From: Mark Thompson <sw@jkqxz.net> To: ffmpeg-devel@ffmpeg.org Subject: Re: [FFmpeg-devel] [PATCH 5/6] lavc/apv: AVX2 transquant for x86-64 Date: Sat, 19 Apr 2025 21:34:59 +0100 Message-ID: <2b794e37-9a5f-4904-bbaa-e4e999f69a48@jkqxz.net> (raw) In-Reply-To: <20250419190712.1265201-6-sw@jkqxz.net> On 19/04/2025 20:07, Mark Thompson wrote: > Typical checkasm result on Alder Lake: > > decode_transquant_8_c: 408.7 ( 1.00x) > decode_transquant_8_avx2: 94.2 ( 4.34x) > decode_transquant_10_c: 413.1 ( 1.00x) > decode_transquant_10_avx2: 87.5 ( 4.72x) Interestingly this is signficantly better on an older Skylake. Typical result there: decode_transquant_8_c: 892.5 ( 1.00x) decode_transquant_8_avx2: 143.4 ( 6.22x) decode_transquant_10_c: 938.7 ( 1.00x) decode_transquant_10_avx2: 137.9 ( 6.81x) > --- > libavcodec/apv_dsp.c | 4 + > libavcodec/apv_dsp.h | 2 + > libavcodec/x86/Makefile | 2 + > libavcodec/x86/apv_dsp.asm | 243 ++++++++++++++++++++++++++++++++++ > libavcodec/x86/apv_dsp_init.c | 41 ++++++ > tests/checkasm/Makefile | 1 + > tests/checkasm/apv_dsp.c | 113 ++++++++++++++++ > tests/checkasm/checkasm.c | 3 + > tests/checkasm/checkasm.h | 1 + > 9 files changed, 410 insertions(+) > create mode 100644 libavcodec/x86/apv_dsp.asm > create mode 100644 libavcodec/x86/apv_dsp_init.c > create mode 100644 tests/checkasm/apv_dsp.c > > diff --git a/libavcodec/apv_dsp.c b/libavcodec/apv_dsp.c > index 6b8b6ce590..108a3cd370 100644 > --- a/libavcodec/apv_dsp.c > +++ b/libavcodec/apv_dsp.c > @@ -133,4 +133,8 @@ static void apv_decode_transquant_c(void *output, > av_cold void ff_apv_dsp_init(APVDSPContext *dsp) > { > dsp->decode_transquant = apv_decode_transquant_c; > + > +#if ARCH_X86_64 > + ff_apv_dsp_init_x86_64(dsp); > +#endif > } > diff --git a/libavcodec/apv_dsp.h b/libavcodec/apv_dsp.h > index 13f401c4cc..a7b5377357 100644 > --- a/libavcodec/apv_dsp.h > +++ b/libavcodec/apv_dsp.h > @@ -34,4 +34,6 @@ typedef struct APVDSPContext { > > void ff_apv_dsp_init(APVDSPContext *dsp); > > +void ff_apv_dsp_init_x86_64(APVDSPContext *dsp); > + > #endif /* AVCODEC_APV_DSP_H */ > diff --git a/libavcodec/x86/Makefile b/libavcodec/x86/Makefile > index 5d53515381..821c410a0f 100644 > --- a/libavcodec/x86/Makefile > +++ b/libavcodec/x86/Makefile > @@ -44,6 +44,7 @@ OBJS-$(CONFIG_ADPCM_G722_DECODER) += x86/g722dsp_init.o > OBJS-$(CONFIG_ADPCM_G722_ENCODER) += x86/g722dsp_init.o > OBJS-$(CONFIG_ALAC_DECODER) += x86/alacdsp_init.o > OBJS-$(CONFIG_APNG_DECODER) += x86/pngdsp_init.o > +OBJS-$(CONFIG_APV_DECODER) += x86/apv_dsp_init.o > OBJS-$(CONFIG_CAVS_DECODER) += x86/cavsdsp.o > OBJS-$(CONFIG_CFHD_DECODER) += x86/cfhddsp_init.o > OBJS-$(CONFIG_CFHD_ENCODER) += x86/cfhdencdsp_init.o > @@ -149,6 +150,7 @@ X86ASM-OBJS-$(CONFIG_ADPCM_G722_DECODER) += x86/g722dsp.o > X86ASM-OBJS-$(CONFIG_ADPCM_G722_ENCODER) += x86/g722dsp.o > X86ASM-OBJS-$(CONFIG_ALAC_DECODER) += x86/alacdsp.o > X86ASM-OBJS-$(CONFIG_APNG_DECODER) += x86/pngdsp.o > +X86ASM-OBJS-$(CONFIG_APV_DECODER) += x86/apv_dsp.o > X86ASM-OBJS-$(CONFIG_CAVS_DECODER) += x86/cavsidct.o > X86ASM-OBJS-$(CONFIG_CFHD_ENCODER) += x86/cfhdencdsp.o > X86ASM-OBJS-$(CONFIG_CFHD_DECODER) += x86/cfhddsp.o > diff --git a/libavcodec/x86/apv_dsp.asm b/libavcodec/x86/apv_dsp.asm > new file mode 100644 > index 0000000000..0329089f45 > --- /dev/null > +++ b/libavcodec/x86/apv_dsp.asm > @@ -0,0 +1,243 @@ > +;************************************************************************ > +;* 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 > +;* 51, Inc., Foundation Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA > +;****************************************************************************** > + > +%include "libavutil/x86/x86util.asm" > + > +SECTION .text > + > +align 32 > +const tmatrixh > + dw 64, 89, 84, 75, 64, 50, 35, 18 > + dw 64, 75, 35, -18, -64, -89, -84, -50 > + dw 64, 50, -35, -89, -64, 18, 84, 75 > + dw 64, 18, -84, -50, 64, 75, -35, -89 > + dw 64, -18, -84, 50, 64, -75, -35, 89 > + dw 64, -50, -35, 89, -64, -18, 84, -75 > + dw 64, -75, 35, 18, -64, 89, -84, 50 > + dw 64, -89, 84, -75, 64, -50, 35, -18 > +const tmatrixv > + dw 64, 89, 84, 75, 64, 50, 35, 18 > + dw 64, -18, -84, 50, 64, -75, -35, 89 > + dw 64, 75, 35, -18, -64, -89, -84, -50 > + dw 64, -50, -35, 89, -64, -18, 84, -75 > + dw 64, 50, -35, -89, -64, 18, 84, 75 > + dw 64, -75, 35, 18, -64, 89, -84, 50 > + dw 64, 18, -84, -50, 64, 75, -35, -89 > + dw 64, -89, 84, -75, 64, -50, 35, -18 > + > +; Memory targets for vpbroadcastd (register version requires AVX512). > +const one > + dd 1 > +const sixtyfour > + dd 64 > + > +; void ff_apv_decode_transquant_avx2(void *output, > +; ptrdiff_t pitch, > +; const int16_t *input, > +; const int16_t *qmatrix, > +; int64_t bit_depth, > +; int64_t qp_shift); > + > +INIT_YMM avx2 > + > +cglobal apv_decode_transquant, 6, 6, 16, output, pitch, input, qmatrix, bit_depth, qp_shift > + > + ; Load input and dequantise > + > + lea rax, [bit_depthq - 2] > + movq xm8, qp_shiftq > + movq xm9, rax > + vpbroadcastd m10, [one] > + vpslld m10, m10, xm9 > + vpsrld m10, m10, 1 > + > + ; m8 = scalar qp_shift > + ; m9 = scalar bd_shift > + ; m10 = vector 1 << (bd_shift - 1) > + ; m11 = qmatrix load > +%macro LOAD_AND_DEQUANT 2 ; (xmm input, constant offset) > + vpmovsxwd m%1, [inputq + %2] > + vpmovsxwd m11, [qmatrixq + %2] > + vpmulld m%1, m%1, m11 > + vpslld m%1, m%1, xm8 > + vpaddd m%1, m%1, m10 > + vpsrad m%1, m%1, xm9 > + vpackssdw m%1, m%1, m%1 > +%endmacro > + > + LOAD_AND_DEQUANT 0, 0x00 > + LOAD_AND_DEQUANT 1, 0x10 > + LOAD_AND_DEQUANT 2, 0x20 > + LOAD_AND_DEQUANT 3, 0x30 > + LOAD_AND_DEQUANT 4, 0x40 > + LOAD_AND_DEQUANT 5, 0x50 > + LOAD_AND_DEQUANT 