Skip to content

F
ffmpeg.wasm-core
Commit 9bfa5363 authored by Daniel Kang's avatar Daniel Kang Committed by Ronald S. Bultje
Browse files
Options

H.264: Add x86 assembly for 10-bit H.264 qpel functions. 

Mainly ported from 8-bit H.264 qpel.

Some code ported from x264. LGPL ok by author.
Signed-off-by: 's avatarRonald S. Bultje <rsbultje@gmail.com>
parent 10dde477
Expand all Hide whitespace changes
Inline Side-by-side
Showing with 1076 additions and 50 deletions
libavcodec/x86/Makefile
View file @ 9bfa5363
......@@ -46,6 +46,7 @@ MMX-OBJS-$(HAVE_YASM) += x86/dsputil_yasm.o \
x86/fmtconvert.o \
x86/h264_chromamc.o \
x86/h264_chromamc_10bit.o \
x86/h264_qpel_10bit.o \
$(YASM-OBJS-yes)
MMX-OBJS-$(CONFIG_FFT) += x86/fft.o
......
libavcodec/x86/dsputil_mmx.c
View file @ 9bfa5363
......@@ -2627,44 +2627,56 @@ void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx)
c->put_no_rnd_pixels_tab[1][2] = put_no_rnd_pixels8_y2_exact_mmx2;
}
#define SET_QPEL_FUNCS(PFX, IDX, SIZE, CPU) \
c->PFX ## _pixels_tab[IDX][ 0] = PFX ## SIZE ## _mc00_ ## CPU; \
c->PFX ## _pixels_tab[IDX][ 1] = PFX ## SIZE ## _mc10_ ## CPU; \
c->PFX ## _pixels_tab[IDX][ 2] = PFX ## SIZE ## _mc20_ ## CPU; \
c->PFX ## _pixels_tab[IDX][ 3] = PFX ## SIZE ## _mc30_ ## CPU; \
c->PFX ## _pixels_tab[IDX][ 4] = PFX ## SIZE ## _mc01_ ## CPU; \
c->PFX ## _pixels_tab[IDX][ 5] = PFX ## SIZE ## _mc11_ ## CPU; \
c->PFX ## _pixels_tab[IDX][ 6] = PFX ## SIZE ## _mc21_ ## CPU; \
c->PFX ## _pixels_tab[IDX][ 7] = PFX ## SIZE ## _mc31_ ## CPU; \
c->PFX ## _pixels_tab[IDX][ 8] = PFX ## SIZE ## _mc02_ ## CPU; \
c->PFX ## _pixels_tab[IDX][ 9] = PFX ## SIZE ## _mc12_ ## CPU; \
c->PFX ## _pixels_tab[IDX][10] = PFX ## SIZE ## _mc22_ ## CPU; \
c->PFX ## _pixels_tab[IDX][11] = PFX ## SIZE ## _mc32_ ## CPU; \
c->PFX ## _pixels_tab[IDX][12] = PFX ## SIZE ## _mc03_ ## CPU; \
c->PFX ## _pixels_tab[IDX][13] = PFX ## SIZE ## _mc13_ ## CPU; \
c->PFX ## _pixels_tab[IDX][14] = PFX ## SIZE ## _mc23_ ## CPU; \
c->PFX ## _pixels_tab[IDX][15] = PFX ## SIZE ## _mc33_ ## CPU
SET_QPEL_FUNCS(put_qpel, 0, 16, mmx2);
SET_QPEL_FUNCS(put_qpel, 1, 8, mmx2);
SET_QPEL_FUNCS(put_no_rnd_qpel, 0, 16, mmx2);
SET_QPEL_FUNCS(put_no_rnd_qpel, 1, 8, mmx2);
SET_QPEL_FUNCS(avg_qpel, 0, 16, mmx2);
