[FFmpeg-devel] [PATCH 1/3] x86/hevc: add ff_hevc_sao_band_filter_{8, 10, 12}_{sse2, avx, avx2}
James Almer
jamrial at gmail.com
Sun Feb 1 20:00:09 CET 2015
Original x86 intrinsics code and initial 8bit yasm port by Pierre-Edouard Lepere.
10/12bit yasm ports, refactoring and optimizations by James Almer
Benchmarks of BQTerrace_1920x1080_60_qp22.bin with an Intel Core i5-4200U
width 32
40338 decicycles in sao_band_filter_0_8, 2048 runs, 0 skips
8056 decicycles in ff_hevc_sao_band_filter_8_32_sse2, 2048 runs, 0 skips
7458 decicycles in ff_hevc_sao_band_filter_8_32_avx, 2048 runs, 0 skips
4504 decicycles in ff_hevc_sao_band_filter_8_32_avx2, 2048 runs, 0 skips
width 64
136046 decicycles in sao_band_filter_0_8, 16384 runs, 0 skips
28576 decicycles in ff_hevc_sao_band_filter_8_32_sse2, 16384 runs, 0 skips
26707 decicycles in ff_hevc_sao_band_filter_8_32_avx, 16384 runs, 0 skips
14387 decicycles in ff_hevc_sao_band_filter_8_32_avx2, 16384 runs, 0 skips
Signed-off-by: James Almer <jamrial at gmail.com>
---
libavcodec/hevc_filter.c | 8 +-
libavcodec/hevcdsp.c | 6 +-
libavcodec/hevcdsp.h | 4 +-
libavcodec/x86/Makefile | 3 +-
libavcodec/x86/hevc_sao.asm | 238 ++++++++++++++++++++++++++++++++++++++++++
libavcodec/x86/hevcdsp_init.c | 49 +++++++++
6 files changed, 301 insertions(+), 7 deletions(-)
create mode 100644 libavcodec/x86/hevc_sao.asm
diff --git a/libavcodec/hevc_filter.c b/libavcodec/hevc_filter.c
index fb1aa37..9354c14 100644
--- a/libavcodec/hevc_filter.c
+++ b/libavcodec/hevc_filter.c
@@ -187,6 +187,7 @@ static void restore_tqb_pixels(HEVCContext *s, int x0, int y0, int width, int he
static void sao_filter_CTB(HEVCContext *s, int x, int y)
{
+ static const uint8_t band_tab[8] = { 0, 1, 2, 2, 3, 3, 4, 4 };
int c_idx;
int edges[4]; // 0 left 1 top 2 right 3 bottom
int x_ctb = x >> s->sps->log2_ctb_size;
@@ -252,15 +253,16 @@ static void sao_filter_CTB(HEVCContext *s, int x, int y)
int ctb_size_v = (1 << (s->sps->log2_ctb_size)) >> s->sps->vshift[c_idx];
int width = FFMIN(ctb_size_h, (s->sps->width >> s->sps->hshift[c_idx]) - x0);
int height = FFMIN(ctb_size_v, (s->sps->height >> s->sps->vshift[c_idx]) - y0);
+ int tab = band_tab[(FFALIGN(width, 8) >> 3) - 1];
uint8_t *src = &s->frame->data[c_idx][y0 * stride_src + (x0 << s->sps->pixel_shift)];
uint8_t *dst = &s->sao_frame->data[c_idx][y0 * stride_dst + (x0 << s->sps->pixel_shift)];
switch (sao->type_idx[c_idx]) {
case SAO_BAND:
copy_CTB(dst, src, width << s->sps->pixel_shift, height, stride_dst, stride_src);
- s->hevcdsp.sao_band_filter(src, dst, stride_src, stride_dst,
- sao->offset_val[c_idx], sao->band_position[c_idx],
- width, height);
+ s->hevcdsp.sao_band_filter[tab](src, dst, stride_src, stride_dst,
+ sao->offset_val[c_idx], sao->band_position[c_idx],
+ width, height);
restore_tqb_pixels(s, x, y, width, height, c_idx);
sao->type_idx[c_idx] = SAO_APPLIED;
break;
