114,126, 90,
+    };
+    static const float fractional_delay_links[] = { 0.43f, 0.75f, 0.347f };
+    const float fractional_delay_gain = 0.39f;
+
+    for (pd0 = 0; pd0 < 8; pd0++) {
+        float pd0_re = ipdopd_cos[pd0];
+        float pd0_im = ipdopd_sin[pd0];
+        for (pd1 = 0; pd1 < 8; pd1++) {
+            float pd1_re = ipdopd_cos[pd1];
+            float pd1_im = ipdopd_sin[pd1];
+            for (pd2 = 0; pd2 < 8; pd2++) {
+                float pd2_re = ipdopd_cos[pd2];
+                float pd2_im = ipdopd_sin[pd2];
+                float re_smooth = 0.25f * pd0_re + 0.5f * pd1_re + pd2_re;
+                float im_smooth = 0.25f * pd0_im + 0.5f * pd1_im + pd2_im;
+                float pd_mag = 1 / hypot(im_smooth, re_smooth);
+                pd_re_smooth[pd0*64+pd1*8+pd2] = re_smooth * pd_mag;
+                pd_im_smooth[pd0*64+pd1*8+pd2] = im_smooth * pd_mag;
+            }
+        }
+    }
+
+    for (iid = 0; iid < 46; iid++) {
+        float c = iid_par_dequant[iid]; ///< Linear Inter-channel Intensity Difference
+        float c1 = (float)M_SQRT2 / sqrtf(1.0f + c*c);
+        float c2 = c * c1;
+        for (icc = 0; icc < 8; icc++) {
+            /*if (PS_BASELINE || ps->icc_mode < 3)*/ {
+                float alpha = 0.5f * acos_icc_invq[icc];
+                float beta  = alpha * (c1 - c2) * (float)M_SQRT1_2;
+                HA[iid][icc][0] = c2 * cosf(beta + alpha);
+                HA[iid][icc][1] = c1 * cosf(beta - alpha);
+                HA[iid][icc][2] = c2 * sinf(beta + alpha);
+                HA[iid][icc][3] = c1 * sinf(beta - alpha);
+            } /* else */ {
+                float alpha, gamma, mu, rho;
+                float alpha_c, alpha_s, gamma_c, gamma_s;
+                rho = FFMAX(icc_invq[icc], 0.05f);
+                alpha = 0.5f * atan2f(2.0f * c * rho, c*c - 1.0f);
+                mu = c + 1.0f / c;
+                mu = sqrtf(1 + (4 * rho * rho - 4)/(mu * mu));
+                gamma = atanf(sqrtf((1.0f - mu)/(1.0f + mu)));
+                if (alpha < 0) alpha += M_PI/2;
+                alpha_c = cosf(alpha);
+                alpha_s = sinf(alpha);
+                gamma_c = cosf(gamma);
+                gamma_s = sinf(gamma);
+                HB[iid][icc][0] =  M_SQRT2 * alpha_c * gamma_c;
+                HB[iid][icc][1] =  M_SQRT2 * alpha_s * gamma_c;
+                HB[iid][icc][2] = -M_SQRT2 * alpha_s * gamma_s;
+                HB[iid][icc][3] =  M_SQRT2 * alpha_c * gamma_s;
+            }
+        }
+    }
+
+    for (k = 0; k < NR_ALLPASS_BANDS20; k++) {
+        double f_center, theta;
+        if (k < FF_ARRAY_ELEMS(f_center_20))
+            f_center = f_center_20[k] * 0.125;
+        else
+            f_center = k - 6.5f;
+        for (m = 0; m < PS_AP_LINKS; m++) {
+            theta = -M_PI * fractional_delay_links[m] * f_center;
+            Q_fract_allpass[0][k][m][0] = cos(theta);
+            Q_fract_allpass[0][k][m][1] = sin(theta);
+        }
+        theta = -M_PI*fractional_delay_gain*f_center;
+        phi_fract[0][k][0] = cos(theta);
+        phi_fract[0][k][1] = sin(theta);
+    }
+    for (k = 0; k < NR_ALLPASS_BANDS34; k++) {
+        double f_center, theta;
+        if (k < FF_ARRAY_ELEMS(f_center_34))
+            f_center = f_center_34[k] / 24.0;
+        else
+            f_center = k - 26.5f;
+        for (m = 0; m < PS_AP_LINKS; m++) {
+            theta = -M_PI * fractional_delay_links[m] * f_center;
+            Q_fract_allpass[1][k][m][0] = cos(theta);
+            Q_fract_allpass[1][k][m][1] = sin(theta);
+        }
+        theta = -M_PI*fractional_delay_gain*f_center;
+        phi_fract[1][k][0] = cos(theta);
+        phi_fract[1][k][1] = sin(theta);
+    }
+
+    make_filters_from_proto(f20_0_8,  g0_Q8,   8);
+    make_filters_from_proto(f34_0_12, g0_Q12, 12);
+    make_filters_from_proto(f34_1_8,  g1_Q8,   8);
+    make_filters_from_proto(f34_2_4,  g2_Q4,   4);
+}
+#endif /* CONFIG_HARDCODED_TABLES */
+
+#endif /* AVCODEC_AACPS_TABLEGEN_H */

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aacps_tablegen_template.c

@@ -0,0 +1,107 @@
+/*
+ * Generate a header file for hardcoded Parametric Stereo tables
+ *
+ * Copyright (c) 2010 Alex Converse <alex.converse@gmail.com>
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include <stdlib.h>
+#define CONFIG_HARDCODED_TABLES 0
+#include "aac_defines.h"
+
+#if USE_FIXED
+#define TYPE_NAME "int32_t"
+typedef int32_t INT32FLOAT;
+#define ARRAY_RENAME(x) write_int32_t_ ## x
+#define ARRAY_URENAME(x) write_uint32_t_ ## x
+#include "aacps_fixed_tablegen.h"
+#else
+#define TYPE_NAME "float"
+typedef float INT32FLOAT;
+#define ARRAY_RENAME(x) write_float_ ## x
+#define ARRAY_URENAME(x) write_float_ ## x
+#include "aacps_tablegen.h"
+#endif /* USE_FIXED */
+#include "tableprint.h"
+
+void ARRAY_RENAME(3d_array) (const void *p, int b, int c, int d)
+{
+    int i;
+    const INT32FLOAT *f = p;
+    for (i = 0; i < b; i++) {
+        printf("{\n");
+        ARRAY_URENAME(2d_array)(f, c, d);
+        printf("},\n");
+        f += c * d;
+    }
+}
+
+void ARRAY_RENAME(4d_array) (const void *p, int a, int b, int c, int d)
+{
+    int i;
+    const INT32FLOAT *f = p;
+    for (i = 0; i < a; i++) {
+        printf("{\n");
+        ARRAY_RENAME(3d_array)(f, b, c, d);
+        printf("},\n");
+        f += b * c * d;
+    }
+}
+
+int main(void)
+{
+    ps_tableinit();
+
+    write_fileheader();
+
+    printf("static const %s pd_re_smooth[8*8*8] = {\n", TYPE_NAME);
+    ARRAY_RENAME(array)(pd_re_smooth, 8*8*8);
+    printf("};\n");
+    printf("static const %s pd_im_smooth[8*8*8] = {\n", TYPE_NAME);
+    ARRAY_RENAME(array)(pd_im_smooth, 8*8*8);
+    printf("};\n");
+
+    printf("static const %s HA[46][8][4] = {\n", TYPE_NAME);
+    ARRAY_RENAME(3d_array)(HA, 46, 8, 4);
+    printf("};\n");
+    printf("static const %s HB[46][8][4] = {\n", TYPE_NAME);
+    ARRAY_RENAME(3d_array)(HB, 46, 8, 4);
+    printf("};\n");
+
+    printf("static const DECLARE_ALIGNED(16, %s, f20_0_8)[8][8][2] = {\n", TYPE_NAME);
+    ARRAY_RENAME(3d_array)(f20_0_8, 8, 8, 2);
+    printf("};\n");
+    printf("static const DECLARE_ALIGNED(16, %s, f34_0_12)[12][8][2] = {\n", TYPE_NAME);
+    ARRAY_RENAME(3d_array)(f34_0_12, 12, 8, 2);
+    printf("};\n");
+    printf("static const DECLARE_ALIGNED(16, %s, f34_1_8)[8][8][2] = {\n", TYPE_NAME);
+    ARRAY_RENAME(3d_array)(f34_1_8, 8, 8, 2);
+    printf("};\n");
+    printf("static const DECLARE_ALIGNED(16, %s, f34_2_4)[4][8][2] = {\n", TYPE_NAME);
+    ARRAY_RENAME(3d_array)(f34_2_4, 4, 8, 2);
+    printf("};\n");
+
+    printf("static const DECLARE_ALIGNED(16, %s, Q_fract_allpass)[2][50][3][2] = {\n", TYPE_NAME);
+    ARRAY_RENAME(4d_array)(Q_fract_allpass, 2, 50, 3, 2);
+    printf("};\n");
+    printf("static const DECLARE_ALIGNED(16, %s, phi_fract)[2][50][2] = {\n", TYPE_NAME);
+    ARRAY_RENAME(3d_array)(phi_fract, 2, 50, 2);
+    printf("};\n");
+
+    return 0;
+}

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aacpsdata.c

@@ -0,0 +1,163 @@
+/*
+ * MPEG-4 Parametric Stereo data tables
+ * Copyright (c) 2010 Alex Converse <alex.converse@gmail.com>
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+static const uint8_t huff_iid_df1_bits[] = {
+    18, 18, 18, 18, 18, 18, 18, 18, 18, 17, 18, 17, 17, 16, 16, 15, 14, 14,
+    13, 12, 12, 11, 10, 10,  8,  7,  6,  5,  4,  3,  1,  3,  4,  5,  6,  7,
+     8,  9, 10, 11, 11, 12, 13, 14, 14, 15, 16, 16, 17, 17, 18, 17, 18, 18,
+    18, 18, 18, 18, 18, 18, 18,
+};
+
+static const uint32_t huff_iid_df1_codes[] = {
+    0x01FEB4, 0x01FEB5, 0x01FD76, 0x01FD77, 0x01FD74, 0x01FD75, 0x01FE8A,
+    0x01FE8B, 0x01FE88, 0x00FE80, 0x01FEB6, 0x00FE82, 0x00FEB8, 0x007F42,
+    0x007FAE, 0x003FAF, 0x001FD1, 0x001FE9, 0x000FE9, 0x0007EA, 0x0007FB,
+    0x0003FB, 0x0001FB, 0x0001FF, 0x00007C, 0x00003C, 0x00001C, 0x00000C,
+    0x000000, 0x000001, 0x000001, 0x000002, 0x000001, 0x00000D, 0x00001D,
+    0x00003D, 0x00007D, 0x0000FC, 0x0001FC, 0x0003FC, 0x0003F4, 0x0007EB,
+    0x000FEA, 0x001FEA, 0x001FD6, 0x003FD0, 0x007FAF, 0x007F43, 0x00FEB9,
+    0x00FE83, 0x01FEB7, 0x00FE81, 0x01FE89, 0x01FE8E, 0x01FE8F, 0x01FE8C,
+    0x01FE8D, 0x01FEB2, 0x01FEB3, 0x01FEB0, 0x01FEB1,
+};
+
+static const uint8_t huff_iid_dt1_bits[] = {
+    16, 16, 16, 16, 16, 16, 16, 16, 16, 15, 15, 15, 15, 15, 15, 14, 14, 13,
+    13, 13, 12, 12, 11, 10,  9,  9,  7,  6,  5,  3,  1,  2,  5,  6,  7,  8,
+     9, 10, 11, 11, 12, 12, 13, 13, 14, 14, 15, 15, 15, 15, 16, 16, 16, 16,
+    16, 16, 16, 16, 16, 16, 16,
+};
+
+static const uint16_t huff_iid_dt1_codes[] = {
+    0x004ED4, 0x004ED5, 0x004ECE, 0x004ECF, 0x004ECC, 0x004ED6, 0x004ED8,
+    0x004F46, 0x004F60, 0x002718, 0x002719, 0x002764, 0x002765, 0x00276D,
+    0x0027B1, 0x0013B7, 0x0013D6, 0x0009C7, 0x0009E9, 0x0009ED, 0x0004EE,
+    0x0004F7, 0x000278, 0x000139, 0x00009A, 0x00009F, 0x000020, 0x000011,
+    0x00000A, 0x000003, 0x000001, 0x000000, 0x00000B, 0x000012, 0x000021,
+    0x00004C, 0x00009B, 0x00013A, 0x000279, 0x000270, 0x0004EF, 0x0004E2,
+    0x0009EA, 0x0009D8, 0x0013D7, 0x0013D0, 0x0027B2, 0x0027A2, 0x00271A,
+    0x00271B, 0x004F66, 0x004F67, 0x004F61, 0x004F47, 0x004ED9, 0x004ED7,
+    0x004ECD, 0x004ED2, 0x004ED3, 0x004ED0, 0x004ED1,
+};
+
+static const uint8_t huff_iid_df0_bits[] = {
+    17, 17, 17, 17, 16, 15, 13, 10,  9,  7,  6,  5,  4,  3,  1,  3,  4,  5,
+     6,  6,  8, 11, 13, 14, 14, 15, 17, 18, 18,
+};
+
+static const uint32_t huff_iid_df0_codes[] = {
+    0x01FFFB, 0x01FFFC, 0x01FFFD, 0x01FFFA, 0x00FFFC, 0x007FFC, 0x001FFD,
+    0x0003FE, 0x0001FE, 0x00007E, 0x00003C, 0x00001D, 0x00000D, 0x000005,
+    0x000000, 0x000004, 0x00000C, 0x00001C, 0x00003D, 0x00003E, 0x0000FE,
+    0x0007FE, 0x001FFC, 0x003FFC, 0x003FFD, 0x007FFD, 0x01FFFE, 0x03FFFE,
+    0x03FFFF,
+};
+
+static const uint8_t huff_iid_dt0_bits[] = {
+    19, 19, 19, 20, 20, 20, 17, 15, 12, 10,  8,  6,  4,  2,  1,  3,  5,  7,
+     9, 11, 13, 14, 17, 19, 20, 20, 20, 20, 20,
+};
+
+static const uint32_t huff_iid_dt0_codes[] = {
+    0x07FFF9, 0x07FFFA, 0x07FFFB, 0x0FFFF8, 0x0FFFF9, 0x0FFFFA, 0x01FFFD,
+    0x007FFE, 0x000FFE, 0x0003FE, 0x0000FE, 0x00003E, 0x00000E, 0x000002,
+    0x000000, 0x000006, 0x00001E, 0x00007E, 0x0001FE, 0x0007FE, 0x001FFE,
+    0x003FFE, 0x01FFFC, 0x07FFF8, 0x0FFFFB, 0x0FFFFC, 0x0FFFFD, 0x0FFFFE,
+    0x0FFFFF,
+};
+
+static const uint8_t huff_icc_df_bits[] = {
+    14, 14, 12, 10, 7, 5, 3, 1, 2, 4, 6, 8, 9, 11, 13,
+};
+
+static const uint16_t huff_icc_df_codes[] = {
+    0x3FFF, 0x3FFE, 0x0FFE, 0x03FE, 0x007E, 0x001E, 0x0006, 0x0000,
+    0x0002, 0x000E, 0x003E, 0x00FE, 0x01FE, 0x07FE, 0x1FFE,
+};
+
+static const uint8_t huff_icc_dt_bits[] = {
+    14, 13, 11, 9, 7, 5, 3, 1, 2, 4, 6, 8, 10, 12, 14,
+};
+
+static const uint16_t huff_icc_dt_codes[] = {
+    0x3FFE, 0x1FFE, 0x07FE, 0x01FE, 0x007E, 0x001E, 0x0006, 0x0000,
+    0x0002, 0x000E, 0x003E, 0x00FE, 0x03FE, 0x0FFE, 0x3FFF,
+};
+
+static const uint8_t huff_ipd_df_bits[] = {
+    1, 3, 4, 4, 4, 4, 4, 4,
+};
+
+static const uint8_t huff_ipd_df_codes[] = {
+    0x01, 0x00, 0x06, 0x04, 0x02, 0x03, 0x05, 0x07,
+};
+
+static const uint8_t huff_ipd_dt_bits[] = {
+    1, 3, 4, 5, 5, 4, 4, 3,
+};
+
+static const uint8_t huff_ipd_dt_codes[] = {
+    0x01, 0x02, 0x02, 0x03, 0x02, 0x00, 0x03, 0x03,
+};
+
+static const uint8_t huff_opd_df_bits[] = {
+    1, 3, 4, 4, 5, 5, 4, 3,
+};
+
+static const uint8_t huff_opd_df_codes[] = {
+    0x01, 0x01, 0x06, 0x04, 0x0F, 0x0E, 0x05, 0x00,
+};
+
+static const uint8_t huff_opd_dt_bits[] = {
+    1, 3, 4, 5, 5, 4, 4, 3,
+};
+
+static const uint8_t huff_opd_dt_codes[] = {
+    0x01, 0x02, 0x01, 0x07, 0x06, 0x00, 0x02, 0x03,
+};
+
+static const int8_t huff_offset[] = {
+    30, 30,
+    14, 14,
+    7, 7,
+    0, 0,
+    0, 0,
+};
+
+///Table 8.48
+static const int8_t k_to_i_20[] = {
+     1,  0,  0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 14, 15,
+    15, 15, 16, 16, 16, 16, 17, 17, 17, 17, 17, 18, 18, 18, 18, 18, 18, 18, 18,
+    18, 18, 18, 18, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19,
+    19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19
+};
+///Table 8.49
+static const int8_t k_to_i_34[] = {
+     0,  1,  2,  3,  4,  5,  6,  6,  7,  2,  1,  0, 10, 10,  4,  5,  6,  7,  8,
+     9, 10, 11, 12,  9, 14, 11, 12, 13, 14, 15, 16, 13, 16, 17, 18, 19, 20, 21,
+    22, 22, 23, 23, 24, 24, 25, 25, 26, 26, 27, 27, 27, 28, 28, 28, 29, 29, 29,
+    30, 30, 30, 31, 31, 31, 31, 32, 32, 32, 32, 33, 33, 33, 33, 33, 33, 33, 33,
+    33, 33, 33, 33, 33, 33, 33, 33, 33, 33, 33, 33, 33, 33, 33
+};
+
+static const INTFLOAT g1_Q2[] = {
+    Q31(0.0f),  Q31(0.01899487526049f), Q31(0.0f), Q31(-0.07293139167538f),
+    Q31(0.0f),  Q31(0.30596630545168f), Q31(0.5f)
+};

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aacpsdsp.asm

@@ -0,0 +1,487 @@
+;******************************************************************************
+;* SIMD optimized MPEG-4 Parametric Stereo decoding functions
+;*
+;* Copyright (C) 2015 James Almer
+;*
+;* This file is part of FFmpeg.
+;*
+;* FFmpeg is free software; you can redistribute it and/or
+;* modify it under the terms of the GNU Lesser General Public
+;* License as published by the Free Software Foundation; either
+;* version 2.1 of the License, or (at your option) any later version.
+;*
+;* FFmpeg is distributed in the hope that it will be useful,
+;* but WITHOUT ANY WARRANTY; without even the implied warranty of
+;* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+;* Lesser General Public License for more details.
+;*
+;* You should have received a copy of the GNU Lesser General Public
+;* License along with FFmpeg; if not, write to the Free Software
+;* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+;******************************************************************************
+
+%include "libavutil/x86/x86util.asm"
+
+SECTION_RODATA
+
+ps_p1m1p1m1: dd 0, 0x80000000, 0, 0x80000000
+
+SECTION .text
+
+;*************************************************************************
+;void ff_ps_add_squares_<opt>(float *dst, const float (*src)[2], int n);
+;*************************************************************************
+%macro PS_ADD_SQUARES 1
+cglobal ps_add_squares, 3, 3, %1, dst, src, n
+    shl    nd, 3
+    add  srcq, nq
+    neg    nq
+
+align 16
+.loop:
+    movaps m0, [srcq+nq]
+    movaps m1, [srcq+nq+mmsize]
+    mulps  m0, m0
+    mulps  m1, m1
+    HADDPS m0, m1, m2
+    addps  m0, [dstq]
+    movaps [dstq], m0
+    add  dstq, mmsize
+    add    nq, mmsize*2
+    jl .loop
+    REP_RET
+%endmacro
+
+INIT_XMM sse
+PS_ADD_SQUARES 2
+INIT_XMM sse3
+PS_ADD_SQUARES 3
+
+;*******************************************************************
+;void ff_ps_mul_pair_single_sse(float (*dst)[2], float (*src0)[2],
+;                                   float *src1, int n);
+;*******************************************************************
+INIT_XMM sse
+cglobal ps_mul_pair_single, 4, 4, 4, dst, src1, src2, n
+    shl      nd, 3
+    add   src1q, nq
+    add    dstq, nq
+    neg      nq
+
+align 16
+.loop:
+    movu     m0, [src1q+nq]
+    movu     m1, [src1q+nq+mmsize]
+    mova     m2, [src2q]
+    mova     m3, m2
+    unpcklps m2, m2
+    unpckhps m3, m3
+    mulps    m0, m2
+    mulps    m1, m3
+    mova [dstq+nq], m0
+    mova [dstq+nq+mmsize], m1
+    add   src2q, mmsize
+    add      nq, mmsize*2
+    jl .loop
+    REP_RET
+
+;***********************************************************************
+;void ff_ps_stereo_interpolate_sse3(float (*l)[2], float (*r)[2],
+;                                   float h[2][4], float h_step[2][4],
+;                                   int len);
+;***********************************************************************
+INIT_XMM sse3
+cglobal ps_stereo_interpolate, 5, 5, 6, l, r, h, h_step, n
+    movaps   m0, [hq]
+    movaps   m1, [h_stepq]
+    unpcklps m4, m0, m0
+    unpckhps m0, m0
+    unpcklps m5, m1, m1
+    unpckhps m1, m1
+    shl      nd, 3
+    add      lq, nq
+    add      rq, nq
+    neg      nq
+
+align 16
+.loop:
+    addps    m4, m5
+    addps    m0, m1
+    movddup  m2, [lq+nq]
+    movddup  m3, [rq+nq]
+    mulps    m2, m4
+    mulps    m3, m0
+    addps    m2, m3
+    movsd  [lq+nq], m2
+    movhps [rq+nq], m2
+    add      nq, 8
+    jl .loop
+    REP_RET
+
+;***************************************************************************
+;void ps_stereo_interpolate_ipdopd_sse3(float (*l)[2], float (*r)[2],
+;                                       float h[2][4], float h_step[2][4],
+;                                       int len);
+;***************************************************************************
+INIT_XMM sse3
+cglobal ps_stereo_interpolate_ipdopd, 5, 5, 10, l, r, h, h_step, n
+    movaps   m0, [hq]
+    movaps   m1, [hq+mmsize]
+%if ARCH_X86_64
+    movaps   m8, [h_stepq]
+    movaps   m9, [h_stepq+mmsize]
+    %define  H_STEP0 m8
+    %define  H_STEP1 m9
+%else
+    %define  H_STEP0 [h_stepq]
+    %define  H_STEP1 [h_stepq+mmsize]
+%endif
+    shl      nd, 3
+    add      lq, nq
+    add      rq, nq
+    neg      nq
+
+align 16
+.loop:
+    addps    m0, H_STEP0
+    addps    m1, H_STEP1
+    movddup  m2, [lq+nq]
+    movddup  m3, [rq+nq]
+    shufps   m4, m2, m2, q2301
+    shufps   m5, m3, m3, q2301
+    unpcklps m6, m0, m0
+    unpckhps m7, m0, m0
+    mulps    m2, m6
+    mulps    m3, m7
+    unpcklps m6, m1, m1
+    unpckhps m7, m1, m1
+    mulps    m4, m6
+    mulps    m5, m7
+    addps    m2, m3
+    addsubps m2, m4
+    addsubps m2, m5
+    movsd  [lq+nq], m2
+    movhps [rq+nq], m2
+    add      nq, 8
+    jl .loop
+    REP_RET
+
+;**********************************************************
+;void ps_hybrid_analysis_ileave_sse(float out[2][38][64],
+;                                   float (*in)[32][2],
+;                                   int i, int len)
+;**********************************************************
+INIT_XMM sse
+cglobal ps_hybrid_analysis_ileave, 3, 7, 5, out, in, i, len, in0, in1, tmp
+    movsxdifnidn        iq, id
+    mov               lend, 32 << 3
+    lea                inq, [inq+iq*4]
+    mov               tmpd, id
+    shl               tmpd, 8
+    add               outq, tmpq
+    mov               tmpd, 64
+    sub               tmpd, id
+    mov                 id, tmpd
+
+    test                id, 1
+    jne .loop4
+    test                id, 2
+    jne .loop8
+
+align 16
+.loop16:
+    mov               in0q, inq
+    mov               in1q, 38*64*4
+    add               in1q, in0q
+    mov               tmpd, lend
+
+.inner_loop16:
+    movaps              m0, [in0q]
+    movaps              m1, [in1q]
+    movaps              m2, [in0q+lenq]
+    movaps              m3, [in1q+lenq]
+    TRANSPOSE4x4PS 0, 1, 2, 3, 4
+    movaps          [outq], m0
+    movaps     [outq+lenq], m1
+    movaps   [outq+lenq*2], m2
+    movaps [outq+3*32*2*4], m3
+    lea               in0q, [in0q+lenq*2]
+    lea               in1q, [in1q+lenq*2]
+    add               outq, mmsize
+    sub               tmpd, mmsize
+    jg .inner_loop16
+    add                inq, 16
+    add               outq, 3*32*2*4
+    sub                 id, 4
+    jg .loop16
+    RET
+
+align 16
+.loop8:
+    mov               in0q, inq
+    mov               in1q, 38*64*4
+    add               in1q, in0q
+    mov               tmpd, lend
+
+.inner_loop8:
+    movlps              m0, [in0q]
+    movlps              m1, [in1q]
+    movhps              m0, [in0q+lenq]
+    movhps              m1, [in1q+lenq]
+    SBUTTERFLYPS 0, 1, 2
+    SBUTTERFLYPD 0, 1, 2
+    movaps          [outq], m0
+    movaps     [outq+lenq], m1
+    lea               in0q, [in0q+lenq*2]
+    lea               in1q, [in1q+lenq*2]
+    add               outq, mmsize
+    sub               tmpd, mmsize
+    jg .inner_loop8
+    add                inq, 8
+    add               outq, lenq
+    sub                 id, 2
+    jg .loop16
+    RET
+
+align 16
+.loop4:
+    mov               in0q, inq
+    mov               in1q, 38*64*4
+    add               in1q, in0q
+    mov               tmpd, lend
+
+.inner_loop4:
+    movss               m0, [in0q]
+    movss               m1, [in1q]
+    movss               m2, [in0q+lenq]
+    movss               m3, [in1q+lenq]
+    movlhps             m0, m1
+    movlhps             m2, m3
+    shufps              m0, m2, q2020
+    movaps          [outq], m0
+    lea               in0q, [in0q+lenq*2]
+    lea               in1q, [in1q+lenq*2]
+    add               outq, mmsize
+    sub               tmpd, mmsize
+    jg .inner_loop4
+    add                inq, 4
+    sub                 id, 1
+    test                id, 2
+    jne .loop8
+    cmp                 id, 4
+    jge .loop16
+    RET
+
+;***********************************************************
+;void ps_hybrid_synthesis_deint_sse4(float out[2][38][64],
+;                                    float (*in)[32][2],
+;                                    int i, int len)
+;***********************************************************
+%macro HYBRID_SYNTHESIS_DEINT 0
+cglobal ps_hybrid_synthesis_deint, 3, 7, 5, out, in, i, len, out0, out1, tmp
+%if cpuflag(sse4)
+%define MOVH movsd
+%else
+%define MOVH movlps
+%endif
+    movsxdifnidn        iq, id
+    mov               lend, 32 << 3
+    lea               outq, [outq+iq*4]
+    mov               tmpd, id
+    shl               tmpd, 8
+    add                inq, tmpq
+    mov               tmpd, 64
+    sub               tmpd, id
+    mov                 id, tmpd
+
+    test                id, 1
+    jne .loop4
+    test                id, 2
+    jne .loop8
+
+align 16
+.loop16:
+    mov              out0q, outq
+    mov              out1q, 38*64*4
+    add              out1q, out0q
+    mov               tmpd, lend
+
+.inner_loop16:
+    movaps              m0, [inq]
+    movaps              m1, [inq+lenq]
+    movaps              m2, [inq+lenq*2]
+    movaps              m3, [inq+3*32*2*4]
+    TRANSPOSE4x4PS 0, 1, 2, 3, 4
+    movaps         [out0q], m0
+    movaps         [out1q], m1
+    movaps    [out0q+lenq], m2
+    movaps    [out1q+lenq], m3
+    lea              out0q, [out0q+lenq*2]
+    lea              out1q, [out1q+lenq*2]
+    add                inq, mmsize
+    sub               tmpd, mmsize
+    jg .inner_loop16
+    add               outq, 16
+    add                inq, 3*32*2*4
+    sub                 id, 4
+    jg .loop16
+    RET
+
+align 16
+.loop8:
+    mov              out0q, outq
+    mov              out1q, 38*64*4
+    add              out1q, out0q
+    mov               tmpd, lend
+
+.inner_loop8:
+    movaps              m0, [inq]
+    movaps              m1, [inq+lenq]
+    SBUTTERFLYPS 0, 1, 2
+    SBUTTERFLYPD 0, 1, 2
+    MOVH           [out0q], m0
+    MOVH           [out1q], m1
+    movhps    [out0q+lenq], m0
+    movhps    [out1q+lenq], m1
+    lea              out0q, [out0q+lenq*2]
+    lea              out1q, [out1q+lenq*2]
+    add                inq, mmsize
+    sub               tmpd, mmsize
+    jg .inner_loop8
+    add               outq, 8
+    add                inq, lenq
+    sub                 id, 2
+    jg .loop16
+    RET
+
+align 16
+.loop4:
+    mov              out0q, outq
+    mov              out1q, 38*64*4
+    add              out1q, out0q
+    mov               tmpd, lend
+
+.inner_loop4:
+    movaps              m0, [inq]
+    movss          [out0q], m0
+%if cpuflag(sse4)
+    extractps      [out1q], m0, 1
+    extractps [out0q+lenq], m0, 2
+    extractps [out1q+lenq], m0, 3
+%else
+    movhlps             m1, m0
+    movss     [out0q+lenq], m1
+    shufps              m0, m0, 0xb1
+    movss          [out1q], m0
+    movhlps             m1, m0
+    movss     [out1q+lenq], m1
+%endif
+    lea              out0q, [out0q+lenq*2]
+    lea              out1q, [out1q+lenq*2]
+    add                inq, mmsize
+    sub               tmpd, mmsize
+    jg .inner_loop4
+    add               outq, 4
+    sub                 id, 1
+    test                id, 2
+    jne .loop8
+    cmp                 id, 4
+    jge .loop16
+    RET
+%endmacro
+
+INIT_XMM sse
+HYBRID_SYNTHESIS_DEINT
+INIT_XMM sse4
+HYBRID_SYNTHESIS_DEINT
+
+;*******************************************************************
+;void ff_ps_hybrid_analysis_<opt>(float (*out)[2], float (*in)[2],
+;                                 const float (*filter)[8][2],
+;                                 ptrdiff_t stride, int n);
+;*******************************************************************
+%macro PS_HYBRID_ANALYSIS_LOOP 3
+    movu     %1, [inq+mmsize*%3]
+    movu     m1, [inq+mmsize*(5-%3)+8]
+%if cpuflag(sse3)
+    pshufd   %2, %1, q2301
+    pshufd   m4, m1, q0123
+    pshufd   m1, m1, q1032
+    pshufd   m2, [filterq+nq+mmsize*%3], q2301
+    addsubps %2, m4
+    addsubps %1, m1
+%else
+    mova     m2, [filterq+nq+mmsize*%3]
+    mova     %2, %1
+    mova     m4, m1
+    shufps   %2, %2, q2301
+    shufps   m4, m4, q0123
+    shufps   m1, m1, q1032
+    shufps   m2, m2, q2301
+    xorps    m4, m7
+    xorps    m1, m7
+    subps    %2, m4
+    subps    %1, m1
+%endif
+    mulps    %2, m2
+    mulps    %1, m2
+%if %3
+    addps    m3, %2
+    addps    m0, %1
+%endif
+%endmacro
+
+%macro PS_HYBRID_ANALYSIS 0
+cglobal ps_hybrid_analysis, 5, 5, 8, out, in, filter, stride, n
+%if cpuflag(sse3)
+%define MOVH movsd
+%else
+%define MOVH movlps
+%endif
+    shl strideq, 3
+    shl nd, 6
+    add filterq, nq
+    neg nq
+    mova m7, [ps_p1m1p1m1]
+
+align 16
+.loop:
+    PS_HYBRID_ANALYSIS_LOOP m0, m3, 0
+    PS_HYBRID_ANALYSIS_LOOP m5, m6, 1
+    PS_HYBRID_ANALYSIS_LOOP m5, m6, 2
+
+%if cpuflag(sse3)
+    pshufd   m3, m3, q2301
+    xorps    m0, m7
+    hsubps   m3, m0
+    pshufd   m1, m3, q0020
+    pshufd   m3, m3, q0031
+    addps    m1, m3
+    movsd    m2, [inq+6*8]
+%else
+    mova     m1, m3
+    mova     m2, m0
+    shufps   m1, m1, q2301
+    shufps   m2, m2, q2301
+    subps    m1, m3
+    addps    m2, m0
+    unpcklps m3, m1, m2
+    unpckhps m1, m2
+    addps    m1, m3
+    movu     m2, [inq+6*8] ; faster than movlps and no risk of overread
+%endif
+    movss    m3, [filterq+nq+8*6]
+    SPLATD   m3
+    mulps    m2, m3
+    addps    m1, m2
+    MOVH [outq], m1
+    add    outq, strideq
+    add      nq, 64
+    jl .loop
+    REP_RET
+%endmacro
+
+INIT_XMM sse
+PS_HYBRID_ANALYSIS
+INIT_XMM sse3
+PS_HYBRID_ANALYSIS

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aacpsdsp.h

@@ -0,0 +1,60 @@
+/*
+ * Copyright (c) 2012 Mans Rullgard
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#ifndef AVCODEC_AACPSDSP_H
+#define AVCODEC_AACPSDSP_H
+
+#include <stddef.h>
+
+#include "aac_defines.h"
+
+#define PS_QMF_TIME_SLOTS 32
+#define PS_AP_LINKS 3
+#define PS_MAX_AP_DELAY 5
+
+typedef struct PSDSPContext {
+    void (*add_squares)(INTFLOAT *dst, const INTFLOAT (*src)[2], int n);
+    void (*mul_pair_single)(INTFLOAT (*dst)[2], INTFLOAT (*src0)[2], INTFLOAT *src1,
+                            int n);
+    void (*hybrid_analysis)(INTFLOAT (*out)[2], INTFLOAT (*in)[2],
+                            const INTFLOAT (*filter)[8][2],
+                            ptrdiff_t stride, int n);
+    void (*hybrid_analysis_ileave)(INTFLOAT (*out)[32][2], INTFLOAT L[2][38][64],
+                                   int i, int len);
+    void (*hybrid_synthesis_deint)(INTFLOAT out[2][38][64], INTFLOAT (*in)[32][2],
+                                   int i, int len);
+    void (*decorrelate)(INTFLOAT (*out)[2], INTFLOAT (*delay)[2],
+                        INTFLOAT (*ap_delay)[PS_QMF_TIME_SLOTS+PS_MAX_AP_DELAY][2],
+                        const INTFLOAT phi_fract[2], const INTFLOAT (*Q_fract)[2],
+                        const INTFLOAT *transient_gain,
+                        INTFLOAT g_decay_slope,
+                        int len);
+    void (*stereo_interpolate[2])(INTFLOAT (*l)[2], INTFLOAT (*r)[2],
+                                  INTFLOAT h[2][4], INTFLOAT h_step[2][4],
+                                  int len);
+} PSDSPContext;
+
+void AAC_RENAME(ff_psdsp_init)(PSDSPContext *s);
+void ff_psdsp_init_arm(PSDSPContext *s);
+void ff_psdsp_init_aarch64(PSDSPContext *s);
+void ff_psdsp_init_mips(PSDSPContext *s);
+void ff_psdsp_init_x86(PSDSPContext *s);
+
+#endif /* AVCODEC_AACPSDSP_H */