SET_QPEL_FUNCS(avg_qpel, 1, 8, mmx2);
#define SET_QPEL_FUNCS(PFX, IDX, SIZE, CPU, PREFIX) \
c->PFX ## _pixels_tab[IDX][ 0] = PREFIX ## PFX ## SIZE ## _mc00_ ## CPU; \
c->PFX ## _pixels_tab[IDX][ 1] = PREFIX ## PFX ## SIZE ## _mc10_ ## CPU; \
c->PFX ## _pixels_tab[IDX][ 2] = PREFIX ## PFX ## SIZE ## _mc20_ ## CPU; \
c->PFX ## _pixels_tab[IDX][ 3] = PREFIX ## PFX ## SIZE ## _mc30_ ## CPU; \
c->PFX ## _pixels_tab[IDX][ 4] = PREFIX ## PFX ## SIZE ## _mc01_ ## CPU; \
c->PFX ## _pixels_tab[IDX][ 5] = PREFIX ## PFX ## SIZE ## _mc11_ ## CPU; \
c->PFX ## _pixels_tab[IDX][ 6] = PREFIX ## PFX ## SIZE ## _mc21_ ## CPU; \
c->PFX ## _pixels_tab[IDX][ 7] = PREFIX ## PFX ## SIZE ## _mc31_ ## CPU; \
c->PFX ## _pixels_tab[IDX][ 8] = PREFIX ## PFX ## SIZE ## _mc02_ ## CPU; \
c->PFX ## _pixels_tab[IDX][ 9] = PREFIX ## PFX ## SIZE ## _mc12_ ## CPU; \
c->PFX ## _pixels_tab[IDX][10] = PREFIX ## PFX ## SIZE ## _mc22_ ## CPU; \
c->PFX ## _pixels_tab[IDX][11] = PREFIX ## PFX ## SIZE ## _mc32_ ## CPU; \
c->PFX ## _pixels_tab[IDX][12] = PREFIX ## PFX ## SIZE ## _mc03_ ## CPU; \
c->PFX ## _pixels_tab[IDX][13] = PREFIX ## PFX ## SIZE ## _mc13_ ## CPU; \
c->PFX ## _pixels_tab[IDX][14] = PREFIX ## PFX ## SIZE ## _mc23_ ## CPU; \
c->PFX ## _pixels_tab[IDX][15] = PREFIX ## PFX ## SIZE ## _mc33_ ## CPU
SET_QPEL_FUNCS(put_qpel, 0, 16, mmx2, );
SET_QPEL_FUNCS(put_qpel, 1, 8, mmx2, );
SET_QPEL_FUNCS(put_no_rnd_qpel, 0, 16, mmx2, );
SET_QPEL_FUNCS(put_no_rnd_qpel, 1, 8, mmx2, );
SET_QPEL_FUNCS(avg_qpel, 0, 16, mmx2, );
SET_QPEL_FUNCS(avg_qpel, 1, 8, mmx2, );
if (!high_bit_depth) {
SET_QPEL_FUNCS(put_h264_qpel, 0, 16, mmx2);
SET_QPEL_FUNCS(put_h264_qpel, 1, 8, mmx2);
SET_QPEL_FUNCS(put_h264_qpel, 2, 4, mmx2);
SET_QPEL_FUNCS(avg_h264_qpel, 0, 16, mmx2);
SET_QPEL_FUNCS(avg_h264_qpel, 1, 8, mmx2);
SET_QPEL_FUNCS(avg_h264_qpel, 2, 4, mmx2);
SET_QPEL_FUNCS(put_h264_qpel, 0, 16, mmx2, );
SET_QPEL_FUNCS(put_h264_qpel, 1, 8, mmx2, );
SET_QPEL_FUNCS(put_h264_qpel, 2, 4, mmx2, );
SET_QPEL_FUNCS(avg_h264_qpel, 0, 16, mmx2, );
SET_QPEL_FUNCS(avg_h264_qpel, 1, 8, mmx2, );
SET_QPEL_FUNCS(avg_h264_qpel, 2, 4, mmx2, );
}
#if HAVE_YASM
else if (bit_depth == 10) {
#if !ARCH_X86_64
SET_QPEL_FUNCS(avg_h264_qpel, 0, 16, 10_mmxext, ff_);