diff --git a/libavcodec/hevcdsp.c b/libavcodec/hevcdsp.c
index 3eae541..8acad1b 100644
--- a/libavcodec/hevcdsp.c
+++ b/libavcodec/hevcdsp.c
@@ -212,7 +212,11 @@ void ff_hevc_dsp_init(HEVCDSPContext *hevcdsp, int bit_depth)
hevcdsp->idct_dc[2] = FUNC(idct_16x16_dc, depth); \
hevcdsp->idct_dc[3] = FUNC(idct_32x32_dc, depth); \
\
- hevcdsp->sao_band_filter = FUNC(sao_band_filter_0, depth); \
+ hevcdsp->sao_band_filter[0] = \
+ hevcdsp->sao_band_filter[1] = \
+ hevcdsp->sao_band_filter[2] = \
+ hevcdsp->sao_band_filter[3] = \
+ hevcdsp->sao_band_filter[4] = FUNC(sao_band_filter_0, depth); \
hevcdsp->sao_edge_filter[0] = FUNC(sao_edge_filter_0, depth); \
hevcdsp->sao_edge_filter[1] = FUNC(sao_edge_filter_1, depth); \
\
diff --git a/libavcodec/hevcdsp.h b/libavcodec/hevcdsp.h
index a798fb1..6fb161b 100644
--- a/libavcodec/hevcdsp.h
+++ b/libavcodec/hevcdsp.h
@@ -58,8 +58,8 @@ typedef struct HEVCDSPContext {
void (*idct_dc[4])(int16_t *coeffs);
- void (*sao_band_filter)(uint8_t *_dst, uint8_t *_src, ptrdiff_t _stride_dst, ptrdiff_t _stride_src,
- int16_t *sao_offset_val, int sao_left_class, int width, int height);
+ void (*sao_band_filter[5])(uint8_t *_dst, uint8_t *_src, ptrdiff_t _stride_dst, ptrdiff_t _stride_src,
+ int16_t *sao_offset_val, int sao_left_class, int width, int height);
void (*sao_edge_filter[2])(uint8_t *_dst, uint8_t *_src, ptrdiff_t _stride_dst, ptrdiff_t _stride_src,
struct SAOParams *sao, int *borders, int _width,
diff --git a/libavcodec/x86/Makefile b/libavcodec/x86/Makefile
index 7c8e7aa..00dacda 100644
--- a/libavcodec/x86/Makefile
+++ b/libavcodec/x86/Makefile
@@ -134,7 +134,8 @@ YASM-OBJS-$(CONFIG_DCA_DECODER) += x86/dcadsp.o
YASM-OBJS-$(CONFIG_HEVC_DECODER) += x86/hevc_mc.o \
x86/hevc_deblock.o \
x86/hevc_idct.o \
- x86/hevc_res_add.o
+ x86/hevc_res_add.o \
+ x86/hevc_sao.o
YASM-OBJS-$(CONFIG_MLP_DECODER) += x86/mlpdsp.o
YASM-OBJS-$(CONFIG_PNG_DECODER) += x86/pngdsp.o
YASM-OBJS-$(CONFIG_PRORES_DECODER) += x86/proresdsp.o
diff --git a/libavcodec/x86/hevc_sao.asm b/libavcodec/x86/hevc_sao.asm
new file mode 100644
index 0000000..7f36fd0
--- /dev/null
+++ b/libavcodec/x86/hevc_sao.asm
@@ -0,0 +1,238 @@
+;******************************************************************************
+;* SIMD optimized SAO functions for HEVC decoding
+;*
+;* Copyright (c) 2013 Pierre-Edouard LEPERE
+;* Copyright (c) 2014 James Almer
+;*
+;* This file is part of FFmpeg.
+;*
+;* FFmpeg is free software; you can redistribute it and/or
+;* modify it under the terms of the GNU Lesser General Public
+;* License as published by the Free Software Foundation; either
+;* version 2.1 of the License, or (at your option) any later version.
+;*
+;* FFmpeg is distributed in the hope that it will be useful,
+;* but WITHOUT ANY WARRANTY; without even the implied warranty of
+;* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+;* Lesser General Public License for more details.
