breeze/FFmpeg
1
0
Fork 0
mirror of https://github.com/FFmpeg/FFmpeg.git synced 2026-02-04 14:30:55 +08:00
Code Issues Packages Projects Releases Wiki Activity

avcodec/x86/fpel: Add blocksize x blocksize avg/put functions

Browse Source 
This commit deduplicates the wrappers around the fpel functions
for copying whole blocks (i.e. height equaling width). It does
this in a manner which avoids having push/pop function arguments
when the calling convention forces one to pass them on the stack
(as in 32bit systems).

Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
This commit is contained in:
Andreas Rheinhardt
2025-10-05 05:33:21 +02:00
parent ad498f9759
commit 378d5bb08a
5 changed files with 24 additions and 67 deletions
Download Patch File Download Diff File Expand all files Collapse all files

34
libavcodec/x86/cavsdsp.c
View File

@@ -46,36 +46,10 @@ static void cavs_idct8_add_sse2(uint8_t *dst, int16_t *block, ptrdiff_t stride)
#endif /* HAVE_SSE2_EXTERNAL */
#if HAVE_MMX_EXTERNAL
static void put_cavs_qpel8_mc00_mmx(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride)
{
ff_put_pixels8_mmx(dst, src, stride, 8);
}
static void avg_cavs_qpel8_mc00_mmxext(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride)
{
ff_avg_pixels8_mmxext(dst, src, stride, 8);
}
static void put_cavs_qpel16_mc00_sse2(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride)
{
ff_put_pixels16_sse2(dst, src, stride, 16);
}
static void avg_cavs_qpel16_mc00_sse2(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride)
{
ff_avg_pixels16_sse2(dst, src, stride, 16);
}
#endif
static av_cold void cavsdsp_init_mmx(CAVSDSPContext *c)
{
#if HAVE_MMX_EXTERNAL
c->put_cavs_qpel_pixels_tab[1][0] = put_cavs_qpel8_mc00_mmx;
c->put_cavs_qpel_pixels_tab[1][0] = ff_put_pixels8x8_mmx;
#endif /* HAVE_MMX_EXTERNAL */
}
@@ -129,12 +103,12 @@ av_cold void ff_cavsdsp_init_x86(CAVSDSPContext *c)
#if HAVE_MMX_EXTERNAL
if (EXTERNAL_MMXEXT(cpu_flags)) {
c->avg_cavs_qpel_pixels_tab[1][0] = avg_cavs_qpel8_mc00_mmxext;
c->avg_cavs_qpel_pixels_tab[1][0] = ff_avg_pixels8x8_mmxext;
}
#endif
#if HAVE_SSE2_EXTERNAL
if (EXTERNAL_SSE2(cpu_flags)) {
c->put_cavs_qpel_pixels_tab[0][ 0] = put_cavs_qpel16_mc00_sse2;
c->put_cavs_qpel_pixels_tab[0][ 0] = ff_put_pixels16x16_sse2;
c->put_cavs_qpel_pixels_tab[0][ 2] = put_cavs_qpel16_mc20_sse2;
c->put_cavs_qpel_pixels_tab[0][ 4] = put_cavs_qpel16_mc01_sse2;
c->put_cavs_qpel_pixels_tab[0][ 8] = put_cavs_qpel16_mc02_sse2;
@@ -144,7 +118,7 @@ av_cold void ff_cavsdsp_init_x86(CAVSDSPContext *c)
c->put_cavs_qpel_pixels_tab[1][ 8] = ff_put_cavs_qpel8_mc02_sse2;
c->put_cavs_qpel_pixels_tab[1][12] = ff_put_cavs_qpel8_mc03_sse2;
c->avg_cavs_qpel_pixels_tab[0][ 0] = avg_cavs_qpel16_mc00_sse2;
c->avg_cavs_qpel_pixels_tab[0][ 0] = ff_avg_pixels16x16_sse2;
c->avg_cavs_qpel_pixels_tab[0][ 2] = avg_cavs_qpel16_mc20_sse2;
c->avg_cavs_qpel_pixels_tab[0][ 4] = avg_cavs_qpel16_mc01_sse2;
c->avg_cavs_qpel_pixels_tab[0][ 8] = avg_cavs_qpel16_mc02_sse2;

