| Message ID | 20210106231308.2952217-2-andreas.rheinhardt@gmail.com |
|---|---|
| State | Superseded |
| Headers | show |
| Series | [FFmpeg-devel,1/5] avcodec/tableprint: Don't include mem_internal.h | expand |
| Context | Check | Description |
|---|---|---|
| andriy/x86_make | success | Make finished |
| andriy/x86_make_fate | success | Make fate finished |
| andriy/PPC64_make | success | Make finished |
| andriy/PPC64_make_fate | success | Make fate finished |
Jan 7, 2021, 00:13 by andreas.rheinhardt@gmail.com: > There are three types of FFTs: floating-point, 32-bit fixed-point and > 16-bit fixed-point. The latter has exactly one user: The fixed-point > AC-3-encoder; the cosine tables used by it use up to seven bits. The > tables corresponding to eight to seventeen bits are unused, as are the > FFT functions for these bits. > > Therefore this commit removes these tables and functions. This is > especially beneficial when using hardcoded tables as they take up more > than 255 KiB. But even without it one saves said unused functions as > well as entries in corresponding tables (this also saves relocations). > > Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com> > --- > Thee changes to ARM assembly are honstely untested. I hope someone can > test them. Btw: It seems that the ARM assembly code wouldn't be able to > deal with an FFT with more than 16 bits (no function for this has been > defined), which only worked because no one ever used that many bits with > the fixed-point FFT. > > libavcodec/arm/fft_fixed_neon.S | 18 ------------------ > libavcodec/cos_tablegen.c | 4 ++-- > libavcodec/fft.h | 4 +++- > libavcodec/fft_fixed.c | 1 + > libavcodec/fft_template.c | 31 +++++++++++++++++++++++-------- > tests/fate/fft.mak | 8 ++++++-- > 6 files changed, 35 insertions(+), 31 deletions(-) > The whole comment was hard to make sense of, since you keep mixing fixed-point FFT precision bits (16 and 32) and FFT length (confusingly also in bits). I'd rather have a blank comment with just the code or just no references to the 16-bit fixed-point FFT. LGTM. Thankfully nothing of the eldritchian fixed-point FFT monstrosity is exposed to API users, so as long as FATE passes on ARM, should be okay.
On Thu, Jan 07, 2021 at 12:13:05AM +0100, Andreas Rheinhardt wrote: > There are three types of FFTs: floating-point, 32-bit fixed-point and > 16-bit fixed-point. The latter has exactly one user: The fixed-point > AC-3-encoder; the cosine tables used by it use up to seven bits. The > tables corresponding to eight to seventeen bits are unused, as are the > FFT functions for these bits. > > Therefore this commit removes these tables and functions. This is > especially beneficial when using hardcoded tables as they take up more > than 255 KiB. But even without it one saves said unused functions as > well as entries in corresponding tables (this also saves relocations). > > Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com> > --- > Thee changes to ARM assembly are honstely untested. I hope someone can > test them. Btw: It seems that the ARM assembly code wouldn't be able to > deal with an FFT with more than 16 bits (no function for this has been > defined), which only worked because no one ever used that many bits with > the fixed-point FFT. > > libavcodec/arm/fft_fixed_neon.S | 18 ------------------ > libavcodec/cos_tablegen.c | 4 ++-- > libavcodec/fft.h | 4 +++- > libavcodec/fft_fixed.c | 1 + > libavcodec/fft_template.c | 31 +++++++++++++++++++++++-------- > tests/fate/fft.mak | 8 ++++++-- > 6 files changed, 35 insertions(+), 31 deletions(-) make -j32 libavcodec/tests/fft-fixed && libavcodec/tests/fft-fixed Segmentation fault (core dumped) (if you cant repro say so and ill rebuild with debug symbols ...) thx [...]