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aacpsdsp_fixed.c

@@ -0,0 +1,23 @@
+/*
+ * Copyright (c) 2010 Alex Converse <alex.converse@gmail.com>
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#define USE_FIXED 1
+
+#include "aacpsdsp_template.c"

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aacpsdsp_float.c

@@ -0,0 +1,23 @@
+/*
+ * Copyright (c) 2010 Alex Converse <alex.converse@gmail.com>
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#define USE_FIXED 0
+
+#include "aacpsdsp_template.c"

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aacpsdsp_init.c

@@ -0,0 +1,72 @@
+/*
+ * SIMD optimized MPEG-4 Parametric Stereo decoding functions
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include <stddef.h>
+
+#include "config.h"
+
+#include "libavutil/x86/cpu.h"
+#include "libavutil/attributes.h"
+#include "libavcodec/aacpsdsp.h"
+
+void ff_ps_add_squares_sse  (float *dst, const float (*src)[2], int n);
+void ff_ps_add_squares_sse3 (float *dst, const float (*src)[2], int n);
+void ff_ps_mul_pair_single_sse (float (*dst)[2], float (*src0)[2],
+                                float *src1, int n);
+void ff_ps_hybrid_analysis_sse (float (*out)[2], float (*in)[2],
+                                const float (*filter)[8][2],
+                                ptrdiff_t stride, int n);
+void ff_ps_hybrid_analysis_sse3(float (*out)[2], float (*in)[2],
+                                const float (*filter)[8][2],
+                                ptrdiff_t stride, int n);
+void ff_ps_stereo_interpolate_sse3(float (*l)[2], float (*r)[2],
+                                   float h[2][4], float h_step[2][4],
+                                   int len);
+void ff_ps_stereo_interpolate_ipdopd_sse3(float (*l)[2], float (*r)[2],
+                                          float h[2][4], float h_step[2][4],
+                                          int len);
+void ff_ps_hybrid_synthesis_deint_sse(float out[2][38][64], float (*in)[32][2],
+                                      int i, int len);
+void ff_ps_hybrid_synthesis_deint_sse4(float out[2][38][64], float (*in)[32][2],
+                                       int i, int len);
+void ff_ps_hybrid_analysis_ileave_sse(float (*out)[32][2], float L[2][38][64],
+                                      int i, int len);
+
+av_cold void ff_psdsp_init_x86(PSDSPContext *s)
+{
+    int cpu_flags = av_get_cpu_flags();
+
+    if (EXTERNAL_SSE(cpu_flags)) {
+        s->add_squares            = ff_ps_add_squares_sse;
+        s->mul_pair_single        = ff_ps_mul_pair_single_sse;
+        s->hybrid_analysis_ileave = ff_ps_hybrid_analysis_ileave_sse;
+        s->hybrid_synthesis_deint = ff_ps_hybrid_synthesis_deint_sse;
+        s->hybrid_analysis        = ff_ps_hybrid_analysis_sse;
+    }
+    if (EXTERNAL_SSE3(cpu_flags)) {
+        s->add_squares            = ff_ps_add_squares_sse3;
+        s->stereo_interpolate[0]  = ff_ps_stereo_interpolate_sse3;
+        s->stereo_interpolate[1]  = ff_ps_stereo_interpolate_ipdopd_sse3;
+        s->hybrid_analysis        = ff_ps_hybrid_analysis_sse3;
+    }
+    if (EXTERNAL_SSE4(cpu_flags)) {
+        s->hybrid_synthesis_deint = ff_ps_hybrid_synthesis_deint_sse4;
+    }
+}

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aacpsdsp_template.c

@@ -0,0 +1,233 @@
+/*
+ * Copyright (c) 2010 Alex Converse <alex.converse@gmail.com>
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ *
+ * Note: Rounding-to-nearest used unless otherwise stated
+ *
+ */
+#include <stdint.h>
+
+#include "config.h"
+#include "libavutil/attributes.h"
+#include "aacpsdsp.h"
+
+static void ps_add_squares_c(INTFLOAT *dst, const INTFLOAT (*src)[2], int n)
+{
+    int i;
+    for (i = 0; i < n; i++)
+        dst[i] += (UINTFLOAT)AAC_MADD28(src[i][0], src[i][0], src[i][1], src[i][1]);
+}
+
+static void ps_mul_pair_single_c(INTFLOAT (*dst)[2], INTFLOAT (*src0)[2], INTFLOAT *src1,
+                                 int n)
+{
+    int i;
+    for (i = 0; i < n; i++) {
+        dst[i][0] = AAC_MUL16(src0[i][0], src1[i]);
+        dst[i][1] = AAC_MUL16(src0[i][1], src1[i]);
+    }
+}
+
+static void ps_hybrid_analysis_c(INTFLOAT (*out)[2], INTFLOAT (*in)[2],
+                                 const INTFLOAT (*filter)[8][2],
+                                 ptrdiff_t stride, int n)
+{
+    int i, j;
+
+    for (i = 0; i < n; i++) {
+        INT64FLOAT sum_re = (INT64FLOAT)filter[i][6][0] * in[6][0];
+        INT64FLOAT sum_im = (INT64FLOAT)filter[i][6][0] * in[6][1];
+
+        for (j = 0; j < 6; j++) {
+            INT64FLOAT in0_re = in[j][0];
+            INT64FLOAT in0_im = in[j][1];
+            INT64FLOAT in1_re = in[12-j][0];
+            INT64FLOAT in1_im = in[12-j][1];
+            sum_re += (INT64FLOAT)filter[i][j][0] * (in0_re + in1_re) -
+                      (INT64FLOAT)filter[i][j][1] * (in0_im - in1_im);
+            sum_im += (INT64FLOAT)filter[i][j][0] * (in0_im + in1_im) +
+                      (INT64FLOAT)filter[i][j][1] * (in0_re - in1_re);
+        }
+#if USE_FIXED
+        out[i * stride][0] = (int)((sum_re + 0x40000000) >> 31);
+        out[i * stride][1] = (int)((sum_im + 0x40000000) >> 31);
+#else
+        out[i * stride][0] = sum_re;
+        out[i * stride][1] = sum_im;
+#endif /* USE_FIXED */
+    }
+}
+
+static void ps_hybrid_analysis_ileave_c(INTFLOAT (*out)[32][2], INTFLOAT L[2][38][64],
+                                      int i, int len)
+{
+    int j;
+
+    for (; i < 64; i++) {
+        for (j = 0; j < len; j++) {
+            out[i][j][0] = L[0][j][i];
+            out[i][j][1] = L[1][j][i];
+        }
+    }
+}
+
+static void ps_hybrid_synthesis_deint_c(INTFLOAT out[2][38][64],
+                                      INTFLOAT (*in)[32][2],
+                                      int i, int len)
+{
+    int n;
+
+    for (; i < 64; i++) {
+        for (n = 0; n < len; n++) {
+            out[0][n][i] = in[i][n][0];
+            out[1][n][i] = in[i][n][1];
+        }
+    }
+}
+
+static void ps_decorrelate_c(INTFLOAT (*out)[2], INTFLOAT (*delay)[2],
+                             INTFLOAT (*ap_delay)[PS_QMF_TIME_SLOTS + PS_MAX_AP_DELAY][2],
+                             const INTFLOAT phi_fract[2], const INTFLOAT (*Q_fract)[2],
+                             const INTFLOAT *transient_gain,
+                             INTFLOAT g_decay_slope,
+                             int len)
+{
+    static const INTFLOAT a[] = { Q31(0.65143905753106f),
+                               Q31(0.56471812200776f),
+                               Q31(0.48954165955695f) };
+    INTFLOAT ag[PS_AP_LINKS];
+    int m, n;
+
+    for (m = 0; m < PS_AP_LINKS; m++)
+        ag[m] = AAC_MUL30(a[m], g_decay_slope);
+
+    for (n = 0; n < len; n++) {
+        INTFLOAT in_re = AAC_MSUB30(delay[n][0], phi_fract[0], delay[n][1], phi_fract[1]);
+        INTFLOAT in_im = AAC_MADD30(delay[n][0], phi_fract[1], delay[n][1], phi_fract[0]);
+        for (m = 0; m < PS_AP_LINKS; m++) {
+            INTFLOAT a_re                = AAC_MUL31(ag[m], in_re);
+            INTFLOAT a_im                = AAC_MUL31(ag[m], in_im);
+            INTFLOAT link_delay_re       = ap_delay[m][n+2-m][0];
+            INTFLOAT link_delay_im       = ap_delay[m][n+2-m][1];
+            INTFLOAT fractional_delay_re = Q_fract[m][0];
+            INTFLOAT fractional_delay_im = Q_fract[m][1];
+            INTFLOAT apd_re = in_re;
+            INTFLOAT apd_im = in_im;
+            in_re = AAC_MSUB30(link_delay_re, fractional_delay_re,
+                    link_delay_im, fractional_delay_im);
+            in_re -= (UINTFLOAT)a_re;
+            in_im = AAC_MADD30(link_delay_re, fractional_delay_im,
+                    link_delay_im, fractional_delay_re);
+            in_im -= (UINTFLOAT)a_im;
+            ap_delay[m][n+5][0] = apd_re + (UINTFLOAT)AAC_MUL31(ag[m], in_re);
+            ap_delay[m][n+5][1] = apd_im + (UINTFLOAT)AAC_MUL31(ag[m], in_im);
+        }
+        out[n][0] = AAC_MUL16(transient_gain[n], in_re);
+        out[n][1] = AAC_MUL16(transient_gain[n], in_im);
+    }
+}
+
+static void ps_stereo_interpolate_c(INTFLOAT (*l)[2], INTFLOAT (*r)[2],
+                                    INTFLOAT h[2][4], INTFLOAT h_step[2][4],
+                                    int len)
+{
+    INTFLOAT h0 = h[0][0];
+    INTFLOAT h1 = h[0][1];
+    INTFLOAT h2 = h[0][2];
+    INTFLOAT h3 = h[0][3];
+    UINTFLOAT hs0 = h_step[0][0];
+    UINTFLOAT hs1 = h_step[0][1];
+    UINTFLOAT hs2 = h_step[0][2];
+    UINTFLOAT hs3 = h_step[0][3];
+    int n;
+
+    for (n = 0; n < len; n++) {
+        //l is s, r is d
+        INTFLOAT l_re = l[n][0];
+        INTFLOAT l_im = l[n][1];
+        INTFLOAT r_re = r[n][0];
+        INTFLOAT r_im = r[n][1];
+        h0 += hs0;
+        h1 += hs1;
+        h2 += hs2;
+        h3 += hs3;
+        l[n][0] = AAC_MADD30(h0, l_re, h2, r_re);
+        l[n][1] = AAC_MADD30(h0, l_im, h2, r_im);
+        r[n][0] = AAC_MADD30(h1, l_re, h3, r_re);
+        r[n][1] = AAC_MADD30(h1, l_im, h3, r_im);
+    }
+}
+
+static void ps_stereo_interpolate_ipdopd_c(INTFLOAT (*l)[2], INTFLOAT (*r)[2],
+                                           INTFLOAT h[2][4], INTFLOAT h_step[2][4],
+                                           int len)
+{
+    INTFLOAT h00  = h[0][0],      h10  = h[1][0];
+    INTFLOAT h01  = h[0][1],      h11  = h[1][1];
+    INTFLOAT h02  = h[0][2],      h12  = h[1][2];
+    INTFLOAT h03  = h[0][3],      h13  = h[1][3];
+    UINTFLOAT hs00 = h_step[0][0], hs10 = h_step[1][0];
+    UINTFLOAT hs01 = h_step[0][1], hs11 = h_step[1][1];
+    UINTFLOAT hs02 = h_step[0][2], hs12 = h_step[1][2];
+    UINTFLOAT hs03 = h_step[0][3], hs13 = h_step[1][3];
+    int n;
+
+    for (n = 0; n < len; n++) {
+        //l is s, r is d
+        INTFLOAT l_re = l[n][0];
+        INTFLOAT l_im = l[n][1];
+        INTFLOAT r_re = r[n][0];
+        INTFLOAT r_im = r[n][1];
+        h00 += hs00;
+        h01 += hs01;
+        h02 += hs02;
+        h03 += hs03;
+        h10 += hs10;
+        h11 += hs11;
+        h12 += hs12;
+        h13 += hs13;
+
+        l[n][0] = AAC_MSUB30_V8(h00, l_re, h02, r_re, h10, l_im, h12, r_im);
+        l[n][1] = AAC_MADD30_V8(h00, l_im, h02, r_im, h10, l_re, h12, r_re);
+        r[n][0] = AAC_MSUB30_V8(h01, l_re, h03, r_re, h11, l_im, h13, r_im);
+        r[n][1] = AAC_MADD30_V8(h01, l_im, h03, r_im, h11, l_re, h13, r_re);
+    }
+}
+
+av_cold void AAC_RENAME(ff_psdsp_init)(PSDSPContext *s)
+{
+    s->add_squares            = ps_add_squares_c;
+    s->mul_pair_single        = ps_mul_pair_single_c;
+    s->hybrid_analysis        = ps_hybrid_analysis_c;
+    s->hybrid_analysis_ileave = ps_hybrid_analysis_ileave_c;
+    s->hybrid_synthesis_deint = ps_hybrid_synthesis_deint_c;
+    s->decorrelate            = ps_decorrelate_c;
+    s->stereo_interpolate[0]  = ps_stereo_interpolate_c;
+    s->stereo_interpolate[1]  = ps_stereo_interpolate_ipdopd_c;
+
+#if !USE_FIXED
+    if (ARCH_ARM)
+        ff_psdsp_init_arm(s);
+    if (ARCH_AARCH64)
+        ff_psdsp_init_aarch64(s);
+    if (ARCH_MIPS)
+        ff_psdsp_init_mips(s);
+    if (ARCH_X86)
+        ff_psdsp_init_x86(s);
+#endif /* !USE_FIXED */
+}

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aacsbr.c

@@ -0,0 +1,370 @@
+/*
+ * AAC Spectral Band Replication decoding functions
+ * Copyright (c) 2008-2009 Robert Swain ( rob opendot cl )
+ * Copyright (c) 2009-2010 Alex Converse <alex.converse@gmail.com>
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+/**
+ * @file
+ * AAC Spectral Band Replication decoding functions
+ * @author Robert Swain ( rob opendot cl )
+ */
+#define USE_FIXED 0
+
+#include "aac.h"
+#include "sbr.h"
+#include "aacsbr.h"
+#include "aacsbrdata.h"
+#include "aacsbr_tablegen.h"
+#include "fft.h"
+#include "internal.h"
+#include "aacps.h"
+#include "sbrdsp.h"
+#include "libavutil/internal.h"
+#include "libavutil/libm.h"
+#include "libavutil/avassert.h"
+
+#include <stdint.h>
+#include <float.h>
+#include <math.h>
+
+#if ARCH_MIPS
+#include "mips/aacsbr_mips.h"
+#endif /* ARCH_MIPS */
+
+static VLC vlc_sbr[10];
+static void aacsbr_func_ptr_init(AACSBRContext *c);
+
+static void make_bands(int16_t* bands, int start, int stop, int num_bands)
+{
+    int k, previous, present;
+    float base, prod;
+
+    base = powf((float)stop / start, 1.0f / num_bands);
+    prod = start;
+    previous = start;
+
+    for (k = 0; k < num_bands-1; k++) {
+        prod *= base;
+        present  = lrintf(prod);
+        bands[k] = present - previous;
+        previous = present;
+    }
+    bands[num_bands-1] = stop - previous;
+}
+
+/// Dequantization and stereo decoding (14496-3 sp04 p203)
+static void sbr_dequant(SpectralBandReplication *sbr, int id_aac)
+{
+    int k, e;
+    int ch;
+    static const double exp2_tab[2] = {1, M_SQRT2};
+    if (id_aac == TYPE_CPE && sbr->bs_coupling) {
+        int pan_offset = sbr->data[0].bs_amp_res ? 12 : 24;
+        for (e = 1; e <= sbr->data[0].bs_num_env; e++) {
+            for (k = 0; k < sbr->n[sbr->data[0].bs_freq_res[e]]; k++) {
+                float temp1, temp2, fac;
+                if (sbr->data[0].bs_amp_res) {
+                    temp1 = ff_exp2fi(sbr->data[0].env_facs_q[e][k] + 7);
+                    temp2 = ff_exp2fi(pan_offset - sbr->data[1].env_facs_q[e][k]);
+                }
+                else {
+                    temp1 = ff_exp2fi((sbr->data[0].env_facs_q[e][k]>>1) + 7) *
+                            exp2_tab[sbr->data[0].env_facs_q[e][k] & 1];
+                    temp2 = ff_exp2fi((pan_offset - sbr->data[1].env_facs_q[e][k])>>1) *
+                            exp2_tab[(pan_offset - sbr->data[1].env_facs_q[e][k]) & 1];
+                }
+                if (temp1 > 1E20) {
+                    av_log(NULL, AV_LOG_ERROR, "envelope scalefactor overflow in dequant\n");
+                    temp1 = 1;
+                }
+                fac   = temp1 / (1.0f + temp2);
+                sbr->data[0].env_facs[e][k] = fac;
+                sbr->data[1].env_facs[e][k] = fac * temp2;
+            }
+        }
+        for (e = 1; e <= sbr->data[0].bs_num_noise; e++) {
+            for (k = 0; k < sbr->n_q; k++) {
+                float temp1 = ff_exp2fi(NOISE_FLOOR_OFFSET - sbr->data[0].noise_facs_q[e][k] + 1);
+                float temp2 = ff_exp2fi(12 - sbr->data[1].noise_facs_q[e][k]);
+                float fac;
+                av_assert0(temp1 <= 1E20);
+                fac = temp1 / (1.0f + temp2);
+                sbr->data[0].noise_facs[e][k] = fac;
+                sbr->data[1].noise_facs[e][k] = fac * temp2;
+            }
+        }
+    } else { // SCE or one non-coupled CPE
+        for (ch = 0; ch < (id_aac == TYPE_CPE) + 1; ch++) {
+            for (e = 1; e <= sbr->data[ch].bs_num_env; e++)
+                for (k = 0; k < sbr->n[sbr->data[ch].bs_freq_res[e]]; k++){
+                    if (sbr->data[ch].bs_amp_res)
+                        sbr->data[ch].env_facs[e][k] = ff_exp2fi(sbr->data[ch].env_facs_q[e][k] + 6);
+                    else
+                        sbr->data[ch].env_facs[e][k] = ff_exp2fi((sbr->data[ch].env_facs_q[e][k]>>1) + 6)
+                                                       * exp2_tab[sbr->data[ch].env_facs_q[e][k] & 1];
+                    if (sbr->data[ch].env_facs[e][k] > 1E20) {
+                        av_log(NULL, AV_LOG_ERROR, "envelope scalefactor overflow in dequant\n");
+                        sbr->data[ch].env_facs[e][k] = 1;
+                    }
+                }
+
+            for (e = 1; e <= sbr->data[ch].bs_num_noise; e++)
+                for (k = 0; k < sbr->n_q; k++)
+                    sbr->data[ch].noise_facs[e][k] =
+                        ff_exp2fi(NOISE_FLOOR_OFFSET - sbr->data[ch].noise_facs_q[e][k]);
+        }
+    }
+}
+
+/** High Frequency Generation (14496-3 sp04 p214+) and Inverse Filtering
+ * (14496-3 sp04 p214)
+ * Warning: This routine does not seem numerically stable.
+ */
+static void sbr_hf_inverse_filter(SBRDSPContext *dsp,
+                                  float (*alpha0)[2], float (*alpha1)[2],
+                                  const float X_low[32][40][2], int k0)
+{
+    int k;
+    for (k = 0; k < k0; k++) {
+        LOCAL_ALIGNED_16(float, phi, [3], [2][2]);
+        float dk;
+
+        dsp->autocorrelate(X_low[k], phi);
+
+        dk =  phi[2][1][0] * phi[1][0][0] -
+             (phi[1][1][0] * phi[1][1][0] + phi[1][1][1] * phi[1][1][1]) / 1.000001f;
+
+        if (!dk) {
+            alpha1[k][0] = 0;
+            alpha1[k][1] = 0;
+        } else {
+            float temp_real, temp_im;
+            temp_real = phi[0][0][0] * phi[1][1][0] -
+                        phi[0][0][1] * phi[1][1][1] -
+                        phi[0][1][0] * phi[1][0][0];
+            temp_im   = phi[0][0][0] * phi[1][1][1] +
+                        phi[0][0][1] * phi[1][1][0] -
+                        phi[0][1][1] * phi[1][0][0];
+
+            alpha1[k][0] = temp_real / dk;
+            alpha1[k][1] = temp_im   / dk;
+        }
+
+        if (!phi[1][0][0]) {
+            alpha0[k][0] = 0;
+            alpha0[k][1] = 0;
+        } else {
+            float temp_real, temp_im;
+            temp_real = phi[0][0][0] + alpha1[k][0] * phi[1][1][0] +
+                                       alpha1[k][1] * phi[1][1][1];
+            temp_im   = phi[0][0][1] + alpha1[k][1] * phi[1][1][0] -
+                                       alpha1[k][0] * phi[1][1][1];
+
+            alpha0[k][0] = -temp_real / phi[1][0][0];
+            alpha0[k][1] = -temp_im   / phi[1][0][0];
+        }
+
+        if (alpha1[k][0] * alpha1[k][0] + alpha1[k][1] * alpha1[k][1] >= 16.0f ||
+           alpha0[k][0] * alpha0[k][0] + alpha0[k][1] * alpha0[k][1] >= 16.0f) {
+            alpha1[k][0] = 0;
+            alpha1[k][1] = 0;
+            alpha0[k][0] = 0;
+            alpha0[k][1] = 0;
+        }
+    }
+}
+
+/// Chirp Factors (14496-3 sp04 p214)
+static void sbr_chirp(SpectralBandReplication *sbr, SBRData *ch_data)
+{
+    int i;
+    float new_bw;
+    static const float bw_tab[] = { 0.0f, 0.75f, 0.9f, 0.98f };
+
+    for (i = 0; i < sbr->n_q; i++) {
+        if (ch_data->bs_invf_mode[0][i] + ch_data->bs_invf_mode[1][i] == 1) {
+            new_bw = 0.6f;
+        } else
+            new_bw = bw_tab[ch_data->bs_invf_mode[0][i]];
+
+        if (new_bw < ch_data->bw_array[i]) {
+            new_bw = 0.75f    * new_bw + 0.25f    * ch_data->bw_array[i];
+        } else
+            new_bw = 0.90625f * new_bw + 0.09375f * ch_data->bw_array[i];
+        ch_data->bw_array[i] = new_bw < 0.015625f ? 0.0f : new_bw;
+    }
+}
+
+/**
+ * Calculation of levels of additional HF signal components (14496-3 sp04 p219)
+ * and Calculation of gain (14496-3 sp04 p219)
+ */
+static void sbr_gain_calc(AACContext *ac, SpectralBandReplication *sbr,
+                          SBRData *ch_data, const int e_a[2])
+{
+    int e, k, m;
+    // max gain limits : -3dB, 0dB, 3dB, inf dB (limiter off)
+    static const float limgain[4] = { 0.70795, 1.0, 1.41254, 10000000000 };
+
+    for (e = 0; e < ch_data->bs_num_env; e++) {
+        int delta = !((e == e_a[1]) || (e == e_a[0]));
+        for (k = 0; k < sbr->n_lim; k++) {
+            float gain_boost, gain_max;
+            float sum[2] = { 0.0f, 0.0f };
+            for (m = sbr->f_tablelim[k] - sbr->kx[1]; m < sbr->f_tablelim[k + 1] - sbr->kx[1]; m++) {
+                const float temp = sbr->e_origmapped[e][m] / (1.0f + sbr->q_mapped[e][m]);
+                sbr->q_m[e][m] = sqrtf(temp * sbr->q_mapped[e][m]);
+                sbr->s_m[e][m] = sqrtf(temp * ch_data->s_indexmapped[e + 1][m]);
+                if (!sbr->s_mapped[e][m]) {
+                    sbr->gain[e][m] = sqrtf(sbr->e_origmapped[e][m] /
+                                            ((1.0f + sbr->e_curr[e][m]) *
+                                             (1.0f + sbr->q_mapped[e][m] * delta)));
+                } else {
+                    sbr->gain[e][m] = sqrtf(sbr->e_origmapped[e][m] * sbr->q_mapped[e][m] /
+                                            ((1.0f + sbr->e_curr[e][m]) *
+                                             (1.0f + sbr->q_mapped[e][m])));
+                }
+                sbr->gain[e][m] += FLT_MIN;
+            }
+            for (m = sbr->f_tablelim[k] - sbr->kx[1]; m < sbr->f_tablelim[k + 1] - sbr->kx[1]; m++) {
+                sum[0] += sbr->e_origmapped[e][m];
+                sum[1] += sbr->e_curr[e][m];
+            }
+            gain_max = limgain[sbr->bs_limiter_gains] * sqrtf((FLT_EPSILON + sum[0]) / (FLT_EPSILON + sum[1]));
+            gain_max = FFMIN(100000.f, gain_max);
+            for (m = sbr->f_tablelim[k] - sbr->kx[1]; m < sbr->f_tablelim[k + 1] - sbr->kx[1]; m++) {
+                float q_m_max   = sbr->q_m[e][m] * gain_max / sbr->gain[e][m];
+                sbr->q_m[e][m]  = FFMIN(sbr->q_m[e][m], q_m_max);
+                sbr->gain[e][m] = FFMIN(sbr->gain[e][m], gain_max);
+            }
+            sum[0] = sum[1] = 0.0f;
+            for (m = sbr->f_tablelim[k] - sbr->kx[1]; m < sbr->f_tablelim[k + 1] - sbr->kx[1]; m++) {
+                sum[0] += sbr->e_origmapped[e][m];
+                sum[1] += sbr->e_curr[e][m] * sbr->gain[e][m] * sbr->gain[e][m]
+                          + sbr->s_m[e][m] * sbr->s_m[e][m]
+                          + (delta && !sbr->s_m[e][m]) * sbr->q_m[e][m] * sbr->q_m[e][m];
+            }
+            gain_boost = sqrtf((FLT_EPSILON + sum[0]) / (FLT_EPSILON + sum[1]));
+            gain_boost = FFMIN(1.584893192f, gain_boost);
+            for (m = sbr->f_tablelim[k] - sbr->kx[1]; m < sbr->f_tablelim[k + 1] - sbr->kx[1]; m++) {
+                sbr->gain[e][m] *= gain_boost;
+                sbr->q_m[e][m]  *= gain_boost;
+                sbr->s_m[e][m]  *= gain_boost;
+            }
+        }
+    }
+}
+
+/// Assembling HF Signals (14496-3 sp04 p220)
+static void sbr_hf_assemble(float Y1[38][64][2],
+                            const float X_high[64][40][2],
+                            SpectralBandReplication *sbr, SBRData *ch_data,
+                            const int e_a[2])
+{
+    int e, i, j, m;
+    const int h_SL = 4 * !sbr->bs_smoothing_mode;
+    const int kx = sbr->kx[1];
+    const int m_max = sbr->m[1];
+    static const float h_smooth[5] = {
+        0.33333333333333,
+        0.30150283239582,
+        0.21816949906249,
+        0.11516383427084,
+        0.03183050093751,
+    };
+    float (*g_temp)[48] = ch_data->g_temp, (*q_temp)[48] = ch_data->q_temp;
+    int indexnoise = ch_data->f_indexnoise;
+    int indexsine  = ch_data->f_indexsine;
+
+    if (sbr->reset) {
+        for (i = 0; i < h_SL; i++) {
+            memcpy(g_temp[i + 2*ch_data->t_env[0]], sbr->gain[0], m_max * sizeof(sbr->gain[0][0]));
+            memcpy(q_temp[i + 2*ch_data->t_env[0]], sbr->q_m[0],  m_max * sizeof(sbr->q_m[0][0]));
+        }
+    } else if (h_SL) {
+        for (i = 0; i < 4; i++) {
+            memcpy(g_temp[i + 2 * ch_data->t_env[0]],
+                   g_temp[i + 2 * ch_data->t_env_num_env_old],
+                   sizeof(g_temp[0]));
+            memcpy(q_temp[i + 2 * ch_data->t_env[0]],
+                   q_temp[i + 2 * ch_data->t_env_num_env_old],
+                   sizeof(q_temp[0]));
+        }
+    }
+
+    for (e = 0; e < ch_data->bs_num_env; e++) {
+        for (i = 2 * ch_data->t_env[e]; i < 2 * ch_data->t_env[e + 1]; i++) {
+            memcpy(g_temp[h_SL + i], sbr->gain[e], m_max * sizeof(sbr->gain[0][0]));
+            memcpy(q_temp[h_SL + i], sbr->q_m[e],  m_max * sizeof(sbr->q_m[0][0]));
+        }
+    }
+
+    for (e = 0; e < ch_data->bs_num_env; e++) {
+        for (i = 2 * ch_data->t_env[e]; i < 2 * ch_data->t_env[e + 1]; i++) {
+            LOCAL_ALIGNED_16(float, g_filt_tab, [48]);
+            LOCAL_ALIGNED_16(float, q_filt_tab, [48]);
+            float *g_filt, *q_filt;
+
+            if (h_SL && e != e_a[0] && e != e_a[1]) {
+                g_filt = g_filt_tab;
+                q_filt = q_filt_tab;
+                for (m = 0; m < m_max; m++) {
+                    const int idx1 = i + h_SL;
+                    g_filt[m] = 0.0f;
+                    q_filt[m] = 0.0f;
+                    for (j = 0; j <= h_SL; j++) {
+                        g_filt[m] += g_temp[idx1 - j][m] * h_smooth[j];
+                        q_filt[m] += q_temp[idx1 - j][m] * h_smooth[j];
+                    }
+                }
+            } else {
+                g_filt = g_temp[i + h_SL];
+                q_filt = q_temp[i];
+            }
+
+            sbr->dsp.hf_g_filt(Y1[i] + kx, X_high + kx, g_filt, m_max,
+                               i + ENVELOPE_ADJUSTMENT_OFFSET);
+
+            if (e != e_a[0] && e != e_a[1]) {
+                sbr->dsp.hf_apply_noise[indexsine](Y1[i] + kx, sbr->s_m[e],
+                                                   q_filt, indexnoise,
+                                                   kx, m_max);
+            } else {
+                int idx = indexsine&1;
+                int A = (1-((indexsine+(kx & 1))&2));
+                int B = (A^(-idx)) + idx;
+                float *out = &Y1[i][kx][idx];
+                float *in  = sbr->s_m[e];
+                for (m = 0; m+1 < m_max; m+=2) {
+                    out[2*m  ] += in[m  ] * A;
+                    out[2*m+2] += in[m+1] * B;
+                }
+                if(m_max&1)
+                    out[2*m  ] += in[m  ] * A;
+            }
+            indexnoise = (indexnoise + m_max) & 0x1ff;
+            indexsine = (indexsine + 1) & 3;
+        }
+    }
+    ch_data->f_indexnoise = indexnoise;
+    ch_data->f_indexsine  = indexsine;
+}
+
+#include "aacsbr_template.c"

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aacsbr.h

@@ -0,0 +1,96 @@
+/*
+ * AAC Spectral Band Replication function declarations
+ * Copyright (c) 2008-2009 Robert Swain ( rob opendot cl )
+ * Copyright (c) 2010      Alex Converse <alex.converse@gmail.com>
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+/**
+ * @file
+ * AAC Spectral Band Replication function declarations
+ * @author Robert Swain ( rob opendot cl )
+ */
+
+#ifndef AVCODEC_AACSBR_H
+#define AVCODEC_AACSBR_H
+
+#include "get_bits.h"
+#include "aac.h"
+#include "sbr.h"
+
+#define ENVELOPE_ADJUSTMENT_OFFSET 2
+#define NOISE_FLOOR_OFFSET 6
+
+/**
+ * SBR VLC tables
+ */
+enum {
+    T_HUFFMAN_ENV_1_5DB,
+    F_HUFFMAN_ENV_1_5DB,
+    T_HUFFMAN_ENV_BAL_1_5DB,
+    F_HUFFMAN_ENV_BAL_1_5DB,
+    T_HUFFMAN_ENV_3_0DB,
+    F_HUFFMAN_ENV_3_0DB,
+    T_HUFFMAN_ENV_BAL_3_0DB,
+    F_HUFFMAN_ENV_BAL_3_0DB,
+    T_HUFFMAN_NOISE_3_0DB,
+    T_HUFFMAN_NOISE_BAL_3_0DB,
+};
+
+/**
+ * bs_frame_class - frame class of current SBR frame (14496-3 sp04 p98)
+ */
+enum {
+    FIXFIX,
+    FIXVAR,
+    VARFIX,
+    VARVAR,
+};
+
+enum {
+    EXTENSION_ID_PS = 2,
+};
+
+static const int8_t vlc_sbr_lav[10] =
+    { 60, 60, 24, 24, 31, 31, 12, 12, 31, 12 };
+
+#define SBR_INIT_VLC_STATIC(num, size) \
+    INIT_VLC_STATIC(&vlc_sbr[num], 9, sbr_tmp[num].table_size / sbr_tmp[num].elem_size,     \
+                    sbr_tmp[num].sbr_bits ,                      1,                      1, \
+                    sbr_tmp[num].sbr_codes, sbr_tmp[num].elem_size, sbr_tmp[num].elem_size, \
+                    size)
+
+#define SBR_VLC_ROW(name) \
+    { name ## _codes, name ## _bits, sizeof(name ## _codes), sizeof(name ## _codes[0]) }
+
+/** Initialize SBR. */
+void AAC_RENAME(ff_aac_sbr_init)(void);
+/** Initialize one SBR context. */
+void AAC_RENAME(ff_aac_sbr_ctx_init)(AACContext *ac, SpectralBandReplication *sbr, int id_aac);
+/** Close one SBR context. */
+void AAC_RENAME(ff_aac_sbr_ctx_close)(SpectralBandReplication *sbr);
+/** Decode one SBR element. */
+int AAC_RENAME(ff_decode_sbr_extension)(AACContext *ac, SpectralBandReplication *sbr,
+                            GetBitContext *gb, int crc, int cnt, int id_aac);
+/** Apply one SBR element to one AAC element. */
+void AAC_RENAME(ff_sbr_apply)(AACContext *ac, SpectralBandReplication *sbr, int id_aac,
+                  INTFLOAT* L, INTFLOAT *R);
+
+void ff_aacsbr_func_ptr_init_mips(AACSBRContext *c);
+
+#endif /* AVCODEC_AACSBR_H */