+;*
+;* You should have received a copy of the GNU Lesser General Public
+;* License along with FFmpeg; if not, write to the Free Software
+;* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+;******************************************************************************
+
+%include "libavutil/x86/x86util.asm"
+
+%if ARCH_X86_64
+SECTION_RODATA 32
+
+pw_mask10: times 16 dw 0x03FF
+pw_mask12: times 16 dw 0x0FFF
+
+SECTION_TEXT
+
+%macro HEVC_SAO_BAND_FILTER_INIT 1
+ and leftq, 31
+ movd xm0, leftd
+ add leftq, 1
+ and leftq, 31
+ movd xm1, leftd
+ add leftq, 1
+ and leftq, 31
+ movd xm2, leftd
+ add leftq, 1
+ and leftq, 31
+ movd xm3, leftd
+
+ SPLATW m0, xm0
+ SPLATW m1, xm1
+ SPLATW m2, xm2
+ SPLATW m3, xm3
+%if mmsize > 16
+ SPLATW m4, [offsetq + 2]
+ SPLATW m5, [offsetq + 4]
+ SPLATW m6, [offsetq + 6]
+ SPLATW m7, [offsetq + 8]
+%else
+ movq m7, [offsetq + 2]
+ SPLATW m4, m7, 0
+ SPLATW m5, m7, 1
+ SPLATW m6, m7, 2
+ SPLATW m7, m7, 3
+%endif
+
+%if %1 > 8
+ mova m13, [pw_mask %+ %1]
+%endif
+ pxor m14, m14
+
+DEFINE_ARGS dst, src, dststride, srcstride, offset, height
+ mov heightd, r7m
+%endmacro
+
+%macro HEVC_SAO_BAND_FILTER_COMPUTE 3
+ psraw %2, %3, %1-5
+ pcmpeqw m10, %2, m0
+ pcmpeqw m11, %2, m1
+ pcmpeqw m12, %2, m2
+ pcmpeqw %2, m3
+ pand m10, m4
+ pand m11, m5
+ pand m12, m6
+ pand %2, m7
+ por m10, m11
+ por m12, %2
+ por m10, m12
+ paddw %3, m10
+%endmacro
+
+;void ff_hevc_sao_band_filter_<width>_8_<opt>(uint8_t *_dst, uint8_t *_src, ptrdiff_t _stride_dst, ptrdiff_t _stride_src,
+; int16_t *sao_offset_val, int sao_left_class, int width, int height);
+%macro HEVC_SAO_BAND_FILTER_8 2
+cglobal hevc_sao_band_filter_%1_8, 6, 6, 15, dst, src, dststride, srcstride, offset, left
+ HEVC_SAO_BAND_FILTER_INIT 8
+
+align 16
+.loop
+%if %1 == 8
+ movq m8, [srcq]
+ punpcklbw m8, m14
+ HEVC_SAO_BAND_FILTER_COMPUTE 8, m9, m8
+ packuswb m8, m14
+ movq [dstq], m8
+%endif ; %1 == 8
+
+%assign i 0
+%rep %2
+ movu m13, [srcq + i]
+ punpcklbw m8, m13, m14
+ HEVC_SAO_BAND_FILTER_COMPUTE 8, m9, m8
+ punpckhbw m13, m14
+ HEVC_SAO_BAND_FILTER_COMPUTE 8, m9, m13
+ packuswb m8, m13
+ movu [dstq + i], m8
+%assign i i+mmsize
+%endrep
+
+%if %1 == 48
+INIT_XMM cpuname
+
+ movu m13, [srcq + i]
+ punpcklbw m8, m13, m14
+ HEVC_SAO_BAND_FILTER_COMPUTE 8, m9, m8
+ punpckhbw m13, m14
+ HEVC_SAO_BAND_FILTER_COMPUTE 8, m9, m13
+ packuswb m8, m13
+ movu [dstq + i], m8
+%assign i i+16
+%endif ; %1 == 48
+
+ add dstq, dststrideq ; dst += dststride
+ add srcq, srcstrideq ; src += srcstride
+ dec heightd ; cmp height