Michael Niedermayer: > On Thu, Jan 07, 2021 at 12:13:05AM +0100, Andreas Rheinhardt wrote: >> There are three types of FFTs: floating-point, 32-bit fixed-point and >> 16-bit fixed-point. The latter has exactly one user: The fixed-point >> AC-3-encoder; the cosine tables used by it use up to seven bits. The >> tables corresponding to eight to seventeen bits are unused, as are the >> FFT functions for these bits. >> >> Therefore this commit removes these tables and functions. This is >> especially beneficial when using hardcoded tables as they take up moreFirst, >> than 255 KiB. But even without it one saves said unused functions as >> well as entries in corresponding tables (this also saves relocations). >> >> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com> >> --- >> Thee changes to ARM assembly are honstely untested. I hope someone can >> test them. Btw: It seems that the ARM assembly code wouldn't be able to >> deal with an FFT with more than 16 bits (no function for this has been >> defined), which only worked because no one ever used that many bits with >> the fixed-point FFT. >> >> libavcodec/arm/fft_fixed_neon.S | 18 ------------------ >> libavcodec/cos_tablegen.c | 4 ++-- >> libavcodec/fft.h | 4 +++- >> libavcodec/fft_fixed.c | 1 + >> libavcodec/fft_template.c | 31 +++++++++++++++++++++++-------- >> tests/fate/fft.mak | 8 ++++++-- >> 6 files changed, 35 insertions(+), 31 deletions(-) > > make -j32 libavcodec/tests/fft-fixed && libavcodec/tests/fft-fixed > Segmentation fault (core dumped) > > (if you cant repro say so and ill rebuild with debug symbols ...) > > thx > [...] > 1. Lynne has an alternative patchset that makes the only user of fft_fixed use fft_fixed_32 instead, so this is not important any more. 2. Are you testing the ARM assembly code (for which I ask for a test) or not? If not, then this surprises me. Did you apply the changes to fft.mak (some of the tests have been removed as they tested functionality that was unused (apart from the tests) and has therefore been removed). - Andreas
On Sun, Jan 10, 2021 at 01:56:21AM +0100, Andreas Rheinhardt wrote: > Michael Niedermayer: > > On Thu, Jan 07, 2021 at 12:13:05AM +0100, Andreas Rheinhardt wrote: > >> There are three types of FFTs: floating-point, 32-bit fixed-point and > >> 16-bit fixed-point. The latter has exactly one user: The fixed-point > >> AC-3-encoder; the cosine tables used by it use up to seven bits. The > >> tables corresponding to eight to seventeen bits are unused, as are the > >> FFT functions for these bits. > >> > >> Therefore this commit removes these tables and functions. This is > >> especially beneficial when using hardcoded tables as they take up moreFirst, > >> than 255 KiB. But even without it one saves said unused functions as > >> well as entries in corresponding tables (this also saves relocations). > >> > >> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com> > >> --- > >> Thee changes to ARM assembly are honstely untested. I hope someone can > >> test them. Btw: It seems that the ARM assembly code wouldn't be able to > >> deal with an FFT with more than 16 bits (no function for this has been > >> defined), which only worked because no one ever used that many bits with > >> the fixed-point FFT. > >> > >> libavcodec/arm/fft_fixed_neon.S | 18 ------------------ > >> libavcodec/cos_tablegen.c | 4 ++-- > >> libavcodec/fft.h | 4 +++- > >> libavcodec/fft_fixed.c | 1 + > >> libavcodec/fft_template.c | 31 +++++++++++++++++++++++-------- > >> tests/fate/fft.mak | 8 ++++++-- > >> 6 files changed, 35 insertions(+), 31 deletions(-) > > > > make -j32 libavcodec/tests/fft-fixed && libavcodec/tests/fft-fixed > > Segmentation fault (core dumped) > > > > (if you cant repro say so and ill rebuild with debug symbols ...) > > > > thx > > [...] > > > 1. Lynne has an alternative patchset that makes the only user of > fft_fixed use fft_fixed_32 instead, so this is not important any more. > 2. Are you testing the ARM assembly code (for which I ask for a test) or x86-64 > not? If not, then this surprises me. Did you apply the changes to > fft.mak (some of the tests have been removed as they tested > functionality that was unused (apart from the tests) and has therefore > been removed). i applied the changes from this patchset up to and including the patch and also did a make distclean FFT 512 test Checking... ==18069== Jump to the invalid address stated on the next line ==18069== at 0x0: ??? ==18069== by 0x10F5F9: main (fft.c:529) ==18069== Address 0x0 is not stack'd, malloc'd or (recently) free'd ==18069== ==18069== ==18069== Process terminating with default action of signal 11 (SIGSEGV) ==18069== Bad permissions for mapped region at address 0x0 ==18069== at 0x0: ??? ==18069== by 0x10F5F9: main (fft.c:529) commit 6c532480712d395f5973063adcefce62fc75f2e1 (HEAD) Author: Andreas Rheinhardt <andreas.rheinhardt@gmail.com> Date: Thu Jan 7 00:13:05 2021 +0100 avcodec/fft_template: Remove unused fixed-point cosine tables There are three types of FFTs: floating-point, 32-bit fixed-point and 16-bit fixed-point. The latter has exactly one user: The fixed-point AC-3-encoder; the cosine tables used by it use up to seven bits. The tables corresponding to eight to seventeen bits are unused, as are the FFT functions for these bits. Therefore this commit removes these tables and functions. This is especially beneficial when using hardcoded tables as they take up more than 255 KiB. But even without it one saves said unused functions as well as entries in corresponding tables (this also saves relocations). Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc> libavcodec/arm/fft_fixed_neon.S | 18 ------------------ libavcodec/cos_tablegen.c | 4 ++-- libavcodec/fft.h | 4 +++- libavcodec/fft_fixed.c | 1 + libavcodec/fft_template.c | 31 +++++++++++++++++++++++-------- tests/fate/fft.mak | 8 ++++++-- 6 files changed, 35 insertions(+), 31 deletions(-) commit c592684681700a7d8b41e75a11104f8c1bdd13d9 Author: Andreas Rheinhardt <andreas.rheinhardt@gmail.com> Date: Thu Jan 7 00:13:04 2021 +0100 avcodec/tableprint: Don't include mem_internal.h tableprint.h does not declare anything as aligned; it just prints DECLARE_ALIGNED. So it can be removed; in fact, it needs to be removed, because mem_internal.h includes config.h which leads to warnings when building with hardcoded tables enabled because of redefinitions of CONFIG_HARDCODED_TABLES. (Furthermore, config.h is only valid for the target, not the host, so HAVE_LOCAL_ALIGNED might even be wrong here.) Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc> libavcodec/tableprint.h | 1 - 1 file changed, 1 deletion(-) commit 91e1625db15fe8853ceedca9eed14307aaa514c7 (origin/master, origin/HEAD, refs/bisect/good-91e1625db15fe8853ceedca9eed14307aaa514c7) [...]
