generic-library/ffmpeg
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

x86: dsputil: Move avg_pixels8_mmx() out of rnd_template.c

Browse Source 
The function is only instantiated once, so there is no point
in keeping it in a template file.
This commit is contained in:
Diego Biurrun 2013-04-27 22:01:07 +02:00
parent 9b3a04d306
commit 9c112a6158
5 changed files with 25 additions and 23 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
libavcodec/x86/dsputil_mmx.c
View File

@ -1018,7 +1018,7 @@ void ff_put_cavs_qpel8_mc00_mmx(uint8_t *dst, uint8_t *src, ptrdiff_t stride)
void ff_avg_cavs_qpel8_mc00_mmx(uint8_t *dst, uint8_t *src, ptrdiff_t stride) void ff_avg_cavs_qpel8_mc00_mmx(uint8_t *dst, uint8_t *src, ptrdiff_t stride)
{ {
avg_pixels8_mmx(dst, src, stride, 8); ff_avg_pixels8_mmx(dst, src, stride, 8);
} }
void ff_put_cavs_qpel16_mc00_mmx(uint8_t *dst, uint8_t *src, ptrdiff_t stride) void ff_put_cavs_qpel16_mc00_mmx(uint8_t *dst, uint8_t *src, ptrdiff_t stride)

2
libavcodec/x86/dsputil_mmx.h
View File

@ -156,6 +156,8 @@ void ff_put_pixels_clamped_mmx(const int16_t *block, uint8_t *pixels, int line_s
void ff_put_signed_pixels_clamped_mmx(const int16_t *block, uint8_t *pixels, int line_size); void ff_put_signed_pixels_clamped_mmx(const int16_t *block, uint8_t *pixels, int line_size);
void ff_avg_pixels8_mmx(uint8_t *block, const uint8_t *pixels,
ptrdiff_t line_size, int h);
void ff_put_pixels8_mmx(uint8_t *block, const uint8_t *pixels, void ff_put_pixels8_mmx(uint8_t *block, const uint8_t *pixels,
ptrdiff_t line_size, int h); ptrdiff_t line_size, int h);
void ff_put_pixels16_mmx(uint8_t *block, const uint8_t *pixels, void ff_put_pixels16_mmx(uint8_t *block, const uint8_t *pixels,

21
libavcodec/x86/fpel_mmx.c
View File

@ -29,6 +29,27 @@
#if HAVE_MMX_INLINE #if HAVE_MMX_INLINE
// in case more speed is needed - unroling would certainly help
void ff_avg_pixels8_mmx(uint8_t *block, const uint8_t *pixels,
ptrdiff_t line_size, int h)
{
MOVQ_BFE(mm6);
JUMPALIGN();
do {
__asm__ volatile(
"movq %0, %%mm0 \n\t"
"movq %1, %%mm1 \n\t"
PAVGB_MMX(%%mm0, %%mm1, %%mm2, %%mm6)
"movq %%mm2, %0 \n\t"
:"+m"(*block)
:"m"(*pixels)
:"memory");
pixels += line_size;
block += line_size;
}
while (--h);
}
void ff_put_pixels8_mmx(uint8_t *block, const uint8_t *pixels, void ff_put_pixels8_mmx(uint8_t *block, const uint8_t *pixels,
ptrdiff_t line_size, int h) ptrdiff_t line_size, int h)
{ {

1
libavcodec/x86/hpeldsp_init.c
View File

@ -74,6 +74,7 @@ void ff_avg_pixels8_xy2_mmxext(uint8_t *block, const uint8_t *pixels,
void ff_avg_pixels8_xy2_3dnow(uint8_t *block, const uint8_t *pixels, void ff_avg_pixels8_xy2_3dnow(uint8_t *block, const uint8_t *pixels,
ptrdiff_t line_size, int h); ptrdiff_t line_size, int h);
#define avg_pixels8_mmx ff_avg_pixels8_mmx
#define put_pixels8_mmx ff_put_pixels8_mmx #define put_pixels8_mmx ff_put_pixels8_mmx
#define put_pixels16_mmx ff_put_pixels16_mmx #define put_pixels16_mmx ff_put_pixels16_mmx
#define put_no_rnd_pixels8_mmx ff_put_pixels8_mmx #define put_no_rnd_pixels8_mmx ff_put_pixels8_mmx

22
libavcodec/x86/rnd_template.c
View File

@ -92,28 +92,6 @@ static void DEF(put, pixels8_xy2)(uint8_t *block, const uint8_t *pixels, ptrdiff
} }
// avg_pixels // avg_pixels
#ifndef NO_RND
// in case more speed is needed - unroling would certainly help
static void DEF(avg, pixels8)(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h)
{
MOVQ_BFE(mm6);
JUMPALIGN();
do {
__asm__ volatile(
"movq %0, %%mm0 \n\t"
"movq %1, %%mm1 \n\t"
OP_AVG(%%mm0, %%mm1, %%mm2, %%mm6)
"movq %%mm2, %0 \n\t"
:"+m"(*block)
:"m"(*pixels)
:"memory");
pixels += line_size;
block += line_size;
}
while (--h);
}
#endif /* NO_RND */
static void DEF(avg, pixels16)(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h) static void DEF(avg, pixels16)(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h)
{ {
MOVQ_BFE(mm6); MOVQ_BFE(mm6);