From fcc456b8293b829d868f55778d76f8da3d54e846 Mon Sep 17 00:00:00 2001
From: Vitor Sessak <vitor1001@gmail.com>
Date: Sat, 12 May 2012 08:59:00 +0200
Subject: [PATCH] x86: use more standard construct for setting ASM functions in
FFT code
Signed-off-by: Diego Biurrun <diego@biurrun.de>
---
libavcodec/x86/fft.c | 45 ++++++++++++++++++++++----------------------
1 file changed, 23 insertions(+), 22 deletions(-)
diff --git a/libavcodec/x86/fft.c b/libavcodec/x86/fft.c
index 3e0c42fb1e5..5495821e5b2 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
--
GitLab