Michael Niedermayer: > On Sun, Jan 10, 2021 at 01:56:21AM +0100, Andreas Rheinhardt wrote: >> Michael Niedermayer: >>> On Thu, Jan 07, 2021 at 12:13:05AM +0100, Andreas Rheinhardt wrote: >>>> There are three types of FFTs: floating-point, 32-bit fixed-point and >>>> 16-bit fixed-point. The latter has exactly one user: The fixed-point >>>> AC-3-encoder; the cosine tables used by it use up to seven bits. The >>>> tables corresponding to eight to seventeen bits are unused, as are the >>>> FFT functions for these bits. >>>> >>>> Therefore this commit removes these tables and functions. This is >>>> especially beneficial when using hardcoded tables as they take up moreFirst, >>>> than 255 KiB. But even without it one saves said unused functions as >>>> well as entries in corresponding tables (this also saves relocations). >>>> >>>> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com> >>>> --- >>>> Thee changes to ARM assembly are honstely untested. I hope someone can >>>> test them. Btw: It seems that the ARM assembly code wouldn't be able to >>>> deal with an FFT with more than 16 bits (no function for this has been >>>> defined), which only worked because no one ever used that many bits with >>>> the fixed-point FFT. >>>> >>>> libavcodec/arm/fft_fixed_neon.S | 18 ------------------ >>>> libavcodec/cos_tablegen.c | 4 ++-- >>>> libavcodec/fft.h | 4 +++- >>>> libavcodec/fft_fixed.c | 1 + >>>> libavcodec/fft_template.c | 31 +++++++++++++++++++++++-------- >>>> tests/fate/fft.mak | 8 ++++++-- >>>> 6 files changed, 35 insertions(+), 31 deletions(-) >>> >>> make -j32 libavcodec/tests/fft-fixed && libavcodec/tests/fft-fixed >>> Segmentation fault (core dumped) >>> >>> (if you cant repro say so and ill rebuild with debug symbols ...) >>> >>> thx >>> [...] >>> >> 1. Lynne has an alternative patchset that makes the only user of >> fft_fixed use fft_fixed_32 instead, so this is not important any more. > >> 2. Are you testing the ARM assembly code (for which I ask for a test) or > > x86-64 > >> not? If not, then this surprises me. Did you apply the changes to >> fft.mak (some of the tests have been removed as they tested >> functionality that was unused (apart from the tests) and has therefore >> been removed). > > i applied the changes from this patchset up to and including the patch > and also did a make distclean > FFT 512 test Now I see what's the problem. You are not running the fft-tests from the FATE-suite (where I disabled unused and newly unsupported tests); instead you are directly using the underlying tests in libavcodec/tests and these tests default to nbits = 9, which is unsupported for fft-fixed with this patch applied. ff_fft_init_fixed is correctly erroring out, yet said error code is ignored in the test's fft_init (therefore testing an unsupported nbits already leads to segfaults on master (try libavcodec/tests/fft -n 20)). So the error checking in the tests needs to be improved (there are other unchecked allocations, too). Furthermore, if this patch were to be applied, one should set the default number of bits for fft-fixed to something supported; but that point is moot given that this patch has been superseded by Lynne's (who wants to nuke fft-fixed). - Andreas
diff --git a/libavcodec/arm/fft_fixed_neon.S b/libavcodec/arm/fft_fixed_neon.S index 2651607544..c94da56f80 100644 --- a/libavcodec/arm/fft_fixed_neon.S +++ b/libavcodec/arm/fft_fixed_neon.S @@ -223,15 +223,6 @@ endfunc def_fft 32, 16, 8 def_fft 64, 32, 16 def_fft 128, 64, 32 - def_fft 256, 128, 64 - def_fft 512, 256, 128 - def_fft 1024, 512, 256 - def_fft 2048, 1024, 512 - def_fft 4096, 2048, 1024 - def_fft 8192, 4096, 2048 - def_fft 16384, 8192, 4096 - def_fft 32768, 16384, 8192 - def_fft 65536, 32768, 16384 function ff_fft_fixed_calc_neon, export=1 ldr r2, [r0] @@ -249,13 +240,4 @@ const fft_fixed_tab_neon, relocate=1 .word fft32_neon .word fft64_neon .word fft128_neon - .word fft256_neon - .word fft512_neon - .word fft1024_neon - .word fft2048_neon - .word fft4096_neon - .word fft8192_neon - .word fft16384_neon - .word fft32768_neon - .word fft65536_neon endconst diff --git a/libavcodec/cos_tablegen.c b/libavcodec/cos_tablegen.c index 7206aad5dd..5929c29e1a 100644 --- a/libavcodec/cos_tablegen.c +++ b/libavcodec/cos_tablegen.c @@ -26,7 +26,6 @@ #include "libavutil/mathematics.h" -#define BITS 17 #define FLOATFMT "%.18e" #define FIXEDFMT "%6d" @@ -56,12 +55,13 @@ int main(int argc, char *argv[]) int i, j; int do_sin = argc > 1 && !strcmp(argv[1], "sin"); int fixed = argc > 1 && strstr(argv[1], "fixed"); + int bits = fixed ? 