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aacsbr_fixed.c

@@ -0,0 +1,613 @@
+/*
+ * Copyright (c) 2013
+ *      MIPS Technologies, Inc., California.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ * 3. Neither the name of the MIPS Technologies, Inc., nor the names of its
+ *    contributors may be used to endorse or promote products derived from
+ *    this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE MIPS TECHNOLOGIES, INC. ``AS IS'' AND
+ * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE MIPS TECHNOLOGIES, INC. BE LIABLE
+ * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
+ * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
+ * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
+ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
+ * SUCH DAMAGE.
+ *
+ * AAC Spectral Band Replication decoding functions (fixed-point)
+ * Copyright (c) 2008-2009 Robert Swain ( rob opendot cl )
+ * Copyright (c) 2009-2010 Alex Converse <alex.converse@gmail.com>
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+/**
+ * @file
+ * AAC Spectral Band Replication decoding functions (fixed-point)
+ * Note: Rounding-to-nearest used unless otherwise stated
+ * @author Robert Swain ( rob opendot cl )
+ * @author Stanislav Ocovaj ( stanislav.ocovaj imgtec com )
+ */
+#define USE_FIXED 1
+
+#include "aac.h"
+#include "sbr.h"
+#include "aacsbr.h"
+#include "aacsbrdata.h"
+#include "aacsbr_fixed_tablegen.h"
+#include "fft.h"
+#include "aacps.h"
+#include "sbrdsp.h"
+#include "libavutil/internal.h"
+#include "libavutil/libm.h"
+#include "libavutil/avassert.h"
+
+#include <stdint.h>
+#include <float.h>
+#include <math.h>
+
+static VLC vlc_sbr[10];
+static void aacsbr_func_ptr_init(AACSBRContext *c);
+static const int CONST_LN2       = Q31(0.6931471806/256);  // ln(2)/256
+static const int CONST_RECIP_LN2 = Q31(0.7213475204);      // 0.5/ln(2)
+static const int CONST_076923    = Q31(0.76923076923076923077f);
+
+static const int fixed_log_table[10] =
+{
+    Q31(1.0/2), Q31(1.0/3), Q31(1.0/4), Q31(1.0/5), Q31(1.0/6),
+    Q31(1.0/7), Q31(1.0/8), Q31(1.0/9), Q31(1.0/10), Q31(1.0/11)
+};
+
+static int fixed_log(int x)
+{
+    int i, ret, xpow, tmp;
+
+    ret = x;
+    xpow = x;
+    for (i=0; i<10; i+=2){
+        xpow = (int)(((int64_t)xpow * x + 0x40000000) >> 31);
+        tmp = (int)(((int64_t)xpow * fixed_log_table[i] + 0x40000000) >> 31);
+        ret -= tmp;
+
+        xpow = (int)(((int64_t)xpow * x + 0x40000000) >> 31);
+        tmp = (int)(((int64_t)xpow * fixed_log_table[i+1] + 0x40000000) >> 31);
+        ret += tmp;
+    }
+
+    return ret;
+}
+
+static const int fixed_exp_table[7] =
+{
+    Q31(1.0/2), Q31(1.0/6), Q31(1.0/24), Q31(1.0/120),
+    Q31(1.0/720), Q31(1.0/5040), Q31(1.0/40320)
+};
+
+static int fixed_exp(int x)
+{
+    int i, ret, xpow, tmp;
+
+    ret = 0x800000 + x;
+    xpow = x;
+    for (i=0; i<7; i++){
+        xpow = (int)(((int64_t)xpow * x + 0x400000) >> 23);
+        tmp = (int)(((int64_t)xpow * fixed_exp_table[i] + 0x40000000) >> 31);
+        ret += tmp;
+    }
+
+    return ret;
+}
+
+static void make_bands(int16_t* bands, int start, int stop, int num_bands)
+{
+    int k, previous, present;
+    int base, prod, nz = 0;
+
+    base = (stop << 23) / start;
+    while (base < 0x40000000){
+        base <<= 1;
+        nz++;
+    }
+    base = fixed_log(base - 0x80000000);
+    base = (((base + 0x80) >> 8) + (8-nz)*CONST_LN2) / num_bands;
+    base = fixed_exp(base);
+
+    previous = start;
+    prod = start << 23;
+
+    for (k = 0; k < num_bands-1; k++) {
+        prod = (int)(((int64_t)prod * base + 0x400000) >> 23);
+        present = (prod + 0x400000) >> 23;
+        bands[k] = present - previous;
+        previous = present;
+    }
+    bands[num_bands-1] = stop - previous;
+}
+
+/// Dequantization and stereo decoding (14496-3 sp04 p203)
+static void sbr_dequant(SpectralBandReplication *sbr, int id_aac)
+{
+    int k, e;
+    int ch;
+
+    if (id_aac == TYPE_CPE && sbr->bs_coupling) {
+        int alpha      = sbr->data[0].bs_amp_res ?  2 :  1;
+        int pan_offset = sbr->data[0].bs_amp_res ? 12 : 24;
+        for (e = 1; e <= sbr->data[0].bs_num_env; e++) {
+            for (k = 0; k < sbr->n[sbr->data[0].bs_freq_res[e]]; k++) {
+                SoftFloat temp1, temp2, fac;
+
+                temp1.exp = sbr->data[0].env_facs_q[e][k] * alpha + 14;
+                if (temp1.exp & 1)
+                  temp1.mant = 759250125;
+                else
+                  temp1.mant = 0x20000000;
+                temp1.exp = (temp1.exp >> 1) + 1;
+                if (temp1.exp > 66) { // temp1 > 1E20
+                    av_log(NULL, AV_LOG_ERROR, "envelope scalefactor overflow in dequant\n");
+                    temp1 = FLOAT_1;
+                }
+
+                temp2.exp = (pan_offset - sbr->data[1].env_facs_q[e][k]) * alpha;
+                if (temp2.exp & 1)
+                  temp2.mant = 759250125;
+                else
+                  temp2.mant = 0x20000000;
+                temp2.exp = (temp2.exp >> 1) + 1;
+                fac   = av_div_sf(temp1, av_add_sf(FLOAT_1, temp2));
+                sbr->data[0].env_facs[e][k] = fac;
+                sbr->data[1].env_facs[e][k] = av_mul_sf(fac, temp2);
+            }
+        }
+        for (e = 1; e <= sbr->data[0].bs_num_noise; e++) {
+            for (k = 0; k < sbr->n_q; k++) {
+                SoftFloat temp1, temp2, fac;
+
+                temp1.exp = NOISE_FLOOR_OFFSET - \
+                    sbr->data[0].noise_facs_q[e][k] + 2;
+                temp1.mant = 0x20000000;
+                av_assert0(temp1.exp <= 66);
+                temp2.exp = 12 - sbr->data[1].noise_facs_q[e][k] + 1;
+                temp2.mant = 0x20000000;
+                fac   = av_div_sf(temp1, av_add_sf(FLOAT_1, temp2));
+                sbr->data[0].noise_facs[e][k] = fac;
+                sbr->data[1].noise_facs[e][k] = av_mul_sf(fac, temp2);
+            }
+        }
+    } else { // SCE or one non-coupled CPE
+        for (ch = 0; ch < (id_aac == TYPE_CPE) + 1; ch++) {
+            int alpha = sbr->data[ch].bs_amp_res ? 2 : 1;
+            for (e = 1; e <= sbr->data[ch].bs_num_env; e++)
+                for (k = 0; k < sbr->n[sbr->data[ch].bs_freq_res[e]]; k++){
+                    SoftFloat temp1;
+
+                    temp1.exp = alpha * sbr->data[ch].env_facs_q[e][k] + 12;
+                    if (temp1.exp & 1)
+                        temp1.mant = 759250125;
+                    else
+                        temp1.mant = 0x20000000;
+                    temp1.exp = (temp1.exp >> 1) + 1;
+                    if (temp1.exp > 66) { // temp1 > 1E20
+                        av_log(NULL, AV_LOG_ERROR, "envelope scalefactor overflow in dequant\n");
+                        temp1 = FLOAT_1;
+                    }
+                    sbr->data[ch].env_facs[e][k] = temp1;
+                }
+            for (e = 1; e <= sbr->data[ch].bs_num_noise; e++)
+                for (k = 0; k < sbr->n_q; k++){
+                    sbr->data[ch].noise_facs[e][k].exp = NOISE_FLOOR_OFFSET - \
+                        sbr->data[ch].noise_facs_q[e][k] + 1;
+                    sbr->data[ch].noise_facs[e][k].mant = 0x20000000;
+                }
+        }
+    }
+}
+
+/** High Frequency Generation (14496-3 sp04 p214+) and Inverse Filtering
+ * (14496-3 sp04 p214)
+ * Warning: This routine does not seem numerically stable.
+ */
+static void sbr_hf_inverse_filter(SBRDSPContext *dsp,
+                                  int (*alpha0)[2], int (*alpha1)[2],
+                                  const int X_low[32][40][2], int k0)
+{
+    int k;
+    int shift, round;
+
+    for (k = 0; k < k0; k++) {
+        SoftFloat phi[3][2][2];
+        SoftFloat a00, a01, a10, a11;
+        SoftFloat dk;
+
+        dsp->autocorrelate(X_low[k], phi);
+
+        dk = av_sub_sf(av_mul_sf(phi[2][1][0], phi[1][0][0]),
+             av_mul_sf(av_add_sf(av_mul_sf(phi[1][1][0], phi[1][1][0]),
+             av_mul_sf(phi[1][1][1], phi[1][1][1])), FLOAT_0999999));
+
+        if (!dk.mant) {
+            a10 = FLOAT_0;
+            a11 = FLOAT_0;
+        } else {
+            SoftFloat temp_real, temp_im;
+            temp_real = av_sub_sf(av_sub_sf(av_mul_sf(phi[0][0][0], phi[1][1][0]),
+                                            av_mul_sf(phi[0][0][1], phi[1][1][1])),
+                                  av_mul_sf(phi[0][1][0], phi[1][0][0]));
+            temp_im   = av_sub_sf(av_add_sf(av_mul_sf(phi[0][0][0], phi[1][1][1]),
+                                            av_mul_sf(phi[0][0][1], phi[1][1][0])),
+                                  av_mul_sf(phi[0][1][1], phi[1][0][0]));
+
+            a10 = av_div_sf(temp_real, dk);
+            a11 = av_div_sf(temp_im,   dk);
+        }
+
+        if (!phi[1][0][0].mant) {
+            a00 = FLOAT_0;
+            a01 = FLOAT_0;
+        } else {
+            SoftFloat temp_real, temp_im;
+            temp_real = av_add_sf(phi[0][0][0],
+                                  av_add_sf(av_mul_sf(a10, phi[1][1][0]),
+                                            av_mul_sf(a11, phi[1][1][1])));
+            temp_im   = av_add_sf(phi[0][0][1],
+                                  av_sub_sf(av_mul_sf(a11, phi[1][1][0]),
+                                            av_mul_sf(a10, phi[1][1][1])));
+
+            temp_real.mant = -temp_real.mant;
+            temp_im.mant   = -temp_im.mant;
+            a00 = av_div_sf(temp_real, phi[1][0][0]);
+            a01 = av_div_sf(temp_im,   phi[1][0][0]);
+        }
+
+        shift = a00.exp;
+        if (shift >= 3)
+            alpha0[k][0] = 0x7fffffff;
+        else if (shift <= -30)
+            alpha0[k][0] = 0;
+        else {
+            shift = 1-shift;
+            if (shift <= 0)
+                alpha0[k][0] = a00.mant * (1<<-shift);
+            else {
+                round = 1 << (shift-1);
+                alpha0[k][0] = (a00.mant + round) >> shift;
+            }
+        }
+
+        shift = a01.exp;
+        if (shift >= 3)
+            alpha0[k][1] = 0x7fffffff;
+        else if (shift <= -30)
+            alpha0[k][1] = 0;
+        else {
+            shift = 1-shift;
+            if (shift <= 0)
+                alpha0[k][1] = a01.mant * (1<<-shift);
+            else {
+                round = 1 << (shift-1);
+                alpha0[k][1] = (a01.mant + round) >> shift;
+            }
+        }
+        shift = a10.exp;
+        if (shift >= 3)
+            alpha1[k][0] = 0x7fffffff;
+        else if (shift <= -30)
+            alpha1[k][0] = 0;
+        else {
+            shift = 1-shift;
+            if (shift <= 0)
+                alpha1[k][0] = a10.mant * (1<<-shift);
+            else {
+                round = 1 << (shift-1);
+                alpha1[k][0] = (a10.mant + round) >> shift;
+            }
+        }
+
+        shift = a11.exp;
+        if (shift >= 3)
+            alpha1[k][1] = 0x7fffffff;
+        else if (shift <= -30)
+            alpha1[k][1] = 0;
+        else {
+            shift = 1-shift;
+            if (shift <= 0)
+                alpha1[k][1] = a11.mant * (1<<-shift);
+            else {
+                round = 1 << (shift-1);
+                alpha1[k][1] = (a11.mant + round) >> shift;
+            }
+        }
+
+        shift = (int)(((int64_t)(alpha1[k][0]>>1) * (alpha1[k][0]>>1) + \
+                       (int64_t)(alpha1[k][1]>>1) * (alpha1[k][1]>>1) + \
+                       0x40000000) >> 31);
+        if (shift >= 0x20000000){
+            alpha1[k][0] = 0;
+            alpha1[k][1] = 0;
+            alpha0[k][0] = 0;
+            alpha0[k][1] = 0;
+        }
+
+        shift = (int)(((int64_t)(alpha0[k][0]>>1) * (alpha0[k][0]>>1) + \
+                       (int64_t)(alpha0[k][1]>>1) * (alpha0[k][1]>>1) + \
+                       0x40000000) >> 31);
+        if (shift >= 0x20000000){
+            alpha1[k][0] = 0;
+            alpha1[k][1] = 0;
+            alpha0[k][0] = 0;
+            alpha0[k][1] = 0;
+        }
+    }
+}
+
+/// Chirp Factors (14496-3 sp04 p214)
+static void sbr_chirp(SpectralBandReplication *sbr, SBRData *ch_data)
+{
+    int i;
+    int new_bw;
+    static const int bw_tab[] = { 0, 1610612736, 1932735283, 2104533975 };
+    int64_t accu;
+
+    for (i = 0; i < sbr->n_q; i++) {
+        if (ch_data->bs_invf_mode[0][i] + ch_data->bs_invf_mode[1][i] == 1)
+            new_bw = 1288490189;
+        else
+            new_bw = bw_tab[ch_data->bs_invf_mode[0][i]];
+
+        if (new_bw < ch_data->bw_array[i]){
+            accu  = (int64_t)new_bw * 1610612736;
+            accu += (int64_t)ch_data->bw_array[i] * 0x20000000;
+            new_bw = (int)((accu + 0x40000000) >> 31);
+        } else {
+            accu  = (int64_t)new_bw * 1946157056;
+            accu += (int64_t)ch_data->bw_array[i] * 201326592;
+            new_bw = (int)((accu + 0x40000000) >> 31);
+        }
+        ch_data->bw_array[i] = new_bw < 0x2000000 ? 0 : new_bw;
+    }
+}
+
+/**
+ * Calculation of levels of additional HF signal components (14496-3 sp04 p219)
+ * and Calculation of gain (14496-3 sp04 p219)
+ */
+static void sbr_gain_calc(AACContext *ac, SpectralBandReplication *sbr,
+                          SBRData *ch_data, const int e_a[2])
+{
+    int e, k, m;
+    // max gain limits : -3dB, 0dB, 3dB, inf dB (limiter off)
+    static const SoftFloat limgain[4] = { { 760155524,  0 }, { 0x20000000,  1 },
+                                            { 758351638,  1 }, { 625000000, 34 } };
+
+    for (e = 0; e < ch_data->bs_num_env; e++) {
+        int delta = !((e == e_a[1]) || (e == e_a[0]));
+        for (k = 0; k < sbr->n_lim; k++) {
+            SoftFloat gain_boost, gain_max;
+            SoftFloat sum[2];
+            sum[0] = sum[1] = FLOAT_0;
+            for (m = sbr->f_tablelim[k] - sbr->kx[1]; m < sbr->f_tablelim[k + 1] - sbr->kx[1]; m++) {
+                const SoftFloat temp = av_div_sf(sbr->e_origmapped[e][m],
+                                            av_add_sf(FLOAT_1, sbr->q_mapped[e][m]));
+                sbr->q_m[e][m] = av_sqrt_sf(av_mul_sf(temp, sbr->q_mapped[e][m]));
+                sbr->s_m[e][m] = av_sqrt_sf(av_mul_sf(temp, av_int2sf(ch_data->s_indexmapped[e + 1][m], 0)));
+                if (!sbr->s_mapped[e][m]) {
+                    if (delta) {
+                      sbr->gain[e][m] = av_sqrt_sf(av_div_sf(sbr->e_origmapped[e][m],
+                                            av_mul_sf(av_add_sf(FLOAT_1, sbr->e_curr[e][m]),
+                                            av_add_sf(FLOAT_1, sbr->q_mapped[e][m]))));
+                    } else {
+                      sbr->gain[e][m] = av_sqrt_sf(av_div_sf(sbr->e_origmapped[e][m],
+                                            av_add_sf(FLOAT_1, sbr->e_curr[e][m])));
+                    }
+                } else {
+                    sbr->gain[e][m] = av_sqrt_sf(
+                                        av_div_sf(
+                                            av_mul_sf(sbr->e_origmapped[e][m], sbr->q_mapped[e][m]),
+                                            av_mul_sf(
+                                                av_add_sf(FLOAT_1, sbr->e_curr[e][m]),
+                                                av_add_sf(FLOAT_1, sbr->q_mapped[e][m]))));
+                }
+                sbr->gain[e][m] = av_add_sf(sbr->gain[e][m], FLOAT_MIN);
+            }
+            for (m = sbr->f_tablelim[k] - sbr->kx[1]; m < sbr->f_tablelim[k + 1] - sbr->kx[1]; m++) {
+                sum[0] = av_add_sf(sum[0], sbr->e_origmapped[e][m]);
+                sum[1] = av_add_sf(sum[1], sbr->e_curr[e][m]);
+            }
+            gain_max = av_mul_sf(limgain[sbr->bs_limiter_gains],
+                            av_sqrt_sf(
+                                av_div_sf(
+                                    av_add_sf(FLOAT_EPSILON, sum[0]),
+                                    av_add_sf(FLOAT_EPSILON, sum[1]))));
+            if (av_gt_sf(gain_max, FLOAT_100000))
+              gain_max = FLOAT_100000;
+            for (m = sbr->f_tablelim[k] - sbr->kx[1]; m < sbr->f_tablelim[k + 1] - sbr->kx[1]; m++) {
+                SoftFloat q_m_max = av_div_sf(
+                                        av_mul_sf(sbr->q_m[e][m], gain_max),
+                                        sbr->gain[e][m]);
+                if (av_gt_sf(sbr->q_m[e][m], q_m_max))
+                  sbr->q_m[e][m] = q_m_max;
+                if (av_gt_sf(sbr->gain[e][m], gain_max))
+                  sbr->gain[e][m] = gain_max;
+            }
+            sum[0] = sum[1] = FLOAT_0;
+            for (m = sbr->f_tablelim[k] - sbr->kx[1]; m < sbr->f_tablelim[k + 1] - sbr->kx[1]; m++) {
+                sum[0] = av_add_sf(sum[0], sbr->e_origmapped[e][m]);
+                sum[1] = av_add_sf(sum[1],
+                            av_mul_sf(
+                                av_mul_sf(sbr->e_curr[e][m],
+                                          sbr->gain[e][m]),
+                                sbr->gain[e][m]));
+                sum[1] = av_add_sf(sum[1],
+                            av_mul_sf(sbr->s_m[e][m], sbr->s_m[e][m]));
+                if (delta && !sbr->s_m[e][m].mant)
+                  sum[1] = av_add_sf(sum[1],
+                                av_mul_sf(sbr->q_m[e][m], sbr->q_m[e][m]));
+            }
+            gain_boost = av_sqrt_sf(
+                            av_div_sf(
+                                av_add_sf(FLOAT_EPSILON, sum[0]),
+                                av_add_sf(FLOAT_EPSILON, sum[1])));
+            if (av_gt_sf(gain_boost, FLOAT_1584893192))
+              gain_boost = FLOAT_1584893192;
+
+            for (m = sbr->f_tablelim[k] - sbr->kx[1]; m < sbr->f_tablelim[k + 1] - sbr->kx[1]; m++) {
+                sbr->gain[e][m] = av_mul_sf(sbr->gain[e][m], gain_boost);
+                sbr->q_m[e][m]  = av_mul_sf(sbr->q_m[e][m], gain_boost);
+                sbr->s_m[e][m]  = av_mul_sf(sbr->s_m[e][m], gain_boost);
+            }
+        }
+    }
+}
+
+/// Assembling HF Signals (14496-3 sp04 p220)
+static void sbr_hf_assemble(int Y1[38][64][2],
+                            const int X_high[64][40][2],
+                            SpectralBandReplication *sbr, SBRData *ch_data,
+                            const int e_a[2])
+{
+    int e, i, j, m;
+    const int h_SL = 4 * !sbr->bs_smoothing_mode;
+    const int kx = sbr->kx[1];
+    const int m_max = sbr->m[1];
+    static const SoftFloat h_smooth[5] = {
+      { 715827883, -1 },
+      { 647472402, -1 },
+      { 937030863, -2 },
+      { 989249804, -3 },
+      { 546843842, -4 },
+    };
+    SoftFloat (*g_temp)[48] = ch_data->g_temp, (*q_temp)[48] = ch_data->q_temp;
+    int indexnoise = ch_data->f_indexnoise;
+    int indexsine  = ch_data->f_indexsine;
+
+    if (sbr->reset) {
+        for (i = 0; i < h_SL; i++) {
+            memcpy(g_temp[i + 2*ch_data->t_env[0]], sbr->gain[0], m_max * sizeof(sbr->gain[0][0]));
+            memcpy(q_temp[i + 2*ch_data->t_env[0]], sbr->q_m[0],  m_max * sizeof(sbr->q_m[0][0]));
+        }
+    } else if (h_SL) {
+        for (i = 0; i < 4; i++) {
+            memcpy(g_temp[i + 2 * ch_data->t_env[0]],
+                   g_temp[i + 2 * ch_data->t_env_num_env_old],
+                   sizeof(g_temp[0]));
+            memcpy(q_temp[i + 2 * ch_data->t_env[0]],
+                   q_temp[i + 2 * ch_data->t_env_num_env_old],
+                   sizeof(q_temp[0]));
+        }
+    }
+
+    for (e = 0; e < ch_data->bs_num_env; e++) {
+        for (i = 2 * ch_data->t_env[e]; i < 2 * ch_data->t_env[e + 1]; i++) {
+            memcpy(g_temp[h_SL + i], sbr->gain[e], m_max * sizeof(sbr->gain[0][0]));
+            memcpy(q_temp[h_SL + i], sbr->q_m[e],  m_max * sizeof(sbr->q_m[0][0]));
+        }
+    }
+
+    for (e = 0; e < ch_data->bs_num_env; e++) {
+        for (i = 2 * ch_data->t_env[e]; i < 2 * ch_data->t_env[e + 1]; i++) {
+            SoftFloat g_filt_tab[48];
+            SoftFloat q_filt_tab[48];
+            SoftFloat *g_filt, *q_filt;
+
+            if (h_SL && e != e_a[0] && e != e_a[1]) {
+                g_filt = g_filt_tab;
+                q_filt = q_filt_tab;
+                for (m = 0; m < m_max; m++) {
+                    const int idx1 = i + h_SL;
+                    g_filt[m].mant = g_filt[m].exp = 0;
+                    q_filt[m].mant = q_filt[m].exp = 0;
+                    for (j = 0; j <= h_SL; j++) {
+                        g_filt[m] = av_add_sf(g_filt[m],
+                                        av_mul_sf(g_temp[idx1 - j][m],
+                                            h_smooth[j]));
+                        q_filt[m] = av_add_sf(q_filt[m],
+                                        av_mul_sf(q_temp[idx1 - j][m],
+                                            h_smooth[j]));
+                    }
+                }
+            } else {
+                g_filt = g_temp[i + h_SL];
+                q_filt = q_temp[i];
+            }
+
+            sbr->dsp.hf_g_filt(Y1[i] + kx, X_high + kx, g_filt, m_max,
+                               i + ENVELOPE_ADJUSTMENT_OFFSET);
+
+            if (e != e_a[0] && e != e_a[1]) {
+                sbr->dsp.hf_apply_noise[indexsine](Y1[i] + kx, sbr->s_m[e],
+                                                   q_filt, indexnoise,
+                                                   kx, m_max);
+            } else {
+                int idx = indexsine&1;
+                int A = (1-((indexsine+(kx & 1))&2));
+                int B = (A^(-idx)) + idx;
+                unsigned *out = &Y1[i][kx][idx];
+                int shift;
+                unsigned round;
+
+                SoftFloat *in  = sbr->s_m[e];
+                for (m = 0; m+1 < m_max; m+=2) {
+                    int shift2;
+                    shift = 22 - in[m  ].exp;
+                    shift2= 22 - in[m+1].exp;
+                    if (shift < 1 || shift2 < 1) {
+                        av_log(NULL, AV_LOG_ERROR, "Overflow in sbr_hf_assemble, shift=%d,%d\n", shift, shift2);
+                        return;
+                    }
+                    if (shift < 32) {
+                        round = 1 << (shift-1);
+                        out[2*m  ] += (int)(in[m  ].mant * A + round) >> shift;
+                    }
+
+                    if (shift2 < 32) {
+                        round = 1 << (shift2-1);
+                        out[2*m+2] += (int)(in[m+1].mant * B + round) >> shift2;
+                    }
+                }
+                if(m_max&1)
+                {
+                    shift = 22 - in[m  ].exp;
+                    if (shift < 1) {
+                        av_log(NULL, AV_LOG_ERROR, "Overflow in sbr_hf_assemble, shift=%d\n", shift);
+                        return;
+                    } else if (shift < 32) {
+                        round = 1 << (shift-1);
+                        out[2*m  ] += (int)(in[m  ].mant * A + round) >> shift;
+                    }
+                }
+            }
+            indexnoise = (indexnoise + m_max) & 0x1ff;
+            indexsine = (indexsine + 1) & 3;
+        }
+    }
+    ch_data->f_indexnoise = indexnoise;
+    ch_data->f_indexsine  = indexsine;
+}
+
+#include "aacsbr_template.c"

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aacsbr_fixed_tablegen.h

@@ -0,0 +1,28 @@
+/*
+ * Header file for hardcoded AAC SBR windows
+ *
+ * Copyright (c) 2014 Reimar Döffinger <Reimar.Doeffinger@gmx.de>
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#ifndef AVCODEC_AACSBR_FIXED_TABLEGEN_H
+#define AVCODEC_AACSBR_FIXED_TABLEGEN_H
+
+#include "aacsbr_tablegen_common.h"
+
+#endif /* AVCODEC_AACSBR_FIXED_TABLEGEN_H */

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aacsbr_tablegen.h

@@ -0,0 +1,28 @@
+/*
+ * Header file for hardcoded AAC SBR windows
+ *
+ * Copyright (c) 2014 Reimar Döffinger <Reimar.Doeffinger@gmx.de>
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#ifndef AVCODEC_AACSBR_TABLEGEN_H
+#define AVCODEC_AACSBR_TABLEGEN_H
+
+#include "aacsbr_tablegen_common.h"
+
+#endif /* AVCODEC_AACSBR_TABLEGEN_H */

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aacsbr_tablegen_common.h

@@ -0,0 +1,114 @@
+/*
+ * Header file for hardcoded AAC SBR windows
+ *
+ * Copyright (c) 2014 Reimar Döffinger <Reimar.Doeffinger@gmx.de>
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#ifndef AVCODEC_AACSBR_TABLEGEN_COMMON_H
+#define AVCODEC_AACSBR_TABLEGEN_COMMON_H
+#include "aac_defines.h"
+#include "libavutil/mem.h"
+
+///< window coefficients for analysis/synthesis QMF banks
+static DECLARE_ALIGNED(32, INTFLOAT, sbr_qmf_window_ds)[320];
+static DECLARE_ALIGNED(32, INTFLOAT, sbr_qmf_window_us)[640] = {
+    Q31( 0.0000000000f), Q31(-0.0005525286f), Q31(-0.0005617692f), Q31(-0.0004947518f),
+    Q31(-0.0004875227f), Q31(-0.0004893791f), Q31(-0.0005040714f), Q31(-0.0005226564f),
+    Q31(-0.0005466565f), Q31(-0.0005677802f), Q31(-0.0005870930f), Q31(-0.0006132747f),
+    Q31(-0.0006312493f), Q31(-0.0006540333f), Q31(-0.0006777690f), Q31(-0.0006941614f),
+    Q31(-0.0007157736f), Q31(-0.0007255043f), Q31(-0.0007440941f), Q31(-0.0007490598f),
+    Q31(-0.0007681371f), Q31(-0.0007724848f), Q31(-0.0007834332f), Q31(-0.0007779869f),
+    Q31(-0.0007803664f), Q31(-0.0007801449f), Q31(-0.0007757977f), Q31(-0.0007630793f),
+    Q31(-0.0007530001f), Q31(-0.0007319357f), Q31(-0.0007215391f), Q31(-0.0006917937f),
+    Q31(-0.0006650415f), Q31(-0.0006341594f), Q31(-0.0005946118f), Q31(-0.0005564576f),
+    Q31(-0.0005145572f), Q31(-0.0004606325f), Q31(-0.0004095121f), Q31(-0.0003501175f),
+    Q31(-0.0002896981f), Q31(-0.0002098337f), Q31(-0.0001446380f), Q31(-0.0000617334f),
+    Q31( 0.0000134949f), Q31( 0.0001094383f), Q31( 0.0002043017f), Q31( 0.0002949531f),
+    Q31( 0.0004026540f), Q31( 0.0005107388f), Q31( 0.0006239376f), Q31( 0.0007458025f),
+    Q31( 0.0008608443f), Q31( 0.0009885988f), Q31( 0.0011250155f), Q31( 0.0012577884f),
+    Q31( 0.0013902494f), Q31( 0.0015443219f), Q31( 0.0016868083f), Q31( 0.0018348265f),
+    Q31( 0.0019841140f), Q31( 0.0021461583f), Q31( 0.0023017254f), Q31( 0.0024625616f),
+    Q31( 0.0026201758f), Q31( 0.0027870464f), Q31( 0.0029469447f), Q31( 0.0031125420f),
+    Q31( 0.0032739613f), Q31( 0.0034418874f), Q31( 0.0036008268f), Q31( 0.0037603922f),
+    Q31( 0.0039207432f), Q31( 0.0040819753f), Q31( 0.0042264269f), Q31( 0.0043730719f),
+    Q31( 0.0045209852f), Q31( 0.0046606460f), Q31( 0.0047932560f), Q31( 0.0049137603f),
+    Q31( 0.0050393022f), Q31( 0.0051407353f), Q31( 0.0052461166f), Q31( 0.0053471681f),
+    Q31( 0.0054196775f), Q31( 0.0054876040f), Q31( 0.0055475714f), Q31( 0.0055938023f),
+    Q31( 0.0056220643f), Q31( 0.0056455196f), Q31( 0.0056389199f), Q31( 0.0056266114f),
+    Q31( 0.0055917128f), Q31( 0.0055404363f), Q31( 0.0054753783f), Q31( 0.0053838975f),
+    Q31( 0.0052715758f), Q31( 0.0051382275f), Q31( 0.0049839687f), Q31( 0.0048109469f),
+    Q31( 0.0046039530f), Q31( 0.0043801861f), Q31( 0.0041251642f), Q31( 0.0038456408f),
+    Q31( 0.0035401246f), Q31( 0.0032091885f), Q31( 0.0028446757f), Q31( 0.0024508540f),
+    Q31( 0.0020274176f), Q31( 0.0015784682f), Q31( 0.0010902329f), Q31( 0.0005832264f),
+    Q31( 0.0000276045f), Q31(-0.0005464280f), Q31(-0.0011568135f), Q31(-0.0018039472f),
+    Q31(-0.0024826723f), Q31(-0.0031933778f), Q31(-0.0039401124f), Q31(-0.0047222596f),
+    Q31(-0.0055337211f), Q31(-0.0063792293f), Q31(-0.0072615816f), Q31(-0.0081798233f),
+    Q31(-0.0091325329f), Q31(-0.0101150215f), Q31(-0.0111315548f), Q31(-0.0121849995f),
+    Q31( 0.0132718220f), Q31( 0.0143904666f), Q31( 0.0155405553f), Q31( 0.0167324712f),
+    Q31( 0.0179433381f), Q31( 0.0191872431f), Q31( 0.0204531793f), Q31( 0.0217467550f),
+    Q31( 0.0230680169f), Q31( 0.0244160992f), Q31( 0.0257875847f), Q31( 0.0271859429f),
+    Q31( 0.0286072173f), Q31( 0.0300502657f), Q31( 0.0315017608f), Q31( 0.0329754081f),
+    Q31( 0.0344620948f), Q31( 0.0359697560f), Q31( 0.0374812850f), Q31( 0.0390053679f),
+    Q31( 0.0405349170f), Q31( 0.0420649094f), Q31( 0.0436097542f), Q31( 0.0451488405f),
+    Q31( 0.0466843027f), Q31( 0.0482165720f), Q31( 0.0497385755f), Q31( 0.0512556155f),
+    Q31( 0.0527630746f), Q31( 0.0542452768f), Q31( 0.0557173648f), Q31( 0.0571616450f),
+    Q31( 0.0585915683f), Q31( 0.0599837480f), Q31( 0.0613455171f), Q31( 0.0626857808f),
+    Q31( 0.0639715898f), Q31( 0.0652247106f), Q31( 0.0664367512f), Q31( 0.0676075985f),
+    Q31( 0.0687043828f), Q31( 0.0697630244f), Q31( 0.0707628710f), Q31( 0.0717002673f),
+    Q31( 0.0725682583f), Q31( 0.0733620255f), Q31( 0.0741003642f), Q31( 0.0747452558f),
+    Q31( 0.0753137336f), Q31( 0.0758008358f), Q31( 0.0761992479f), Q31( 0.0764992170f),
+    Q31( 0.0767093490f), Q31( 0.0768173975f), Q31( 0.0768230011f), Q31( 0.0767204924f),
+    Q31( 0.0765050718f), Q31( 0.0761748321f), Q31( 0.0757305756f), Q31( 0.0751576255f),
+    Q31( 0.0744664394f), Q31( 0.0736406005f), Q31( 0.0726774642f), Q31( 0.0715826364f),
+    Q31( 0.0703533073f), Q31( 0.0689664013f), Q31( 0.0674525021f), Q31( 0.0657690668f),
+    Q31( 0.0639444805f), Q31( 0.0619602779f), Q31( 0.0598166570f), Q31( 0.0575152691f),
+    Q31( 0.0550460034f), Q31( 0.0524093821f), Q31( 0.0495978676f), Q31( 0.0466303305f),
+    Q31( 0.0434768782f), Q31( 0.0401458278f), Q31( 0.0366418116f), Q31( 0.0329583930f),
+    Q31( 0.0290824006f), Q31( 0.0250307561f), Q31( 0.0207997072f), Q31( 0.0163701258f),
+    Q31( 0.0117623832f), Q31( 0.0069636862f), Q31( 0.0019765601f), Q31(-0.0032086896f),
+    Q31(-0.0085711749f), Q31(-0.0141288827f), Q31(-0.0198834129f), Q31(-0.0258227288f),
+    Q31(-0.0319531274f), Q31(-0.0382776572f), Q31(-0.0447806821f), Q31(-0.0514804176f),
+    Q31(-0.0583705326f), Q31(-0.0654409853f), Q31(-0.0726943300f), Q31(-0.0801372934f),
+    Q31(-0.0877547536f), Q31(-0.0955533352f), Q31(-0.1035329531f), Q31(-0.1116826931f),
+    Q31(-0.1200077984f), Q31(-0.1285002850f), Q31(-0.1371551761f), Q31(-0.1459766491f),
+    Q31(-0.1549607071f), Q31(-0.1640958855f), Q31(-0.1733808172f), Q31(-0.1828172548f),
+    Q31(-0.1923966745f), Q31(-0.2021250176f), Q31(-0.2119735853f), Q31(-0.2219652696f),
+    Q31(-0.2320690870f), Q31(-0.2423016884f), Q31(-0.2526480309f), Q31(-0.2631053299f),
+    Q31(-0.2736634040f), Q31(-0.2843214189f), Q31(-0.2950716717f), Q31(-0.3059098575f),
+    Q31(-0.3168278913f), Q31(-0.3278113727f), Q31(-0.3388722693f), Q31(-0.3499914122f),
+    Q31( 0.3611589903f), Q31( 0.3723795546f), Q31( 0.3836350013f), Q31( 0.3949211761f),
+    Q31( 0.4062317676f), Q31( 0.4175696896f), Q31( 0.4289119920f), Q31( 0.4402553754f),
+    Q31( 0.4515996535f), Q31( 0.4629308085f), Q31( 0.4742453214f), Q31( 0.4855253091f),
+    Q31( 0.4967708254f), Q31( 0.5079817500f), Q31( 0.5191234970f), Q31( 0.5302240895f),
+    Q31( 0.5412553448f), Q31( 0.5522051258f), Q31( 0.5630789140f), Q31( 0.5738524131f),
+    Q31( 0.5845403235f), Q31( 0.5951123086f), Q31( 0.6055783538f), Q31( 0.6159109932f),
+    Q31( 0.6261242695f), Q31( 0.6361980107f), Q31( 0.6461269695f), Q31( 0.6559016302f),
+    Q31( 0.6655139880f), Q31( 0.6749663190f), Q31( 0.6842353293f), Q31( 0.6933282376f),
+    Q31( 0.7022388719f), Q31( 0.7109410426f), Q31( 0.7194462634f), Q31( 0.7277448900f),
+    Q31( 0.7358211758f), Q31( 0.7436827863f), Q31( 0.7513137456f), Q31( 0.7587080760f),
+    Q31( 0.7658674865f), Q31( 0.7727780881f), Q31( 0.7794287519f), Q31( 0.7858353120f),
+    Q31( 0.7919735841f), Q31( 0.7978466413f), Q31( 0.8034485751f), Q31( 0.8087695004f),
+    Q31( 0.8138191270f), Q31( 0.8185776004f), Q31( 0.8230419890f), Q31( 0.8272275347f),
+    Q31( 0.8311038457f), Q31( 0.8346937361f), Q31( 0.8379717337f), Q31( 0.8409541392f),
+    Q31( 0.8436238281f), Q31( 0.8459818469f), Q31( 0.8480315777f), Q31( 0.8497805198f),
+    Q31( 0.8511971524f), Q31( 0.8523047035f), Q31( 0.8531020949f), Q31( 0.8535720573f),
+    Q31( 0.8537385600f),
+};
+
+#endif /* AVCODEC_AACSBR_TABLEGEN_COMMON_H */

