diff options
author | Michael Niedermayer <michaelni@gmx.at> | 2012-05-14 22:12:47 +0400 |
---|---|---|
committer | Michael Niedermayer <michaelni@gmx.at> | 2012-05-14 22:17:24 +0400 |
commit | 7e944159c63c4d4504cf76f90bb668519c3109af (patch) | |
tree | 99c916277e5523ae607f50cdde69030513b70d39 /libavcodec/x86 | |
parent | 281bde27894f994d0982ab9283f15d6073ae352c (diff) | |
parent | 100c70b0481b889d522b4fc2aac5b948ddb05c70 (diff) |
Merge remote-tracking branch 'qatar/master'
* qatar/master: (25 commits)
vcr1: Add vcr1_ prefixes to all static functions with generic names.
vcr1: Fix return type of common_init to match the function pointer signature.
vcr1enc: Replace obsolete get_bit_count by put_bits_count/flush_put_bits.
motion-test: remove disabled code
gxfenc: remove disabled half-implemented MJPEG tag
x86: use more standard construct for setting ASM functions in FFT code
fate: westwood-aud: disable decoding
fate: caf: disable decoding
fate: film-cvid: drop pcm audio and rename test
fate: d-cinema-demux: drop unnecessary flags
fate: split off dpcm-interplay from interplay-mve tests
fate: rename funcom-iss to adpcm-ima-iss
fate: rename cryo-apc to adpcm-ima-apc
fate: rename adpcm-psx-str-v3 to adpcm-xa
fate: split off adpcm-ms-mono test from dxa-feeble
fate: split off adpcm-ima-ws test from vqa-cc
fate: add adpcm-ima-smjpeg test
fate: split off adpcm-ima-amv from amv test
fate: separate bmv audio and video tests
fate: separate delphine-cin audio and video tests
...
Conflicts:
doc/platform.texi
libavcodec/vcr1.c
tests/fate/audio.mak
tests/fate/demux.mak
tests/fate/video.mak
tests/ref/fate/ea-mad-pcm-planar
tests/ref/fate/interplay-mve-16bit
tests/ref/fate/interplay-mve-8bit
tests/ref/fate/mtv
tests/ref/fate/qtrle-1bit
tests/ref/fate/qtrle-2bit
tests/ref/fate/truemotion1-15
tests/ref/fate/truemotion1-24
tests/ref/fate/vqa-cc
Merged-by: Michael Niedermayer <michaelni@gmx.at>
Diffstat (limited to 'libavcodec/x86')
-rw-r--r-- | libavcodec/x86/fft.c | 45 |
1 files changed, 23 insertions, 22 deletions
diff --git a/libavcodec/x86/fft.c b/libavcodec/x86/fft.c index 8544e322c4..7c21335834 100644 --- a/libavcodec/x86/fft.c +++ b/libavcodec/x86/fft.c @@ -25,30 +25,31 @@ av_cold void ff_fft_init_mmx(FFTContext *s) { #if HAVE_YASM int has_vectors = av_get_cpu_flags(); - if (has_vectors & AV_CPU_FLAG_AVX && HAVE_AVX && s->nbits >= 5) { - /* AVX for SB */ - s->imdct_calc = ff_imdct_calc_sse; - s->imdct_half = ff_imdct_half_avx; - s->fft_permute = ff_fft_permute_sse; - s->fft_calc = ff_fft_calc_avx; - s->fft_permutation = FF_FFT_PERM_AVX; - } else if (has_vectors & AV_CPU_FLAG_SSE && HAVE_SSE) { + if (has_vectors & AV_CPU_FLAG_3DNOW && HAVE_AMD3DNOW) { + /* 3DNow! for K6-2/3 */ + s->imdct_calc = ff_imdct_calc_3dn; + s->imdct_half = ff_imdct_half_3dn; + s->fft_calc = ff_fft_calc_3dn; + } + if (has_vectors & AV_CPU_FLAG_3DNOWEXT && HAVE_AMD3DNOWEXT) { + /* 3DNowEx for K7 */ + s->imdct_calc = ff_imdct_calc_3dn2; + s->imdct_half = ff_imdct_half_3dn2; + s->fft_calc = ff_fft_calc_3dn2; + } + if (has_vectors & AV_CPU_FLAG_SSE && HAVE_SSE) { /* SSE for P3/P4/K8 */ s->imdct_calc = ff_imdct_calc_sse; s->imdct_half = ff_imdct_half_sse; s->fft_permute = ff_fft_permute_sse; s->fft_calc = ff_fft_calc_sse; s->fft_permutation = FF_FFT_PERM_SWAP_LSBS; - } else if (has_vectors & AV_CPU_FLAG_3DNOWEXT && HAVE_AMD3DNOWEXT) { - /* 3DNowEx for K7 */ - s->imdct_calc = ff_imdct_calc_3dn2; - s->imdct_half = ff_imdct_half_3dn2; - s->fft_calc = ff_fft_calc_3dn2; - } else if (has_vectors & AV_CPU_FLAG_3DNOW && HAVE_AMD3DNOW) { - /* 3DNow! for K6-2/3 */ - s->imdct_calc = ff_imdct_calc_3dn; - s->imdct_half = ff_imdct_half_3dn; - s->fft_calc = ff_fft_calc_3dn; + } + if (has_vectors & AV_CPU_FLAG_AVX && HAVE_AVX && s->nbits >= 5) { + /* AVX for SB */ + s->imdct_half = ff_imdct_half_avx; + s->fft_calc = ff_fft_calc_avx; + s->fft_permutation = FF_FFT_PERM_AVX; } #endif } @@ -58,12 +59,12 @@ av_cold void ff_dct_init_mmx(DCTContext *s) { #if HAVE_YASM int has_vectors = av_get_cpu_flags(); + if (has_vectors & AV_CPU_FLAG_SSE && HAVE_SSE) + s->dct32 = ff_dct32_float_sse; + if (has_vectors & AV_CPU_FLAG_SSE2 && HAVE_SSE) + s->dct32 = ff_dct32_float_sse2; if (has_vectors & AV_CPU_FLAG_AVX && HAVE_AVX) s->dct32 = ff_dct32_float_avx; - else if (has_vectors & AV_CPU_FLAG_SSE2 && HAVE_SSE) - s->dct32 = ff_dct32_float_sse2; - else if (has_vectors & AV_CPU_FLAG_SSE && HAVE_SSE) - s->dct32 = ff_dct32_float_sse; #endif } #endif |