SET_QPEL_FUNCS(put_h264_qpel, 0, 16, 10_mmxext, ff_);
SET_QPEL_FUNCS(put_h264_qpel, 1, 8, 10_mmxext, ff_);
SET_QPEL_FUNCS(avg_h264_qpel, 1, 8, 10_mmxext, ff_);
#endif
SET_QPEL_FUNCS(put_h264_qpel, 2, 4, 10_mmxext, ff_);
SET_QPEL_FUNCS(avg_h264_qpel, 2, 4, 10_mmxext, ff_);
}
#endif
SET_QPEL_FUNCS(put_2tap_qpel, 0, 16, mmx2);
SET_QPEL_FUNCS(put_2tap_qpel, 1, 8, mmx2);
SET_QPEL_FUNCS(avg_2tap_qpel, 0, 16, mmx2);
SET_QPEL_FUNCS(avg_2tap_qpel, 1, 8, mmx2);
SET_QPEL_FUNCS(put_2tap_qpel, 0, 16, mmx2, );
SET_QPEL_FUNCS(put_2tap_qpel, 1, 8, mmx2, );
SET_QPEL_FUNCS(avg_2tap_qpel, 0, 16, mmx2, );
SET_QPEL_FUNCS(avg_2tap_qpel, 1, 8, mmx2, );
#if HAVE_YASM
c->avg_rv40_chroma_pixels_tab[0]= ff_avg_rv40_chroma_mc8_mmx2;
......@@ -2725,26 +2737,26 @@ void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx)
c->put_no_rnd_pixels_tab[1][2] = put_no_rnd_pixels8_y2_exact_3dnow;
}
SET_QPEL_FUNCS(put_qpel, 0, 16, 3dnow);
SET_QPEL_FUNCS(put_qpel, 1, 8, 3dnow);
SET_QPEL_FUNCS(put_no_rnd_qpel, 0, 16, 3dnow);
SET_QPEL_FUNCS(put_no_rnd_qpel, 1, 8, 3dnow);
SET_QPEL_FUNCS(avg_qpel, 0, 16, 3dnow);
SET_QPEL_FUNCS(avg_qpel, 1, 8, 3dnow);
SET_QPEL_FUNCS(put_qpel, 0, 16, 3dnow, );
SET_QPEL_FUNCS(put_qpel, 1, 8, 3dnow, );
SET_QPEL_FUNCS(put_no_rnd_qpel, 0, 16, 3dnow, );
SET_QPEL_FUNCS(put_no_rnd_qpel, 1, 8, 3dnow, );
SET_QPEL_FUNCS(avg_qpel, 0, 16, 3dnow, );
SET_QPEL_FUNCS(avg_qpel, 1, 8, 3dnow, );
if (!high_bit_depth) {
SET_QPEL_FUNCS(put_h264_qpel, 0, 16, 3dnow);
SET_QPEL_FUNCS(put_h264_qpel, 1, 8, 3dnow);
SET_QPEL_FUNCS(put_h264_qpel, 2, 4, 3dnow);
SET_QPEL_FUNCS(avg_h264_qpel, 0, 16, 3dnow);
SET_QPEL_FUNCS(avg_h264_qpel, 1, 8, 3dnow);
SET_QPEL_FUNCS(avg_h264_qpel, 2, 4, 3dnow);
SET_QPEL_FUNCS(put_h264_qpel, 0, 16, 3dnow, );
SET_QPEL_FUNCS(put_h264_qpel, 1, 8, 3dnow, );
SET_QPEL_FUNCS(put_h264_qpel, 2, 4, 3dnow, );
SET_QPEL_FUNCS(avg_h264_qpel, 0, 16, 3dnow, );
SET_QPEL_FUNCS(avg_h264_qpel, 1, 8, 3dnow, );
SET_QPEL_FUNCS(avg_h264_qpel, 2, 4, 3dnow, );
}
SET_QPEL_FUNCS(put_2tap_qpel, 0, 16, 3dnow);
SET_QPEL_FUNCS(put_2tap_qpel, 1, 8, 3dnow);
SET_QPEL_FUNCS(avg_2tap_qpel, 0, 16, 3dnow);
SET_QPEL_FUNCS(avg_2tap_qpel, 1, 8, 3dnow);
SET_QPEL_FUNCS(put_2tap_qpel, 0, 16, 3dnow, );
SET_QPEL_FUNCS(put_2tap_qpel, 1, 8, 3dnow, );
