diff options
Diffstat (limited to 'libavcodec/x86/dsputil_mmx.c')
| -rw-r--r-- | libavcodec/x86/dsputil_mmx.c | 81 |
1 files changed, 0 insertions, 81 deletions
diff --git a/libavcodec/x86/dsputil_mmx.c b/libavcodec/x86/dsputil_mmx.c index 738bbd152b..73001d8677 100644 --- a/libavcodec/x86/dsputil_mmx.c +++ b/libavcodec/x86/dsputil_mmx.c @@ -2071,72 +2071,6 @@ static void vorbis_inverse_coupling_sse(float *mag, float *ang, int blocksize) } } -#if HAVE_6REGS -static void vector_fmul_window_3dnowext(float *dst, const float *src0, - const float *src1, const float *win, - int len) -{ - x86_reg i = -len * 4; - x86_reg j = len * 4 - 8; - __asm__ volatile ( - "1: \n" - "pswapd (%5, %1), %%mm1 \n" - "movq (%5, %0), %%mm0 \n" - "pswapd (%4, %1), %%mm5 \n" - "movq (%3, %0), %%mm4 \n" - "movq %%mm0, %%mm2 \n" - "movq %%mm1, %%mm3 \n" - "pfmul %%mm4, %%mm2 \n" // src0[len + i] * win[len + i] - "pfmul %%mm5, %%mm3 \n" // src1[j] * win[len + j] - "pfmul %%mm4, %%mm1 \n" // src0[len + i] * win[len + j] - "pfmul %%mm5, %%mm0 \n" // src1[j] * win[len + i] - "pfadd %%mm3, %%mm2 \n" - "pfsub %%mm0, %%mm1 \n" - "pswapd %%mm2, %%mm2 \n" - "movq %%mm1, (%2, %0) \n" - "movq %%mm2, (%2, %1) \n" - "sub $8, %1 \n" - "add $8, %0 \n" - "jl 1b \n" - "femms \n" - : "+r"(i), "+r"(j) - : "r"(dst + len), "r"(src0 + len), "r"(src1), "r"(win + len) - ); -} - -static void vector_fmul_window_sse(float *dst, const float *src0, - const float *src1, const float *win, int len) -{ - x86_reg i = -len * 4; - x86_reg j = len * 4 - 16; - __asm__ volatile ( - "1: \n" - "movaps (%5, %1), %%xmm1 \n" - "movaps (%5, %0), %%xmm0 \n" - "movaps (%4, %1), %%xmm5 \n" - "movaps (%3, %0), %%xmm4 \n" - "shufps $0x1b, %%xmm1, %%xmm1 \n" - "shufps $0x1b, %%xmm5, %%xmm5 \n" - "movaps %%xmm0, %%xmm2 \n" - "movaps %%xmm1, %%xmm3 \n" - "mulps %%xmm4, %%xmm2 \n" // src0[len + i] * win[len + i] - "mulps %%xmm5, %%xmm3 \n" // src1[j] * win[len + j] - "mulps %%xmm4, %%xmm1 \n" // src0[len + i] * win[len + j] - "mulps %%xmm5, %%xmm0 \n" // src1[j] * win[len + i] - "addps %%xmm3, %%xmm2 \n" - "subps %%xmm0, %%xmm1 \n" - "shufps $0x1b, %%xmm2, %%xmm2 \n" - "movaps %%xmm1, (%2, %0) \n" - "movaps %%xmm2, (%2, %1) \n" - "sub $16, %1 \n" - "add $16, %0 \n" - "jl 1b \n" - : "+r"(i), "+r"(j) - : "r"(dst + len), "r"(src0 + len), "r"(src1), "r"(win + len) - ); -} -#endif /* HAVE_6REGS */ - static void vector_clipf_sse(float *dst, const float *src, float min, float max, int len) { @@ -2483,14 +2417,6 @@ static void dsputil_init_3dnow(DSPContext *c, AVCodecContext *avctx, #endif /* HAVE_YASM */ } -static void dsputil_init_3dnowext(DSPContext *c, AVCodecContext *avctx, - int mm_flags) -{ -#if HAVE_AMD3DNOWEXT_INLINE && HAVE_6REGS - c->vector_fmul_window = vector_fmul_window_3dnowext; -#endif -} - static void dsputil_init_sse(DSPContext *c, AVCodecContext *avctx, int mm_flags) { const int high_bit_depth = avctx->bits_per_raw_sample > 8; @@ -2506,10 +2432,6 @@ static void dsputil_init_sse(DSPContext *c, AVCodecContext *avctx, int mm_flags) c->vorbis_inverse_coupling = vorbis_inverse_coupling_sse; -#if HAVE_6REGS - c->vector_fmul_window = vector_fmul_window_sse; -#endif - c->vector_clipf = vector_clipf_sse; #endif /* HAVE_INLINE_ASM */ @@ -2739,9 +2661,6 @@ void ff_dsputil_init_mmx(DSPContext *c, AVCodecContext *avctx) if (mm_flags & AV_CPU_FLAG_3DNOW) dsputil_init_3dnow(c, avctx, mm_flags); - if (mm_flags & AV_CPU_FLAG_3DNOWEXT) - dsputil_init_3dnowext(c, avctx, mm_flags); - if (mm_flags & AV_CPU_FLAG_SSE) dsputil_init_sse(c, avctx, mm_flags); |