5
libavcodec/x86/fpel.asm
View File

@@ -35,7 +35,12 @@ SECTION .text
%define LOAD movu
%define SAVE mova
%endif
cglobal %1_pixels%2x%2, 3,5,4
mov r3d, %2
jmp %1_pixels%2_after_prologue
cglobal %1_pixels%2, 4,5,4
%1_pixels%2_after_prologue:
lea r4, [r2*3]
.loop:
LOAD m0, [r1]

8
libavcodec/x86/fpel.h
View File

@@ -24,12 +24,20 @@
void ff_avg_pixels8_mmxext(uint8_t *block, const uint8_t *pixels,
ptrdiff_t line_size, int h);
void ff_avg_pixels8x8_mmxext(uint8_t *block, const uint8_t *pixels,
ptrdiff_t line_size);
void ff_avg_pixels16_sse2(uint8_t *block, const uint8_t *pixels,
ptrdiff_t line_size, int h);
void ff_avg_pixels16x16_sse2(uint8_t *block, const uint8_t *pixels,
ptrdiff_t line_size);
void ff_put_pixels8_mmx(uint8_t *block, const uint8_t *pixels,
ptrdiff_t line_size, int h);
void ff_put_pixels8x8_mmx(uint8_t *block, const uint8_t *pixels,
ptrdiff_t line_size);
void ff_put_pixels16_sse2(uint8_t *block, const uint8_t *pixels,
ptrdiff_t line_size, int h);
void ff_put_pixels16x16_sse2(uint8_t *block, const uint8_t *pixels,
ptrdiff_t line_size);
#endif /* AVCODEC_X86_FPEL_H */

23
libavcodec/x86/h264_qpel.c
View File

@@ -163,23 +163,6 @@ H264_MC_V(OPNAME, SIZE, MMX, ALIGN, SHIFT5_EXT)\
H264_MC_H(OPNAME, SIZE, MMX, ALIGN, SHIFT5_EXT)\
H264_MC_HV(OPNAME, SIZE, MMX, ALIGN, SHIFT5_EXT)\
static void put_h264_qpel16_mc00_sse2 (uint8_t *dst, const uint8_t *src,
ptrdiff_t stride)
{
ff_put_pixels16_sse2(dst, src, stride, 16);
}
static void avg_h264_qpel16_mc00_sse2 (uint8_t *dst, const uint8_t *src,
ptrdiff_t stride)
{
ff_avg_pixels16_sse2(dst, src, stride, 16);
}
static void avg_h264_qpel8_mc00_mmxext(uint8_t *dst, const uint8_t *src,
ptrdiff_t stride)
{
ff_avg_pixels8_mmxext(dst, src, stride, 8);
}
#define H264_MC_H(OPNAME, SIZE, MMX, ALIGN, UNUSED) \
static void OPNAME ## h264_qpel ## SIZE ## _mc10_ ## MMX(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
{\
@@ -424,7 +407,7 @@ av_cold void ff_h264qpel_init_x86(H264QpelContext *c, int bit_depth)
if (EXTERNAL_MMXEXT(cpu_flags)) {
if (!high_bit_depth) {
SET_QPEL_FUNCS_1PP(put_h264_qpel, 2, 4, mmxext, );
c->avg_h264_qpel_pixels_tab[1][0] = avg_h264_qpel8_mc00_mmxext;
c->avg_h264_qpel_pixels_tab[1][0] = ff_avg_pixels8x8_mmxext;
SET_QPEL_FUNCS_1PP(avg_h264_qpel, 2, 4, mmxext, );
c->avg_h264_qpel_pixels_tab[2][0] = ff_avg_pixels4_mmxext;
} else if (bit_depth == 10) {
@@ -447,8 +430,8 @@ av_cold void ff_h264qpel_init_x86(H264QpelContext *c, int bit_depth)
H264_QPEL_FUNCS(3, 1, sse2);
H264_QPEL_FUNCS(3, 2, sse2);
H264_QPEL_FUNCS(3, 3, sse2);
c->put_h264_qpel_pixels_tab[0][0] = put_h264_qpel16_mc00_sse2;
c->avg_h264_qpel_pixels_tab[0][0] = avg_h264_qpel16_mc00_sse2;
c->put_h264_qpel_pixels_tab[0][0] = ff_put_pixels16x16_sse2;
c->avg_h264_qpel_pixels_tab[0][0] = ff_avg_pixels16x16_sse2;
}
if (bit_depth == 10) {