6, 0x60 > + LOAD_AND_DEQUANT 7, 0x70 > + > + ; Transform columns > + ; This produces the output in rows to avoid needing any > + ; transpose operations > + > + vpunpcklwd m12, m0, m1 > + vpunpcklwd m13, m2, m3 > + vpunpcklwd m14, m4, m5 > + vpunpcklwd m15, m6, m7 > + > +%macro TRANS_COL_STEP 2 > + vpbroadcastd m8, [tmatrixh + %2 + 0x00] > + vpbroadcastd m9, [tmatrixh + %2 + 0x04] > + vpbroadcastd m10, [tmatrixh + %2 + 0x08] > + vpbroadcastd m11, [tmatrixh + %2 + 0x0c] > + vpmaddwd m8, m8, m12 > + vpmaddwd m9, m9, m13 > + vpmaddwd m10, m10, m14 > + vpmaddwd m11, m11, m15 > + vpaddd m8, m8, m9 > + vpaddd m10, m10, m11 > + vpaddd m%1, m8, m10 > +%endmacro > + > + TRANS_COL_STEP 0, 0x00 > + TRANS_COL_STEP 1, 0x10 > + TRANS_COL_STEP 2, 0x20 > + TRANS_COL_STEP 3, 0x30 > + TRANS_COL_STEP 4, 0x40 > + TRANS_COL_STEP 5, 0x50 > + TRANS_COL_STEP 6, 0x60 > + TRANS_COL_STEP 7, 0x70 > + > + ; Mid-transform normalisation > + ; Note that outputs here are fitted to 16 bits > + > + vpbroadcastd m8, [sixtyfour] > + > +%macro NORMALISE 1 > + vpaddd m%1, m%1, m8 > + vpsrad m%1, m%1, 7 > + vpackssdw m%1, m%1, m%1 > + vpermq m%1, m%1, q3120 > +%endmacro > + > + NORMALISE 0 > + NORMALISE 1 > + NORMALISE 2 > + NORMALISE 3 > + NORMALISE 4 > + NORMALISE 5 > + NORMALISE 6 > + NORMALISE 7 > + > + ; Vertical transform > + > + mova m12, [tmatrixv + 0x00] > + mova m13, [tmatrixv + 0x20] > + mova m14, [tmatrixv + 0x40] > + mova m15, [tmatrixv + 0x60] > + > +%macro TRANS_ROW_STEP 1 > + vperm2i128 m%1, m%1, m%1, 1 > + vpmaddwd m8, m%1, m12 > + vpmaddwd m9, m%1, m13 > + vpmaddwd m10, m%1, m14 > + vpmaddwd m11, m%1, m15 > + vphaddd m8, m8, m9 > + vphaddd m10, m10, m11 > + vphaddd m%1, m8, m10 > +%endmacro > + > + TRANS_ROW_STEP 0 > + TRANS_ROW_STEP 1 > + TRANS_ROW_STEP 2 > + TRANS_ROW_STEP 3 > + TRANS_ROW_STEP 4 > + TRANS_ROW_STEP 5 > + TRANS_ROW_STEP 6 > + TRANS_ROW_STEP 7 > + > + ; Renormalise, clip and store output > + > + vpbroadcastd m14, [one] > + mov rax, 20 > + sub rax, bit_depthq ; bd_shift > + movq xm9, rax > + dec rax > + movq xm13, rax > + vpslld m8, m14, xm13 > + movq xm13, bit_depthq > + vpslld m12, m14, xm13 > + vpsrld m10, m12, 1 > + vpsubd m12, m12, m14 > + vpxor m11, m11, m11 > + > + cmp bit_depthq, 8 > + jne store_10 > + > + ; m8 = vector 1 << (bd_shift - 1) > + ; m9 = scalar bd_shift > + ; m10 = vector 1 << (bit_depth - 1) > + ; m11 = zero > + ; m12 = vector (1 << bit_depth) - 1 > + > +%macro NORMALISE_AND_STORE_8 1 > + vpaddd m%1, m%1, m8 > + vpsrad m%1, m%1, xm9 > + vpaddd m%1, m%1, m10 > + vextracti128 xm13, m%1, 0 > + vextracti128 xm14, m%1, 1 > + vpackusdw xm%1, xm13, xm14 > + vpackuswb xm%1, xm%1, xm%1 > + movq [outputq], xm%1 > + add outputq, pitchq > +%endmacro > + > + NORMALISE_AND_STORE_8 0 > + NORMALISE_AND_STORE_8 1 > + NORMALISE_AND_STORE_8 2 > + NORMALISE_AND_STORE_8 3 > + NORMALISE_AND_STORE_8 4 > + NORMALISE_AND_STORE_8 5 > + NORMALISE_AND_STORE_8 6 > + NORMALISE_AND_STORE_8 7 > + > + RET > + > +store_10: > + > +%macro NORMALISE_AND_STORE_10 1 > + vpaddd m%1, m%1, m8 > + vpsrad m%1, m%1, xm9 > + vpaddd m%1, m%1, m10 > + vpmaxsd m%1, m%1, m11 > + vpminsd m%1, m%1, m12 > + vextracti128 xm13, m%1, 0 > + vextracti128 xm14, m%1, 1 > + vpackusdw xm%1, xm13, xm14 > + mova [outputq], xm%1 > + add outputq, pitchq > +%endmacro > + > + NORMALISE_AND_STORE_10 0 > + NORMALISE_AND_STORE_10 1 > + NORMALISE_AND_STORE_10 2 > + NORMALISE_AND_STORE_10 3 > + NORMALISE_AND_STORE_10 4 > + NORMALISE_AND_STORE_10 5 > + NORMALISE_AND_STORE_10 6 > + NORMALISE_AND_STORE_10 7 > + > + RET > diff --git a/libavcodec/x86/apv_dsp_init.c b/libavcodec/x86/apv_dsp_init.c > new file mode 100644 > index 0000000000..f7f209cea8 > --- /dev/null > +++ b/libavcodec/x86/apv_dsp_init.c > @@ -0,0 +1,41 @@ > +/* > + * 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 "config.h" > +#include "libavutil/attributes.h" > +#include "libavutil/cpu.h" > +#include "libavutil/x86/asm.h" > +#include "libavutil/x86/cpu.h" > +#include "libavcodec/apv_dsp.h" > + > +void ff_apv_decode_transquant_avx2(void *output, > + ptrdiff_t pitch, > + const int16_t *input, > + const int16_t *qmatrix, > + int64_t bit_depth, > + int64_t qp_shift); > + > + > +av_cold void ff_apv_dsp_init_x86_64(APVDSPContext *dsp) > +{ > + int cpu_flags = av_get_cpu_flags(); > + > + if (EXTERNAL_AVX2_FAST(cpu_flags)) { > + dsp->decode_transquant = ff_apv_decode_transquant_avx2; > + } > +} > diff --git a/tests/checkasm/Makefile b/tests/checkasm/Makefile > index d5c50e5599..193c1e4633 100644 > --- a/tests/checkasm/Makefile > +++ b/tests/checkasm/Makefile > @@ -28,6 +28,7 @@ AVCODECOBJS-$(CONFIG_AAC_DECODER) += aacpsdsp.o \ > sbrdsp.o > AVCODECOBJS-$(CONFIG_AAC_ENCODER) += aacencdsp.o > AVCODECOBJS-$(CONFIG_ALAC_DECODER) += alacdsp.o > +AVCODECOBJS-$(CONFIG_APV_DECODER) += apv_dsp.o > AVCODECOBJS-$(CONFIG_DCA_DECODER) += synth_filter.o > AVCODECOBJS-$(CONFIG_DIRAC_DECODER) += diracdsp.o > AVCODECOBJS-$(CONFIG_EXR_DECODER) += exrdsp.o > diff --git a/tests/checkasm/apv_dsp.c b/tests/checkasm/apv_dsp.c > new file mode 100644 > index 0000000000..a0272d8edc > --- /dev/null > +++ b/tests/checkasm/apv_dsp.c > @@ -0,0 +1,113 @@ > +/* > + * 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 <stdint.h> > + > +#include "checkasm.h" > + > +#include "libavutil/attributes.h" > +#include "libavutil/mem_internal.h" > +#include "libavcodec/apv_dsp.h" > + > + > +static void check_decode_transquant_8(void) > +{ > + LOCAL_ALIGNED_16(int16_t, input, [64]); > + LOCAL_ALIGNED_16(int16_t, qmatrix, [64]); > + LOCAL_ALIGNED_16(uint8_t, new_output, [64]); > + LOCAL_ALIGNED_16(uint8_t, ref_output, [64]); > + > + declare_func(void, > + uint8_t *output, > + ptrdiff_t pitch, > + const int16_t *input, > + const int16_t *qmatrix, > + int64_t bit_depth, > + int64_t qp_shift); > + > + for (int i = 0; i < 64; i++) { > + // Any signed 12-bit integer. > + input[i] = rnd() % 2048 - 1024; > + > + // qmatrix input is premultiplied by level_scale, so > + // range is 1 to 255 * 71. Interesting values are all > + // at the low end of that, though. > + qmatrix[i] = rnd() % 16 + 16; > + } > + > + call_ref(ref_output, 8, input, qmatrix, 8, 4); > + call_new(new_output, 8, input, qmatrix, 8, 4); > + > + for (int i = 0; i < 64; i++) { > + if (ref_output[i] != new_output[i]) > + fail(); > + } > + > + bench_new(new_output, 8, input, qmatrix, 8, 4); > +} > + > +static void check_decode_transquant_10(void) > +{ > + LOCAL_ALIGNED_16( int16_t, input, [64]); > + LOCAL_ALIGNED_16( int16_t, qmatrix, [64]); > + LOCAL_ALIGNED_16(uint16_t, new_output, [64]); > + LOCAL_ALIGNED_16(uint16_t, ref_output, [64]); > + > + declare_func(void, > + uint16_t *output, > + ptrdiff_t pitch, > + const int16_t *input, > + const int16_t *qmatrix, > + int64_t bit_depth, > + int64_t qp_shift); > + > + for (int i = 0; i < 64; i++) { > + // Any signed 14-bit integer. > + input[i] = rnd() % 16384 - 8192; > + > + // qmatrix input is premultiplied by level_scale, so > + // range is 1 to 255 * 71. Interesting values are all > + // at the low end of that, though. > + qmatrix[i] = 16; //rnd() % 16 + 16; > + } > + > + call_ref(ref_output, 16, input, qmatrix, 10, 4); > + call_new(new_output, 16, input, qmatrix, 10, 4); > + > + for (int i = 0; i < 64; i++) { > + if (ref_output[i] != new_output[i]) > + fail(); > + } > + > + bench_new(new_output, 16, input, qmatrix, 10, 4); > +} > + > +void checkasm_check_apv_dsp(void) > +{ > + APVDSPContext dsp; > + > + ff_apv_dsp_init(&dsp); > + > + if (check_func(dsp.decode_transquant, "decode_transquant_8")) > + check_decode_transquant_8(); > + > + if (check_func(dsp.decode_transquant, "decode_transquant_10")) > + check_decode_transquant_10(); > + > + report("apv_dsp"); > +} > diff --git a/tests/checkasm/checkasm.c b/tests/checkasm/checkasm.c > index 412b8b2cd1..3bb82ed0e5 100644 > --- a/tests/checkasm/checkasm.c > +++ b/tests/checkasm/checkasm.c > @@ -129,6 +129,9 @@ static const struct { > #if CONFIG_ALAC_DECODER > { "alacdsp", checkasm_check_alacdsp }, > #endif > + #if CONFIG_APV_DECODER > + { "apv_dsp", checkasm_check_apv_dsp }, > + #endif > #if CONFIG_AUDIODSP > { "audiodsp", checkasm_check_audiodsp }, > #endif > diff --git a/tests/checkasm/checkasm.h b/tests/checkasm/checkasm.h > index ad239fb2a4..a6b5965e02 100644 > --- a/tests/checkasm/checkasm.h > +++ b/tests/checkasm/checkasm.h > @@ -83,6 +83,7 @@ void checkasm_check_ac3dsp(void); > void checkasm_check_aes(void); > void checkasm_check_afir(void); > void checkasm_check_alacdsp(void); > +void checkasm_check_apv_dsp(void); > void checkasm_check_audiodsp(void); > void checkasm_check_av_tx(void); > void checkasm_check_blend(void); _______________________________________________ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-devel To unsubscribe, visit link above, or email ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".
next prev parent reply other threads:[~2025-04-19 20:35 UTC|newest] Thread overview: 19+ messages / expand[flat|nested] mbox.gz Atom feed top 2025-04-19 19:06 [FFmpeg-devel] [PATCH 0/6] APV support Mark Thompson 2025-04-19 19:06 ` [FFmpeg-devel] [PATCH 1/6] lavc: APV codec ID and descriptor Mark Thompson 2025-04-19 19:07 ` [FFmpeg-devel] [PATCH 2/6] lavf: APV demuxer Mark Thompson 2025-04-20 16:07 ` Derek Buitenhuis 2025-04-20 16:20 ` James Almer 2025-04-20 16:57 ` Mark Thompson 2025-04-20 18:59 ` James Almer 2025-04-21 0:54 ` Michael Niedermayer 2025-04-21 14:59 ` Mark Thompson 2025-04-21 15:22 ` Andreas Rheinhardt 2025-04-21 21:30 ` Michael Niedermayer 2025-04-19 19:07 ` [FFmpeg-devel] [PATCH 3/6] lavc/cbs: APV support Mark Thompson 2025-04-19 19:07 ` [FFmpeg-devel] [PATCH 4/6] lavc: APV decoder Mark Thompson 2025-04-21 14:09 ` James Almer 2025-04-19 19:07 ` [FFmpeg-devel] [PATCH 5/6] lavc/apv: AVX2 transquant for x86-64 Mark Thompson 2025-04-19 20:34 ` Mark Thompson [this message] 2025-04-19 21:16 ` James Almer 2025-04-20 1:48 ` James Almer 2025-04-19 19:07 ` [FFmpeg-devel] [PATCH 6/6] lavc: APV metadata bitstream filter Mark Thompson
Reply instructions: You may reply publicly to this message via plain-text email using any one of the following methods: * Save the following mbox file, import it into your mail client, and reply-to-all from there: mbox Avoid top-posting and favor interleaved quoting: https://en.wikipedia.org/wiki/Posting_style#Interleaved_style * Reply using the --to, --cc, and --in-reply-to switches of git-send-email(1): git send-email \ --in-reply-to=2b794e37-9a5f-4904-bbaa-e4e999f69a48@jkqxz.net \ --to=sw@jkqxz.net \ --cc=ffmpeg-devel@ffmpeg.org \ /path/to/YOUR_REPLY https://kernel.org/pub/software/scm/git/docs/git-send-email.html * If your mail client supports setting the In-Reply-To header via mailto: links, try the mailto: link
Git Inbox Mirror of the ffmpeg-devel mailing list - see https://ffmpeg.org/mailman/listinfo/ffmpeg-devel This inbox may be cloned and mirrored by anyone: git clone --mirror https://master.gitmailbox.com/ffmpegdev/0 ffmpegdev/git/0.git # If you have public-inbox 1.1+ installed, you may # initialize and index your mirror using the following commands: public-inbox-init -V2 ffmpegdev ffmpegdev/ https://master.gitmailbox.com/ffmpegdev \ ffmpegdev@gitmailbox.com public-inbox-index ffmpegdev Example config snippet for mirrors. AGPL code for this site: git clone https://public-inbox.org/public-inbox.git