+ jnz .loop ; height loop
+ REP_RET
+%endmacro
+
+;void ff_hevc_sao_band_filter_<width>_<depth>_<opt>(uint8_t *_dst, uint8_t *_src, ptrdiff_t _stride_dst, ptrdiff_t _stride_src,
+; int16_t *sao_offset_val, int sao_left_class, int width, int height);
+%macro HEVC_SAO_BAND_FILTER_16 3
+cglobal hevc_sao_band_filter_%2_%1, 6, 6, 15, dst, src, dststride, srcstride, offset, left
+ HEVC_SAO_BAND_FILTER_INIT %1
+
+align 16
+.loop
+%if %2 == 8
+ movu m8, [srcq]
+ HEVC_SAO_BAND_FILTER_COMPUTE %1, m9, m8
+ CLIPW m8, m14, m13
+ movu [dstq], m8
+%endif
+
+%assign i 0
+%rep %3
+ movu m8, [srcq + i]
+ HEVC_SAO_BAND_FILTER_COMPUTE %1, m9, m8
+ CLIPW m8, m14, m13
+ movu [dstq + i], m8
+
+ movu m9, [srcq + i + mmsize]
+ HEVC_SAO_BAND_FILTER_COMPUTE %1, m8, m9
+ CLIPW m9, m14, m13
+ movu [dstq + i + mmsize], m9
+%assign i i+mmsize*2
+%endrep
+
+%if %2 == 48
+INIT_XMM cpuname
+ movu m8, [srcq + i]
+ HEVC_SAO_BAND_FILTER_COMPUTE %1, m9, m8
+ CLIPW m8, m14, m13
+ movu [dstq + i], m8
+
+ movu m9, [srcq + i + mmsize]
+ HEVC_SAO_BAND_FILTER_COMPUTE %1, m8, m9
+ CLIPW m9, m14, m13
+ movu [dstq + i + mmsize], m9
+%assign i i+32
+%endif ; %1 == 48
+
+ add dstq, dststrideq
+ add srcq, srcstrideq
+ dec heightd
+ jg .loop
+ REP_RET
+%endmacro
+
+%macro HEVC_SAO_BAND_FILTER_FUNCS 0
+HEVC_SAO_BAND_FILTER_8 8, 0
+HEVC_SAO_BAND_FILTER_8 16, 1
+HEVC_SAO_BAND_FILTER_8 32, 2
+HEVC_SAO_BAND_FILTER_8 48, 2
+HEVC_SAO_BAND_FILTER_8 64, 4
+
+HEVC_SAO_BAND_FILTER_16 10, 8, 0
+HEVC_SAO_BAND_FILTER_16 10, 16, 1
+HEVC_SAO_BAND_FILTER_16 10, 32, 2
+HEVC_SAO_BAND_FILTER_16 10, 48, 2
+HEVC_SAO_BAND_FILTER_16 10, 64, 4
+
+HEVC_SAO_BAND_FILTER_16 12, 8, 0
+HEVC_SAO_BAND_FILTER_16 12, 16, 1
+HEVC_SAO_BAND_FILTER_16 12, 32, 2
+HEVC_SAO_BAND_FILTER_16 12, 48, 2
+HEVC_SAO_BAND_FILTER_16 12, 64, 4
+%endmacro
+
+INIT_XMM sse2
+HEVC_SAO_BAND_FILTER_FUNCS
+INIT_XMM avx
+HEVC_SAO_BAND_FILTER_FUNCS
+
+INIT_XMM avx2
+HEVC_SAO_BAND_FILTER_8 8, 0
+HEVC_SAO_BAND_FILTER_8 16, 1
+INIT_YMM avx2
+HEVC_SAO_BAND_FILTER_8 32, 1
+HEVC_SAO_BAND_FILTER_8 48, 1
+INIT_YMM avx2
+HEVC_SAO_BAND_FILTER_8 64, 2
+
+INIT_XMM avx2
+HEVC_SAO_BAND_FILTER_16 10, 8, 0
+HEVC_SAO_BAND_FILTER_16 10, 16, 1
+INIT_YMM avx2
+HEVC_SAO_BAND_FILTER_16 10, 32, 1
+HEVC_SAO_BAND_FILTER_16 10, 48, 1
+INIT_YMM avx2
+HEVC_SAO_BAND_FILTER_16 10, 64, 2
+
+INIT_XMM avx2
+HEVC_SAO_BAND_FILTER_16 12, 8, 0
+HEVC_SAO_BAND_FILTER_16 12, 16, 1
+INIT_YMM avx2
+HEVC_SAO_BAND_FILTER_16 12, 32, 1
+HEVC_SAO_BAND_FILTER_16 12, 48, 1
+INIT_YMM avx2
+HEVC_SAO_BAND_FILTER_16 12, 64, 2
+%endif