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aacsbrdata.h

@@ -0,0 +1,535 @@
+/*
+ * AAC Spectral Band Replication decoding data
+ * Copyright (c) 2008-2009 Robert Swain ( rob opendot cl )
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+/**
+ * @file
+ * AAC Spectral Band Replication decoding data
+ * @author Robert Swain ( rob opendot cl )
+ */
+
+#ifndef AVCODEC_AACSBRDATA_H
+#define AVCODEC_AACSBRDATA_H
+
+#include <stdint.h>
+#include "libavutil/mem.h"
+#include "aac_defines.h"
+
+///< Huffman tables for SBR
+
+static const uint8_t t_huffman_env_1_5dB_bits[121] = {
+    18, 18, 18, 18, 18, 18, 19, 19,
+    19, 19, 19, 19, 19, 19, 19, 19,
+    19, 19, 19, 19, 19, 19, 19, 19,
+    19, 19, 19, 19, 19, 19, 19, 19,
+    19, 19, 17, 18, 16, 17, 18, 17,
+    16, 16, 16, 16, 15, 14, 14, 13,
+    13, 12, 11, 10,  9,  8,  7,  6,
+     5,  4,  3,  2,  2,  3,  4,  5,
+     6,  7,  8,  9, 10, 12, 13, 14,
+    14, 15, 16, 17, 16, 19, 19, 19,
+    19, 19, 19, 19, 19, 19, 19, 19,
+    19, 19, 19, 19, 19, 19, 19, 19,
+    19, 19, 19, 19, 19, 19, 19, 19,
+    19, 19, 19, 19, 19, 19, 19, 19,
+    19, 19, 19, 19, 19, 19, 19, 19,
+    19,
+};
+
+static const uint32_t t_huffman_env_1_5dB_codes[121] = {
+    0x3ffd6, 0x3ffd7, 0x3ffd8, 0x3ffd9, 0x3ffda, 0x3ffdb, 0x7ffb8, 0x7ffb9,
+    0x7ffba, 0x7ffbb, 0x7ffbc, 0x7ffbd, 0x7ffbe, 0x7ffbf, 0x7ffc0, 0x7ffc1,
+    0x7ffc2, 0x7ffc3, 0x7ffc4, 0x7ffc5, 0x7ffc6, 0x7ffc7, 0x7ffc8, 0x7ffc9,
+    0x7ffca, 0x7ffcb, 0x7ffcc, 0x7ffcd, 0x7ffce, 0x7ffcf, 0x7ffd0, 0x7ffd1,
+    0x7ffd2, 0x7ffd3, 0x1ffe6, 0x3ffd4, 0x0fff0, 0x1ffe9, 0x3ffd5, 0x1ffe7,
+    0x0fff1, 0x0ffec, 0x0ffed, 0x0ffee, 0x07ff4, 0x03ff9, 0x03ff7, 0x01ffa,
+    0x01ff9, 0x00ffb, 0x007fc, 0x003fc, 0x001fd, 0x000fd, 0x0007d, 0x0003d,
+    0x0001d, 0x0000d, 0x00005, 0x00001, 0x00000, 0x00004, 0x0000c, 0x0001c,
+    0x0003c, 0x0007c, 0x000fc, 0x001fc, 0x003fd, 0x00ffa, 0x01ff8, 0x03ff6,
+    0x03ff8, 0x07ff5, 0x0ffef, 0x1ffe8, 0x0fff2, 0x7ffd4, 0x7ffd5, 0x7ffd6,
+    0x7ffd7, 0x7ffd8, 0x7ffd9, 0x7ffda, 0x7ffdb, 0x7ffdc, 0x7ffdd, 0x7ffde,
+    0x7ffdf, 0x7ffe0, 0x7ffe1, 0x7ffe2, 0x7ffe3, 0x7ffe4, 0x7ffe5, 0x7ffe6,
+    0x7ffe7, 0x7ffe8, 0x7ffe9, 0x7ffea, 0x7ffeb, 0x7ffec, 0x7ffed, 0x7ffee,
+    0x7ffef, 0x7fff0, 0x7fff1, 0x7fff2, 0x7fff3, 0x7fff4, 0x7fff5, 0x7fff6,
+    0x7fff7, 0x7fff8, 0x7fff9, 0x7fffa, 0x7fffb, 0x7fffc, 0x7fffd, 0x7fffe,
+    0x7ffff,
+};
+
+static const uint8_t f_huffman_env_1_5dB_bits[121] = {
+    19, 19, 20, 20, 20, 20, 20, 20,
+    20, 19, 20, 20, 20, 20, 19, 20,
+    19, 19, 20, 18, 20, 20, 20, 19,
+    20, 20, 20, 19, 20, 19, 18, 19,
+    18, 18, 17, 18, 17, 17, 17, 16,
+    16, 16, 15, 15, 14, 13, 13, 12,
+    12, 11, 10,  9,  9,  8,  7,  6,
+     5,  4,  3,  2,  2,  3,  4,  5,
+     6,  8,  8,  9, 10, 11, 11, 11,
+    12, 12, 13, 13, 14, 14, 16, 16,
+    17, 17, 18, 18, 18, 18, 18, 18,
+    18, 20, 19, 20, 20, 20, 20, 20,
+    20, 19, 20, 20, 20, 20, 19, 20,
+    18, 20, 20, 19, 19, 20, 20, 20,
+    20, 20, 20, 20, 20, 20, 20, 20,
+    20,
+};
+
+static const uint32_t f_huffman_env_1_5dB_codes[121] = {
+    0x7ffe7, 0x7ffe8, 0xfffd2, 0xfffd3, 0xfffd4, 0xfffd5, 0xfffd6, 0xfffd7,
+    0xfffd8, 0x7ffda, 0xfffd9, 0xfffda, 0xfffdb, 0xfffdc, 0x7ffdb, 0xfffdd,
+    0x7ffdc, 0x7ffdd, 0xfffde, 0x3ffe4, 0xfffdf, 0xfffe0, 0xfffe1, 0x7ffde,
+    0xfffe2, 0xfffe3, 0xfffe4, 0x7ffdf, 0xfffe5, 0x7ffe0, 0x3ffe8, 0x7ffe1,
+    0x3ffe0, 0x3ffe9, 0x1ffef, 0x3ffe5, 0x1ffec, 0x1ffed, 0x1ffee, 0x0fff4,
+    0x0fff3, 0x0fff0, 0x07ff7, 0x07ff6, 0x03ffa, 0x01ffa, 0x01ff9, 0x00ffa,
+    0x00ff8, 0x007f9, 0x003fb, 0x001fc, 0x001fa, 0x000fb, 0x0007c, 0x0003c,
+    0x0001c, 0x0000c, 0x00005, 0x00001, 0x00000, 0x00004, 0x0000d, 0x0001d,
+    0x0003d, 0x000fa, 0x000fc, 0x001fb, 0x003fa, 0x007f8, 0x007fa, 0x007fb,
+    0x00ff9, 0x00ffb, 0x01ff8, 0x01ffb, 0x03ff8, 0x03ff9, 0x0fff1, 0x0fff2,
+    0x1ffea, 0x1ffeb, 0x3ffe1, 0x3ffe2, 0x3ffea, 0x3ffe3, 0x3ffe6, 0x3ffe7,
+    0x3ffeb, 0xfffe6, 0x7ffe2, 0xfffe7, 0xfffe8, 0xfffe9, 0xfffea, 0xfffeb,
+    0xfffec, 0x7ffe3, 0xfffed, 0xfffee, 0xfffef, 0xffff0, 0x7ffe4, 0xffff1,
+    0x3ffec, 0xffff2, 0xffff3, 0x7ffe5, 0x7ffe6, 0xffff4, 0xffff5, 0xffff6,
+    0xffff7, 0xffff8, 0xffff9, 0xffffa, 0xffffb, 0xffffc, 0xffffd, 0xffffe,
+    0xfffff,
+};
+
+static const uint8_t t_huffman_env_bal_1_5dB_bits[49] = {
+    16, 16, 16, 16, 16, 16, 16, 16,
+    16, 16, 16, 16, 16, 16, 16, 16,
+    16, 16, 12, 11,  9,  7,  5,  3,
+     1,  2,  4,  6,  8, 11, 12, 15,
+    16, 16, 16, 16, 16, 16, 16, 17,
+    17, 17, 17, 17, 17, 17, 17, 17,
+    17,
+};
+
+static const uint32_t t_huffman_env_bal_1_5dB_codes[49] = {
+    0x0ffe4, 0x0ffe5, 0x0ffe6, 0x0ffe7, 0x0ffe8, 0x0ffe9, 0x0ffea, 0x0ffeb,
+    0x0ffec, 0x0ffed, 0x0ffee, 0x0ffef, 0x0fff0, 0x0fff1, 0x0fff2, 0x0fff3,
+    0x0fff4, 0x0ffe2, 0x00ffc, 0x007fc, 0x001fe, 0x0007e, 0x0001e, 0x00006,
+    0x00000, 0x00002, 0x0000e, 0x0003e, 0x000fe, 0x007fd, 0x00ffd, 0x07ff0,
+    0x0ffe3, 0x0fff5, 0x0fff6, 0x0fff7, 0x0fff8, 0x0fff9, 0x0fffa, 0x1fff6,
+    0x1fff7, 0x1fff8, 0x1fff9, 0x1fffa, 0x1fffb, 0x1fffc, 0x1fffd, 0x1fffe,
+    0x1ffff,
+};
+
+static const uint8_t f_huffman_env_bal_1_5dB_bits[49] = {
+    18, 18, 18, 18, 18, 18, 18, 18,
+    18, 18, 18, 18, 18, 18, 18, 16,
+    17, 14, 11, 11,  8,  7,  4,  2,
+     1,  3,  5,  6,  9, 11, 12, 15,
+    16, 18, 18, 18, 18, 18, 18, 18,
+    18, 18, 18, 18, 18, 18, 18, 19,
+    19,
+};
+
+static const uint32_t f_huffman_env_bal_1_5dB_codes[49] = {
+    0x3ffe2, 0x3ffe3, 0x3ffe4, 0x3ffe5, 0x3ffe6, 0x3ffe7, 0x3ffe8, 0x3ffe9,
+    0x3ffea, 0x3ffeb, 0x3ffec, 0x3ffed, 0x3ffee, 0x3ffef, 0x3fff0, 0x0fff7,
+    0x1fff0, 0x03ffc, 0x007fe, 0x007fc, 0x000fe, 0x0007e, 0x0000e, 0x00002,
+    0x00000, 0x00006, 0x0001e, 0x0003e, 0x001fe, 0x007fd, 0x00ffe, 0x07ffa,
+    0x0fff6, 0x3fff1, 0x3fff2, 0x3fff3, 0x3fff4, 0x3fff5, 0x3fff6, 0x3fff7,
+    0x3fff8, 0x3fff9, 0x3fffa, 0x3fffb, 0x3fffc, 0x3fffd, 0x3fffe, 0x7fffe,
+    0x7ffff,
+};
+
+static const uint8_t t_huffman_env_3_0dB_bits[63] = {
+    18, 18, 19, 19, 19, 19, 19, 19,
+    19, 19, 19, 19, 19, 19, 19, 19,
+    19, 17, 16, 16, 16, 14, 14, 14,
+    13, 12, 11,  8,  6,  4,  2,  1,
+     3,  5,  7,  9, 11, 13, 14, 14,
+    15, 16, 17, 18, 19, 19, 19, 19,
+    19, 19, 19, 19, 19, 19, 19, 19,
+    19, 19, 19, 19, 19, 19, 19,
+};
+
+static const uint32_t t_huffman_env_3_0dB_codes[63] = {
+    0x3ffed, 0x3ffee, 0x7ffde, 0x7ffdf, 0x7ffe0, 0x7ffe1, 0x7ffe2, 0x7ffe3,
+    0x7ffe4, 0x7ffe5, 0x7ffe6, 0x7ffe7, 0x7ffe8, 0x7ffe9, 0x7ffea, 0x7ffeb,
+    0x7ffec, 0x1fff4, 0x0fff7, 0x0fff9, 0x0fff8, 0x03ffb, 0x03ffa, 0x03ff8,
+    0x01ffa, 0x00ffc, 0x007fc, 0x000fe, 0x0003e, 0x0000e, 0x00002, 0x00000,
+    0x00006, 0x0001e, 0x0007e, 0x001fe, 0x007fd, 0x01ffb, 0x03ff9, 0x03ffc,
+    0x07ffa, 0x0fff6, 0x1fff5, 0x3ffec, 0x7ffed, 0x7ffee, 0x7ffef, 0x7fff0,
+    0x7fff1, 0x7fff2, 0x7fff3, 0x7fff4, 0x7fff5, 0x7fff6, 0x7fff7, 0x7fff8,
+    0x7fff9, 0x7fffa, 0x7fffb, 0x7fffc, 0x7fffd, 0x7fffe, 0x7ffff,
+};
+
+static const uint8_t f_huffman_env_3_0dB_bits[63] = {
+    20, 20, 20, 20, 20, 20, 20, 18,
+    19, 19, 19, 19, 18, 18, 20, 19,
+    17, 18, 17, 16, 16, 15, 14, 12,
+    11, 10,  9,  8,  6,  4,  2,  1,
+     3,  5,  8,  9, 10, 11, 12, 13,
+    14, 15, 15, 16, 16, 17, 17, 18,
+    18, 18, 20, 19, 19, 19, 20, 19,
+    19, 20, 20, 20, 20, 20, 20,
+};
+
+static const uint32_t f_huffman_env_3_0dB_codes[63] = {
+    0xffff0, 0xffff1, 0xffff2, 0xffff3, 0xffff4, 0xffff5, 0xffff6, 0x3fff3,
+    0x7fff5, 0x7ffee, 0x7ffef, 0x7fff6, 0x3fff4, 0x3fff2, 0xffff7, 0x7fff0,
+    0x1fff5, 0x3fff0, 0x1fff4, 0x0fff7, 0x0fff6, 0x07ff8, 0x03ffb, 0x00ffd,
+    0x007fd, 0x003fd, 0x001fd, 0x000fd, 0x0003e, 0x0000e, 0x00002, 0x00000,
+    0x00006, 0x0001e, 0x000fc, 0x001fc, 0x003fc, 0x007fc, 0x00ffc, 0x01ffc,
+    0x03ffa, 0x07ff9, 0x07ffa, 0x0fff8, 0x0fff9, 0x1fff6, 0x1fff7, 0x3fff5,
+    0x3fff6, 0x3fff1, 0xffff8, 0x7fff1, 0x7fff2, 0x7fff3, 0xffff9, 0x7fff7,
+    0x7fff4, 0xffffa, 0xffffb, 0xffffc, 0xffffd, 0xffffe, 0xfffff,
+};
+
+static const uint8_t t_huffman_env_bal_3_0dB_bits[25] = {
+    13, 13, 13, 13, 13, 13, 13, 12,
+     8,  7,  4,  3,  1,  2,  5,  6,
+     9, 13, 13, 13, 13, 13, 13, 14,
+    14,
+};
+
+static const uint16_t t_huffman_env_bal_3_0dB_codes[25] = {
+    0x1ff2, 0x1ff3, 0x1ff4, 0x1ff5, 0x1ff6, 0x1ff7, 0x1ff8, 0x0ff8,
+    0x00fe, 0x007e, 0x000e, 0x0006, 0x0000, 0x0002, 0x001e, 0x003e,
+    0x01fe, 0x1ff9, 0x1ffa, 0x1ffb, 0x1ffc, 0x1ffd, 0x1ffe, 0x3ffe,
+    0x3fff,
+};
+
+static const uint8_t f_huffman_env_bal_3_0dB_bits[25] = {
+    13, 13, 13, 13, 13, 14, 14, 11,
+     8,  7,  4,  2,  1,  3,  5,  6,
+     9, 12, 13, 14, 14, 14, 14, 14,
+    14,
+};
+
+static const uint16_t f_huffman_env_bal_3_0dB_codes[25] = {
+    0x1ff7, 0x1ff8, 0x1ff9, 0x1ffa, 0x1ffb, 0x3ff8, 0x3ff9, 0x07fc,
+    0x00fe, 0x007e, 0x000e, 0x0002, 0x0000, 0x0006, 0x001e, 0x003e,
+    0x01fe, 0x0ffa, 0x1ff6, 0x3ffa, 0x3ffb, 0x3ffc, 0x3ffd, 0x3ffe,
+    0x3fff,
+};
+
+static const uint8_t t_huffman_noise_3_0dB_bits[63] = {
+    13, 13, 13, 13, 13, 13, 13, 13,
+    13, 13, 13, 13, 13, 13, 13, 13,
+    13, 13, 13, 13, 13, 13, 13, 13,
+    13, 13, 11,  8,  6,  4,  3,  1,
+     2,  5,  8, 10, 13, 13, 13, 13,
+    13, 13, 13, 13, 13, 13, 13, 13,
+    13, 13, 13, 13, 13, 13, 13, 13,
+    13, 13, 13, 13, 13, 14, 14,
+};
+
+static const uint16_t t_huffman_noise_3_0dB_codes[63] = {
+    0x1fce, 0x1fcf, 0x1fd0, 0x1fd1, 0x1fd2, 0x1fd3, 0x1fd4, 0x1fd5,
+    0x1fd6, 0x1fd7, 0x1fd8, 0x1fd9, 0x1fda, 0x1fdb, 0x1fdc, 0x1fdd,
+    0x1fde, 0x1fdf, 0x1fe0, 0x1fe1, 0x1fe2, 0x1fe3, 0x1fe4, 0x1fe5,
+    0x1fe6, 0x1fe7, 0x07f2, 0x00fd, 0x003e, 0x000e, 0x0006, 0x0000,
+    0x0002, 0x001e, 0x00fc, 0x03f8, 0x1fcc, 0x1fe8, 0x1fe9, 0x1fea,
+    0x1feb, 0x1fec, 0x1fcd, 0x1fed, 0x1fee, 0x1fef, 0x1ff0, 0x1ff1,
+    0x1ff2, 0x1ff3, 0x1ff4, 0x1ff5, 0x1ff6, 0x1ff7, 0x1ff8, 0x1ff9,
+    0x1ffa, 0x1ffb, 0x1ffc, 0x1ffd, 0x1ffe, 0x3ffe, 0x3fff,
+};
+
+static const uint8_t t_huffman_noise_bal_3_0dB_bits[25] = {
+    8, 8, 8, 8, 8, 8, 8, 8,
+    8, 8, 5, 2, 1, 3, 6, 8,
+    8, 8, 8, 8, 8, 8, 8, 8,
+    8,
+};
+
+static const uint8_t t_huffman_noise_bal_3_0dB_codes[25] = {
+    0xec, 0xed, 0xee, 0xef, 0xf0, 0xf1, 0xf2, 0xf3,
+    0xf4, 0xf5, 0x1c, 0x02, 0x00, 0x06, 0x3a, 0xf6,
+    0xf7, 0xf8, 0xf9, 0xfa, 0xfb, 0xfc, 0xfd, 0xfe,
+    0xff,
+};
+
+static const int8_t sbr_offset[6][16] = {
+    {-8, -7, -6, -5, -4, -3, -2, -1,  0,  1,  2,  3,  4,  5,  6,  7}, //             fs_sbr  = 16000 Hz
+    {-5, -4, -3, -2, -1,  0,  1,  2,  3,  4,  5,  6,  7,  9, 11, 13}, //             fs_sbr  = 22050 Hz
+    {-5, -3, -2, -1,  0,  1,  2,  3,  4,  5,  6,  7,  9, 11, 13, 16}, //             fs_sbr  = 24000 Hz
+    {-6, -4, -2, -1,  0,  1,  2,  3,  4,  5,  6,  7,  9, 11, 13, 16}, //             fs_sbr  = 32000 Hz
+    {-4, -2, -1,  0,  1,  2,  3,  4,  5,  6,  7,  9, 11, 13, 16, 20}, // 44100 Hz <= fs_sbr <= 64000 Hz
+    {-2, -1,  0,  1,  2,  3,  4,  5,  6,  7,  9, 11, 13, 16, 20, 24}, // 64000 Hz <  fs_sbr
+};
+
+/* First eight entries repeated at end to simplify SIMD implementations. */
+const DECLARE_ALIGNED(16, INTFLOAT, AAC_RENAME(ff_sbr_noise_table))[][2] = {
+{Q31(-0.99948153278296f), Q31(-0.59483417516607f)}, {Q31( 0.97113454393991f), Q31(-0.67528515225647f)},
+{Q31( 0.14130051758487f), Q31(-0.95090983575689f)}, {Q31(-0.47005496701697f), Q31(-0.37340549728647f)},
+{Q31( 0.80705063769351f), Q31( 0.29653668284408f)}, {Q31(-0.38981478896926f), Q31( 0.89572605717087f)},
+{Q31(-0.01053049862020f), Q31(-0.66959058036166f)}, {Q31(-0.91266367957293f), Q31(-0.11522938140034f)},
+{Q31( 0.54840422910309f), Q31( 0.75221367176302f)}, {Q31( 0.40009252867955f), Q31(-0.98929400334421f)},
+{Q31(-0.99867974711855f), Q31(-0.88147068645358f)}, {Q31(-0.95531076805040f), Q31( 0.90908757154593f)},
+{Q31(-0.45725933317144f), Q31(-0.56716323646760f)}, {Q31(-0.72929675029275f), Q31(-0.98008272727324f)},
+{Q31( 0.75622801399036f), Q31( 0.20950329995549f)}, {Q31( 0.07069442601050f), Q31(-0.78247898470706f)},
+{Q31( 0.74496252926055f), Q31(-0.91169004445807f)}, {Q31(-0.96440182703856f), Q31(-0.94739918296622f)},
+{Q31( 0.30424629369539f), Q31(-0.49438267012479f)}, {Q31( 0.66565033746925f), Q31( 0.64652935542491f)},
+{Q31( 0.91697008020594f), Q31( 0.17514097332009f)}, {Q31(-0.70774918760427f), Q31( 0.52548653416543f)},
+{Q31(-0.70051415345560f), Q31(-0.45340028808763f)}, {Q31(-0.99496513054797f), Q31(-0.90071908066973f)},
+{Q31( 0.98164490790123f), Q31(-0.77463155528697f)}, {Q31(-0.54671580548181f), Q31(-0.02570928536004f)},
+{Q31(-0.01689629065389f), Q31( 0.00287506445732f)}, {Q31(-0.86110349531986f), Q31( 0.42548583726477f)},
+{Q31(-0.98892980586032f), Q31(-0.87881132267556f)}, {Q31( 0.51756627678691f), Q31( 0.66926784710139f)},
+{Q31(-0.99635026409640f), Q31(-0.58107730574765f)}, {Q31(-0.99969370862163f), Q31( 0.98369989360250f)},
+{Q31( 0.55266258627194f), Q31( 0.59449057465591f)}, {Q31( 0.34581177741673f), Q31( 0.94879421061866f)},
+{Q31( 0.62664209577999f), Q31(-0.74402970906471f)}, {Q31(-0.77149701404973f), Q31(-0.33883658042801f)},
+{Q31(-0.91592244254432f), Q31( 0.03687901376713f)}, {Q31(-0.76285492357887f), Q31(-0.91371867919124f)},
+{Q31( 0.79788337195331f), Q31(-0.93180971199849f)}, {Q31( 0.54473080610200f), Q31(-0.11919206037186f)},
+{Q31(-0.85639281671058f), Q31( 0.42429854760451f)}, {Q31(-0.92882402971423f), Q31( 0.27871809078609f)},
+{Q31(-0.11708371046774f), Q31(-0.99800843444966f)}, {Q31( 0.21356749817493f), Q31(-0.90716295627033f)},
+{Q31(-0.76191692573909f), Q31( 0.99768118356265f)}, {Q31( 0.98111043100884f), Q31(-0.95854459734407f)},
+{Q31(-0.85913269895572f), Q31( 0.95766566168880f)}, {Q31(-0.93307242253692f), Q31( 0.49431757696466f)},
+{Q31( 0.30485754879632f), Q31(-0.70540034357529f)}, {Q31( 0.85289650925190f), Q31( 0.46766131791044f)},
+{Q31( 0.91328082618125f), Q31(-0.99839597361769f)}, {Q31(-0.05890199924154f), Q31( 0.70741827819497f)},
+{Q31( 0.28398686150148f), Q31( 0.34633555702188f)}, {Q31( 0.95258164539612f), Q31(-0.54893416026939f)},
+{Q31(-0.78566324168507f), Q31(-0.75568541079691f)}, {Q31(-0.95789495447877f), Q31(-0.20423194696966f)},
+{Q31( 0.82411158711197f), Q31( 0.96654618432562f)}, {Q31(-0.65185446735885f), Q31(-0.88734990773289f)},
+{Q31(-0.93643603134666f), Q31( 0.99870790442385f)}, {Q31( 0.91427159529618f), Q31(-0.98290505544444f)},
+{Q31(-0.70395684036886f), Q31( 0.58796798221039f)}, {Q31( 0.00563771969365f), Q31( 0.61768196727244f)},
+{Q31( 0.89065051931895f), Q31( 0.52783352697585f)}, {Q31(-0.68683707712762f), Q31( 0.80806944710339f)},
+{Q31( 0.72165342518718f), Q31(-0.69259857349564f)}, {Q31(-0.62928247730667f), Q31( 0.13627037407335f)},
+{Q31( 0.29938434065514f), Q31(-0.46051329682246f)}, {Q31(-0.91781958879280f), Q31(-0.74012716684186f)},
+{Q31( 0.99298717043688f), Q31( 0.40816610075661f)}, {Q31( 0.82368298622748f), Q31(-0.74036047190173f)},
+{Q31(-0.98512833386833f), Q31(-0.99972330709594f)}, {Q31(-0.95915368242257f), Q31(-0.99237800466040f)},
+{Q31(-0.21411126572790f), Q31(-0.93424819052545f)}, {Q31(-0.68821476106884f), Q31(-0.26892306315457f)},
+{Q31( 0.91851997982317f), Q31( 0.09358228901785f)}, {Q31(-0.96062769559127f), Q31( 0.36099095133739f)},
+{Q31( 0.51646184922287f), Q31(-0.71373332873917f)}, {Q31( 0.61130721139669f), Q31( 0.46950141175917f)},
+{Q31( 0.47336129371299f), Q31(-0.27333178296162f)}, {Q31( 0.90998308703519f), Q31( 0.96715662938132f)},
+{Q31( 0.44844799194357f), Q31( 0.99211574628306f)}, {Q31( 0.66614891079092f), Q31( 0.96590176169121f)},
+{Q31( 0.74922239129237f), Q31(-0.89879858826087f)}, {Q31(-0.99571588506485f), Q31( 0.52785521494349f)},
+{Q31( 0.97401082477563f), Q31(-0.16855870075190f)}, {Q31( 0.72683747733879f), Q31(-0.48060774432251f)},
+{Q31( 0.95432193457128f), Q31( 0.68849603408441f)}, {Q31(-0.72962208425191f), Q31(-0.76608443420917f)},
+{Q31(-0.85359479233537f), Q31( 0.88738125901579f)}, {Q31(-0.81412430338535f), Q31(-0.97480768049637f)},
+{Q31(-0.87930772356786f), Q31( 0.74748307690436f)}, {Q31(-0.71573331064977f), Q31(-0.98570608178923f)},
+{Q31( 0.83524300028228f), Q31( 0.83702537075163f)}, {Q31(-0.48086065601423f), Q31(-0.98848504923531f)},
+{Q31( 0.97139128574778f), Q31( 0.80093621198236f)}, {Q31( 0.51992825347895f), Q31( 0.80247631400510f)},
+{Q31(-0.00848591195325f), Q31(-0.76670128000486f)}, {Q31(-0.70294374303036f), Q31( 0.55359910445577f)},
+{Q31(-0.95894428168140f), Q31(-0.43265504344783f)}, {Q31( 0.97079252950321f), Q31( 0.09325857238682f)},
+{Q31(-0.92404293670797f), Q31( 0.85507704027855f)}, {Q31(-0.69506469500450f), Q31( 0.98633412625459f)},
+{Q31( 0.26559203620024f), Q31( 0.73314307966524f)}, {Q31( 0.28038443336943f), Q31( 0.14537913654427f)},
+{Q31(-0.74138124825523f), Q31( 0.99310339807762f)}, {Q31(-0.01752795995444f), Q31(-0.82616635284178f)},
+{Q31(-0.55126773094930f), Q31(-0.98898543862153f)}, {Q31( 0.97960898850996f), Q31(-0.94021446752851f)},
+{Q31(-0.99196309146936f), Q31( 0.67019017358456f)}, {Q31(-0.67684928085260f), Q31( 0.12631491649378f)},
+{Q31( 0.09140039465500f), Q31(-0.20537731453108f)}, {Q31(-0.71658965751996f), Q31(-0.97788200391224f)},
+{Q31( 0.81014640078925f), Q31( 0.53722648362443f)}, {Q31( 0.40616991671205f), Q31(-0.26469008598449f)},
+{Q31(-0.67680188682972f), Q31( 0.94502052337695f)}, {Q31( 0.86849774348749f), Q31(-0.18333598647899f)},
+{Q31(-0.99500381284851f), Q31(-0.02634122068550f)}, {Q31( 0.84329189340667f), Q31( 0.10406957462213f)},
+{Q31(-0.09215968531446f), Q31( 0.69540012101253f)}, {Q31( 0.99956173327206f), Q31(-0.12358542001404f)},
+{Q31(-0.79732779473535f), Q31(-0.91582524736159f)}, {Q31( 0.96349973642406f), Q31( 0.96640458041000f)},
+{Q31(-0.79942778496547f), Q31( 0.64323902822857f)}, {Q31(-0.11566039853896f), Q31( 0.28587846253726f)},
+{Q31(-0.39922954514662f), Q31( 0.94129601616966f)}, {Q31( 0.99089197565987f), Q31(-0.92062625581587f)},
+{Q31( 0.28631285179909f), Q31(-0.91035047143603f)}, {Q31(-0.83302725605608f), Q31(-0.67330410892084f)},
+{Q31( 0.95404443402072f), Q31( 0.49162765398743f)}, {Q31(-0.06449863579434f), Q31( 0.03250560813135f)},
+{Q31(-0.99575054486311f), Q31( 0.42389784469507f)}, {Q31(-0.65501142790847f), Q31( 0.82546114655624f)},
+{Q31(-0.81254441908887f), Q31(-0.51627234660629f)}, {Q31(-0.99646369485481f), Q31( 0.84490533520752f)},
+{Q31( 0.00287840603348f), Q31( 0.64768261158166f)}, {Q31( 0.70176989408455f), Q31(-0.20453028573322f)},
+{Q31( 0.96361882270190f), Q31( 0.40706967140989f)}, {Q31(-0.68883758192426f), Q31( 0.91338958840772f)},
+{Q31(-0.34875585502238f), Q31( 0.71472290693300f)}, {Q31( 0.91980081243087f), Q31( 0.66507455644919f)},
+{Q31(-0.99009048343881f), Q31( 0.85868021604848f)}, {Q31( 0.68865791458395f), Q31( 0.55660316809678f)},
+{Q31(-0.99484402129368f), Q31(-0.20052559254934f)}, {Q31( 0.94214511408023f), Q31(-0.99696425367461f)},
+{Q31(-0.67414626793544f), Q31( 0.49548221180078f)}, {Q31(-0.47339353684664f), Q31(-0.85904328834047f)},
+{Q31( 0.14323651387360f), Q31(-0.94145598222488f)}, {Q31(-0.29268293575672f), Q31( 0.05759224927952f)},
+{Q31( 0.43793861458754f), Q31(-0.78904969892724f)}, {Q31(-0.36345126374441f), Q31( 0.64874435357162f)},
+{Q31(-0.08750604656825f), Q31( 0.97686944362527f)}, {Q31(-0.96495267812511f), Q31(-0.53960305946511f)},
+{Q31( 0.55526940659947f), Q31( 0.78891523734774f)}, {Q31( 0.73538215752630f), Q31( 0.96452072373404f)},
+{Q31(-0.30889773919437f), Q31(-0.80664389776860f)}, {Q31( 0.03574995626194f), Q31(-0.97325616900959f)},
+{Q31( 0.98720684660488f), Q31( 0.48409133691962f)}, {Q31(-0.81689296271203f), Q31(-0.90827703628298f)},
+{Q31( 0.67866860118215f), Q31( 0.81284503870856f)}, {Q31(-0.15808569732583f), Q31( 0.85279555024382f)},
+{Q31( 0.80723395114371f), Q31(-0.24717418514605f)}, {Q31( 0.47788757329038f), Q31(-0.46333147839295f)},
+{Q31( 0.96367554763201f), Q31( 0.38486749303242f)}, {Q31(-0.99143875716818f), Q31(-0.24945277239809f)},
+{Q31( 0.83081876925833f), Q31(-0.94780851414763f)}, {Q31(-0.58753191905341f), Q31( 0.01290772389163f)},
+{Q31( 0.95538108220960f), Q31(-0.85557052096538f)}, {Q31(-0.96490920476211f), Q31(-0.64020970923102f)},
+{Q31(-0.97327101028521f), Q31( 0.12378128133110f)}, {Q31( 0.91400366022124f), Q31( 0.57972471346930f)},
+{Q31(-0.99925837363824f), Q31( 0.71084847864067f)}, {Q31(-0.86875903507313f), Q31(-0.20291699203564f)},
+{Q31(-0.26240034795124f), Q31(-0.68264554369108f)}, {Q31(-0.24664412953388f), Q31(-0.87642273115183f)},
+{Q31( 0.02416275806869f), Q31( 0.27192914288905f)}, {Q31( 0.82068619590515f), Q31(-0.85087787994476f)},
+{Q31( 0.88547373760759f), Q31(-0.89636802901469f)}, {Q31(-0.18173078152226f), Q31(-0.26152145156800f)},
+{Q31( 0.09355476558534f), Q31( 0.54845123045604f)}, {Q31(-0.54668414224090f), Q31( 0.95980774020221f)},
+{Q31( 0.37050990604091f), Q31(-0.59910140383171f)}, {Q31(-0.70373594262891f), Q31( 0.91227665827081f)},
+{Q31(-0.34600785879594f), Q31(-0.99441426144200f)}, {Q31(-0.68774481731008f), Q31(-0.30238837956299f)},
+{Q31(-0.26843291251234f), Q31( 0.83115668004362f)}, {Q31( 0.49072334613242f), Q31(-0.45359708737775f)},
+{Q31( 0.38975993093975f), Q31( 0.95515358099121f)}, {Q31(-0.97757125224150f), Q31( 0.05305894580606f)},
+{Q31(-0.17325552859616f), Q31(-0.92770672250494f)}, {Q31( 0.99948035025744f), Q31( 0.58285545563426f)},
+{Q31(-0.64946246527458f), Q31( 0.68645507104960f)}, {Q31(-0.12016920576437f), Q31(-0.57147322153312f)},
+{Q31(-0.58947456517751f), Q31(-0.34847132454388f)}, {Q31(-0.41815140454465f), Q31( 0.16276422358861f)},
+{Q31( 0.99885650204884f), Q31( 0.11136095490444f)}, {Q31(-0.56649614128386f), Q31(-0.90494866361587f)},
+{Q31( 0.94138021032330f), Q31( 0.35281916733018f)}, {Q31(-0.75725076534641f), Q31( 0.53650549640587f)},
+{Q31( 0.20541973692630f), Q31(-0.94435144369918f)}, {Q31( 0.99980371023351f), Q31( 0.79835913565599f)},
+{Q31( 0.29078277605775f), Q31( 0.35393777921520f)}, {Q31(-0.62858772103030f), Q31( 0.38765693387102f)},
+{Q31( 0.43440904467688f), Q31(-0.98546330463232f)}, {Q31(-0.98298583762390f), Q31( 0.21021524625209f)},
+{Q31( 0.19513029146934f), Q31(-0.94239832251867f)}, {Q31(-0.95476662400101f), Q31( 0.98364554179143f)},
+{Q31( 0.93379635304810f), Q31(-0.70881994583682f)}, {Q31(-0.85235410573336f), Q31(-0.08342347966410f)},
+{Q31(-0.86425093011245f), Q31(-0.45795025029466f)}, {Q31( 0.38879779059045f), Q31( 0.97274429344593f)},
+{Q31( 0.92045124735495f), Q31(-0.62433652524220f)}, {Q31( 0.89162532251878f), Q31( 0.54950955570563f)},
+{Q31(-0.36834336949252f), Q31( 0.96458298020975f)}, {Q31( 0.93891760988045f), Q31(-0.89968353740388f)},
+{Q31( 0.99267657565094f), Q31(-0.03757034316958f)}, {Q31(-0.94063471614176f), Q31( 0.41332338538963f)},
+{Q31( 0.99740224117019f), Q31(-0.16830494996370f)}, {Q31(-0.35899413170555f), Q31(-0.46633226649613f)},
+{Q31( 0.05237237274947f), Q31(-0.25640361602661f)}, {Q31( 0.36703583957424f), Q31(-0.38653265641875f)},
+{Q31( 0.91653180367913f), Q31(-0.30587628726597f)}, {Q31( 0.69000803499316f), Q31( 0.90952171386132f)},
+{Q31(-0.38658751133527f), Q31( 0.99501571208985f)}, {Q31(-0.29250814029851f), Q31( 0.37444994344615f)},
+{Q31(-0.60182204677608f), Q31( 0.86779651036123f)}, {Q31(-0.97418588163217f), Q31( 0.96468523666475f)},
+{Q31( 0.88461574003963f), Q31( 0.57508405276414f)}, {Q31( 0.05198933055162f), Q31( 0.21269661669964f)},
+{Q31(-0.53499621979720f), Q31( 0.97241553731237f)}, {Q31(-0.49429560226497f), Q31( 0.98183865291903f)},
+{Q31(-0.98935142339139f), Q31(-0.40249159006933f)}, {Q31(-0.98081380091130f), Q31(-0.72856895534041f)},
+{Q31(-0.27338148835532f), Q31( 0.99950922447209f)}, {Q31( 0.06310802338302f), Q31(-0.54539587529618f)},
+{Q31(-0.20461677199539f), Q31(-0.14209977628489f)}, {Q31( 0.66223843141647f), Q31( 0.72528579940326f)},
+{Q31(-0.84764345483665f), Q31( 0.02372316801261f)}, {Q31(-0.89039863483811f), Q31( 0.88866581484602f)},
+{Q31( 0.95903308477986f), Q31( 0.76744927173873f)}, {Q31( 0.73504123909879f), Q31(-0.03747203173192f)},
+{Q31(-0.31744434966056f), Q31(-0.36834111883652f)}, {Q31(-0.34110827591623f), Q31( 0.40211222807691f)},
+{Q31( 0.47803883714199f), Q31(-0.39423219786288f)}, {Q31( 0.98299195879514f), Q31( 0.01989791390047f)},
+{Q31(-0.30963073129751f), Q31(-0.18076720599336f)}, {Q31( 0.99992588229018f), Q31(-0.26281872094289f)},
+{Q31(-0.93149731080767f), Q31(-0.98313162570490f)}, {Q31( 0.99923472302773f), Q31(-0.80142993767554f)},
+{Q31(-0.26024169633417f), Q31(-0.75999759855752f)}, {Q31(-0.35712514743563f), Q31( 0.19298963768574f)},
+{Q31(-0.99899084509530f), Q31( 0.74645156992493f)}, {Q31( 0.86557171579452f), Q31( 0.55593866696299f)},
+{Q31( 0.33408042438752f), Q31( 0.86185953874709f)}, {Q31( 0.99010736374716f), Q31( 0.04602397576623f)},
+{Q31(-0.66694269691195f), Q31(-0.91643611810148f)}, {Q31( 0.64016792079480f), Q31( 0.15649530836856f)},
+{Q31( 0.99570534804836f), Q31( 0.45844586038111f)}, {Q31(-0.63431466947340f), Q31( 0.21079116459234f)},
+{Q31(-0.07706847005931f), Q31(-0.89581437101329f)}, {Q31( 0.98590090577724f), Q31( 0.88241721133981f)},
+{Q31( 0.80099335254678f), Q31(-0.36851896710853f)}, {Q31( 0.78368131392666f), Q31( 0.45506999802597f)},
+{Q31( 0.08707806671691f), Q31( 0.80938994918745f)}, {Q31(-0.86811883080712f), Q31( 0.39347308654705f)},
+{Q31(-0.39466529740375f), Q31(-0.66809432114456f)}, {Q31( 0.97875325649683f), Q31(-0.72467840967746f)},
+{Q31(-0.95038560288864f), Q31( 0.89563219587625f)}, {Q31( 0.17005239424212f), Q31( 0.54683053962658f)},
+{Q31(-0.76910792026848f), Q31(-0.96226617549298f)}, {Q31( 0.99743281016846f), Q31( 0.42697157037567f)},
+{Q31( 0.95437383549973f), Q31( 0.97002324109952f)}, {Q31( 0.99578905365569f), Q31(-0.54106826257356f)},
+{Q31( 0.28058259829990f), Q31(-0.85361420634036f)}, {Q31( 0.85256524470573f), Q31(-0.64567607735589f)},
+{Q31(-0.50608540105128f), Q31(-0.65846015480300f)}, {Q31(-0.97210735183243f), Q31(-0.23095213067791f)},
+{Q31( 0.95424048234441f), Q31(-0.99240147091219f)}, {Q31(-0.96926570524023f), Q31( 0.73775654896574f)},
+{Q31( 0.30872163214726f), Q31( 0.41514960556126f)}, {Q31(-0.24523839572639f), Q31( 0.63206633394807f)},
+{Q31(-0.33813265086024f), Q31(-0.38661779441897f)}, {Q31(-0.05826828420146f), Q31(-0.06940774188029f)},
+{Q31(-0.22898461455054f), Q31( 0.97054853316316f)}, {Q31(-0.18509915019881f), Q31( 0.47565762892084f)},
+{Q31(-0.10488238045009f), Q31(-0.87769947402394f)}, {Q31(-0.71886586182037f), Q31( 0.78030982480538f)},
+{Q31( 0.99793873738654f), Q31( 0.90041310491497f)}, {Q31( 0.57563307626120f), Q31(-0.91034337352097f)},
+{Q31( 0.28909646383717f), Q31( 0.96307783970534f)}, {Q31( 0.42188998312520f), Q31( 0.48148651230437f)},
+{Q31( 0.93335049681047f), Q31(-0.43537023883588f)}, {Q31(-0.97087374418267f), Q31( 0.86636445711364f)},
+{Q31( 0.36722871286923f), Q31( 0.65291654172961f)}, {Q31(-0.81093025665696f), Q31( 0.08778370229363f)},
+{Q31(-0.26240603062237f), Q31(-0.92774095379098f)}, {Q31( 0.83996497984604f), Q31( 0.55839849139647f)},
+{Q31(-0.99909615720225f), Q31(-0.96024605713970f)}, {Q31( 0.74649464155061f), Q31( 0.12144893606462f)},
+{Q31(-0.74774595569805f), Q31(-0.26898062008959f)}, {Q31( 0.95781667469567f), Q31(-0.79047927052628f)},
+{Q31( 0.95472308713099f), Q31(-0.08588776019550f)}, {Q31( 0.48708332746299f), Q31( 0.99999041579432f)},
+{Q31( 0.46332038247497f), Q31( 0.10964126185063f)}, {Q31(-0.76497004940162f), Q31( 0.89210929242238f)},
+{Q31( 0.57397389364339f), Q31( 0.35289703373760f)}, {Q31( 0.75374316974495f), Q31( 0.96705214651335f)},
+{Q31(-0.59174397685714f), Q31(-0.89405370422752f)}, {Q31( 0.75087906691890f), Q31(-0.29612672982396f)},
+{Q31(-0.98607857336230f), Q31( 0.25034911730023f)}, {Q31(-0.40761056640505f), Q31(-0.90045573444695f)},
+{Q31( 0.66929266740477f), Q31( 0.98629493401748f)}, {Q31(-0.97463695257310f), Q31(-0.00190223301301f)},
+{Q31( 0.90145509409859f), Q31( 0.99781390365446f)}, {Q31(-0.87259289048043f), Q31( 0.99233587353666f)},
+{Q31(-0.91529461447692f), Q31(-0.15698707534206f)}, {Q31(-0.03305738840705f), Q31(-0.37205262859764f)},
+{Q31( 0.07223051368337f), Q31(-0.88805001733626f)}, {Q31( 0.99498012188353f), Q31( 0.97094358113387f)},
+{Q31(-0.74904939500519f), Q31( 0.99985483641521f)}, {Q31( 0.04585228574211f), Q31( 0.99812337444082f)},
+{Q31(-0.89054954257993f), Q31(-0.31791913188064f)}, {Q31(-0.83782144651251f), Q31( 0.97637632547466f)},
+{Q31( 0.33454804933804f), Q31(-0.86231516800408f)}, {Q31(-0.99707579362824f), Q31( 0.93237990079441f)},
+{Q31(-0.22827527843994f), Q31( 0.18874759397997f)}, {Q31( 0.67248046289143f), Q31(-0.03646211390569f)},
+{Q31(-0.05146538187944f), Q31(-0.92599700120679f)}, {Q31( 0.99947295749905f), Q31( 0.93625229707912f)},
+{Q31( 0.66951124390363f), Q31( 0.98905825623893f)}, {Q31(-0.99602956559179f), Q31(-0.44654715757688f)},
+{Q31( 0.82104905483590f), Q31( 0.99540741724928f)}, {Q31( 0.99186510988782f), Q31( 0.72023001312947f)},
+{Q31(-0.65284592392918f), Q31( 0.52186723253637f)}, {Q31( 0.93885443798188f), Q31(-0.74895312615259f)},
+{Q31( 0.96735248738388f), Q31( 0.90891816978629f)}, {Q31(-0.22225968841114f), Q31( 0.57124029781228f)},
+{Q31(-0.44132783753414f), Q31(-0.92688840659280f)}, {Q31(-0.85694974219574f), Q31( 0.88844532719844f)},
+{Q31( 0.91783042091762f), Q31(-0.46356892383970f)}, {Q31( 0.72556974415690f), Q31(-0.99899555770747f)},
+{Q31(-0.99711581834508f), Q31( 0.58211560180426f)}, {Q31( 0.77638976371966f), Q31( 0.94321834873819f)},
+{Q31( 0.07717324253925f), Q31( 0.58638399856595f)}, {Q31(-0.56049829194163f), Q31( 0.82522301569036f)},
+{Q31( 0.98398893639988f), Q31( 0.39467440420569f)}, {Q31( 0.47546946844938f), Q31( 0.68613044836811f)},
+{Q31( 0.65675089314631f), Q31( 0.18331637134880f)}, {Q31( 0.03273375457980f), Q31(-0.74933109564108f)},
+{Q31(-0.38684144784738f), Q31( 0.51337349030406f)}, {Q31(-0.97346267944545f), Q31(-0.96549364384098f)},
+{Q31(-0.53282156061942f), Q31(-0.91423265091354f)}, {Q31( 0.99817310731176f), Q31( 0.61133572482148f)},
+{Q31(-0.50254500772635f), Q31(-0.88829338134294f)}, {Q31( 0.01995873238855f), Q31( 0.85223515096765f)},
+{Q31( 0.99930381973804f), Q31( 0.94578896296649f)}, {Q31( 0.82907767600783f), Q31(-0.06323442598128f)},
+{Q31(-0.58660709669728f), Q31( 0.96840773806582f)}, {Q31(-0.17573736667267f), Q31(-0.48166920859485f)},
+{Q31( 0.83434292401346f), Q31(-0.13023450646997f)}, {Q31( 0.05946491307025f), Q31( 0.20511047074866f)},
+{Q31( 0.81505484574602f), Q31(-0.94685947861369f)}, {Q31(-0.44976380954860f), Q31( 0.40894572671545f)},
+{Q31(-0.89746474625671f), Q31( 0.99846578838537f)}, {Q31( 0.39677256130792f), Q31(-0.74854668609359f)},
+{Q31(-0.07588948563079f), Q31( 0.74096214084170f)}, {Q31( 0.76343198951445f), Q31( 0.41746629422634f)},
+{Q31(-0.74490104699626f), Q31( 0.94725911744610f)}, {Q31( 0.64880119792759f), Q31( 0.41336660830571f)},
+{Q31( 0.62319537462542f), Q31(-0.93098313552599f)}, {Q31( 0.42215817594807f), Q31(-0.07712787385208f)},
+{Q31( 0.02704554141885f), Q31(-0.05417518053666f)}, {Q31( 0.80001773566818f), Q31( 0.91542195141039f)},
+{Q31(-0.79351832348816f), Q31(-0.36208897989136f)}, {Q31( 0.63872359151636f), Q31( 0.08128252493444f)},
+{Q31( 0.52890520960295f), Q31( 0.60048872455592f)}, {Q31( 0.74238552914587f), Q31( 0.04491915291044f)},
+{Q31( 0.99096131449250f), Q31(-0.19451182854402f)}, {Q31(-0.80412329643109f), Q31(-0.88513818199457f)},
+{Q31(-0.64612616129736f), Q31( 0.72198674804544f)}, {Q31( 0.11657770663191f), Q31(-0.83662833815041f)},
+{Q31(-0.95053182488101f), Q31(-0.96939905138082f)}, {Q31(-0.62228872928622f), Q31( 0.82767262846661f)},
+{Q31( 0.03004475787316f), Q31(-0.99738896333384f)}, {Q31(-0.97987214341034f), Q31( 0.36526129686425f)},
+{Q31(-0.99986980746200f), Q31(-0.36021610299715f)}, {Q31( 0.89110648599879f), Q31(-0.97894250343044f)},
+{Q31( 0.10407960510582f), Q31( 0.77357793811619f)}, {Q31( 0.95964737821728f), Q31(-0.35435818285502f)},
+{Q31( 0.50843233159162f), Q31( 0.96107691266205f)}, {Q31( 0.17006334670615f), Q31(-0.76854025314829f)},
+{Q31( 0.25872675063360f), Q31( 0.99893303933816f)}, {Q31(-0.01115998681937f), Q31( 0.98496019742444f)},
+{Q31(-0.79598702973261f), Q31( 0.97138411318894f)}, {Q31(-0.99264708948101f), Q31(-0.99542822402536f)},
+{Q31(-0.99829663752818f), Q31( 0.01877138824311f)}, {Q31(-0.70801016548184f), Q31( 0.33680685948117f)},
+{Q31(-0.70467057786826f), Q31( 0.93272777501857f)}, {Q31( 0.99846021905254f), Q31(-0.98725746254433f)},
+{Q31(-0.63364968534650f), Q31(-0.16473594423746f)}, {Q31(-0.16258217500792f), Q31(-0.95939125400802f)},
+{Q31(-0.43645594360633f), Q31(-0.94805030113284f)}, {Q31(-0.99848471702976f), Q31( 0.96245166923809f)},
+{Q31(-0.16796458968998f), Q31(-0.98987511890470f)}, {Q31(-0.87979225745213f), Q31(-0.71725725041680f)},
+{Q31( 0.44183099021786f), Q31(-0.93568974498761f)}, {Q31( 0.93310180125532f), Q31(-0.99913308068246f)},
+{Q31(-0.93941931782002f), Q31(-0.56409379640356f)}, {Q31(-0.88590003188677f), Q31( 0.47624600491382f)},
+{Q31( 0.99971463703691f), Q31(-0.83889954253462f)}, {Q31(-0.75376385639978f), Q31( 0.00814643438625f)},
+{Q31( 0.93887685615875f), Q31(-0.11284528204636f)}, {Q31( 0.85126435782309f), Q31( 0.52349251543547f)},
+{Q31( 0.39701421446381f), Q31( 0.81779634174316f)}, {Q31(-0.37024464187437f), Q31(-0.87071656222959f)},
+{Q31(-0.36024828242896f), Q31( 0.34655735648287f)}, {Q31(-0.93388812549209f), Q31(-0.84476541096429f)},
+{Q31(-0.65298804552119f), Q31(-0.18439575450921f)}, {Q31( 0.11960319006843f), Q31( 0.99899346780168f)},
+{Q31( 0.94292565553160f), Q31( 0.83163906518293f)}, {Q31( 0.75081145286948f), Q31(-0.35533223142265f)},
+{Q31( 0.56721979748394f), Q31(-0.24076836414499f)}, {Q31( 0.46857766746029f), Q31(-0.30140233457198f)},
+{Q31( 0.97312313923635f), Q31(-0.99548191630031f)}, {Q31(-0.38299976567017f), Q31( 0.98516909715427f)},
+{Q31( 0.41025800019463f), Q31( 0.02116736935734f)}, {Q31( 0.09638062008048f), Q31( 0.04411984381457f)},
+{Q31(-0.85283249275397f), Q31( 0.91475563922421f)}, {Q31( 0.88866808958124f), Q31(-0.99735267083226f)},
+{Q31(-0.48202429536989f), Q31(-0.96805608884164f)}, {Q31( 0.27572582416567f), Q31( 0.58634753335832f)},
+{Q31(-0.65889129659168f), Q31( 0.58835634138583f)}, {Q31( 0.98838086953732f), Q31( 0.99994349600236f)},
+{Q31(-0.20651349620689f), Q31( 0.54593044066355f)}, {Q31(-0.62126416356920f), Q31(-0.59893681700392f)},
+{Q31( 0.20320105410437f), Q31(-0.86879180355289f)}, {Q31(-0.97790548600584f), Q31( 0.96290806999242f)},
+{Q31( 0.11112534735126f), Q31( 0.21484763313301f)}, {Q31(-0.41368337314182f), Q31( 0.28216837680365f)},
+{Q31( 0.24133038992960f), Q31( 0.51294362630238f)}, {Q31(-0.66393410674885f), Q31(-0.08249679629081f)},
+{Q31(-0.53697829178752f), Q31(-0.97649903936228f)}, {Q31(-0.97224737889348f), Q31( 0.22081333579837f)},
+{Q31( 0.87392477144549f), Q31(-0.12796173740361f)}, {Q31( 0.19050361015753f), Q31( 0.01602615387195f)},
+{Q31(-0.46353441212724f), Q31(-0.95249041539006f)}, {Q31(-0.07064096339021f), Q31(-0.94479803205886f)},
+{Q31(-0.92444085484466f), Q31(-0.10457590187436f)}, {Q31(-0.83822593578728f), Q31(-0.01695043208885f)},
+{Q31( 0.75214681811150f), Q31(-0.99955681042665f)}, {Q31(-0.42102998829339f), Q31( 0.99720941999394f)},
+{Q31(-0.72094786237696f), Q31(-0.35008961934255f)}, {Q31( 0.78843311019251f), Q31( 0.52851398958271f)},
+{Q31( 0.97394027897442f), Q31(-0.26695944086561f)}, {Q31( 0.99206463477946f), Q31(-0.57010120849429f)},
+{Q31( 0.76789609461795f), Q31(-0.76519356730966f)}, {Q31(-0.82002421836409f), Q31(-0.73530179553767f)},
+{Q31( 0.81924990025724f), Q31( 0.99698425250579f)}, {Q31(-0.26719850873357f), Q31( 0.68903369776193f)},
+{Q31(-0.43311260380975f), Q31( 0.85321815947490f)}, {Q31( 0.99194979673836f), Q31( 0.91876249766422f)},
+{Q31(-0.80692001248487f), Q31(-0.32627540663214f)}, {Q31( 0.43080003649976f), Q31(-0.21919095636638f)},
+{Q31( 0.67709491937357f), Q31(-0.95478075822906f)}, {Q31( 0.56151770568316f), Q31(-0.70693811747778f)},
+{Q31( 0.10831862810749f), Q31(-0.08628837174592f)}, {Q31( 0.91229417540436f), Q31(-0.65987351408410f)},
+{Q31(-0.48972893932274f), Q31( 0.56289246362686f)}, {Q31(-0.89033658689697f), Q31(-0.71656563987082f)},
+{Q31( 0.65269447475094f), Q31( 0.65916004833932f)}, {Q31( 0.67439478141121f), Q31(-0.81684380846796f)},
+{Q31(-0.47770832416973f), Q31(-0.16789556203025f)}, {Q31(-0.99715979260878f), Q31(-0.93565784007648f)},
+{Q31(-0.90889593602546f), Q31( 0.62034397054380f)}, {Q31(-0.06618622548177f), Q31(-0.23812217221359f)},
+{Q31( 0.99430266919728f), Q31( 0.18812555317553f)}, {Q31( 0.97686402381843f), Q31(-0.28664534366620f)},
+{Q31( 0.94813650221268f), Q31(-0.97506640027128f)}, {Q31(-0.95434497492853f), Q31(-0.79607978501983f)},
+{Q31(-0.49104783137150f), Q31( 0.32895214359663f)}, {Q31( 0.99881175120751f), Q31( 0.88993983831354f)},
+{Q31( 0.50449166760303f), Q31(-0.85995072408434f)}, {Q31( 0.47162891065108f), Q31(-0.18680204049569f)},
+{Q31(-0.62081581361840f), Q31( 0.75000676218956f)}, {Q31(-0.43867015250812f), Q31( 0.99998069244322f)},
+{Q31( 0.98630563232075f), Q31(-0.53578899600662f)}, {Q31(-0.61510362277374f), Q31(-0.89515019899997f)},
+{Q31(-0.03841517601843f), Q31(-0.69888815681179f)}, {Q31(-0.30102157304644f), Q31(-0.07667808922205f)},
+{Q31( 0.41881284182683f), Q31( 0.02188098922282f)}, {Q31(-0.86135454941237f), Q31( 0.98947480909359f)},
+{Q31( 0.67226861393788f), Q31(-0.13494389011014f)}, {Q31(-0.70737398842068f), Q31(-0.76547349325992f)},
+{Q31( 0.94044946687963f), Q31( 0.09026201157416f)}, {Q31(-0.82386352534327f), Q31( 0.08924768823676f)},
+{Q31(-0.32070666698656f), Q31( 0.50143421908753f)}, {Q31( 0.57593163224487f), Q31(-0.98966422921509f)},
+{Q31(-0.36326018419965f), Q31( 0.07440243123228f)}, {Q31( 0.99979044674350f), Q31(-0.14130287347405f)},
+{Q31(-0.92366023326932f), Q31(-0.97979298068180f)}, {Q31(-0.44607178518598f), Q31(-0.54233252016394f)},
+{Q31( 0.44226800932956f), Q31( 0.71326756742752f)}, {Q31( 0.03671907158312f), Q31( 0.63606389366675f)},
+{Q31( 0.52175424682195f), Q31(-0.85396826735705f)}, {Q31(-0.94701139690956f), Q31(-0.01826348194255f)},
+{Q31(-0.98759606946049f), Q31( 0.82288714303073f)}, {Q31( 0.87434794743625f), Q31( 0.89399495655433f)},
+{Q31(-0.93412041758744f), Q31( 0.41374052024363f)}, {Q31( 0.96063943315511f), Q31( 0.93116709541280f)},
+{Q31( 0.97534253457837f), Q31( 0.86150930812689f)}, {Q31( 0.99642466504163f), Q31( 0.70190043427512f)},
+{Q31(-0.94705089665984f), Q31(-0.29580042814306f)}, {Q31( 0.91599807087376f), Q31(-0.98147830385781f)},
+// Start of duplicated table
+{Q31(-0.99948153278296f), Q31(-0.59483417516607f)}, {Q31( 0.97113454393991f), Q31(-0.67528515225647f)},
+{Q31( 0.14130051758487f), Q31(-0.95090983575689f)}, {Q31(-0.47005496701697f), Q31(-0.37340549728647f)},
+{Q31( 0.80705063769351f), Q31( 0.29653668284408f)}, {Q31(-0.38981478896926f), Q31( 0.89572605717087f)},
+{Q31(-0.01053049862020f), Q31(-0.66959058036166f)}, {Q31(-0.91266367957293f), Q31(-0.11522938140034f)},
+};
+
+#endif /* AVCODEC_AACSBRDATA_H */