SET_QPEL_FUNCS(avg_2tap_qpel, 0, 16, 3dnow, );
SET_QPEL_FUNCS(avg_2tap_qpel, 1, 8, 3dnow, );
#if HAVE_YASM
if (!high_bit_depth) {
......@@ -2788,7 +2800,20 @@ void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx)
H264_QPEL_FUNCS(3, 3, sse2);
}
#if HAVE_YASM
#define H264_QPEL_FUNCS_10(x, y, CPU)\
c->put_h264_qpel_pixels_tab[0][x+y*4] = ff_put_h264_qpel16_mc##x##y##_10_##CPU;\
c->put_h264_qpel_pixels_tab[1][x+y*4] = ff_put_h264_qpel8_mc##x##y##_10_##CPU;\
c->avg_h264_qpel_pixels_tab[0][x+y*4] = ff_avg_h264_qpel16_mc##x##y##_10_##CPU;\
c->avg_h264_qpel_pixels_tab[1][x+y*4] = ff_avg_h264_qpel8_mc##x##y##_10_##CPU;
if (bit_depth == 10) {
SET_QPEL_FUNCS(put_h264_qpel, 0, 16, 10_sse2, ff_);
SET_QPEL_FUNCS(put_h264_qpel, 1, 8, 10_sse2, ff_);
SET_QPEL_FUNCS(avg_h264_qpel, 0, 16, 10_sse2, ff_);
SET_QPEL_FUNCS(avg_h264_qpel, 1, 8, 10_sse2, ff_);
H264_QPEL_FUNCS_10(1, 0, sse2_cache64)
H264_QPEL_FUNCS_10(2, 0, sse2_cache64)
H264_QPEL_FUNCS_10(3, 0, sse2_cache64)
c->put_h264_chroma_pixels_tab[0]= ff_put_h264_chroma_mc8_10_sse2;
c->avg_h264_chroma_pixels_tab[0]= ff_avg_h264_chroma_mc8_10_sse2;
}
......@@ -2810,6 +2835,11 @@ void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx)
H264_QPEL_FUNCS(3, 2, ssse3);
H264_QPEL_FUNCS(3, 3, ssse3);
}
else if (bit_depth == 10) {
H264_QPEL_FUNCS_10(1, 0, ssse3_cache64)
H264_QPEL_FUNCS_10(2, 0, ssse3_cache64)
H264_QPEL_FUNCS_10(3, 0, ssse3_cache64)
}
c->add_png_paeth_prediction= add_png_paeth_prediction_ssse3;
#if HAVE_YASM
if (!high_bit_depth) {
......@@ -2906,6 +2936,12 @@ void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx)
#if HAVE_AVX && HAVE_YASM
if (mm_flags & AV_CPU_FLAG_AVX) {
if (bit_depth == 10) {
//AVX implies !cache64.
//TODO: Port cache(32|64) detection from x264.
H264_QPEL_FUNCS_10(1, 0, sse2)
H264_QPEL_FUNCS_10(2, 0, sse2)
H264_QPEL_FUNCS_10(3, 0, sse2)
c->put_h264_chroma_pixels_tab[0]= ff_put_h264_chroma_mc8_10_avx;
c->avg_h264_chroma_pixels_tab[0]= ff_avg_h264_chroma_mc8_10_avx;
}
......
libavcodec/x86/h264_qpel_10bit.asm 0 → 100644
View file @ 9bfa5363
This diff is collapsed.
libavcodec/x86/h264_qpel_mmx.c
View file @ 9bfa5363
/*
* Copyright (c) 2004-2005 Michael Niedermayer, Loren Merritt
* Copyright (c) 2011 Daniel Kang
*
* This file is part of Libav.