7 : 17; double (*func)(double) = do_sin ? sin : cos; printf("/* This file was automatically generated. */\n"); printf("#define FFT_FLOAT %d\n", !fixed); printf("#include \"libavcodec/%s\"\n", do_sin ? "rdft.h" : "fft.h"); - for (i = 4; i <= BITS; i++) { + for (i = 4; i <= bits; i++) { int m = 1 << i; double freq = 2*M_PI/m; printf("%s(%i) = {\n ", do_sin ? "SINTABLE" : "COSTABLE", m); diff --git a/libavcodec/fft.h b/libavcodec/fft.h index 5f67b61f06..fedc0c5ef0 100644 --- a/libavcodec/fft.h +++ b/libavcodec/fft.h @@ -127,6 +127,7 @@ extern COSTABLE(16); extern COSTABLE(32); extern COSTABLE(64); extern COSTABLE(128); +#if FFT_FLOAT extern COSTABLE(256); extern COSTABLE(512); extern COSTABLE(1024); @@ -137,7 +138,8 @@ extern COSTABLE(16384); extern COSTABLE(32768); extern COSTABLE(65536); extern COSTABLE(131072); -extern COSTABLE_CONST FFTSample* const FFT_NAME(ff_cos_tabs)[18]; +#endif /* FFT_FLOAT */ +extern COSTABLE_CONST FFTSample* const FFT_NAME(ff_cos_tabs)[]; #define ff_init_ff_cos_tabs FFT_NAME(ff_init_ff_cos_tabs) diff --git a/libavcodec/fft_fixed.c b/libavcodec/fft_fixed.c index 3d3bd2fca6..52d225ee09 100644 --- a/libavcodec/fft_fixed.c +++ b/libavcodec/fft_fixed.c @@ -18,4 +18,5 @@ #define FFT_FLOAT 0 #define FFT_FIXED_32 0 +#define MAX_BITS 7 #include "fft_template.c" diff --git a/libavcodec/fft_template.c b/libavcodec/fft_template.c index 8825e39f79..7a7d51a6b4 100644 --- a/libavcodec/fft_template.c +++ b/libavcodec/fft_template.c @@ -33,6 +33,10 @@ #include "fft.h" #include "fft-internal.h" +#ifndef MAX_BITS +#define MAX_BITS 17 +#endif + #if FFT_FIXED_32 #include "fft_table.h" #else /* FFT_FIXED_32 */ @@ -43,6 +47,7 @@ COSTABLE(16); COSTABLE(32); COSTABLE(64); COSTABLE(128); +#if FFT_FLOAT COSTABLE(256); COSTABLE(512); COSTABLE(1024); @@ -53,6 +58,7 @@ COSTABLE(16384); COSTABLE(32768); COSTABLE(65536); COSTABLE(131072); +#endif /* FFT_FLOAT */ static av_cold void init_ff_cos_tabs(int index) { @@ -81,6 +87,7 @@ INIT_FF_COS_TABS_FUNC(4, 16) INIT_FF_COS_TABS_FUNC(5, 32) INIT_FF_COS_TABS_FUNC(6, 64) INIT_FF_COS_TABS_FUNC(7, 128) +#if FFT_FLOAT INIT_FF_COS_TABS_FUNC(8, 256) INIT_FF_COS_TABS_FUNC(9, 512) INIT_FF_COS_TABS_FUNC(10, 1024) @@ -91,6 +98,7 @@ INIT_FF_COS_TABS_FUNC(14, 16384) INIT_FF_COS_TABS_FUNC(15, 32768) INIT_FF_COS_TABS_FUNC(16, 65536) INIT_FF_COS_TABS_FUNC(17, 131072) +#endif /* FFT_FLOAT */ static CosTabsInitOnce cos_tabs_init_once[] = { { NULL }, @@ -101,6 +109,7 @@ static CosTabsInitOnce cos_tabs_init_once[] = { { init_ff_cos_tabs_32, AV_ONCE_INIT }, { init_ff_cos_tabs_64, AV_ONCE_INIT }, { init_ff_cos_tabs_128, AV_ONCE_INIT }, +#if FFT_FLOAT { init_ff_cos_tabs_256, AV_ONCE_INIT }, { init_ff_cos_tabs_512, AV_ONCE_INIT }, { init_ff_cos_tabs_1024, AV_ONCE_INIT }, @@ -111,6 +120,7 @@ static CosTabsInitOnce cos_tabs_init_once[] = { { init_ff_cos_tabs_32768, AV_ONCE_INIT }, { init_ff_cos_tabs_65536, AV_ONCE_INIT }, { init_ff_cos_tabs_131072, AV_ONCE_INIT }, +#endif /* FFT_FLOAT */ }; #endif @@ -120,6 +130,7 @@ COSTABLE_CONST FFTSample * const FFT_NAME(ff_cos_tabs)[] = { FFT_NAME(ff_cos_32), FFT_NAME(ff_cos_64), FFT_NAME(ff_cos_128), +#if FFT_FLOAT FFT_NAME(ff_cos_256), FFT_NAME(ff_cos_512), FFT_NAME(ff_cos_1024), @@ -130,6 +141,7 @@ COSTABLE_CONST FFTSample * const FFT_NAME(ff_cos_tabs)[] = { FFT_NAME(ff_cos_32768), FFT_NAME(ff_cos_65536), FFT_NAME(ff_cos_131072), +#endif /* FFT_FLOAT */ }; #endif /* FFT_FIXED_32 */ @@ -200,7 +212,7 @@ av_cold int ff_fft_init(FFTContext *s, int nbits, int inverse) s->revtab = NULL; s->revtab32 = NULL; - if (nbits < 2 || nbits > 17) + if (nbits < 2 || nbits > MAX_BITS) goto fail; s->nbits = nbits; n = 1 << nbits; @@ -537,11 +549,6 @@ static void name(FFTComplex *z, const FFTSample *wre, unsigned int n)\ } PASS(pass) -#if !CONFIG_SMALL -#undef BUTTERFLIES -#define BUTTERFLIES BUTTERFLIES_BIG -PASS(pass_big) -#endif #define DECL_FFT(n,n2,n4)\ static void fft##n(FFTComplex *z)\ @@ -603,9 +610,13 @@ DECL_FFT(16,8,4) DECL_FFT(32,16,8) DECL_FFT(64,32,16) DECL_FFT(128,64,32) +#if FFT_FLOAT DECL_FFT(256,128,64) DECL_FFT(512,256,128) #if !CONFIG_SMALL +#undef BUTTERFLIES +#define BUTTERFLIES BUTTERFLIES_BIG +PASS(pass_big) #define pass pass_big #endif DECL_FFT(1024,512,256) @@ -616,10 +627,14 @@ DECL_FFT(16384,8192,4096) DECL_FFT(32768,16384,8192) DECL_FFT(65536,32768,16384) DECL_FFT(131072,65536,32768) +#endif /* FFT_FLOAT */ static void (* const fft_dispatch[])(FFTComplex*) = { - fft4, fft8, fft16, fft32, fft64, fft128, fft256, fft512, fft1024, - fft2048, fft4096, fft8192, fft16384, fft32768, fft65536, fft131072 + fft4, fft8, fft16, fft32, fft64, fft128, +#if FFT_FLOAT + fft256, fft512, fft1024, fft2048, fft4096, + fft8192, fft16384, fft32768, fft65536, fft131072 +#endif /* FFT_FLOAT */ }; static void fft_calc_c(FFTContext *s, FFTComplex *z) diff --git a/tests/fate/fft.mak b/tests/fate/fft.mak index 5da6e687ec..3eb8450d94 100644 --- a/tests/fate/fft.mak +++ b/tests/fate/fft.mak @@ -28,15 +28,19 @@ $(FATE_FFT_ALL): CMD = run libavcodec/tests/fft$(EXESUF) $(CPUFLAGS:%=-c%) $(ARG define DEF_FFT_FIXED FATE_FFT_FIXED-$(CONFIG_FFT) += fate-fft-fixed-$(1) fate-ifft-fixed-$(1) -FATE_MDCT_FIXED-$(CONFIG_MDCT) += fate-mdct-fixed-$(1) fate-imdct-fixed-$(1) fate-fft-fixed-$(1): ARGS = -n$(1) fate-ifft-fixed-$(1): ARGS = -n$(1) -i +endef +define DEF_MDCT_FIXED +FATE_MDCT_FIXED-$(CONFIG_MDCT) += fate-mdct-fixed-$(1) fate-imdct-fixed-$(1) + fate-mdct-fixed-$(1): ARGS = -n$(1) -m fate-imdct-fixed-$(1): ARGS = -n$(1) -m -i endef -$(foreach N, 4 5 6 7 8 9 10 11 12, $(eval $(call DEF_FFT_FIXED,$(N)))) +$(foreach N, 4 5 6 7, $(eval $(call DEF_FFT_FIXED,$(N)))) +$(foreach N, 4 5 6 7 8 9, $(eval $(call DEF_MDCT_FIXED,$(N)))) fate-fft-fixed: $(FATE_FFT_FIXED-yes) fate-mdct-fixed: $(FATE_MDCT_FIXED-yes)
There are three types of FFTs: floating-point, 32-bit fixed-point and 16-bit fixed-point. The latter has exactly one user: The fixed-point AC-3-encoder; the cosine tables used by it use up to seven bits. The tables corresponding to eight to seventeen bits are unused, as are the FFT functions for these bits. Therefore this commit removes these tables and functions. This is especially beneficial when using hardcoded tables as they take up more than 255 KiB. But even without it one saves said unused functions as well as entries in corresponding tables (this also saves relocations). Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com> --- Thee changes to ARM assembly are honstely untested. I hope someone can test them. Btw: It seems that the ARM assembly code wouldn't be able to deal with an FFT with more than 16 bits (no function for this has been defined), which only worked because no one ever used that many bits with the fixed-point FFT. libavcodec/arm/fft_fixed_neon.S | 18 ------------------ libavcodec/cos_tablegen.c | 4 ++-- libavcodec/fft.h | 4 +++- libavcodec/fft_fixed.c | 1 + libavcodec/fft_template.c | 31 +++++++++++++++++++++++-------- tests/fate/fft.mak | 8 ++++++-- 6 files changed, 35 insertions(+), 31 deletions(-)