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aactab.h

@@ -0,0 +1,186 @@
+/*
+ * AAC data declarations
+ * Copyright (c) 2005-2006 Oded Shimon ( ods15 ods15 dyndns org )
+ * Copyright (c) 2006-2007 Maxim Gavrilov ( maxim.gavrilov gmail com )
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+/**
+ * @file
+ * AAC data declarations
+ * @author Oded Shimon  ( ods15 ods15 dyndns org )
+ * @author Maxim Gavrilov ( maxim.gavrilov gmail com )
+ */
+
+#ifndef AVCODEC_AACTAB_H
+#define AVCODEC_AACTAB_H
+
+#include "libavutil/mem.h"
+#include "aac.h"
+
+#include <stdint.h>
+
+/* NOTE:
+ * Tables in this file are shared by the AAC decoders and encoder
+ */
+
+extern float ff_aac_pow2sf_tab[428];
+extern float ff_aac_pow34sf_tab[428];
+
+static inline void ff_aac_tableinit(void)
+{
+    int i;
+
+    /* 2^(i/16) for 0 <= i <= 15 */
+    static const float exp2_lut[] = {
+        1.00000000000000000000,
+        1.04427378242741384032,
+        1.09050773266525765921,
+        1.13878863475669165370,
+        1.18920711500272106672,
+        1.24185781207348404859,
+        1.29683955465100966593,
+        1.35425554693689272830,
+        1.41421356237309504880,
+        1.47682614593949931139,
+        1.54221082540794082361,
+        1.61049033194925430818,
+        1.68179283050742908606,
+        1.75625216037329948311,
+        1.83400808640934246349,
+        1.91520656139714729387,
+    };
+    float t1 = 8.8817841970012523233890533447265625e-16; // 2^(-50)
+    float t2 = 3.63797880709171295166015625e-12; // 2^(-38)
+    int t1_inc_cur, t2_inc_cur;
+    int t1_inc_prev = 0;
+    int t2_inc_prev = 8;
+
+    for (i = 0; i < 428; i++) {
+        t1_inc_cur = 4 * (i % 4);
+        t2_inc_cur = (8 + 3*i) % 16;
+        if (t1_inc_cur < t1_inc_prev)
+            t1 *= 2;
+        if (t2_inc_cur < t2_inc_prev)
+            t2 *= 2;
+        // A much more efficient and accurate way of doing:
+        // ff_aac_pow2sf_tab[i] = pow(2, (i - POW_SF2_ZERO) / 4.0);
+        // ff_aac_pow34sf_tab[i] = pow(ff_aac_pow2sf_tab[i], 3.0/4.0);
+        ff_aac_pow2sf_tab[i] = t1 * exp2_lut[t1_inc_cur];
+        ff_aac_pow34sf_tab[i] = t2 * exp2_lut[t2_inc_cur];
+        t1_inc_prev = t1_inc_cur;
+        t2_inc_prev = t2_inc_cur;
+    }
+}
+
+/* @name ltp_coef
+ * Table of the LTP coefficients
+ */
+static const INTFLOAT ltp_coef[8] = {
+    Q30(0.570829), Q30(0.696616), Q30(0.813004), Q30(0.911304),
+    Q30(0.984900), Q30(1.067894), Q30(1.194601), Q30(1.369533),
+};
+
+/* @name tns_tmp2_map
+ * Tables of the tmp2[] arrays of LPC coefficients used for TNS.
+ * The suffix _M_N[] indicate the values of coef_compress and coef_res
+ * respectively.
+ * @{
+ */
+static const INTFLOAT tns_tmp2_map_1_3[4] = {
+    Q31(0.00000000), Q31(-0.43388373),  Q31(0.64278758),  Q31(0.34202015),
+};
+
+static const INTFLOAT tns_tmp2_map_0_3[8] = {
+    Q31(0.00000000), Q31(-0.43388373), Q31(-0.78183150), Q31(-0.97492790),
+    Q31(0.98480773), Q31( 0.86602539), Q31( 0.64278758), Q31( 0.34202015),
+};
+
+static const INTFLOAT tns_tmp2_map_1_4[8] = {
+    Q31(0.00000000), Q31(-0.20791170), Q31(-0.40673664), Q31(-0.58778524),
+    Q31(0.67369562), Q31( 0.52643216), Q31( 0.36124167), Q31( 0.18374951),
+};
+
+static const INTFLOAT tns_tmp2_map_0_4[16] = {
+    Q31( 0.00000000), Q31(-0.20791170), Q31(-0.40673664), Q31(-0.58778524),
+    Q31(-0.74314481), Q31(-0.86602539), Q31(-0.95105654), Q31(-0.99452192),
+    Q31( 0.99573416), Q31( 0.96182561), Q31( 0.89516330), Q31( 0.79801720),
+    Q31( 0.67369562), Q31( 0.52643216), Q31( 0.36124167), Q31( 0.18374951),
+};
+
+static const INTFLOAT * const tns_tmp2_map[4] = {
+    tns_tmp2_map_0_3,
+    tns_tmp2_map_0_4,
+    tns_tmp2_map_1_3,
+    tns_tmp2_map_1_4
+};
+// @}
+
+/* @name window coefficients
+ * @{
+ */
+DECLARE_ALIGNED(32, extern float,  ff_aac_kbd_long_1024)[1024];
+DECLARE_ALIGNED(32, extern float,  ff_aac_kbd_short_128)[128];
+DECLARE_ALIGNED(32, extern float,  ff_aac_kbd_long_960)[960];
+DECLARE_ALIGNED(32, extern float,  ff_aac_kbd_short_120)[120];
+DECLARE_ALIGNED(32, extern int,    ff_aac_kbd_long_1024_fixed)[1024];
+DECLARE_ALIGNED(32, extern int,    ff_aac_kbd_long_512_fixed)[512];
+DECLARE_ALIGNED(32, extern int,    ff_aac_kbd_short_128_fixed)[128];
+DECLARE_ALIGNED(32, extern const float, ff_aac_eld_window_512)[1920];
+DECLARE_ALIGNED(32, extern const int,   ff_aac_eld_window_512_fixed)[1920];
+DECLARE_ALIGNED(32, extern const float, ff_aac_eld_window_480)[1800];
+DECLARE_ALIGNED(32, extern const int,   ff_aac_eld_window_480_fixed)[1800];
+// @}
+
+/* @name number of scalefactor window bands for long and short transform windows respectively
+ * @{
+ */
+extern const uint8_t ff_aac_num_swb_1024[];
+extern const uint8_t ff_aac_num_swb_960 [];
+extern const uint8_t ff_aac_num_swb_512 [];
+extern const uint8_t ff_aac_num_swb_480 [];
+extern const uint8_t ff_aac_num_swb_128 [];
+extern const uint8_t ff_aac_num_swb_120 [];
+// @}
+
+extern const uint8_t ff_aac_pred_sfb_max [];
+
+extern const uint32_t ff_aac_scalefactor_code[121];
+extern const uint8_t  ff_aac_scalefactor_bits[121];
+
+extern const uint16_t * const ff_aac_spectral_codes[11];
+extern const uint8_t  * const ff_aac_spectral_bits [11];
+extern const uint16_t  ff_aac_spectral_sizes[11];
+
+extern const float *ff_aac_codebook_vectors[];
+extern const float *ff_aac_codebook_vector_vals[];
+extern const uint16_t *ff_aac_codebook_vector_idx[];
+
+extern const uint16_t * const ff_swb_offset_1024[13];
+extern const uint16_t * const ff_swb_offset_960 [13];
+extern const uint16_t * const ff_swb_offset_512 [13];
+extern const uint16_t * const ff_swb_offset_480 [13];
+extern const uint16_t * const ff_swb_offset_128 [13];
+extern const uint16_t * const ff_swb_offset_120 [13];
+
+extern const uint8_t ff_tns_max_bands_1024[13];
+extern const uint8_t ff_tns_max_bands_512 [13];
+extern const uint8_t ff_tns_max_bands_480 [13];
+extern const uint8_t ff_tns_max_bands_128 [13];
+
+#endif /* AVCODEC_AACTAB_H */