21
libavcodec/x86/qpeldsp_init.c
View File

@@ -489,19 +489,6 @@ QPEL_OP(put_, _, mmxext, PASSTHROUGH)
QPEL_OP(avg_, _, mmxext, STRIP_HEIGHT)
QPEL_OP(put_no_rnd_, _no_rnd_, mmxext, PASSTHROUGH)
#define MC00(OPNAME, SIZE, EXT) \
static void OPNAME ## _qpel ## SIZE ## _mc00_ ## EXT(uint8_t *dst, \
const uint8_t *src,\
ptrdiff_t stride) \
{ \
ff_ ## OPNAME ## _pixels ## SIZE ##_ ## EXT(dst, src, stride, SIZE);\
}
MC00(put, 8, mmx)
MC00(avg, 8, mmxext)
MC00(put, 16, sse2)
MC00(avg, 16, sse2)
#endif /* HAVE_X86ASM */
#define SET_QPEL_FUNCS(PFX, IDX, SIZE, CPU, PREFIX) \
@@ -530,12 +517,12 @@ av_cold void ff_qpeldsp_init_x86(QpelDSPContext *c)
if (X86_MMXEXT(cpu_flags)) {
#if HAVE_MMXEXT_EXTERNAL
SET_QPEL_FUNCS(avg_qpel, 0, 16, mmxext, );
c->avg_qpel_pixels_tab[1][0] = avg_qpel8_mc00_mmxext;
c->avg_qpel_pixels_tab[1][0] = ff_avg_pixels8x8_mmxext;
SET_QPEL_FUNCS(avg_qpel, 1, 8, mmxext, );
SET_QPEL_FUNCS(put_qpel, 0, 16, mmxext, );
c->put_no_rnd_qpel_pixels_tab[1][0] =
c->put_qpel_pixels_tab[1][0] = put_qpel8_mc00_mmx;
c->put_qpel_pixels_tab[1][0] = ff_put_pixels8x8_mmx;
SET_QPEL_FUNCS(put_qpel, 1, 8, mmxext, );
SET_QPEL_FUNCS(put_no_rnd_qpel, 0, 16, mmxext, );
SET_QPEL_FUNCS(put_no_rnd_qpel, 1, 8, mmxext, );
@@ -544,8 +531,8 @@ av_cold void ff_qpeldsp_init_x86(QpelDSPContext *c)
#if HAVE_SSE2_EXTERNAL
if (EXTERNAL_SSE2(cpu_flags)) {
c->put_no_rnd_qpel_pixels_tab[0][0] =
c->put_qpel_pixels_tab[0][0] = put_qpel16_mc00_sse2;
c->avg_qpel_pixels_tab[0][0] = avg_qpel16_mc00_sse2;
c->put_qpel_pixels_tab[0][0] = ff_put_pixels16x16_sse2;
c->avg_qpel_pixels_tab[0][0] = ff_avg_pixels16x16_sse2;
}
#endif
}