diff --git a/libavcodec/x86/hevcdsp_init.c b/libavcodec/x86/hevcdsp_init.c
index eaa97e1..5a01ff6 100644
--- a/libavcodec/x86/hevcdsp_init.c
+++ b/libavcodec/x86/hevcdsp_init.c
@@ -478,6 +478,27 @@ mc_bi_w_funcs(qpel_v, 12, sse4);
mc_bi_w_funcs(qpel_hv, 12, sse4);
#endif //ARCH_X86_64 && HAVE_SSE4_EXTERNAL
+#define SAO_BAND_FILTER_FUNCS(bitd, opt) \
+void ff_hevc_sao_band_filter_8_##bitd##_##opt(uint8_t *_dst, uint8_t *_src, ptrdiff_t _stride_dst, ptrdiff_t _stride_src, \
+ int16_t *sao_offset_val, int sao_left_class, int width, int height); \
+void ff_hevc_sao_band_filter_16_##bitd##_##opt(uint8_t *_dst, uint8_t *_src, ptrdiff_t _stride_dst, ptrdiff_t _stride_src, \
+ int16_t *sao_offset_val, int sao_left_class, int width, int height); \
+void ff_hevc_sao_band_filter_32_##bitd##_##opt(uint8_t *_dst, uint8_t *_src, ptrdiff_t _stride_dst, ptrdiff_t _stride_src, \
+ int16_t *sao_offset_val, int sao_left_class, int width, int height); \
+void ff_hevc_sao_band_filter_48_##bitd##_##opt(uint8_t *_dst, uint8_t *_src, ptrdiff_t _stride_dst, ptrdiff_t _stride_src, \
+ int16_t *sao_offset_val, int sao_left_class, int width, int height); \
+void ff_hevc_sao_band_filter_64_##bitd##_##opt(uint8_t *_dst, uint8_t *_src, ptrdiff_t _stride_dst, ptrdiff_t _stride_src, \
+ int16_t *sao_offset_val, int sao_left_class, int width, int height)
+
+SAO_BAND_FILTER_FUNCS(8, sse2);
+SAO_BAND_FILTER_FUNCS(10, sse2);
+SAO_BAND_FILTER_FUNCS(12, sse2);
+SAO_BAND_FILTER_FUNCS(8, avx);
+SAO_BAND_FILTER_FUNCS(10, avx);
+SAO_BAND_FILTER_FUNCS(12, avx);
+SAO_BAND_FILTER_FUNCS(8, avx2);
+SAO_BAND_FILTER_FUNCS(10, avx2);
+SAO_BAND_FILTER_FUNCS(12, avx2);
#define EPEL_LINKS(pointer, my, mx, fname, bitd, opt ) \
PEL_LINK(pointer, 1, my , mx , fname##4 , bitd, opt ); \
@@ -499,6 +520,13 @@ mc_bi_w_funcs(qpel_hv, 12, sse4);
PEL_LINK(pointer, 8, my , mx , fname##48, bitd, opt ); \
PEL_LINK(pointer, 9, my , mx , fname##64, bitd, opt )
+#define SAO_BAND_INIT(bitd, opt) do { \
+ c->sao_band_filter[0] = ff_hevc_sao_band_filter_8_##bitd##_##opt; \
+ c->sao_band_filter[1] = ff_hevc_sao_band_filter_16_##bitd##_##opt; \
+ c->sao_band_filter[2] = ff_hevc_sao_band_filter_32_##bitd##_##opt; \
+ c->sao_band_filter[3] = ff_hevc_sao_band_filter_48_##bitd##_##opt; \
+ c->sao_band_filter[4] = ff_hevc_sao_band_filter_64_##bitd##_##opt; \
+} while (0)
void ff_hevc_dsp_init_x86(HEVCDSPContext *c, const int bit_depth)
{
@@ -516,6 +544,8 @@ void ff_hevc_dsp_init_x86(HEVCDSPContext *c, const int bit_depth)
if (ARCH_X86_64) {
c->hevc_v_loop_filter_luma = ff_hevc_v_loop_filter_luma_8_sse2;
c->hevc_h_loop_filter_luma = ff_hevc_h_loop_filter_luma_8_sse2;
+
+ SAO_BAND_INIT(8, sse2);
}
c->idct_dc[1] = ff_hevc_idct8x8_dc_8_sse2;
c->idct_dc[2] = ff_hevc_idct16x16_dc_8_sse2;
@@ -547,6 +577,8 @@ void ff_hevc_dsp_init_x86(HEVCDSPContext *c, const int bit_depth)
if (ARCH_X86_64) {
c->hevc_v_loop_filter_luma = ff_hevc_v_loop_filter_luma_8_avx;
c->hevc_h_loop_filter_luma = ff_hevc_h_loop_filter_luma_8_avx;
+
+ SAO_BAND_INIT(8, avx);
}
c->transform_add[1] = ff_hevc_transform_add8_8_avx;
c->transform_add[2] = ff_hevc_transform_add16_8_avx;
@@ -555,6 +587,9 @@ void ff_hevc_dsp_init_x86(HEVCDSPContext *c, const int bit_depth)
if (EXTERNAL_AVX2(cpu_flags)) {
c->idct_dc[2] = ff_hevc_idct16x16_dc_8_avx2;
c->idct_dc[3] = ff_hevc_idct32x32_dc_8_avx2;
+ if (ARCH_X86_64) {
+ SAO_BAND_INIT(8, avx2);
+ }
c->transform_add[3] = ff_hevc_transform_add32_8_avx2;
}
@@ -570,6 +605,8 @@ void ff_hevc_dsp_init_x86(HEVCDSPContext *c, const int bit_depth)
if (ARCH_X86_64) {
c->hevc_v_loop_filter_luma = ff_hevc_v_loop_filter_luma_10_sse2;
c->hevc_h_loop_filter_luma = ff_hevc_h_loop_filter_luma_10_sse2;
+
+ SAO_BAND_INIT(10, sse2);
}
c->idct_dc[1] = ff_hevc_idct8x8_dc_10_sse2;
@@ -601,12 +638,17 @@ void ff_hevc_dsp_init_x86(HEVCDSPContext *c, const int bit_depth)
if (ARCH_X86_64) {
c->hevc_v_loop_filter_luma = ff_hevc_v_loop_filter_luma_10_avx;
c->hevc_h_loop_filter_luma = ff_hevc_h_loop_filter_luma_10_avx;
+
+ SAO_BAND_INIT(10, avx);
}
}
if (EXTERNAL_AVX2(cpu_flags)) {
c->idct_dc[2] = ff_hevc_idct16x16_dc_10_avx2;
c->idct_dc[3] = ff_hevc_idct32x32_dc_10_avx2;
+ if (ARCH_X86_64) {
+ SAO_BAND_INIT(10, avx2);
+ }
c->transform_add[2] = ff_hevc_transform_add16_10_avx2;
c->transform_add[3] = ff_hevc_transform_add32_10_avx2;
@@ -623,6 +665,8 @@ void ff_hevc_dsp_init_x86(HEVCDSPContext *c, const int bit_depth)
if (ARCH_X86_64) {
c->hevc_v_loop_filter_luma = ff_hevc_v_loop_filter_luma_12_sse2;
c->hevc_h_loop_filter_luma = ff_hevc_h_loop_filter_luma_12_sse2;
+
+ SAO_BAND_INIT(12, sse2);
}
c->idct_dc[1] = ff_hevc_idct8x8_dc_12_sse2;
@@ -650,11 +694,16 @@ void ff_hevc_dsp_init_x86(HEVCDSPContext *c, const int bit_depth)
if (ARCH_X86_64) {
c->hevc_v_loop_filter_luma = ff_hevc_v_loop_filter_luma_12_avx;
c->hevc_h_loop_filter_luma = ff_hevc_h_loop_filter_luma_12_avx;
+
+ SAO_BAND_INIT(12, avx);
}
}
if (EXTERNAL_AVX2(cpu_flags)) {
c->idct_dc[2] = ff_hevc_idct16x16_dc_12_avx2;
c->idct_dc[3] = ff_hevc_idct32x32_dc_12_avx2;
+ if (ARCH_X86_64) {
+ SAO_BAND_INIT(12, avx2);
+ }
}
}
}
--
2.2.2
More information about the ffmpeg-devel
mailing list