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aarch64/Makefile

@@ -0,0 +1,60 @@
+# subsystems
+OBJS-$(CONFIG_FFT)                      += aarch64/fft_init_aarch64.o
+OBJS-$(CONFIG_FMTCONVERT)               += aarch64/fmtconvert_init.o
+OBJS-$(CONFIG_H264CHROMA)               += aarch64/h264chroma_init_aarch64.o
+OBJS-$(CONFIG_H264DSP)                  += aarch64/h264dsp_init_aarch64.o
+OBJS-$(CONFIG_H264PRED)                 += aarch64/h264pred_init.o
+OBJS-$(CONFIG_H264QPEL)                 += aarch64/h264qpel_init_aarch64.o
+OBJS-$(CONFIG_HPELDSP)                  += aarch64/hpeldsp_init_aarch64.o
+OBJS-$(CONFIG_MPEGAUDIODSP)             += aarch64/mpegaudiodsp_init.o
+OBJS-$(CONFIG_NEON_CLOBBER_TEST)        += aarch64/neontest.o
+OBJS-$(CONFIG_VIDEODSP)                 += aarch64/videodsp_init.o
+OBJS-$(CONFIG_VP8DSP)                   += aarch64/vp8dsp_init_aarch64.o
+
+# decoders/encoders
+OBJS-$(CONFIG_AAC_DECODER)              += aarch64/aacpsdsp_init_aarch64.o \
+                                           aarch64/sbrdsp_init_aarch64.o
+OBJS-$(CONFIG_DCA_DECODER)              += aarch64/synth_filter_init.o
+OBJS-$(CONFIG_OPUS_DECODER)             += aarch64/opusdsp_init.o
+OBJS-$(CONFIG_RV40_DECODER)             += aarch64/rv40dsp_init_aarch64.o
+OBJS-$(CONFIG_VC1DSP)                   += aarch64/vc1dsp_init_aarch64.o
+OBJS-$(CONFIG_VORBIS_DECODER)           += aarch64/vorbisdsp_init.o
+OBJS-$(CONFIG_VP9_DECODER)              += aarch64/vp9dsp_init_10bpp_aarch64.o \
+                                           aarch64/vp9dsp_init_12bpp_aarch64.o \
+                                           aarch64/vp9dsp_init_aarch64.o
+
+# ARMv8 optimizations
+
+# subsystems
+ARMV8-OBJS-$(CONFIG_VIDEODSP)           += aarch64/videodsp.o
+
+# NEON optimizations
+
+# subsystems
+NEON-OBJS-$(CONFIG_AAC_DECODER)         += aarch64/sbrdsp_neon.o
+NEON-OBJS-$(CONFIG_FFT)                 += aarch64/fft_neon.o
+NEON-OBJS-$(CONFIG_FMTCONVERT)          += aarch64/fmtconvert_neon.o
+NEON-OBJS-$(CONFIG_H264CHROMA)          += aarch64/h264cmc_neon.o
+NEON-OBJS-$(CONFIG_H264DSP)             += aarch64/h264dsp_neon.o              \
+                                           aarch64/h264idct_neon.o
+NEON-OBJS-$(CONFIG_H264PRED)            += aarch64/h264pred_neon.o
+NEON-OBJS-$(CONFIG_H264QPEL)            += aarch64/h264qpel_neon.o             \
+                                           aarch64/hpeldsp_neon.o
+NEON-OBJS-$(CONFIG_HPELDSP)             += aarch64/hpeldsp_neon.o
+NEON-OBJS-$(CONFIG_IDCTDSP)             += aarch64/idctdsp_init_aarch64.o      \
+                                           aarch64/simple_idct_neon.o
+NEON-OBJS-$(CONFIG_MDCT)                += aarch64/mdct_neon.o
+NEON-OBJS-$(CONFIG_MPEGAUDIODSP)        += aarch64/mpegaudiodsp_neon.o
+NEON-OBJS-$(CONFIG_VP8DSP)              += aarch64/vp8dsp_neon.o
+
+# decoders/encoders
+NEON-OBJS-$(CONFIG_AAC_DECODER)         += aarch64/aacpsdsp_neon.o
+NEON-OBJS-$(CONFIG_DCA_DECODER)         += aarch64/synth_filter_neon.o
+NEON-OBJS-$(CONFIG_OPUS_DECODER)        += aarch64/opusdsp_neon.o
+NEON-OBJS-$(CONFIG_VORBIS_DECODER)      += aarch64/vorbisdsp_neon.o
+NEON-OBJS-$(CONFIG_VP9_DECODER)         += aarch64/vp9itxfm_16bpp_neon.o       \
+                                           aarch64/vp9itxfm_neon.o             \
+                                           aarch64/vp9lpf_16bpp_neon.o         \
+                                           aarch64/vp9lpf_neon.o               \
+                                           aarch64/vp9mc_16bpp_neon.o          \
+                                           aarch64/vp9mc_neon.o

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aarch64/aacpsdsp_init_aarch64.c

@@ -0,0 +1,48 @@
+/*
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include "config.h"
+
+#include "libavutil/aarch64/cpu.h"
+#include "libavcodec/aacpsdsp.h"
+
+void ff_ps_add_squares_neon(float *dst, const float (*src)[2], int n);
+void ff_ps_mul_pair_single_neon(float (*dst)[2], float (*src0)[2],
+                                float *src1, int n);
+void ff_ps_hybrid_analysis_neon(float (*out)[2], float (*in)[2],
+                                const float (*filter)[8][2],
+                                ptrdiff_t stride, int n);
+void ff_ps_stereo_interpolate_neon(float (*l)[2], float (*r)[2],
+                                   float h[2][4], float h_step[2][4],
+                                   int len);
+void ff_ps_stereo_interpolate_ipdopd_neon(float (*l)[2], float (*r)[2],
+                                          float h[2][4], float h_step[2][4],
+                                          int len);
+
+av_cold void ff_psdsp_init_aarch64(PSDSPContext *s)
+{
+    int cpu_flags = av_get_cpu_flags();
+
+    if (have_neon(cpu_flags)) {
+        s->add_squares           = ff_ps_add_squares_neon;
+        s->mul_pair_single       = ff_ps_mul_pair_single_neon;
+        s->hybrid_analysis       = ff_ps_hybrid_analysis_neon;
+        s->stereo_interpolate[0] = ff_ps_stereo_interpolate_neon;
+        s->stereo_interpolate[1] = ff_ps_stereo_interpolate_ipdopd_neon;
+    }
+}

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aarch64/asm-offsets.h

@@ -0,0 +1,25 @@
+/*
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#ifndef AVCODEC_AARCH64_ASM_OFFSETS_H
+#define AVCODEC_AARCH64_ASM_OFFSETS_H
+
+/* FFTContext */
+#define IMDCT_HALF                      0x48
+
+#endif /* AVCODEC_AARCH64_ASM_OFFSETS_H */

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aarch64/cabac.h

@@ -0,0 +1,104 @@
+/*
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#ifndef AVCODEC_AARCH64_CABAC_H
+#define AVCODEC_AARCH64_CABAC_H
+
+#include "config.h"
+#if HAVE_INLINE_ASM
+
+#include "libavutil/attributes.h"
+#include "libavutil/internal.h"
+#include "libavcodec/cabac.h"
+
+#define get_cabac_inline get_cabac_inline_aarch64
+static av_always_inline int get_cabac_inline_aarch64(CABACContext *c,
+                                                     uint8_t *const state)
+{
+    int bit;
+    void *reg_a, *reg_b, *reg_c, *tmp;
+
+    __asm__ volatile(
+        "ldrb       %w[bit]       , [%[state]]                  \n\t"
+        "add        %[r_b]        , %[tables]   , %[lps_off]    \n\t"
+        "mov        %w[tmp]       , %w[range]                   \n\t"
+        "and        %w[range]     , %w[range]   , #0xC0         \n\t"
+        "lsl        %w[r_c]       , %w[range]   , #1            \n\t"
+        "add        %[r_b]        , %[r_b]      , %w[bit], UXTW \n\t"
+        "ldrb       %w[range]     , [%[r_b], %w[r_c], SXTW]     \n\t"
+        "sub        %w[r_c]       , %w[tmp]     , %w[range]     \n\t"
+        "lsl        %w[tmp]       , %w[r_c]     , #17           \n\t"
+        "cmp        %w[tmp]       , %w[low]                     \n\t"
+        "csel       %w[tmp]       , %w[tmp]     , wzr      , cc \n\t"
+        "csel       %w[range]     , %w[r_c]     , %w[range], gt \n\t"
+        "cinv       %w[bit]       , %w[bit]     , cc            \n\t"
+        "sub        %w[low]       , %w[low]     , %w[tmp]       \n\t"
+        "add        %[r_b]        , %[tables]   , %[norm_off]   \n\t"
+        "add        %[r_a]        , %[tables]   , %[mlps_off]   \n\t"
+        "ldrb       %w[tmp]       , [%[r_b], %w[range], SXTW]   \n\t"
+        "ldrb       %w[r_a]       , [%[r_a], %w[bit], SXTW]     \n\t"
+        "lsl        %w[low]       , %w[low]     , %w[tmp]       \n\t"
+        "lsl        %w[range]     , %w[range]   , %w[tmp]       \n\t"
+        "uxth       %w[r_c]       , %w[low]                     \n\t"
+        "strb       %w[r_a]       , [%[state]]                  \n\t"
+        "cbnz       %w[r_c]       , 2f                          \n\t"
+        "ldr        %[r_c]        , [%[c], %[byte]]             \n\t"
+        "ldr        %[r_a]        , [%[c], %[end]]              \n\t"
+        "ldrh       %w[tmp]       , [%[r_c]]                    \n\t"
+        "cmp        %[r_c]        , %[r_a]                      \n\t"
+        "b.ge       1f                                          \n\t"
+        "add        %[r_a]        , %[r_c]      , #2            \n\t"
+        "str        %[r_a]        , [%[c], %[byte]]             \n\t"
+        "1:                                                     \n\t"
+        "sub        %w[r_c]       , %w[low]     , #1            \n\t"
+        "eor        %w[r_c]       , %w[r_c]     , %w[low]       \n\t"
+        "rev        %w[tmp]       , %w[tmp]                     \n\t"
+        "lsr        %w[r_c]       , %w[r_c]     , #15           \n\t"
+        "lsr        %w[tmp]       , %w[tmp]     , #15           \n\t"
+        "ldrb       %w[r_c]       , [%[r_b], %w[r_c], SXTW]     \n\t"
+        "mov        %w[r_b]       , #0xFFFF                     \n\t"
+        "mov        %w[r_a]       , #7                          \n\t"
+        "sub        %w[tmp]       , %w[tmp]     , %w[r_b]       \n\t"
+        "sub        %w[r_c]       , %w[r_a]     , %w[r_c]       \n\t"
+        "lsl        %w[tmp]       , %w[tmp]     , %w[r_c]       \n\t"
+        "add        %w[low]       , %w[low]     , %w[tmp]       \n\t"
+        "2:                                                     \n\t"
+        :    [bit]"=&r"(bit),
+             [low]"+&r"(c->low),
+           [range]"+&r"(c->range),
+             [r_a]"=&r"(reg_a),
+             [r_b]"=&r"(reg_b),
+             [r_c]"=&r"(reg_c),
+             [tmp]"=&r"(tmp)
+        :        [c]"r"(c),
+             [state]"r"(state),
+            [tables]"r"(ff_h264_cabac_tables),
+              [byte]"i"(offsetof(CABACContext, bytestream)),
+               [end]"i"(offsetof(CABACContext, bytestream_end)),
+          [norm_off]"I"(H264_NORM_SHIFT_OFFSET),
+           [lps_off]"I"(H264_LPS_RANGE_OFFSET),
+          [mlps_off]"I"(H264_MLPS_STATE_OFFSET + 128)
+        : "memory", "cc"
+        );
+
+    return bit & 1;
+}
+
+#endif /* HAVE_INLINE_ASM */
+
+#endif /* AVCODEC_AARCH64_CABAC_H */

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aarch64/fft_init_aarch64.c

@@ -0,0 +1,50 @@
+/*
+ * Copyright (c) 2009 Mans Rullgard <mans@mansr.com>
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include "config.h"
+
+#include "libavutil/attributes.h"
+#include "libavutil/cpu.h"
+#include "libavutil/aarch64/cpu.h"
+
+#include "libavcodec/fft.h"
+
+void ff_fft_permute_neon(FFTContext *s, FFTComplex *z);
+void ff_fft_calc_neon(FFTContext *s, FFTComplex *z);
+
+void ff_imdct_calc_neon(FFTContext *s, FFTSample *output, const FFTSample *input);
+void ff_imdct_half_neon(FFTContext *s, FFTSample *output, const FFTSample *input);
+void ff_mdct_calc_neon(FFTContext *s, FFTSample *output, const FFTSample *input);
+
+av_cold void ff_fft_init_aarch64(FFTContext *s)
+{
+    int cpu_flags = av_get_cpu_flags();
+
+    if (have_neon(cpu_flags)) {
+        s->fft_permute  = ff_fft_permute_neon;
+        s->fft_calc     = ff_fft_calc_neon;
+#if CONFIG_MDCT
+        s->imdct_calc   = ff_imdct_calc_neon;
+        s->imdct_half   = ff_imdct_half_neon;
+        s->mdct_calc    = ff_mdct_calc_neon;
+        s->mdct_permutation = FF_MDCT_PERM_INTERLEAVE;
+#endif
+    }
+}

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aarch64/fmtconvert_init.c

@@ -0,0 +1,43 @@
+/*
+ * ARM optimized Format Conversion Utils
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include <stdint.h>
+
+#include "libavutil/attributes.h"
+#include "libavutil/aarch64/cpu.h"
+#include "libavcodec/avcodec.h"
+#include "libavcodec/fmtconvert.h"
+
+void ff_int32_to_float_fmul_array8_neon(FmtConvertContext *c, float *dst,
+                                        const int32_t *src, const float *mul,
+                                        int len);
+void ff_int32_to_float_fmul_scalar_neon(float *dst, const int32_t *src,
+                                        float mul, int len);
+
+av_cold void ff_fmt_convert_init_aarch64(FmtConvertContext *c,
+                                         AVCodecContext *avctx)
+{
+    int cpu_flags = av_get_cpu_flags();
+
+    if (have_neon(cpu_flags)) {
+        c->int32_to_float_fmul_array8 = ff_int32_to_float_fmul_array8_neon;
+        c->int32_to_float_fmul_scalar = ff_int32_to_float_fmul_scalar_neon;
+    }
+}

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aarch64/h264chroma_init_aarch64.c

@@ -0,0 +1,59 @@
+/*
+ * ARM NEON optimised H.264 chroma functions
+ * Copyright (c) 2008 Mans Rullgard <mans@mansr.com>
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include <stdint.h>
+
+#include "libavutil/attributes.h"
+#include "libavutil/cpu.h"
+#include "libavutil/aarch64/cpu.h"
+#include "libavcodec/h264chroma.h"
+
+#include "config.h"
+
+void ff_put_h264_chroma_mc8_neon(uint8_t *dst, uint8_t *src, ptrdiff_t stride,
+                                 int h, int x, int y);
+void ff_put_h264_chroma_mc4_neon(uint8_t *dst, uint8_t *src, ptrdiff_t stride,
+                                 int h, int x, int y);
+void ff_put_h264_chroma_mc2_neon(uint8_t *dst, uint8_t *src, ptrdiff_t stride,
+                                 int h, int x, int y);
+
+void ff_avg_h264_chroma_mc8_neon(uint8_t *dst, uint8_t *src, ptrdiff_t stride,
+                                 int h, int x, int y);
+void ff_avg_h264_chroma_mc4_neon(uint8_t *dst, uint8_t *src, ptrdiff_t stride,
+                                 int h, int x, int y);
+void ff_avg_h264_chroma_mc2_neon(uint8_t *dst, uint8_t *src, ptrdiff_t stride,
+                                 int h, int x, int y);
+
+av_cold void ff_h264chroma_init_aarch64(H264ChromaContext *c, int bit_depth)
+{
+    const int high_bit_depth = bit_depth > 8;
+    int cpu_flags = av_get_cpu_flags();
+
+    if (have_neon(cpu_flags) && !high_bit_depth) {
+        c->put_h264_chroma_pixels_tab[0] = ff_put_h264_chroma_mc8_neon;
+        c->put_h264_chroma_pixels_tab[1] = ff_put_h264_chroma_mc4_neon;
+        c->put_h264_chroma_pixels_tab[2] = ff_put_h264_chroma_mc2_neon;
+
+        c->avg_h264_chroma_pixels_tab[0] = ff_avg_h264_chroma_mc8_neon;
+        c->avg_h264_chroma_pixels_tab[1] = ff_avg_h264_chroma_mc4_neon;
+        c->avg_h264_chroma_pixels_tab[2] = ff_avg_h264_chroma_mc2_neon;
+    }
+}

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aarch64/h264dsp_init_aarch64.c

@@ -0,0 +1,129 @@
+/*
+ * Copyright (c) 2010 Mans Rullgard <mans@mansr.com>
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include <stdint.h>
+
+#include "libavutil/attributes.h"
+#include "libavutil/cpu.h"
+#include "libavutil/aarch64/cpu.h"
+#include "libavcodec/h264dsp.h"
+
+void ff_h264_v_loop_filter_luma_neon(uint8_t *pix, ptrdiff_t stride, int alpha,
+                                     int beta, int8_t *tc0);
+void ff_h264_h_loop_filter_luma_neon(uint8_t *pix, ptrdiff_t stride, int alpha,
+                                     int beta, int8_t *tc0);
+void ff_h264_v_loop_filter_luma_intra_neon(uint8_t *pix, ptrdiff_t stride, int alpha,
+                                           int beta);
+void ff_h264_h_loop_filter_luma_intra_neon(uint8_t *pix, ptrdiff_t stride, int alpha,
+                                           int beta);
+void ff_h264_v_loop_filter_chroma_neon(uint8_t *pix, ptrdiff_t stride, int alpha,
+                                       int beta, int8_t *tc0);
+void ff_h264_h_loop_filter_chroma_neon(uint8_t *pix, ptrdiff_t stride, int alpha,
+                                       int beta, int8_t *tc0);
+void ff_h264_h_loop_filter_chroma422_neon(uint8_t *pix, ptrdiff_t stride, int alpha,
+                                          int beta, int8_t *tc0);
+void ff_h264_v_loop_filter_chroma_intra_neon(uint8_t *pix, ptrdiff_t stride,
+                                             int alpha, int beta);
+void ff_h264_h_loop_filter_chroma_intra_neon(uint8_t *pix, ptrdiff_t stride,
+                                             int alpha, int beta);
+void ff_h264_h_loop_filter_chroma422_intra_neon(uint8_t *pix, ptrdiff_t stride,
+                                                int alpha, int beta);
+void ff_h264_h_loop_filter_chroma_mbaff_intra_neon(uint8_t *pix, ptrdiff_t stride,
+                                                   int alpha, int beta);
+
+void ff_weight_h264_pixels_16_neon(uint8_t *dst, ptrdiff_t stride, int height,
+                                   int log2_den, int weight, int offset);
+void ff_weight_h264_pixels_8_neon(uint8_t *dst, ptrdiff_t stride, int height,
+                                  int log2_den, int weight, int offset);
+void ff_weight_h264_pixels_4_neon(uint8_t *dst, ptrdiff_t stride, int height,
+                                  int log2_den, int weight, int offset);
+
+void ff_biweight_h264_pixels_16_neon(uint8_t *dst, uint8_t *src, ptrdiff_t stride,
+                                     int height, int log2_den, int weightd,
+                                     int weights, int offset);
+void ff_biweight_h264_pixels_8_neon(uint8_t *dst, uint8_t *src, ptrdiff_t stride,
+                                    int height, int log2_den, int weightd,
+                                    int weights, int offset);
+void ff_biweight_h264_pixels_4_neon(uint8_t *dst, uint8_t *src, ptrdiff_t stride,
+                                    int height, int log2_den, int weightd,
+                                    int weights, int offset);
+
+void ff_h264_idct_add_neon(uint8_t *dst, int16_t *block, int stride);
+void ff_h264_idct_dc_add_neon(uint8_t *dst, int16_t *block, int stride);
+void ff_h264_idct_add16_neon(uint8_t *dst, const int *block_offset,
+                             int16_t *block, int stride,
+                             const uint8_t nnzc[6*8]);
+void ff_h264_idct_add16intra_neon(uint8_t *dst, const int *block_offset,
+                                  int16_t *block, int stride,
+                                  const uint8_t nnzc[6*8]);
+void ff_h264_idct_add8_neon(uint8_t **dest, const int *block_offset,
+                            int16_t *block, int stride,
+                            const uint8_t nnzc[6*8]);
+
+void ff_h264_idct8_add_neon(uint8_t *dst, int16_t *block, int stride);
+void ff_h264_idct8_dc_add_neon(uint8_t *dst, int16_t *block, int stride);
+void ff_h264_idct8_add4_neon(uint8_t *dst, const int *block_offset,
+                             int16_t *block, int stride,
+                             const uint8_t nnzc[6*8]);
+
+av_cold void ff_h264dsp_init_aarch64(H264DSPContext *c, const int bit_depth,
+                                     const int chroma_format_idc)
+{
+    int cpu_flags = av_get_cpu_flags();
+
+    if (have_neon(cpu_flags) && bit_depth == 8) {
+        c->h264_v_loop_filter_luma   = ff_h264_v_loop_filter_luma_neon;
+        c->h264_h_loop_filter_luma   = ff_h264_h_loop_filter_luma_neon;
+        c->h264_v_loop_filter_luma_intra= ff_h264_v_loop_filter_luma_intra_neon;
+        c->h264_h_loop_filter_luma_intra= ff_h264_h_loop_filter_luma_intra_neon;
+
+        c->h264_v_loop_filter_chroma = ff_h264_v_loop_filter_chroma_neon;
+        c->h264_v_loop_filter_chroma_intra = ff_h264_v_loop_filter_chroma_intra_neon;
+
+        if (chroma_format_idc <= 1) {
+            c->h264_h_loop_filter_chroma = ff_h264_h_loop_filter_chroma_neon;
+            c->h264_h_loop_filter_chroma_intra = ff_h264_h_loop_filter_chroma_intra_neon;
+            c->h264_h_loop_filter_chroma_mbaff_intra = ff_h264_h_loop_filter_chroma_mbaff_intra_neon;
+        } else {
+            c->h264_h_loop_filter_chroma = ff_h264_h_loop_filter_chroma422_neon;
+            c->h264_h_loop_filter_chroma_mbaff = ff_h264_h_loop_filter_chroma_neon;
+            c->h264_h_loop_filter_chroma_intra = ff_h264_h_loop_filter_chroma422_intra_neon;
+            c->h264_h_loop_filter_chroma_mbaff_intra = ff_h264_h_loop_filter_chroma_intra_neon;
+        }
+
+        c->weight_h264_pixels_tab[0] = ff_weight_h264_pixels_16_neon;
+        c->weight_h264_pixels_tab[1] = ff_weight_h264_pixels_8_neon;
+        c->weight_h264_pixels_tab[2] = ff_weight_h264_pixels_4_neon;
+
+        c->biweight_h264_pixels_tab[0] = ff_biweight_h264_pixels_16_neon;
+        c->biweight_h264_pixels_tab[1] = ff_biweight_h264_pixels_8_neon;
+        c->biweight_h264_pixels_tab[2] = ff_biweight_h264_pixels_4_neon;
+
+        c->h264_idct_add        = ff_h264_idct_add_neon;
+        c->h264_idct_dc_add     = ff_h264_idct_dc_add_neon;
+        c->h264_idct_add16      = ff_h264_idct_add16_neon;
+        c->h264_idct_add16intra = ff_h264_idct_add16intra_neon;
+        if (chroma_format_idc <= 1)
+            c->h264_idct_add8   = ff_h264_idct_add8_neon;
+        c->h264_idct8_add       = ff_h264_idct8_add_neon;
+        c->h264_idct8_dc_add    = ff_h264_idct8_dc_add_neon;
+        c->h264_idct8_add4      = ff_h264_idct8_add4_neon;
+    }
+}

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aarch64/h264pred_init.c

@@ -0,0 +1,93 @@
+/*
+ * Copyright (c) 2009 Mans Rullgard <mans@mansr.com>
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include <stdint.h>
+
+#include "libavutil/attributes.h"
+#include "libavutil/aarch64/cpu.h"
+#include "libavcodec/avcodec.h"
+#include "libavcodec/h264pred.h"
+
+void ff_pred16x16_vert_neon(uint8_t *src, ptrdiff_t stride);
+void ff_pred16x16_hor_neon(uint8_t *src, ptrdiff_t stride);
+void ff_pred16x16_plane_neon(uint8_t *src, ptrdiff_t stride);
+void ff_pred16x16_dc_neon(uint8_t *src, ptrdiff_t stride);
+void ff_pred16x16_128_dc_neon(uint8_t *src, ptrdiff_t stride);
+void ff_pred16x16_left_dc_neon(uint8_t *src, ptrdiff_t stride);
+void ff_pred16x16_top_dc_neon(uint8_t *src, ptrdiff_t stride);
+
+void ff_pred8x8_vert_neon(uint8_t *src, ptrdiff_t stride);
+void ff_pred8x8_hor_neon(uint8_t *src, ptrdiff_t stride);
+void ff_pred8x8_plane_neon(uint8_t *src, ptrdiff_t stride);
+void ff_pred8x8_dc_neon(uint8_t *src, ptrdiff_t stride);
+void ff_pred8x8_128_dc_neon(uint8_t *src, ptrdiff_t stride);
+void ff_pred8x8_left_dc_neon(uint8_t *src, ptrdiff_t stride);
+void ff_pred8x8_top_dc_neon(uint8_t *src, ptrdiff_t stride);
+void ff_pred8x8_l0t_dc_neon(uint8_t *src, ptrdiff_t stride);
+void ff_pred8x8_0lt_dc_neon(uint8_t *src, ptrdiff_t stride);
+void ff_pred8x8_l00_dc_neon(uint8_t *src, ptrdiff_t stride);
+void ff_pred8x8_0l0_dc_neon(uint8_t *src, ptrdiff_t stride);
+
+static av_cold void h264_pred_init_neon(H264PredContext *h, int codec_id,
+                                        const int bit_depth,
+                                        const int chroma_format_idc)
+{
+    const int high_depth = bit_depth > 8;
+
+    if (high_depth)
+        return;
+
+    if (chroma_format_idc <= 1) {
+        h->pred8x8[VERT_PRED8x8     ] = ff_pred8x8_vert_neon;
+        h->pred8x8[HOR_PRED8x8      ] = ff_pred8x8_hor_neon;
+        if (codec_id != AV_CODEC_ID_VP7 && codec_id != AV_CODEC_ID_VP8)
+            h->pred8x8[PLANE_PRED8x8] = ff_pred8x8_plane_neon;
+        h->pred8x8[DC_128_PRED8x8   ] = ff_pred8x8_128_dc_neon;
+        if (codec_id != AV_CODEC_ID_RV40 && codec_id != AV_CODEC_ID_VP7 &&
+            codec_id != AV_CODEC_ID_VP8) {
+            h->pred8x8[DC_PRED8x8     ] = ff_pred8x8_dc_neon;
+            h->pred8x8[LEFT_DC_PRED8x8] = ff_pred8x8_left_dc_neon;
+            h->pred8x8[TOP_DC_PRED8x8 ] = ff_pred8x8_top_dc_neon;
+            h->pred8x8[ALZHEIMER_DC_L0T_PRED8x8] = ff_pred8x8_l0t_dc_neon;
+            h->pred8x8[ALZHEIMER_DC_0LT_PRED8x8] = ff_pred8x8_0lt_dc_neon;
+            h->pred8x8[ALZHEIMER_DC_L00_PRED8x8] = ff_pred8x8_l00_dc_neon;
+            h->pred8x8[ALZHEIMER_DC_0L0_PRED8x8] = ff_pred8x8_0l0_dc_neon;
+        }
+    }
+
+    h->pred16x16[DC_PRED8x8     ] = ff_pred16x16_dc_neon;
+    h->pred16x16[VERT_PRED8x8   ] = ff_pred16x16_vert_neon;
+    h->pred16x16[HOR_PRED8x8    ] = ff_pred16x16_hor_neon;
+    h->pred16x16[LEFT_DC_PRED8x8] = ff_pred16x16_left_dc_neon;
+    h->pred16x16[TOP_DC_PRED8x8 ] = ff_pred16x16_top_dc_neon;
+    h->pred16x16[DC_128_PRED8x8 ] = ff_pred16x16_128_dc_neon;
+    if (codec_id != AV_CODEC_ID_SVQ3 && codec_id != AV_CODEC_ID_RV40 &&
+        codec_id != AV_CODEC_ID_VP7 && codec_id != AV_CODEC_ID_VP8)
+        h->pred16x16[PLANE_PRED8x8  ] = ff_pred16x16_plane_neon;
+}
+
+av_cold void ff_h264_pred_init_aarch64(H264PredContext *h, int codec_id,
+                                       int bit_depth, const int chroma_format_idc)
+{
+    int cpu_flags = av_get_cpu_flags();
+
+    if (have_neon(cpu_flags))
+        h264_pred_init_neon(h, codec_id, bit_depth, chroma_format_idc);
+}

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aarch64/h264qpel_init_aarch64.c

@@ -0,0 +1,172 @@
+/*
+ * ARM NEON optimised DSP functions
+ * Copyright (c) 2008 Mans Rullgard <mans@mansr.com>
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include <stdint.h>
+
+#include "config.h"
+#include "libavutil/attributes.h"
+#include "libavutil/cpu.h"
+#include "libavutil/aarch64/cpu.h"
+#include "libavcodec/h264qpel.h"
+
+void ff_put_h264_qpel16_mc00_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+void ff_put_h264_qpel16_mc10_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+void ff_put_h264_qpel16_mc20_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+void ff_put_h264_qpel16_mc30_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+void ff_put_h264_qpel16_mc01_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+void ff_put_h264_qpel16_mc11_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+void ff_put_h264_qpel16_mc21_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+void ff_put_h264_qpel16_mc31_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+void ff_put_h264_qpel16_mc02_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+void ff_put_h264_qpel16_mc12_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+void ff_put_h264_qpel16_mc22_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+void ff_put_h264_qpel16_mc32_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+void ff_put_h264_qpel16_mc03_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+void ff_put_h264_qpel16_mc13_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+void ff_put_h264_qpel16_mc23_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+void ff_put_h264_qpel16_mc33_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+
+void ff_put_h264_qpel8_mc00_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+void ff_put_h264_qpel8_mc10_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+void ff_put_h264_qpel8_mc20_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+void ff_put_h264_qpel8_mc30_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+void ff_put_h264_qpel8_mc01_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+void ff_put_h264_qpel8_mc11_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+void ff_put_h264_qpel8_mc21_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+void ff_put_h264_qpel8_mc31_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+void ff_put_h264_qpel8_mc02_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+void ff_put_h264_qpel8_mc12_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+void ff_put_h264_qpel8_mc22_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+void ff_put_h264_qpel8_mc32_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+void ff_put_h264_qpel8_mc03_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+void ff_put_h264_qpel8_mc13_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+void ff_put_h264_qpel8_mc23_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+void ff_put_h264_qpel8_mc33_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+
+void ff_avg_h264_qpel16_mc00_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+void ff_avg_h264_qpel16_mc10_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+void ff_avg_h264_qpel16_mc20_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+void ff_avg_h264_qpel16_mc30_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+void ff_avg_h264_qpel16_mc01_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+void ff_avg_h264_qpel16_mc11_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+void ff_avg_h264_qpel16_mc21_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+void ff_avg_h264_qpel16_mc31_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+void ff_avg_h264_qpel16_mc02_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+void ff_avg_h264_qpel16_mc12_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+void ff_avg_h264_qpel16_mc22_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+void ff_avg_h264_qpel16_mc32_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+void ff_avg_h264_qpel16_mc03_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+void ff_avg_h264_qpel16_mc13_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+void ff_avg_h264_qpel16_mc23_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+void ff_avg_h264_qpel16_mc33_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+
+void ff_avg_h264_qpel8_mc00_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+void ff_avg_h264_qpel8_mc10_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+void ff_avg_h264_qpel8_mc20_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+void ff_avg_h264_qpel8_mc30_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+void ff_avg_h264_qpel8_mc01_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+void ff_avg_h264_qpel8_mc11_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+void ff_avg_h264_qpel8_mc21_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+void ff_avg_h264_qpel8_mc31_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+void ff_avg_h264_qpel8_mc02_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+void ff_avg_h264_qpel8_mc12_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+void ff_avg_h264_qpel8_mc22_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+void ff_avg_h264_qpel8_mc32_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+void ff_avg_h264_qpel8_mc03_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+void ff_avg_h264_qpel8_mc13_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+void ff_avg_h264_qpel8_mc23_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+void ff_avg_h264_qpel8_mc33_neon(uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
+
+av_cold void ff_h264qpel_init_aarch64(H264QpelContext *c, int bit_depth)
+{
+    const int high_bit_depth = bit_depth > 8;
+    int cpu_flags = av_get_cpu_flags();
+
+    if (have_neon(cpu_flags) && !high_bit_depth) {
+        c->put_h264_qpel_pixels_tab[0][ 0] = ff_put_h264_qpel16_mc00_neon;
+        c->put_h264_qpel_pixels_tab[0][ 1] = ff_put_h264_qpel16_mc10_neon;
+        c->put_h264_qpel_pixels_tab[0][ 2] = ff_put_h264_qpel16_mc20_neon;
+        c->put_h264_qpel_pixels_tab[0][ 3] = ff_put_h264_qpel16_mc30_neon;
+        c->put_h264_qpel_pixels_tab[0][ 4] = ff_put_h264_qpel16_mc01_neon;
+        c->put_h264_qpel_pixels_tab[0][ 5] = ff_put_h264_qpel16_mc11_neon;
+        c->put_h264_qpel_pixels_tab[0][ 6] = ff_put_h264_qpel16_mc21_neon;
+        c->put_h264_qpel_pixels_tab[0][ 7] = ff_put_h264_qpel16_mc31_neon;
+        c->put_h264_qpel_pixels_tab[0][ 8] = ff_put_h264_qpel16_mc02_neon;
+        c->put_h264_qpel_pixels_tab[0][ 9] = ff_put_h264_qpel16_mc12_neon;
+        c->put_h264_qpel_pixels_tab[0][10] = ff_put_h264_qpel16_mc22_neon;
+        c->put_h264_qpel_pixels_tab[0][11] = ff_put_h264_qpel16_mc32_neon;
+        c->put_h264_qpel_pixels_tab[0][12] = ff_put_h264_qpel16_mc03_neon;
+        c->put_h264_qpel_pixels_tab[0][13] = ff_put_h264_qpel16_mc13_neon;
+        c->put_h264_qpel_pixels_tab[0][14] = ff_put_h264_qpel16_mc23_neon;
+        c->put_h264_qpel_pixels_tab[0][15] = ff_put_h264_qpel16_mc33_neon;
+
+        c->put_h264_qpel_pixels_tab[1][ 0] = ff_put_h264_qpel8_mc00_neon;
+        c->put_h264_qpel_pixels_tab[1][ 1] = ff_put_h264_qpel8_mc10_neon;
+        c->put_h264_qpel_pixels_tab[1][ 2] = ff_put_h264_qpel8_mc20_neon;
+        c->put_h264_qpel_pixels_tab[1][ 3] = ff_put_h264_qpel8_mc30_neon;
+        c->put_h264_qpel_pixels_tab[1][ 4] = ff_put_h264_qpel8_mc01_neon;
+        c->put_h264_qpel_pixels_tab[1][ 5] = ff_put_h264_qpel8_mc11_neon;
+        c->put_h264_qpel_pixels_tab[1][ 6] = ff_put_h264_qpel8_mc21_neon;
+        c->put_h264_qpel_pixels_tab[1][ 7] = ff_put_h264_qpel8_mc31_neon;
+        c->put_h264_qpel_pixels_tab[1][ 8] = ff_put_h264_qpel8_mc02_neon;
+        c->put_h264_qpel_pixels_tab[1][ 9] = ff_put_h264_qpel8_mc12_neon;
+        c->put_h264_qpel_pixels_tab[1][10] = ff_put_h264_qpel8_mc22_neon;
+        c->put_h264_qpel_pixels_tab[1][11] = ff_put_h264_qpel8_mc32_neon;
+        c->put_h264_qpel_pixels_tab[1][12] = ff_put_h264_qpel8_mc03_neon;
+        c->put_h264_qpel_pixels_tab[1][13] = ff_put_h264_qpel8_mc13_neon;
+        c->put_h264_qpel_pixels_tab[1][14] = ff_put_h264_qpel8_mc23_neon;
+        c->put_h264_qpel_pixels_tab[1][15] = ff_put_h264_qpel8_mc33_neon;
+
+        c->avg_h264_qpel_pixels_tab[0][ 0] = ff_avg_h264_qpel16_mc00_neon;
+        c->avg_h264_qpel_pixels_tab[0][ 1] = ff_avg_h264_qpel16_mc10_neon;
+        c->avg_h264_qpel_pixels_tab[0][ 2] = ff_avg_h264_qpel16_mc20_neon;
+        c->avg_h264_qpel_pixels_tab[0][ 3] = ff_avg_h264_qpel16_mc30_neon;
+        c->avg_h264_qpel_pixels_tab[0][ 4] = ff_avg_h264_qpel16_mc01_neon;
+        c->avg_h264_qpel_pixels_tab[0][ 5] = ff_avg_h264_qpel16_mc11_neon;
+        c->avg_h264_qpel_pixels_tab[0][ 6] = ff_avg_h264_qpel16_mc21_neon;
+        c->avg_h264_qpel_pixels_tab[0][ 7] = ff_avg_h264_qpel16_mc31_neon;
+        c->avg_h264_qpel_pixels_tab[0][ 8] = ff_avg_h264_qpel16_mc02_neon;
+        c->avg_h264_qpel_pixels_tab[0][ 9] = ff_avg_h264_qpel16_mc12_neon;
+        c->avg_h264_qpel_pixels_tab[0][10] = ff_avg_h264_qpel16_mc22_neon;
+        c->avg_h264_qpel_pixels_tab[0][11] = ff_avg_h264_qpel16_mc32_neon;
+        c->avg_h264_qpel_pixels_tab[0][12] = ff_avg_h264_qpel16_mc03_neon;
+        c->avg_h264_qpel_pixels_tab[0][13] = ff_avg_h264_qpel16_mc13_neon;
+        c->avg_h264_qpel_pixels_tab[0][14] = ff_avg_h264_qpel16_mc23_neon;
+        c->avg_h264_qpel_pixels_tab[0][15] = ff_avg_h264_qpel16_mc33_neon;
+
+        c->avg_h264_qpel_pixels_tab[1][ 0] = ff_avg_h264_qpel8_mc00_neon;
+        c->avg_h264_qpel_pixels_tab[1][ 1] = ff_avg_h264_qpel8_mc10_neon;
+        c->avg_h264_qpel_pixels_tab[1][ 2] = ff_avg_h264_qpel8_mc20_neon;
+        c->avg_h264_qpel_pixels_tab[1][ 3] = ff_avg_h264_qpel8_mc30_neon;
+        c->avg_h264_qpel_pixels_tab[1][ 4] = ff_avg_h264_qpel8_mc01_neon;
+        c->avg_h264_qpel_pixels_tab[1][ 5] = ff_avg_h264_qpel8_mc11_neon;
+        c->avg_h264_qpel_pixels_tab[1][ 6] = ff_avg_h264_qpel8_mc21_neon;
+        c->avg_h264_qpel_pixels_tab[1][ 7] = ff_avg_h264_qpel8_mc31_neon;
+        c->avg_h264_qpel_pixels_tab[1][ 8] = ff_avg_h264_qpel8_mc02_neon;
+        c->avg_h264_qpel_pixels_tab[1][ 9] = ff_avg_h264_qpel8_mc12_neon;
+        c->avg_h264_qpel_pixels_tab[1][10] = ff_avg_h264_qpel8_mc22_neon;
+        c->avg_h264_qpel_pixels_tab[1][11] = ff_avg_h264_qpel8_mc32_neon;
+        c->avg_h264_qpel_pixels_tab[1][12] = ff_avg_h264_qpel8_mc03_neon;
+        c->avg_h264_qpel_pixels_tab[1][13] = ff_avg_h264_qpel8_mc13_neon;
+        c->avg_h264_qpel_pixels_tab[1][14] = ff_avg_h264_qpel8_mc23_neon;
+        c->avg_h264_qpel_pixels_tab[1][15] = ff_avg_h264_qpel8_mc33_neon;
+    }
+}

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aarch64/hpeldsp_init_aarch64.c

@@ -0,0 +1,123 @@
+/*
+ * ARM NEON optimised DSP functions
+ * Copyright (c) 2008 Mans Rullgard <mans@mansr.com>
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include <stddef.h>
+#include <stdint.h>
+
+#include "config.h"
+
+#include "libavutil/attributes.h"
+#include "libavutil/cpu.h"
+#include "libavutil/aarch64/cpu.h"
+#include "libavcodec/hpeldsp.h"
+
+void     ff_put_pixels16_neon(uint8_t *block, const uint8_t *pixels,
+                              ptrdiff_t line_size, int h);
+void  ff_put_pixels16_x2_neon(uint8_t *block, const uint8_t *pixels,
+                              ptrdiff_t line_size, int h);
+void  ff_put_pixels16_y2_neon(uint8_t *block, const uint8_t *pixels,
+                              ptrdiff_t line_size, int h);
+void ff_put_pixels16_xy2_neon(uint8_t *block, const uint8_t *pixels,
+                              ptrdiff_t line_size, int h);
+void      ff_put_pixels8_neon(uint8_t *block, const uint8_t *pixels,
+                              ptrdiff_t line_size, int h);
+void   ff_put_pixels8_x2_neon(uint8_t *block, const uint8_t *pixels,
+                              ptrdiff_t line_size, int h);
+void   ff_put_pixels8_y2_neon(uint8_t *block, const uint8_t *pixels,
+                              ptrdiff_t line_size, int h);
+void  ff_put_pixels8_xy2_neon(uint8_t *block, const uint8_t *pixels,
+                              ptrdiff_t line_size, int h);
+
+void  ff_put_pixels16_x2_no_rnd_neon(uint8_t *block, const uint8_t *pixels,
+                                     ptrdiff_t line_size, int h);
+void  ff_put_pixels16_y2_no_rnd_neon(uint8_t *block, const uint8_t *pixels,
+                                     ptrdiff_t line_size, int h);
+void ff_put_pixels16_xy2_no_rnd_neon(uint8_t *block, const uint8_t *pixels,
+                                     ptrdiff_t line_size, int h);
+void   ff_put_pixels8_x2_no_rnd_neon(uint8_t *block, const uint8_t *pixels,
+                                     ptrdiff_t line_size, int h);
+void   ff_put_pixels8_y2_no_rnd_neon(uint8_t *block, const uint8_t *pixels,
+                                     ptrdiff_t line_size, int h);
+void  ff_put_pixels8_xy2_no_rnd_neon(uint8_t *block, const uint8_t *pixels,
+                                     ptrdiff_t line_size, int h);
+
+void     ff_avg_pixels16_neon(uint8_t *block, const uint8_t *pixels,
+                              ptrdiff_t line_size, int h);
+void  ff_avg_pixels16_x2_neon(uint8_t *block, const uint8_t *pixels,
+                              ptrdiff_t line_size, int h);
+void  ff_avg_pixels16_y2_neon(uint8_t *block, const uint8_t *pixels,
+                              ptrdiff_t line_size, int h);
+void ff_avg_pixels16_xy2_neon(uint8_t *block, const uint8_t *pixels,
+                              ptrdiff_t line_size, int h);
+void      ff_avg_pixels8_neon(uint8_t *block, const uint8_t *pixels,
+                              ptrdiff_t line_size, int h);
+void   ff_avg_pixels8_x2_neon(uint8_t *block, const uint8_t *pixels,
+                              ptrdiff_t line_size, int h);
+void   ff_avg_pixels8_y2_neon(uint8_t *block, const uint8_t *pixels,
+                              ptrdiff_t line_size, int h);
+void  ff_avg_pixels8_xy2_neon(uint8_t *block, const uint8_t *pixels,
+                              ptrdiff_t line_size, int h);
+
+void  ff_avg_pixels16_x2_no_rnd_neon(uint8_t *block, const uint8_t *pixels,
+                                     ptrdiff_t line_size, int h);
+void  ff_avg_pixels16_y2_no_rnd_neon(uint8_t *block, const uint8_t *pixels,
+                                     ptrdiff_t line_size, int h);
+void ff_avg_pixels16_xy2_no_rnd_neon(uint8_t *block, const uint8_t *pixels,
+                                     ptrdiff_t line_size, int h);
+
+av_cold void ff_hpeldsp_init_aarch64(HpelDSPContext *c, int flags)
+{
+    int cpu_flags = av_get_cpu_flags();
+
+    if (have_neon(cpu_flags)) {
+        c->put_pixels_tab[0][0] = ff_put_pixels16_neon;
+        c->put_pixels_tab[0][1] = ff_put_pixels16_x2_neon;
+        c->put_pixels_tab[0][2] = ff_put_pixels16_y2_neon;
+        c->put_pixels_tab[0][3] = ff_put_pixels16_xy2_neon;
+        c->put_pixels_tab[1][0] = ff_put_pixels8_neon;
+        c->put_pixels_tab[1][1] = ff_put_pixels8_x2_neon;
+        c->put_pixels_tab[1][2] = ff_put_pixels8_y2_neon;
+        c->put_pixels_tab[1][3] = ff_put_pixels8_xy2_neon;
+
+        c->put_no_rnd_pixels_tab[0][0] = ff_put_pixels16_neon;
+        c->put_no_rnd_pixels_tab[0][1] = ff_put_pixels16_x2_no_rnd_neon;
+        c->put_no_rnd_pixels_tab[0][2] = ff_put_pixels16_y2_no_rnd_neon;
+        c->put_no_rnd_pixels_tab[0][3] = ff_put_pixels16_xy2_no_rnd_neon;
+        c->put_no_rnd_pixels_tab[1][0] = ff_put_pixels8_neon;
+        c->put_no_rnd_pixels_tab[1][1] = ff_put_pixels8_x2_no_rnd_neon;
+        c->put_no_rnd_pixels_tab[1][2] = ff_put_pixels8_y2_no_rnd_neon;
+        c->put_no_rnd_pixels_tab[1][3] = ff_put_pixels8_xy2_no_rnd_neon;
+
+        c->avg_pixels_tab[0][0] = ff_avg_pixels16_neon;
+        c->avg_pixels_tab[0][1] = ff_avg_pixels16_x2_neon;
+        c->avg_pixels_tab[0][2] = ff_avg_pixels16_y2_neon;
+        c->avg_pixels_tab[0][3] = ff_avg_pixels16_xy2_neon;
+        c->avg_pixels_tab[1][0] = ff_avg_pixels8_neon;
+        c->avg_pixels_tab[1][1] = ff_avg_pixels8_x2_neon;
+        c->avg_pixels_tab[1][2] = ff_avg_pixels8_y2_neon;
+        c->avg_pixels_tab[1][3] = ff_avg_pixels8_xy2_neon;
+
+        c->avg_no_rnd_pixels_tab[0] = ff_avg_pixels16_neon;
+        c->avg_no_rnd_pixels_tab[1] = ff_avg_pixels16_x2_no_rnd_neon;
+        c->avg_no_rnd_pixels_tab[2] = ff_avg_pixels16_y2_no_rnd_neon;
+        c->avg_no_rnd_pixels_tab[3] = ff_avg_pixels16_xy2_no_rnd_neon;
+    }
+}

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aarch64/idct.h

@@ -0,0 +1,28 @@
+/*
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#ifndef AVCODEC_AARCH64_IDCT_H
+#define AVCODEC_AARCH64_IDCT_H
+
+#include <stdint.h>
+
+void ff_simple_idct_neon(int16_t *data);
+void ff_simple_idct_put_neon(uint8_t *dest, ptrdiff_t line_size, int16_t *data);
+void ff_simple_idct_add_neon(uint8_t *dest, ptrdiff_t line_size, int16_t *data);
+
+#endif /* AVCODEC_AARCH64_IDCT_H */

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aarch64/idctdsp_init_aarch64.c

@@ -0,0 +1,41 @@
+/*
+ * ARM-NEON-optimized IDCT functions
+ * Copyright (c) 2008 Mans Rullgard <mans@mansr.com>
+ * Copyright (c) 2017 Matthieu Bouron <matthieu.bouron@gmail.com>
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include "libavutil/attributes.h"
+#include "libavcodec/avcodec.h"
+#include "libavcodec/idctdsp.h"
+#include "idct.h"
+
+av_cold void ff_idctdsp_init_aarch64(IDCTDSPContext *c, AVCodecContext *avctx,
+                                     unsigned high_bit_depth)
+{
+    if (!avctx->lowres && !high_bit_depth) {
+        if (avctx->idct_algo == FF_IDCT_AUTO ||
+            avctx->idct_algo == FF_IDCT_SIMPLEAUTO ||
+            avctx->idct_algo == FF_IDCT_SIMPLENEON) {
+            c->idct_put  = ff_simple_idct_put_neon;
+            c->idct_add  = ff_simple_idct_add_neon;
+            c->idct      = ff_simple_idct_neon;
+            c->perm_type = FF_IDCT_PERM_PARTTRANS;
+        }
+    }
+}

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aarch64/mpegaudiodsp_init.c

@@ -0,0 +1,40 @@
+/*
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include <stddef.h>
+#include <stdint.h>
+
+#include "libavutil/attributes.h"
+#include "libavutil/aarch64/cpu.h"
+#include "libavcodec/mpegaudiodsp.h"
+#include "config.h"
+
+void ff_mpadsp_apply_window_fixed_neon(int32_t *synth_buf, int32_t *window,
+                                       int *dither, int16_t *samples, ptrdiff_t incr);
+void ff_mpadsp_apply_window_float_neon(float *synth_buf, float *window,
+                                       int *dither, float *samples, ptrdiff_t incr);
+
+av_cold void ff_mpadsp_init_aarch64(MPADSPContext *s)
+{
+    int cpu_flags = av_get_cpu_flags();
+
+    if (have_neon(cpu_flags)) {
+        s->apply_window_fixed = ff_mpadsp_apply_window_fixed_neon;
+        s->apply_window_float = ff_mpadsp_apply_window_float_neon;
+    }
+}

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aarch64/opusdsp_init.c

@@ -0,0 +1,35 @@
+/*
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include "config.h"
+
+#include "libavutil/aarch64/cpu.h"
+#include "libavcodec/opusdsp.h"
+
+void ff_opus_postfilter_neon(float *data, int period, float *gains, int len);
+float ff_opus_deemphasis_neon(float *out, float *in, float coeff, int len);
+
+av_cold void ff_opus_dsp_init_aarch64(OpusDSP *ctx)
+{
+    int cpu_flags = av_get_cpu_flags();
+
+    if (have_neon(cpu_flags)) {
+        ctx->postfilter = ff_opus_postfilter_neon;
+        ctx->deemphasis = ff_opus_deemphasis_neon;
+    }
+}

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aarch64/rv40dsp_init_aarch64.c

@@ -0,0 +1,48 @@
+/*
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include <stdint.h>
+
+#include "libavutil/attributes.h"
+#include "libavutil/cpu.h"
+#include "libavutil/aarch64/cpu.h"
+#include "libavcodec/rv34dsp.h"
+
+#include "config.h"
+
+void ff_put_rv40_chroma_mc8_neon(uint8_t *dst, uint8_t *src, ptrdiff_t stride,
+                                 int h, int x, int y);
+void ff_put_rv40_chroma_mc4_neon(uint8_t *dst, uint8_t *src, ptrdiff_t stride,
+                                 int h, int x, int y);
+
+void ff_avg_rv40_chroma_mc8_neon(uint8_t *dst, uint8_t *src, ptrdiff_t stride,
+                                 int h, int x, int y);
+void ff_avg_rv40_chroma_mc4_neon(uint8_t *dst, uint8_t *src, ptrdiff_t stride,
+                                 int h, int x, int y);
+
+av_cold void ff_rv40dsp_init_aarch64(RV34DSPContext *c)
+{
+    int cpu_flags = av_get_cpu_flags();
+
+    if (have_neon(cpu_flags)) {
+        c->put_chroma_pixels_tab[0] = ff_put_rv40_chroma_mc8_neon;
+        c->put_chroma_pixels_tab[1] = ff_put_rv40_chroma_mc4_neon;
+        c->avg_chroma_pixels_tab[0] = ff_avg_rv40_chroma_mc8_neon;
+        c->avg_chroma_pixels_tab[1] = ff_avg_rv40_chroma_mc4_neon;
+    }
+}

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aarch64/sbrdsp_init_aarch64.c

@@ -0,0 +1,70 @@
+/*
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include "config.h"
+#include "libavutil/aarch64/cpu.h"
+#include "libavutil/attributes.h"
+#include "libavcodec/sbrdsp.h"
+
+void ff_sbr_sum64x5_neon(float *z);
+float ff_sbr_sum_square_neon(float (*x)[2], int n);
+void ff_sbr_neg_odd_64_neon(float *x);
+void ff_sbr_qmf_pre_shuffle_neon(float *z);
+void ff_sbr_qmf_post_shuffle_neon(float W[32][2], const float *z);
+void ff_sbr_qmf_deint_neg_neon(float *v, const float *src);
+void ff_sbr_qmf_deint_bfly_neon(float *v, const float *src0, const float *src1);
+void ff_sbr_hf_g_filt_neon(float (*Y)[2], const float (*X_high)[40][2],
+                           const float *g_filt, int m_max, intptr_t ixh);
+void ff_sbr_hf_gen_neon(float (*X_high)[2], const float (*X_low)[2],
+                        const float alpha0[2], const float alpha1[2],
+                        float bw, int start, int end);
+void ff_sbr_autocorrelate_neon(const float x[40][2], float phi[3][2][2]);
+void ff_sbr_hf_apply_noise_0_neon(float Y[64][2], const float *s_m,
+                                  const float *q_filt, int noise,
+                                  int kx, int m_max);
+void ff_sbr_hf_apply_noise_1_neon(float Y[64][2], const float *s_m,
+                                  const float *q_filt, int noise,
+                                  int kx, int m_max);
+void ff_sbr_hf_apply_noise_2_neon(float Y[64][2], const float *s_m,
+                                  const float *q_filt, int noise,
+                                  int kx, int m_max);
+void ff_sbr_hf_apply_noise_3_neon(float Y[64][2], const float *s_m,
+                                  const float *q_filt, int noise,
+                                  int kx, int m_max);
+
+av_cold void ff_sbrdsp_init_aarch64(SBRDSPContext *s)
+{
+    int cpu_flags = av_get_cpu_flags();
+
+    if (have_neon(cpu_flags)) {
+        s->sum64x5 = ff_sbr_sum64x5_neon;
+        s->sum_square = ff_sbr_sum_square_neon;
+        s->neg_odd_64 = ff_sbr_neg_odd_64_neon;
+        s->qmf_pre_shuffle = ff_sbr_qmf_pre_shuffle_neon;
+        s->qmf_post_shuffle = ff_sbr_qmf_post_shuffle_neon;
+        s->qmf_deint_neg = ff_sbr_qmf_deint_neg_neon;
+        s->qmf_deint_bfly = ff_sbr_qmf_deint_bfly_neon;
+        s->hf_g_filt = ff_sbr_hf_g_filt_neon;
+        s->hf_gen = ff_sbr_hf_gen_neon;
+        s->autocorrelate = ff_sbr_autocorrelate_neon;
+        s->hf_apply_noise[0] = ff_sbr_hf_apply_noise_0_neon;
+        s->hf_apply_noise[1] = ff_sbr_hf_apply_noise_1_neon;
+        s->hf_apply_noise[2] = ff_sbr_hf_apply_noise_2_neon;
+        s->hf_apply_noise[3] = ff_sbr_hf_apply_noise_3_neon;
+    }
+}

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aarch64/synth_filter_init.c

@@ -0,0 +1,47 @@
+/*
+ * Copyright (c) 2010 Mans Rullgard <mans@mansr.com>
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include "config.h"
+
+#include "libavutil/aarch64/cpu.h"
+#include "libavutil/attributes.h"
+#include "libavutil/internal.h"
+#include "libavcodec/fft.h"
+#include "libavcodec/synth_filter.h"
+
+#include "asm-offsets.h"
+
+#if HAVE_NEON || HAVE_VFP
+AV_CHECK_OFFSET(FFTContext, imdct_half, IMDCT_HALF);
+#endif
+
+void ff_synth_filter_float_neon(FFTContext *imdct,
+                                float *synth_buf_ptr, int *synth_buf_offset,
+                                float synth_buf2[32], const float window[512],
+                                float out[32], const float in[32],
+                                float scale);
+
+av_cold void ff_synth_filter_init_aarch64(SynthFilterContext *s)
+{
+    int cpu_flags = av_get_cpu_flags();
+
+    if (have_neon(cpu_flags))
+        s->synth_filter_float = ff_synth_filter_float_neon;
+}

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aarch64/vc1dsp_init_aarch64.c

@@ -0,0 +1,47 @@
+/*
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include <stdint.h>
+
+#include "libavutil/attributes.h"
+#include "libavutil/cpu.h"
+#include "libavutil/aarch64/cpu.h"
+#include "libavcodec/vc1dsp.h"
+
+#include "config.h"
+
+void ff_put_vc1_chroma_mc8_neon(uint8_t *dst, uint8_t *src, ptrdiff_t stride,
+                                int h, int x, int y);
+void ff_avg_vc1_chroma_mc8_neon(uint8_t *dst, uint8_t *src, ptrdiff_t stride,
+                                int h, int x, int y);
+void ff_put_vc1_chroma_mc4_neon(uint8_t *dst, uint8_t *src, ptrdiff_t stride,
+                                int h, int x, int y);
+void ff_avg_vc1_chroma_mc4_neon(uint8_t *dst, uint8_t *src, ptrdiff_t stride,
+                                int h, int x, int y);
+
+av_cold void ff_vc1dsp_init_aarch64(VC1DSPContext *dsp)
+{
+    int cpu_flags = av_get_cpu_flags();
+
+    if (have_neon(cpu_flags)) {
+        dsp->put_no_rnd_vc1_chroma_pixels_tab[0] = ff_put_vc1_chroma_mc8_neon;
+        dsp->avg_no_rnd_vc1_chroma_pixels_tab[0] = ff_avg_vc1_chroma_mc8_neon;
+        dsp->put_no_rnd_vc1_chroma_pixels_tab[1] = ff_put_vc1_chroma_mc4_neon;
+        dsp->avg_no_rnd_vc1_chroma_pixels_tab[1] = ff_avg_vc1_chroma_mc4_neon;
+    }
+}

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aarch64/videodsp_init.c

@@ -0,0 +1,32 @@
+/*
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include "libavutil/attributes.h"
+#include "libavutil/cpu.h"
+#include "libavutil/aarch64/cpu.h"
+#include "libavcodec/videodsp.h"
+
+void ff_prefetch_aarch64(uint8_t *mem, ptrdiff_t stride, int h);
+
+av_cold void ff_videodsp_init_aarch64(VideoDSPContext *ctx, int bpc)
+{
+    int cpu_flags = av_get_cpu_flags();
+
+    if (have_armv8(cpu_flags))
+        ctx->prefetch = ff_prefetch_aarch64;
+}

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aarch64/vorbisdsp_init.c

@@ -0,0 +1,34 @@
+/*
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include "libavutil/attributes.h"
+#include "libavutil/cpu.h"
+#include "libavutil/aarch64/cpu.h"
+#include "libavcodec/vorbisdsp.h"
+
+void ff_vorbis_inverse_coupling_neon(float *mag, float *ang,
+                                     intptr_t blocksize);
+
+av_cold void ff_vorbisdsp_init_aarch64(VorbisDSPContext *c)
+{
+    int cpu_flags = av_get_cpu_flags();
+
+    if (have_neon(cpu_flags)) {
+        c->vorbis_inverse_coupling = ff_vorbis_inverse_coupling_neon;
+    }
+}

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aarch64/vp8dsp.h

@@ -0,0 +1,75 @@
+/*
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#ifndef AVCODEC_AARCH64_VP8DSP_H
+#define AVCODEC_AARCH64_VP8DSP_H
+
+#include "libavcodec/vp8dsp.h"
+
+#define VP8_LF_Y(hv, inner, opt)                                             \
+    void ff_vp8_##hv##_loop_filter16##inner##_##opt(uint8_t *dst,            \
+                                                    ptrdiff_t stride,        \
+                                                    int flim_E, int flim_I,  \
+                                                    int hev_thresh)
+
+#define VP8_LF_UV(hv, inner, opt)                                            \
+    void ff_vp8_##hv##_loop_filter8uv##inner##_##opt(uint8_t *dstU,          \
+                                                     uint8_t *dstV,          \
+                                                     ptrdiff_t stride,       \
+                                                     int flim_E, int flim_I, \
+                                                     int hev_thresh)
+
+#define VP8_LF_SIMPLE(hv, opt)                                          \
+    void ff_vp8_##hv##_loop_filter16_simple_##opt(uint8_t *dst,         \
+                                                  ptrdiff_t stride,     \
+                                                  int flim)
+
+#define VP8_LF_HV(inner, opt)                   \
+    VP8_LF_Y(h,  inner, opt);                   \
+    VP8_LF_Y(v,  inner, opt);                   \
+    VP8_LF_UV(h, inner, opt);                   \
+    VP8_LF_UV(v, inner, opt)
+
+#define VP8_LF(opt)                             \
+    VP8_LF_HV(,       opt);                     \
+    VP8_LF_HV(_inner, opt);                     \
+    VP8_LF_SIMPLE(h, opt);                      \
+    VP8_LF_SIMPLE(v, opt)
+
+#define VP8_MC(n, opt)                                                  \
+    void ff_put_vp8_##n##_##opt(uint8_t *dst, ptrdiff_t dststride,      \
+                                uint8_t *src, ptrdiff_t srcstride,      \
+                                int h, int x, int y)
+
+#define VP8_EPEL(w, opt)                        \
+    VP8_MC(pixels ## w, opt);                   \
+    VP8_MC(epel ## w ## _h4, opt);              \
+    VP8_MC(epel ## w ## _h6, opt);              \
+    VP8_MC(epel ## w ## _v4, opt);              \
+    VP8_MC(epel ## w ## _h4v4, opt);            \
+    VP8_MC(epel ## w ## _h6v4, opt);            \
+    VP8_MC(epel ## w ## _v6, opt);              \
+    VP8_MC(epel ## w ## _h4v6, opt);            \
+    VP8_MC(epel ## w ## _h6v6, opt)
+
+#define VP8_BILIN(w, opt)                       \
+    VP8_MC(bilin ## w ## _h, opt);              \
+    VP8_MC(bilin ## w ## _v, opt);              \
+    VP8_MC(bilin ## w ## _hv, opt)
+
+#endif /* AVCODEC_AARCH64_VP8DSP_H */

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aarch64/vp8dsp_init_aarch64.c

@@ -0,0 +1,124 @@
+/*
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include <stdint.h>
+
+#include "libavutil/attributes.h"
+#include "libavutil/aarch64/cpu.h"
+#include "libavcodec/vp8dsp.h"
+#include "vp8dsp.h"
+
+void ff_vp8_luma_dc_wht_neon(int16_t block[4][4][16], int16_t dc[16]);
+
+void ff_vp8_idct_add_neon(uint8_t *dst, int16_t block[16], ptrdiff_t stride);
+void ff_vp8_idct_dc_add_neon(uint8_t *dst, int16_t block[16], ptrdiff_t stride);
+void ff_vp8_idct_dc_add4y_neon(uint8_t *dst, int16_t block[4][16], ptrdiff_t stride);
+void ff_vp8_idct_dc_add4uv_neon(uint8_t *dst, int16_t block[4][16], ptrdiff_t stride);
+
+VP8_LF(neon);
+
+VP8_EPEL(16, neon);
+VP8_EPEL(8,  neon);
+VP8_EPEL(4,  neon);
+
+VP8_BILIN(16, neon);
+VP8_BILIN(8,  neon);
+VP8_BILIN(4,  neon);
+
+av_cold void ff_vp78dsp_init_aarch64(VP8DSPContext *dsp)
+{
+    if (!have_neon(av_get_cpu_flags()))
+        return;
+    dsp->put_vp8_epel_pixels_tab[0][0][0] = ff_put_vp8_pixels16_neon;
+    dsp->put_vp8_epel_pixels_tab[0][0][2] = ff_put_vp8_epel16_h6_neon;
+    dsp->put_vp8_epel_pixels_tab[0][2][0] = ff_put_vp8_epel16_v6_neon;
+    dsp->put_vp8_epel_pixels_tab[0][2][2] = ff_put_vp8_epel16_h6v6_neon;
+
+    dsp->put_vp8_epel_pixels_tab[1][0][0] = ff_put_vp8_pixels8_neon;
+    dsp->put_vp8_epel_pixels_tab[1][0][1] = ff_put_vp8_epel8_h4_neon;
+    dsp->put_vp8_epel_pixels_tab[1][0][2] = ff_put_vp8_epel8_h6_neon;
+    dsp->put_vp8_epel_pixels_tab[1][1][0] = ff_put_vp8_epel8_v4_neon;
+    dsp->put_vp8_epel_pixels_tab[1][1][1] = ff_put_vp8_epel8_h4v4_neon;
+    dsp->put_vp8_epel_pixels_tab[1][1][2] = ff_put_vp8_epel8_h6v4_neon;
+    dsp->put_vp8_epel_pixels_tab[1][2][0] = ff_put_vp8_epel8_v6_neon;
+    dsp->put_vp8_epel_pixels_tab[1][2][1] = ff_put_vp8_epel8_h4v6_neon;
+    dsp->put_vp8_epel_pixels_tab[1][2][2] = ff_put_vp8_epel8_h6v6_neon;
+
+    dsp->put_vp8_epel_pixels_tab[2][0][1] = ff_put_vp8_epel4_h4_neon;
+    dsp->put_vp8_epel_pixels_tab[2][0][2] = ff_put_vp8_epel4_h6_neon;
+    dsp->put_vp8_epel_pixels_tab[2][1][0] = ff_put_vp8_epel4_v4_neon;
+    dsp->put_vp8_epel_pixels_tab[2][1][1] = ff_put_vp8_epel4_h4v4_neon;
+    dsp->put_vp8_epel_pixels_tab[2][1][2] = ff_put_vp8_epel4_h6v4_neon;
+    dsp->put_vp8_epel_pixels_tab[2][2][0] = ff_put_vp8_epel4_v6_neon;
+    dsp->put_vp8_epel_pixels_tab[2][2][1] = ff_put_vp8_epel4_h4v6_neon;
+    dsp->put_vp8_epel_pixels_tab[2][2][2] = ff_put_vp8_epel4_h6v6_neon;
+
+    dsp->put_vp8_bilinear_pixels_tab[0][0][0] = ff_put_vp8_pixels16_neon;
+    dsp->put_vp8_bilinear_pixels_tab[0][0][1] = ff_put_vp8_bilin16_h_neon;
+    dsp->put_vp8_bilinear_pixels_tab[0][0][2] = ff_put_vp8_bilin16_h_neon;
+    dsp->put_vp8_bilinear_pixels_tab[0][1][0] = ff_put_vp8_bilin16_v_neon;
+    dsp->put_vp8_bilinear_pixels_tab[0][1][1] = ff_put_vp8_bilin16_hv_neon;
+    dsp->put_vp8_bilinear_pixels_tab[0][1][2] = ff_put_vp8_bilin16_hv_neon;
+    dsp->put_vp8_bilinear_pixels_tab[0][2][0] = ff_put_vp8_bilin16_v_neon;
+    dsp->put_vp8_bilinear_pixels_tab[0][2][1] = ff_put_vp8_bilin16_hv_neon;
+    dsp->put_vp8_bilinear_pixels_tab[0][2][2] = ff_put_vp8_bilin16_hv_neon;
+
+    dsp->put_vp8_bilinear_pixels_tab[1][0][0] = ff_put_vp8_pixels8_neon;
+    dsp->put_vp8_bilinear_pixels_tab[1][0][1] = ff_put_vp8_bilin8_h_neon;
+    dsp->put_vp8_bilinear_pixels_tab[1][0][2] = ff_put_vp8_bilin8_h_neon;
+    dsp->put_vp8_bilinear_pixels_tab[1][1][0] = ff_put_vp8_bilin8_v_neon;
+    dsp->put_vp8_bilinear_pixels_tab[1][1][1] = ff_put_vp8_bilin8_hv_neon;
+    dsp->put_vp8_bilinear_pixels_tab[1][1][2] = ff_put_vp8_bilin8_hv_neon;
+    dsp->put_vp8_bilinear_pixels_tab[1][2][0] = ff_put_vp8_bilin8_v_neon;
+    dsp->put_vp8_bilinear_pixels_tab[1][2][1] = ff_put_vp8_bilin8_hv_neon;
+    dsp->put_vp8_bilinear_pixels_tab[1][2][2] = ff_put_vp8_bilin8_hv_neon;
+
+    dsp->put_vp8_bilinear_pixels_tab[2][0][1] = ff_put_vp8_bilin4_h_neon;
+    dsp->put_vp8_bilinear_pixels_tab[2][0][2] = ff_put_vp8_bilin4_h_neon;
+    dsp->put_vp8_bilinear_pixels_tab[2][1][0] = ff_put_vp8_bilin4_v_neon;
+    dsp->put_vp8_bilinear_pixels_tab[2][1][1] = ff_put_vp8_bilin4_hv_neon;
+    dsp->put_vp8_bilinear_pixels_tab[2][1][2] = ff_put_vp8_bilin4_hv_neon;
+    dsp->put_vp8_bilinear_pixels_tab[2][2][0] = ff_put_vp8_bilin4_v_neon;
+    dsp->put_vp8_bilinear_pixels_tab[2][2][1] = ff_put_vp8_bilin4_hv_neon;
+    dsp->put_vp8_bilinear_pixels_tab[2][2][2] = ff_put_vp8_bilin4_hv_neon;
+}
+
+av_cold void ff_vp8dsp_init_aarch64(VP8DSPContext *dsp)
+{
+    if (!have_neon(av_get_cpu_flags()))
+        return;
+    dsp->vp8_luma_dc_wht    = ff_vp8_luma_dc_wht_neon;
+
+    dsp->vp8_idct_add       = ff_vp8_idct_add_neon;
+    dsp->vp8_idct_dc_add    = ff_vp8_idct_dc_add_neon;
+    dsp->vp8_idct_dc_add4y  = ff_vp8_idct_dc_add4y_neon;
+    dsp->vp8_idct_dc_add4uv = ff_vp8_idct_dc_add4uv_neon;
+
+    dsp->vp8_v_loop_filter16y = ff_vp8_v_loop_filter16_neon;
+    dsp->vp8_h_loop_filter16y = ff_vp8_h_loop_filter16_neon;
+    dsp->vp8_v_loop_filter8uv = ff_vp8_v_loop_filter8uv_neon;
+    dsp->vp8_h_loop_filter8uv = ff_vp8_h_loop_filter8uv_neon;
+
+    dsp->vp8_v_loop_filter16y_inner = ff_vp8_v_loop_filter16_inner_neon;
+    dsp->vp8_h_loop_filter16y_inner = ff_vp8_h_loop_filter16_inner_neon;
+    dsp->vp8_v_loop_filter8uv_inner = ff_vp8_v_loop_filter8uv_inner_neon;
+    dsp->vp8_h_loop_filter8uv_inner = ff_vp8_h_loop_filter8uv_inner_neon;
+
+    dsp->vp8_v_loop_filter_simple = ff_vp8_v_loop_filter16_simple_neon;
+    dsp->vp8_h_loop_filter_simple = ff_vp8_h_loop_filter16_simple_neon;
+}

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aarch64/vp9dsp_init.h

@@ -0,0 +1,29 @@
+/*
+ * Copyright (c) 2017 Google Inc.
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#ifndef AVCODEC_AARCH64_VP9DSP_INIT_H
+#define AVCODEC_AARCH64_VP9DSP_INIT_H
+
+#include "libavcodec/vp9dsp.h"
+
+void ff_vp9dsp_init_10bpp_aarch64(VP9DSPContext *dsp);
+void ff_vp9dsp_init_12bpp_aarch64(VP9DSPContext *dsp);
+
+#endif /* AVCODEC_AARCH64_VP9DSP_INIT_H */

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aarch64/vp9dsp_init_10bpp_aarch64.c

@@ -0,0 +1,23 @@
+/*
+ * Copyright (c) 2017 Google Inc.
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#define BPP 10
+#define INIT_FUNC ff_vp9dsp_init_10bpp_aarch64
+#include "vp9dsp_init_16bpp_aarch64_template.c"

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aarch64/vp9dsp_init_12bpp_aarch64.c

@@ -0,0 +1,23 @@
+/*
+ * Copyright (c) 2017 Google Inc.
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#define BPP 12
+#define INIT_FUNC ff_vp9dsp_init_12bpp_aarch64
+#include "vp9dsp_init_16bpp_aarch64_template.c"

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aarch64/vp9dsp_init_16bpp_aarch64_template.c

@@ -0,0 +1,273 @@
+/*
+ * Copyright (c) 2017 Google Inc.
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include <stdint.h>
+
+#include "libavutil/attributes.h"
+#include "libavutil/internal.h"
+#include "libavutil/aarch64/cpu.h"
+#include "vp9dsp_init.h"
+
+#define declare_fpel(type, sz, suffix)                                          \
+void ff_vp9_##type##sz##suffix##_neon(uint8_t *dst, ptrdiff_t dst_stride,       \
+                                      const uint8_t *src, ptrdiff_t src_stride, \
+                                      int h, int mx, int my)
+
+#define decl_mc_func(op, filter, dir, sz, bpp)                                                   \
+void ff_vp9_##op##_##filter##sz##_##dir##_##bpp##_neon(uint8_t *dst, ptrdiff_t dst_stride,       \
+                                                       const uint8_t *src, ptrdiff_t src_stride, \
+                                                       int h, int mx, int my)
+
+#define define_8tap_2d_fn(op, filter, sz, bpp)                                      \
+static void op##_##filter##sz##_hv_##bpp##_neon(uint8_t *dst, ptrdiff_t dst_stride, \
+                                                const uint8_t *src,                 \
+                                                ptrdiff_t src_stride,               \
+                                                int h, int mx, int my)              \
+{                                                                                   \
+    LOCAL_ALIGNED_16(uint8_t, temp, [((1 + (sz < 64)) * sz + 8) * sz * 2]);         \
+    /* We only need h + 7 lines, but the horizontal filter assumes an               \
+     * even number of rows, so filter h + 8 lines here. */                          \
+    ff_vp9_put_##filter##sz##_h_##bpp##_neon(temp, 2 * sz,                          \
+                                             src - 3 * src_stride, src_stride,      \
+                                             h + 8, mx, 0);                         \
+    ff_vp9_##op##_##filter##sz##_v_##bpp##_neon(dst, dst_stride,                    \
+                                                temp + 3 * 2 * sz, 2 * sz,          \
+                                                h, 0, my);                          \
+}
+
+#define decl_filter_funcs(op, dir, sz, bpp)  \
+    decl_mc_func(op, regular, dir, sz, bpp); \
+    decl_mc_func(op, sharp,   dir, sz, bpp); \
+    decl_mc_func(op, smooth,  dir, sz, bpp)
+
+#define decl_mc_funcs(sz, bpp)           \
+    decl_filter_funcs(put, h,  sz, bpp); \
+    decl_filter_funcs(avg, h,  sz, bpp); \
+    decl_filter_funcs(put, v,  sz, bpp); \
+    decl_filter_funcs(avg, v,  sz, bpp); \
+    decl_filter_funcs(put, hv, sz, bpp); \
+    decl_filter_funcs(avg, hv, sz, bpp)
+
+#define ff_vp9_copy32_neon  ff_vp9_copy32_aarch64
+#define ff_vp9_copy64_neon  ff_vp9_copy64_aarch64
+#define ff_vp9_copy128_neon ff_vp9_copy128_aarch64
+
+declare_fpel(copy, 128, );
+declare_fpel(copy, 64,  );
+declare_fpel(copy, 32,  );
+declare_fpel(copy, 16,  );
+declare_fpel(copy, 8,   );
+declare_fpel(avg, 64, _16);
+declare_fpel(avg, 32, _16);
+declare_fpel(avg, 16, _16);
+declare_fpel(avg, 8,  _16);
+declare_fpel(avg, 4,  _16);
+
+decl_mc_funcs(64, BPP);
+decl_mc_funcs(32, BPP);
+decl_mc_funcs(16, BPP);
+decl_mc_funcs(8, BPP);
+decl_mc_funcs(4, BPP);
+
+#define define_8tap_2d_funcs(sz, bpp)        \
+    define_8tap_2d_fn(put, regular, sz, bpp) \
+    define_8tap_2d_fn(put, sharp,   sz, bpp) \
+    define_8tap_2d_fn(put, smooth,  sz, bpp) \
+    define_8tap_2d_fn(avg, regular, sz, bpp) \
+    define_8tap_2d_fn(avg, sharp,   sz, bpp) \
+    define_8tap_2d_fn(avg, smooth,  sz, bpp)
+
+define_8tap_2d_funcs(64, BPP)
+define_8tap_2d_funcs(32, BPP)
+define_8tap_2d_funcs(16, BPP)
+define_8tap_2d_funcs(8,  BPP)
+define_8tap_2d_funcs(4,  BPP)
+
+static av_cold void vp9dsp_mc_init_aarch64(VP9DSPContext *dsp)
+{
+    int cpu_flags = av_get_cpu_flags();
+
+#define init_fpel(idx1, idx2, sz, type, suffix)      \
+    dsp->mc[idx1][FILTER_8TAP_SMOOTH ][idx2][0][0] = \
+    dsp->mc[idx1][FILTER_8TAP_REGULAR][idx2][0][0] = \
+    dsp->mc[idx1][FILTER_8TAP_SHARP  ][idx2][0][0] = \
+    dsp->mc[idx1][FILTER_BILINEAR    ][idx2][0][0] = ff_vp9_##type##sz##suffix
+
+#define init_copy(idx, sz, suffix) \
+    init_fpel(idx, 0, sz, copy, suffix)
+
+#define init_avg(idx, sz, suffix) \
+    init_fpel(idx, 1, sz, avg,  suffix)
+
+#define init_copy_avg(idx, sz1, sz2) \
+    init_copy(idx, sz2, _neon);      \
+    init_avg (idx, sz1, _16_neon)
+
+    if (have_armv8(cpu_flags)) {
+        init_copy(0, 128, _aarch64);
+        init_copy(1, 64,  _aarch64);
+        init_copy(2, 32,  _aarch64);
+    }
+
+    if (have_neon(cpu_flags)) {
+#define init_mc_func(idx1, idx2, op, filter, fname, dir, mx, my, sz, pfx, bpp) \
+    dsp->mc[idx1][filter][idx2][mx][my] = pfx##op##_##fname##sz##_##dir##_##bpp##_neon
+
+#define init_mc_funcs(idx, dir, mx, my, sz, pfx, bpp)                                   \
+    init_mc_func(idx, 0, put, FILTER_8TAP_REGULAR, regular, dir, mx, my, sz, pfx, bpp); \
+    init_mc_func(idx, 0, put, FILTER_8TAP_SHARP,   sharp,   dir, mx, my, sz, pfx, bpp); \
+    init_mc_func(idx, 0, put, FILTER_8TAP_SMOOTH,  smooth,  dir, mx, my, sz, pfx, bpp); \
+    init_mc_func(idx, 1, avg, FILTER_8TAP_REGULAR, regular, dir, mx, my, sz, pfx, bpp); \
+    init_mc_func(idx, 1, avg, FILTER_8TAP_SHARP,   sharp,   dir, mx, my, sz, pfx, bpp); \
+    init_mc_func(idx, 1, avg, FILTER_8TAP_SMOOTH,  smooth,  dir, mx, my, sz, pfx, bpp)
+
+#define init_mc_funcs_dirs(idx, sz, bpp)            \
+    init_mc_funcs(idx, v,  0, 1, sz, ff_vp9_, bpp); \
+    init_mc_funcs(idx, h,  1, 0, sz, ff_vp9_, bpp); \
+    init_mc_funcs(idx, hv, 1, 1, sz,        , bpp)
+
+
+        init_avg(0, 64, _16_neon);
+        init_avg(1, 32, _16_neon);
+        init_avg(2, 16, _16_neon);
+        init_copy_avg(3, 8, 16);
+        init_copy_avg(4, 4, 8);
+
+        init_mc_funcs_dirs(0, 64, BPP);
+        init_mc_funcs_dirs(1, 32, BPP);
+        init_mc_funcs_dirs(2, 16, BPP);
+        init_mc_funcs_dirs(3, 8,  BPP);
+        init_mc_funcs_dirs(4, 4,  BPP);
+    }
+}
+
+#define define_itxfm2(type_a, type_b, sz, bpp)                                     \
+void ff_vp9_##type_a##_##type_b##_##sz##x##sz##_add_##bpp##_neon(uint8_t *_dst,    \
+                                                                 ptrdiff_t stride, \
+                                                                 int16_t *_block, int eob)
+#define define_itxfm(type_a, type_b, sz, bpp) define_itxfm2(type_a, type_b, sz, bpp)
+
+#define define_itxfm_funcs(sz, bpp)      \
+    define_itxfm(idct,  idct,  sz, bpp); \
+    define_itxfm(iadst, idct,  sz, bpp); \
+    define_itxfm(idct,  iadst, sz, bpp); \
+    define_itxfm(iadst, iadst, sz, bpp)
+
+define_itxfm_funcs(4,  BPP);
+define_itxfm_funcs(8,  BPP);
+define_itxfm_funcs(16, BPP);
+define_itxfm(idct, idct, 32, BPP);
+define_itxfm(iwht, iwht, 4,  BPP);
+
+
+static av_cold void vp9dsp_itxfm_init_aarch64(VP9DSPContext *dsp)
+{
+    int cpu_flags = av_get_cpu_flags();
+
+    if (have_neon(cpu_flags)) {
+#define init_itxfm2(tx, sz, bpp)                                               \
+    dsp->itxfm_add[tx][DCT_DCT]   = ff_vp9_idct_idct_##sz##_add_##bpp##_neon;  \
+    dsp->itxfm_add[tx][DCT_ADST]  = ff_vp9_iadst_idct_##sz##_add_##bpp##_neon; \
+    dsp->itxfm_add[tx][ADST_DCT]  = ff_vp9_idct_iadst_##sz##_add_##bpp##_neon; \
+    dsp->itxfm_add[tx][ADST_ADST] = ff_vp9_iadst_iadst_##sz##_add_##bpp##_neon
+#define init_itxfm(tx, sz, bpp) init_itxfm2(tx, sz, bpp)
+
+#define init_idct2(tx, nm, bpp)     \
+    dsp->itxfm_add[tx][DCT_DCT]   = \
+    dsp->itxfm_add[tx][ADST_DCT]  = \
+    dsp->itxfm_add[tx][DCT_ADST]  = \
+    dsp->itxfm_add[tx][ADST_ADST] = ff_vp9_##nm##_add_##bpp##_neon
+#define init_idct(tx, nm, bpp) init_idct2(tx, nm, bpp)
+
+        init_itxfm(TX_4X4,   4x4,   BPP);
+        init_itxfm(TX_8X8,   8x8,   BPP);
+        init_itxfm(TX_16X16, 16x16, BPP);
+        init_idct(TX_32X32, idct_idct_32x32, BPP);
+        init_idct(4,        iwht_iwht_4x4,   BPP);
+    }
+}
+
+#define define_loop_filter(dir, wd, size, bpp) \
+void ff_vp9_loop_filter_##dir##_##wd##_##size##_##bpp##_neon(uint8_t *dst, ptrdiff_t stride, int E, int I, int H)
+
+#define define_loop_filters(wd, size, bpp) \
+    define_loop_filter(h, wd, size, bpp);  \
+    define_loop_filter(v, wd, size, bpp)
+
+define_loop_filters(4,  8,  BPP);
+define_loop_filters(8,  8,  BPP);
+define_loop_filters(16, 8,  BPP);
+
+define_loop_filters(16, 16, BPP);
+
+define_loop_filters(44, 16, BPP);
+define_loop_filters(48, 16, BPP);
+define_loop_filters(84, 16, BPP);
+define_loop_filters(88, 16, BPP);
+
+static av_cold void vp9dsp_loopfilter_init_aarch64(VP9DSPContext *dsp)
+{
+    int cpu_flags = av_get_cpu_flags();
+
+    if (have_neon(cpu_flags)) {
+#define init_lpf_func_8(idx1, idx2, dir, wd, bpp) \
+    dsp->loop_filter_8[idx1][idx2] = ff_vp9_loop_filter_##dir##_##wd##_8_##bpp##_neon
+
+#define init_lpf_func_16(idx, dir, bpp) \
+    dsp->loop_filter_16[idx] = ff_vp9_loop_filter_##dir##_16_16_##bpp##_neon
+
+#define init_lpf_func_mix2(idx1, idx2, idx3, dir, wd, bpp) \
+    dsp->loop_filter_mix2[idx1][idx2][idx3] = ff_vp9_loop_filter_##dir##_##wd##_16_##bpp##_neon
+
+#define init_lpf_funcs_8_wd(idx, wd, bpp) \
+    init_lpf_func_8(idx, 0, h, wd, bpp);  \
+    init_lpf_func_8(idx, 1, v, wd, bpp)
+
+#define init_lpf_funcs_16(bpp)   \
+    init_lpf_func_16(0, h, bpp); \
+    init_lpf_func_16(1, v, bpp)
+
+#define init_lpf_funcs_mix2_wd(idx1, idx2, wd, bpp) \
+    init_lpf_func_mix2(idx1, idx2, 0, h, wd, bpp);  \
+    init_lpf_func_mix2(idx1, idx2, 1, v, wd, bpp)
+
+#define init_lpf_funcs_8(bpp)        \
+    init_lpf_funcs_8_wd(0, 4,  bpp); \
+    init_lpf_funcs_8_wd(1, 8,  bpp); \
+    init_lpf_funcs_8_wd(2, 16, bpp)
+
+#define init_lpf_funcs_mix2(bpp)           \
+    init_lpf_funcs_mix2_wd(0, 0, 44, bpp); \
+    init_lpf_funcs_mix2_wd(0, 1, 48, bpp); \
+    init_lpf_funcs_mix2_wd(1, 0, 84, bpp); \
+    init_lpf_funcs_mix2_wd(1, 1, 88, bpp)
+
+        init_lpf_funcs_8(BPP);
+        init_lpf_funcs_16(BPP);
+        init_lpf_funcs_mix2(BPP);
+    }
+}
+
+av_cold void INIT_FUNC(VP9DSPContext *dsp)
+{
+    vp9dsp_mc_init_aarch64(dsp);
+    vp9dsp_loopfilter_init_aarch64(dsp);
+    vp9dsp_itxfm_init_aarch64(dsp);
+}

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/aarch64/vp9dsp_init_aarch64.c

@@ -0,0 +1,258 @@
+/*
+ * Copyright (c) 2016 Google Inc.
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include <stdint.h>
+
+#include "libavutil/attributes.h"
+#include "libavutil/internal.h"
+#include "libavutil/aarch64/cpu.h"
+#include "libavcodec/vp9dsp.h"
+#include "vp9dsp_init.h"
+
+#define declare_fpel(type, sz)                                          \
+void ff_vp9_##type##sz##_neon(uint8_t *dst, ptrdiff_t dst_stride,       \
+                              const uint8_t *src, ptrdiff_t src_stride, \
+                              int h, int mx, int my)
+
+#define declare_copy_avg(sz) \
+    declare_fpel(copy, sz);  \
+    declare_fpel(avg , sz)
+
+#define decl_mc_func(op, filter, dir, sz)                                                \
+void ff_vp9_##op##_##filter##sz##_##dir##_neon(uint8_t *dst, ptrdiff_t dst_stride,       \
+                                               const uint8_t *src, ptrdiff_t src_stride, \
+                                               int h, int mx, int my)
+
+#define define_8tap_2d_fn(op, filter, sz)                                         \
+static void op##_##filter##sz##_hv_neon(uint8_t *dst, ptrdiff_t dst_stride,       \
+                                        const uint8_t *src, ptrdiff_t src_stride, \
+                                        int h, int mx, int my)                    \
+{                                                                                 \
+    LOCAL_ALIGNED_16(uint8_t, temp, [((1 + (sz < 64)) * sz + 8) * sz]);           \
+    /* We only need h + 7 lines, but the horizontal filter assumes an             \
+     * even number of rows, so filter h + 8 lines here. */                        \
+    ff_vp9_put_##filter##sz##_h_neon(temp, sz,                                    \
+                                     src - 3 * src_stride, src_stride,            \
+                                     h + 8, mx, 0);                               \
+    ff_vp9_##op##_##filter##sz##_v_neon(dst, dst_stride,                          \
+                                        temp + 3 * sz, sz,                        \
+                                        h, 0, my);                                \
+}
+
+#define decl_filter_funcs(op, dir, sz)  \
+    decl_mc_func(op, regular, dir, sz); \
+    decl_mc_func(op, sharp,   dir, sz); \
+    decl_mc_func(op, smooth,  dir, sz)
+
+#define decl_mc_funcs(sz)           \
+    decl_filter_funcs(put, h,  sz); \
+    decl_filter_funcs(avg, h,  sz); \
+    decl_filter_funcs(put, v,  sz); \
+    decl_filter_funcs(avg, v,  sz); \
+    decl_filter_funcs(put, hv, sz); \
+    decl_filter_funcs(avg, hv, sz)
+
+#define ff_vp9_copy32_neon ff_vp9_copy32_aarch64
+#define ff_vp9_copy64_neon ff_vp9_copy64_aarch64
+
+declare_copy_avg(64);
+declare_copy_avg(32);
+declare_copy_avg(16);
+declare_copy_avg(8);
+declare_copy_avg(4);
+
+decl_mc_funcs(64);
+decl_mc_funcs(32);
+decl_mc_funcs(16);
+decl_mc_funcs(8);
+decl_mc_funcs(4);
+
+#define define_8tap_2d_funcs(sz)        \
+    define_8tap_2d_fn(put, regular, sz) \
+    define_8tap_2d_fn(put, sharp,   sz) \
+    define_8tap_2d_fn(put, smooth,  sz) \
+    define_8tap_2d_fn(avg, regular, sz) \
+    define_8tap_2d_fn(avg, sharp,   sz) \
+    define_8tap_2d_fn(avg, smooth,  sz)
+
+define_8tap_2d_funcs(64)
+define_8tap_2d_funcs(32)
+define_8tap_2d_funcs(16)
+define_8tap_2d_funcs(8)
+define_8tap_2d_funcs(4)
+
+static av_cold void vp9dsp_mc_init_aarch64(VP9DSPContext *dsp)
+{
+    int cpu_flags = av_get_cpu_flags();
+
+#define init_fpel(idx1, idx2, sz, type, suffix)      \
+    dsp->mc[idx1][FILTER_8TAP_SMOOTH ][idx2][0][0] = \
+    dsp->mc[idx1][FILTER_8TAP_REGULAR][idx2][0][0] = \
+    dsp->mc[idx1][FILTER_8TAP_SHARP  ][idx2][0][0] = \
+    dsp->mc[idx1][FILTER_BILINEAR    ][idx2][0][0] = ff_vp9_##type##sz##suffix
+
+#define init_copy(idx, sz, suffix) \
+    init_fpel(idx, 0, sz, copy, suffix)
+
+#define init_avg(idx, sz, suffix) \
+    init_fpel(idx, 1, sz, avg,  suffix)
+
+#define init_copy_avg(idx, sz) \
+    init_copy(idx, sz, _neon); \
+    init_avg (idx, sz, _neon)
+
+    if (have_armv8(cpu_flags)) {
+        init_copy(0, 64, _aarch64);
+        init_copy(1, 32, _aarch64);
+    }
+
+    if (have_neon(cpu_flags)) {
+#define init_mc_func(idx1, idx2, op, filter, fname, dir, mx, my, sz, pfx) \
+    dsp->mc[idx1][filter][idx2][mx][my] = pfx##op##_##fname##sz##_##dir##_neon
+
+#define init_mc_funcs(idx, dir, mx, my, sz, pfx)                                   \
+    init_mc_func(idx, 0, put, FILTER_8TAP_REGULAR, regular, dir, mx, my, sz, pfx); \
+    init_mc_func(idx, 0, put, FILTER_8TAP_SHARP,   sharp,   dir, mx, my, sz, pfx); \
+    init_mc_func(idx, 0, put, FILTER_8TAP_SMOOTH,  smooth,  dir, mx, my, sz, pfx); \
+    init_mc_func(idx, 1, avg, FILTER_8TAP_REGULAR, regular, dir, mx, my, sz, pfx); \
+    init_mc_func(idx, 1, avg, FILTER_8TAP_SHARP,   sharp,   dir, mx, my, sz, pfx); \
+    init_mc_func(idx, 1, avg, FILTER_8TAP_SMOOTH,  smooth,  dir, mx, my, sz, pfx)
+
+#define init_mc_funcs_dirs(idx, sz)            \
+    init_mc_funcs(idx, h,  1, 0, sz, ff_vp9_); \
+    init_mc_funcs(idx, v,  0, 1, sz, ff_vp9_); \
+    init_mc_funcs(idx, hv, 1, 1, sz,)
+
+        init_avg(0, 64, _neon);
+        init_avg(1, 32, _neon);
+        init_copy_avg(2, 16);
+        init_copy_avg(3, 8);
+        init_copy_avg(4, 4);
+
+        init_mc_funcs_dirs(0, 64);
+        init_mc_funcs_dirs(1, 32);
+        init_mc_funcs_dirs(2, 16);
+        init_mc_funcs_dirs(3, 8);
+        init_mc_funcs_dirs(4, 4);
+    }
+}
+
+#define define_itxfm(type_a, type_b, sz)                                   \
+void ff_vp9_##type_a##_##type_b##_##sz##x##sz##_add_neon(uint8_t *_dst,    \
+                                                         ptrdiff_t stride, \
+                                                         int16_t *_block, int eob)
+
+#define define_itxfm_funcs(sz)      \
+    define_itxfm(idct,  idct,  sz); \
+    define_itxfm(iadst, idct,  sz); \
+    define_itxfm(idct,  iadst, sz); \
+    define_itxfm(iadst, iadst, sz)
+
+define_itxfm_funcs(4);
+define_itxfm_funcs(8);
+define_itxfm_funcs(16);
+define_itxfm(idct, idct, 32);
+define_itxfm(iwht, iwht, 4);
+
+
+static av_cold void vp9dsp_itxfm_init_aarch64(VP9DSPContext *dsp)
+{
+    int cpu_flags = av_get_cpu_flags();
+
+    if (have_neon(cpu_flags)) {
+#define init_itxfm(tx, sz)                                             \
+    dsp->itxfm_add[tx][DCT_DCT]   = ff_vp9_idct_idct_##sz##_add_neon;  \
+    dsp->itxfm_add[tx][DCT_ADST]  = ff_vp9_iadst_idct_##sz##_add_neon; \
+    dsp->itxfm_add[tx][ADST_DCT]  = ff_vp9_idct_iadst_##sz##_add_neon; \
+    dsp->itxfm_add[tx][ADST_ADST] = ff_vp9_iadst_iadst_##sz##_add_neon
+
+#define init_idct(tx, nm)           \
+    dsp->itxfm_add[tx][DCT_DCT]   = \
+    dsp->itxfm_add[tx][ADST_DCT]  = \
+    dsp->itxfm_add[tx][DCT_ADST]  = \
+    dsp->itxfm_add[tx][ADST_ADST] = ff_vp9_##nm##_add_neon
+
+        init_itxfm(TX_4X4, 4x4);
+        init_itxfm(TX_8X8, 8x8);
+        init_itxfm(TX_16X16, 16x16);
+        init_idct(TX_32X32, idct_idct_32x32);
+        init_idct(4, iwht_iwht_4x4);
+    }
+}
+
+#define define_loop_filter(dir, wd, len) \
+void ff_vp9_loop_filter_##dir##_##wd##_##len##_neon(uint8_t *dst, ptrdiff_t stride, int E, int I, int H)
+
+#define define_loop_filters(wd, len) \
+    define_loop_filter(h, wd, len);  \
+    define_loop_filter(v, wd, len)
+
+define_loop_filters(4, 8);
+define_loop_filters(8, 8);
+define_loop_filters(16, 8);
+
+define_loop_filters(16, 16);
+
+define_loop_filters(44, 16);
+define_loop_filters(48, 16);
+define_loop_filters(84, 16);
+define_loop_filters(88, 16);
+
+static av_cold void vp9dsp_loopfilter_init_aarch64(VP9DSPContext *dsp)
+{
+    int cpu_flags = av_get_cpu_flags();
+
+    if (have_neon(cpu_flags)) {
+        dsp->loop_filter_8[0][1] = ff_vp9_loop_filter_v_4_8_neon;
+        dsp->loop_filter_8[0][0] = ff_vp9_loop_filter_h_4_8_neon;
+        dsp->loop_filter_8[1][1] = ff_vp9_loop_filter_v_8_8_neon;
+        dsp->loop_filter_8[1][0] = ff_vp9_loop_filter_h_8_8_neon;
+        dsp->loop_filter_8[2][1] = ff_vp9_loop_filter_v_16_8_neon;
+        dsp->loop_filter_8[2][0] = ff_vp9_loop_filter_h_16_8_neon;
+
+        dsp->loop_filter_16[0] = ff_vp9_loop_filter_h_16_16_neon;
+        dsp->loop_filter_16[1] = ff_vp9_loop_filter_v_16_16_neon;
+
+        dsp->loop_filter_mix2[0][0][0] = ff_vp9_loop_filter_h_44_16_neon;
+        dsp->loop_filter_mix2[0][0][1] = ff_vp9_loop_filter_v_44_16_neon;
+        dsp->loop_filter_mix2[0][1][0] = ff_vp9_loop_filter_h_48_16_neon;
+        dsp->loop_filter_mix2[0][1][1] = ff_vp9_loop_filter_v_48_16_neon;
+        dsp->loop_filter_mix2[1][0][0] = ff_vp9_loop_filter_h_84_16_neon;
+        dsp->loop_filter_mix2[1][0][1] = ff_vp9_loop_filter_v_84_16_neon;
+        dsp->loop_filter_mix2[1][1][0] = ff_vp9_loop_filter_h_88_16_neon;
+        dsp->loop_filter_mix2[1][1][1] = ff_vp9_loop_filter_v_88_16_neon;
+    }
+}
+
+av_cold void ff_vp9dsp_init_aarch64(VP9DSPContext *dsp, int bpp)
+{
+    if (bpp == 10) {
+        ff_vp9dsp_init_10bpp_aarch64(dsp);
+        return;
+    } else if (bpp == 12) {
+        ff_vp9dsp_init_12bpp_aarch64(dsp);
+        return;
+    } else if (bpp != 8)
+        return;
+
+    vp9dsp_mc_init_aarch64(dsp);
+    vp9dsp_loopfilter_init_aarch64(dsp);
+    vp9dsp_itxfm_init_aarch64(dsp);
+}

trunk/3rdparty/ffmpeg-4.2-fit/libavcodec/ac3.c

@@ -0,0 +1,211 @@
+/*
+ * Common code between the AC-3 encoder and decoder
+ * Copyright (c) 2000 Fabrice Bellard
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+/**
+ * @file
+ * Common code between the AC-3 encoder and decoder.
+ */
+
+#include "libavutil/common.h"
+
+#include "avcodec.h"
+#include "ac3.h"
+
+/**
+ * Starting frequency coefficient bin for each critical band.
+ */
+const uint8_t ff_ac3_band_start_tab[AC3_CRITICAL_BANDS+1] = {
+      0,  1,   2,   3,   4,   5,   6,   7,   8,   9,
+     10,  11, 12,  13,  14,  15,  16,  17,  18,  19,
+     20,  21, 22,  23,  24,  25,  26,  27,  28,  31,
+     34,  37, 40,  43,  46,  49,  55,  61,  67,  73,
+     79,  85, 97, 109, 121, 133, 157, 181, 205, 229, 253
+};
+
+/**
+ * Map each frequency coefficient bin to the critical band that contains it.
+ */
+const uint8_t ff_ac3_bin_to_band_tab[253] = {
+     0,
+     1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12,
+    13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24,
+    25, 26, 27, 28, 28, 28, 29, 29, 29, 30, 30, 30,
+    31, 31, 31, 32, 32, 32, 33, 33, 33, 34, 34, 34,
+    35, 35, 35, 35, 35, 35, 36, 36, 36, 36, 36, 36,
+    37, 37, 37, 37, 37, 37, 38, 38, 38, 38, 38, 38,
+    39, 39, 39, 39, 39, 39, 40, 40, 40, 40, 40, 40,
+    41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41,
+    42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42,
+    43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43, 43,
+    44, 44, 44, 44, 44, 44, 44, 44, 44, 44, 44, 44,
+    45, 45, 45, 45, 45, 45, 45, 45, 45, 45, 45, 45,
+    45, 45, 45, 45, 45, 45, 45, 45, 45, 45, 45, 45,
+    46, 46, 46, 46, 46, 46, 46, 46, 46, 46, 46, 46,
+    46, 46, 46, 46, 46, 46, 46, 46, 46, 46, 46, 46,
+    47, 47, 47, 47, 47, 47, 47, 47, 47, 47, 47, 47,
+    47, 47, 47, 47, 47, 47, 47, 47, 47, 47, 47, 47,
+    48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48,
+    48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48,
+    49, 49, 49, 49, 49, 49, 49, 49, 49, 49, 49, 49,
+    49, 49, 49, 49, 49, 49, 49, 49, 49, 49, 49, 49
+};
+
+static inline int calc_lowcomp1(int a, int b0, int b1, int c)
+{
+    if ((b0 + 256) == b1) {
+        a = c;
+    } else if (b0 > b1) {
+        a = FFMAX(a - 64, 0);
+    }
+    return a;
+}
+
+static inline int calc_lowcomp(int a, int b0, int b1, int bin)
+{
+    if (bin < 7) {
+        return calc_lowcomp1(a, b0, b1, 384);
+    } else if (bin < 20) {
+        return calc_lowcomp1(a, b0, b1, 320);
+    } else {
+        return FFMAX(a - 128, 0);
+    }
+}
+
+void ff_ac3_bit_alloc_calc_psd(int8_t *exp, int start, int end, int16_t *psd,
+                               int16_t *band_psd)
+{
+    int bin, band;
+
+    /* exponent mapping to PSD */
+    for (bin = start; bin < end; bin++) {
+        psd[bin]=(3072 - (exp[bin] << 7));
+    }
+
+    /* PSD integration */
+    bin  = start;
+    band = ff_ac3_bin_to_band_tab[start];
+    do {
+        int v = psd[bin++];
+        int band_end = FFMIN(ff_ac3_band_start_tab[band+1], end);
+        for (; bin < band_end; bin++) {
+            int max = FFMAX(v, psd[bin]);
+            /* logadd */
+            int adr = FFMIN(max - ((v + psd[bin] + 1) >> 1), 255);
+            v = max + ff_ac3_log_add_tab[adr];
+        }
+        band_psd[band++] = v;
+    } while (end > ff_ac3_band_start_tab[band]);
+}
+
+int ff_ac3_bit_alloc_calc_mask(AC3BitAllocParameters *s, int16_t *band_psd,
+                               int start, int end, int fast_gain, int is_lfe,
+                               int dba_mode, int dba_nsegs, uint8_t *dba_offsets,
+                               uint8_t *dba_lengths, uint8_t *dba_values,
+                               int16_t *mask)
+{
+    int16_t excite[AC3_CRITICAL_BANDS]; /* excitation */
+    int band;
+    int band_start, band_end, begin, end1;
+    int lowcomp, fastleak, slowleak;
+
+    if (end <= 0)
+        return AVERROR_INVALIDDATA;
+
+    /* excitation function */
+    band_start = ff_ac3_bin_to_band_tab[start];
+    band_end   = ff_ac3_bin_to_band_tab[end-1] + 1;
+
+    if (band_start == 0) {
+        lowcomp = 0;
+        lowcomp = calc_lowcomp1(lowcomp, band_psd[0], band_psd[1], 384);
+        excite[0] = band_psd[0] - fast_gain - lowcomp;
+        lowcomp = calc_lowcomp1(lowcomp, band_psd[1], band_psd[2], 384);
+        excite[1] = band_psd[1] - fast_gain - lowcomp;
+        begin = 7;
+        for (band = 2; band < 7; band++) {
+            if (!(is_lfe && band == 6))
+                lowcomp = calc_lowcomp1(lowcomp, band_psd[band], band_psd[band+1], 384);
+            fastleak = band_psd[band] - fast_gain;
+            slowleak = band_psd[band] - s->slow_gain;
+            excite[band] = fastleak - lowcomp;
+            if (!(is_lfe && band == 6)) {
+                if (band_psd[band] <= band_psd[band+1]) {
+                    begin = band + 1;
+                    break;
+                }
+            }
+        }
+
+        end1 = FFMIN(band_end, 22);
+        for (band = begin; band < end1; band++) {
+            if (!(is_lfe && band == 6))
+                lowcomp = calc_lowcomp(lowcomp, band_psd[band], band_psd[band+1], band);
+            fastleak = FFMAX(fastleak - s->fast_decay, band_psd[band] - fast_gain);
+            slowleak = FFMAX(slowleak - s->slow_decay, band_psd[band] - s->slow_gain);
+            excite[band] = FFMAX(fastleak - lowcomp, slowleak);
+        }
+        begin = 22;
+    } else {
+        /* coupling channel */
+        begin = band_start;
+        fastleak = (s->cpl_fast_leak << 8) + 768;
+        slowleak = (s->cpl_slow_leak << 8) + 768;
+    }
+
+    for (band = begin; band < band_end; band++) {
+        fastleak = FFMAX(fastleak - s->fast_decay, band_psd[band] - fast_gain);
+        slowleak = FFMAX(slowleak - s->slow_decay, band_psd[band] - s->slow_gain);
+        excite[band] = FFMAX(fastleak, slowleak);
+    }
+
+    /* compute masking curve */
+
+    for (band = band_start; band < band_end; band++) {
+        int tmp = s->db_per_bit - band_psd[band];
+        if (tmp > 0) {
+            excite[band] += tmp >> 2;
+        }
+        mask[band] = FFMAX(ff_ac3_hearing_threshold_tab[band >> s->sr_shift][s->sr_code], excite[band]);
+    }
+
+    /* delta bit allocation */
+
+    if (dba_mode == DBA_REUSE || dba_mode == DBA_NEW) {
+        int i, seg, delta;
+        if (dba_nsegs > 8)
+            return -1;
+        band = band_start;
+        for (seg = 0; seg < dba_nsegs; seg++) {
+            band += dba_offsets[seg];
+            if (band >= AC3_CRITICAL_BANDS || dba_lengths[seg] > AC3_CRITICAL_BANDS-band)
+                return -1;
+            if (dba_values[seg] >= 4) {
+                delta = (dba_values[seg] - 3) * 128;
+            } else {
+                delta = (dba_values[seg] - 4) * 128;
+            }
+            for (i = 0; i < dba_lengths[seg]; i++) {
+                mask[band++] += delta;
+            }
+        }
+    }
+    return 0;
+}