*
......@@ -1199,3 +1200,100 @@ H264_MC_816(H264_MC_HV, sse2)
H264_MC_816(H264_MC_H, ssse3)
H264_MC_816(H264_MC_HV, ssse3)
#endif
//10bit
#define LUMA_MC_OP(OP, NUM, DEPTH, TYPE, OPT) \
void ff_ ## OP ## _h264_qpel ## NUM ## _ ## TYPE ## _ ## DEPTH ## _ ## OPT \
(uint8_t *dst, uint8_t *src, int stride);
#define LUMA_MC_ALL(DEPTH, TYPE, OPT) \
LUMA_MC_OP(put, 4, DEPTH, TYPE, OPT) \
LUMA_MC_OP(avg, 4, DEPTH, TYPE, OPT) \
LUMA_MC_OP(put, 8, DEPTH, TYPE, OPT) \
LUMA_MC_OP(avg, 8, DEPTH, TYPE, OPT) \
LUMA_MC_OP(put, 16, DEPTH, TYPE, OPT) \
LUMA_MC_OP(avg, 16, DEPTH, TYPE, OPT)
#define LUMA_MC_816(DEPTH, TYPE, OPT) \
LUMA_MC_OP(put, 8, DEPTH, TYPE, OPT) \
LUMA_MC_OP(avg, 8, DEPTH, TYPE, OPT) \
LUMA_MC_OP(put, 16, DEPTH, TYPE, OPT) \
LUMA_MC_OP(avg, 16, DEPTH, TYPE, OPT)
LUMA_MC_ALL(10, mc00, mmxext)
LUMA_MC_ALL(10, mc10, mmxext)
LUMA_MC_ALL(10, mc20, mmxext)
LUMA_MC_ALL(10, mc30, mmxext)
LUMA_MC_ALL(10, mc01, mmxext)
LUMA_MC_ALL(10, mc11, mmxext)
LUMA_MC_ALL(10, mc21, mmxext)
LUMA_MC_ALL(10, mc31, mmxext)
LUMA_MC_ALL(10, mc02, mmxext)
LUMA_MC_ALL(10, mc12, mmxext)
LUMA_MC_ALL(10, mc22, mmxext)
LUMA_MC_ALL(10, mc32, mmxext)
LUMA_MC_ALL(10, mc03, mmxext)
LUMA_MC_ALL(10, mc13, mmxext)
LUMA_MC_ALL(10, mc23, mmxext)
LUMA_MC_ALL(10, mc33, mmxext)
LUMA_MC_816(10, mc00, sse2)
LUMA_MC_816(10, mc10, sse2)
LUMA_MC_816(10, mc10, sse2_cache64)
LUMA_MC_816(10, mc10, ssse3_cache64)
LUMA_MC_816(10, mc20, sse2)
LUMA_MC_816(10, mc20, sse2_cache64)
LUMA_MC_816(10, mc20, ssse3_cache64)
LUMA_MC_816(10, mc30, sse2)
LUMA_MC_816(10, mc30, sse2_cache64)
LUMA_MC_816(10, mc30, ssse3_cache64)
LUMA_MC_816(10, mc01, sse2)
LUMA_MC_816(10, mc11, sse2)
LUMA_MC_816(10, mc21, sse2)
LUMA_MC_816(10, mc31, sse2)
LUMA_MC_816(10, mc02, sse2)
LUMA_MC_816(10, mc12, sse2)
LUMA_MC_816(10, mc22, sse2)
LUMA_MC_816(10, mc32, sse2)
LUMA_MC_816(10, mc03, sse2)
LUMA_MC_816(10, mc13, sse2)
LUMA_MC_816(10, mc23, sse2)
LUMA_MC_816(10, mc33, sse2)
#define QPEL16_OPMC(OP, MC, MMX)\
void ff_ ## OP ## _h264_qpel16_ ## MC ## _10_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
ff_ ## OP ## _h264_qpel8_ ## MC ## _10_ ## MMX(dst , src , stride);\
ff_ ## OP ## _h264_qpel8_ ## MC ## _10_ ## MMX(dst+16, src+16, stride);\
src += 8*stride;\
dst += 8*stride;\
ff_ ## OP ## _h264_qpel8_ ## MC ## _10_ ## MMX(dst , src , stride);\
ff_ ## OP ## _h264_qpel8_ ## MC ## _10_ ## MMX(dst+16, src+16, stride);\
}
#define QPEL16_OP(MC, MMX)\
QPEL16_OPMC(put, MC, MMX)\
QPEL16_OPMC(avg, MC, MMX)
#define QPEL16(MMX)\
QPEL16_OP(mc00, MMX)\
QPEL16_OP(mc01, MMX)\
QPEL16_OP(mc02, MMX)\
QPEL16_OP(mc03, MMX)\
QPEL16_OP(mc10, MMX)\
QPEL16_OP(mc11, MMX)\
QPEL16_OP(mc12, MMX)\
QPEL16_OP(mc13, MMX)\
QPEL16_OP(mc20, MMX)\
QPEL16_OP(mc21, MMX)\
QPEL16_OP(mc22, MMX)\
QPEL16_OP(mc23, MMX)\
QPEL16_OP(mc30, MMX)\
QPEL16_OP(mc31, MMX)\
QPEL16_OP(mc32, MMX)\
QPEL16_OP(mc33, MMX)
#if ARCH_X86_32 // ARCH_X86_64 implies sse2+
QPEL16(mmxext)
#endif
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment