diff options
| author | Michael Niedermayer <michaelni@gmx.at> | 2013-01-22 14:54:06 +0100 |
|---|---|---|
| committer | Michael Niedermayer <michaelni@gmx.at> | 2013-01-22 15:33:23 +0100 |
| commit | 26345acb0e723d28aa28e09126ea383b2f679f5b (patch) | |
| tree | f8d7a9f62d7557aca5ad56dd244fa415fdab56f2 /libavcodec/x86/vp3dsp_init.c | |
| parent | 9774251c45eadfe0ba7b0381344451fa8840cc0d (diff) | |
| parent | 0881cbf314982cce8448bd12644ce2a6e0b8c576 (diff) | |
| download | ffmpeg-26345acb0e723d28aa28e09126ea383b2f679f5b.tar.gz | |
Merge remote-tracking branch 'qatar/master'
* qatar/master:
proresdec: support mixed interlaced/non-interlaced content
vp3/5: move put_no_rnd_pixels_l2 from dsputil to VP3DSPContext.
Merged-by: Michael Niedermayer <michaelni@gmx.at>
Diffstat (limited to 'libavcodec/x86/vp3dsp_init.c')
| -rw-r--r-- | libavcodec/x86/vp3dsp_init.c | 61 |
1 files changed, 61 insertions, 0 deletions
diff --git a/libavcodec/x86/vp3dsp_init.c b/libavcodec/x86/vp3dsp_init.c index 245b60f6fa..e425f503ac 100644 --- a/libavcodec/x86/vp3dsp_init.c +++ b/libavcodec/x86/vp3dsp_init.c @@ -1,4 +1,6 @@ /* + * Copyright (c) 2009 David Conrad <lessen42@gmail.com> + * * This file is part of FFmpeg. * * FFmpeg is free software; you can redistribute it and/or @@ -21,6 +23,7 @@ #include "libavutil/attributes.h" #include "libavutil/cpu.h" #include "libavutil/x86/cpu.h" +#include "libavutil/x86/asm.h" #include "libavcodec/avcodec.h" #include "libavcodec/vp3dsp.h" #include "config.h" @@ -39,10 +42,68 @@ void ff_vp3_v_loop_filter_mmxext(uint8_t *src, int stride, void ff_vp3_h_loop_filter_mmxext(uint8_t *src, int stride, int *bounding_values); +#if HAVE_INLINE_ASM + +#define MOVQ_BFE(regd) \ + __asm__ volatile ( \ + "pcmpeqd %%"#regd", %%"#regd" \n\t" \ + "paddb %%"#regd", %%"#regd" \n\t" ::) + +#define PAVGBP_MMX_NO_RND(rega, regb, regr, regc, regd, regp) \ + "movq "#rega", "#regr" \n\t" \ + "movq "#regc", "#regp" \n\t" \ + "pand "#regb", "#regr" \n\t" \ + "pand "#regd", "#regp" \n\t" \ + "pxor "#rega", "#regb" \n\t" \ + "pxor "#regc", "#regd" \n\t" \ + "pand %%mm6, "#regb" \n\t" \ + "pand %%mm6, "#regd" \n\t" \ + "psrlq $1, "#regb" \n\t" \ + "psrlq $1, "#regd" \n\t" \ + "paddb "#regb", "#regr" \n\t" \ + "paddb "#regd", "#regp" \n\t" + +static void put_vp_no_rnd_pixels8_l2_mmx(uint8_t *dst, const uint8_t *a, const uint8_t *b, int stride, int h) +{ +// START_TIMER + MOVQ_BFE(mm6); + __asm__ volatile( + "1: \n\t" + "movq (%1), %%mm0 \n\t" + "movq (%2), %%mm1 \n\t" + "movq (%1,%4), %%mm2 \n\t" + "movq (%2,%4), %%mm3 \n\t" + PAVGBP_MMX_NO_RND(%%mm0, %%mm1, %%mm4, %%mm2, %%mm3, %%mm5) + "movq %%mm4, (%3) \n\t" + "movq %%mm5, (%3,%4) \n\t" + + "movq (%1,%4,2), %%mm0 \n\t" + "movq (%2,%4,2), %%mm1 \n\t" + "movq (%1,%5), %%mm2 \n\t" + "movq (%2,%5), %%mm3 \n\t" + "lea (%1,%4,4), %1 \n\t" + "lea (%2,%4,4), %2 \n\t" + PAVGBP_MMX_NO_RND(%%mm0, %%mm1, %%mm4, %%mm2, %%mm3, %%mm5) + "movq %%mm4, (%3,%4,2) \n\t" + "movq %%mm5, (%3,%5) \n\t" + "lea (%3,%4,4), %3 \n\t" + "subl $4, %0 \n\t" + "jnz 1b \n\t" + :"+r"(h), "+r"(a), "+r"(b), "+r"(dst) + :"r"((x86_reg)stride), "r"((x86_reg)3L*stride) + :"memory"); +// STOP_TIMER("put_vp_no_rnd_pixels8_l2_mmx") +} +#endif /* HAVE_INLINE_ASM */ + av_cold void ff_vp3dsp_init_x86(VP3DSPContext *c, int flags) { int cpuflags = av_get_cpu_flags(); +#if HAVE_INLINE_ASM + c->put_no_rnd_pixels_l2 = put_vp_no_rnd_pixels8_l2_mmx; +#endif /* HAVE_INLINE_ASM */ + #if ARCH_X86_32 if (EXTERNAL_MMX(cpuflags)) { c->idct_put = ff_vp3_idct_put_mmx; |