diff options
author | Michael Niedermayer <michaelni@gmx.at> | 2014-01-22 22:55:49 +0100 |
---|---|---|
committer | Michael Niedermayer <michaelni@gmx.at> | 2014-01-22 22:55:49 +0100 |
commit | fad49aae28cf3e47791f03a04d7aad328a6d6fdf (patch) | |
tree | 77a3ce9dbe4baebb9a9492f275c34d64c2b0758a | |
parent | fee97f25fa1275e2a35485cb16283a466c28aadc (diff) | |
download | ffmpeg-fad49aae28cf3e47791f03a04d7aad328a6d6fdf.tar.gz |
avcodec/x86/lossless_videodsp: Port sub_hfyu_median_prediction_mmxext to int16
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
-rw-r--r-- | libavcodec/x86/lossless_videodsp_init.c | 48 |
1 files changed, 48 insertions, 0 deletions
diff --git a/libavcodec/x86/lossless_videodsp_init.c b/libavcodec/x86/lossless_videodsp_init.c index 4eca2a11b1..eac3395eb2 100644 --- a/libavcodec/x86/lossless_videodsp_init.c +++ b/libavcodec/x86/lossless_videodsp_init.c @@ -20,6 +20,8 @@ #include "../lossless_videodsp.h" #include "libavutil/x86/cpu.h" +#include "libavutil/x86/asm.h" +#include "libavcodec/mathops.h" void ff_add_int16_mmx(uint16_t *dst, const uint16_t *src, unsigned mask, int w); void ff_add_int16_sse2(uint16_t *dst, const uint16_t *src, unsigned mask, int w); @@ -29,6 +31,51 @@ int ff_add_hfyu_left_prediction_int16_ssse3(uint16_t *dst, const uint16_t *src, int ff_add_hfyu_left_prediction_int16_sse4(uint16_t *dst, const uint16_t *src, unsigned mask, int w, int acc); void ff_add_hfyu_median_prediction_int16_mmxext(uint16_t *dst, const uint16_t *top, const uint16_t *diff, unsigned mask, int w, int *left, int *left_top); +static void sub_hfyu_median_prediction_int16_mmxext(uint16_t *dst, const uint16_t *src1, + const uint16_t *src2, unsigned mask, int w, + int *left, int *left_top) +{ + x86_reg i=0; + uint16_t l, lt; + + __asm__ volatile( + "movd %5, %%mm7 \n\t" + "pshufw $0, %%mm7, %%mm7 \n\t" + "movq (%1, %0), %%mm0 \n\t" // LT + "psllq $16, %%mm0 \n\t" + "1: \n\t" + "movq (%1, %0), %%mm1 \n\t" // T + "movq -2(%2, %0), %%mm2 \n\t" // L + "movq (%2, %0), %%mm3 \n\t" // X + "movq %%mm2, %%mm4 \n\t" // L + "psubw %%mm0, %%mm2 \n\t" + "paddw %%mm1, %%mm2 \n\t" // L + T - LT + "pand %%mm7, %%mm2 \n\t" + "movq %%mm4, %%mm5 \n\t" // L + "pmaxsw %%mm1, %%mm4 \n\t" // max(T, L) + "pminsw %%mm5, %%mm1 \n\t" // min(T, L) + "pminsw %%mm2, %%mm4 \n\t" + "pmaxsw %%mm1, %%mm4 \n\t" + "psubw %%mm4, %%mm3 \n\t" // dst - pred + "pand %%mm7, %%mm3 \n\t" + "movq %%mm3, (%3, %0) \n\t" + "add $8, %0 \n\t" + "movq -2(%1, %0), %%mm0 \n\t" // LT + "cmp %4, %0 \n\t" + " jb 1b \n\t" + : "+r" (i) + : "r"(src1), "r"(src2), "r"(dst), "r"((x86_reg)2*w), "rm"(mask) + ); + + l= *left; + lt= *left_top; + + dst[0]= src2[0] - mid_pred(l, src1[0], (l + src1[0] - lt)&mask); + + *left_top= src1[w-1]; + *left = src2[w-1]; +} + void ff_llviddsp_init_x86(LLVidDSPContext *c) { int cpu_flags = av_get_cpu_flags(); @@ -40,6 +87,7 @@ void ff_llviddsp_init_x86(LLVidDSPContext *c) if (EXTERNAL_MMXEXT(cpu_flags)) { c->add_hfyu_median_prediction_int16 = ff_add_hfyu_median_prediction_int16_mmxext; + c->sub_hfyu_median_prediction_int16 = sub_hfyu_median_prediction_int16_mmxext; } if (EXTERNAL_SSE2(cpu_flags)) { |