January 2017 - ffmpeg-cvslog - lists.ffmpeg.org

tools/target_dec_fuzzer: Only audio uses the return value to decode packets in pieces, correct the code to match that
by git＠videolan.org 31 Jan '17

31 Jan '17

ffmpeg | branch: master | Michael Niedermayer <michael(a)niedermayer.cc> | Tue Jan 31 22:26:01 2017 +0100| [8bdba1092f50d52e0af90f425811302ec91152f8] | committer: Michael Niedermayer tools/target_dec_fuzzer: Only audio uses the return value to decode packets in pieces, correct the code to match that Fixes: Timeout Fixes: 483/fuzz-0-ffmpeg_VIDEO_AV_CODEC_ID_H263I_fuzzer Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/targets/ffmpeg Signed-off-by: Michael Niedermayer <michael(a)niedermayer.cc> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=8bdba1092f50d52e0… --- tools/target_dec_fuzzer.c | 2 ++ 1 file changed, 2 insertions(+) diff --git a/tools/target_dec_fuzzer.c b/tools/target_dec_fuzzer.c index 533aa37..8761d96 100644 --- a/tools/target_dec_fuzzer.c +++ b/tools/target_dec_fuzzer.c @@ -186,6 +186,8 @@ int LLVMFuzzerTestOneInput(const uint8_t *data, size_t size) { if (ret <= 0 || ret > avpkt.size) break; + if (ctx->codec_type != AVMEDIA_TYPE_AUDIO) + ret = avpkt.size; avpkt.data += ret; avpkt.size -= ret; }

1 0

Merge commit 'b4bb9593834460bbbe0e70823f2c503cb01ad052'
by git＠videolan.org 31 Jan '17

31 Jan '17

ffmpeg | branch: master | James Almer <jamrial(a)gmail.com> | Tue Jan 31 15:50:56 2017 -0300| [1df08cae82d75e01b1250bb9416d6cfbab430e82] | committer: James Almer Merge commit 'b4bb9593834460bbbe0e70823f2c503cb01ad052' * commit 'b4bb9593834460bbbe0e70823f2c503cb01ad052': ratecontrol: Drop commented out cruft Merged-by: James Almer <jamrial(a)gmail.com> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=1df08cae82d75e01b… --- libavcodec/ratecontrol.c | 12 ------------ 1 file changed, 12 deletions(-) diff --git a/libavcodec/ratecontrol.c b/libavcodec/ratecontrol.c index 3a8ac22..507dcbb 100644 --- a/libavcodec/ratecontrol.c +++ b/libavcodec/ratecontrol.c @@ -101,12 +101,6 @@ av_cold int ff_rate_control_init(MpegEncContext *s) "isB", "avgQP", "qComp", -#if 0 - "lastIQP", - "lastPQP", - "lastBQP", - "nextNonBQP", -#endif "avgIITex", "avgPITex", "avgPPTex", @@ -394,12 +388,6 @@ static double get_qscale(MpegEncContext *s, RateControlEntry *rce, rce->pict_type == AV_PICTURE_TYPE_B, rcc->qscale_sum[pict_type] / (double)rcc->frame_count[pict_type], a->qcompress, -#if 0 - rcc->last_qscale_for[AV_PICTURE_TYPE_I], - rcc->last_qscale_for[AV_PICTURE_TYPE_P], - rcc->last_qscale_for[AV_PICTURE_TYPE_B], - rcc->next_non_b_qscale, -#endif rcc->i_cplx_sum[AV_PICTURE_TYPE_I] / (double)rcc->frame_count[AV_PICTURE_TYPE_I], rcc->i_cplx_sum[AV_PICTURE_TYPE_P] / (double)rcc->frame_count[AV_PICTURE_TYPE_P], rcc->p_cplx_sum[AV_PICTURE_TYPE_P] / (double)rcc->frame_count[AV_PICTURE_TYPE_P], ======================================================================

1 0

ratecontrol: Drop commented out cruft
by git＠videolan.org 31 Jan '17

31 Jan '17

ffmpeg | branch: master | Vittorio Giovara <vittorio.giovara(a)gmail.com> | Wed Jul 20 16:12:31 2016 +0200| [b4bb9593834460bbbe0e70823f2c503cb01ad052] | committer: Vittorio Giovara ratecontrol: Drop commented out cruft > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=b4bb9593834460bbb… --- libavcodec/ratecontrol.c | 12 ------------ 1 file changed, 12 deletions(-) diff --git a/libavcodec/ratecontrol.c b/libavcodec/ratecontrol.c index 47a1490..c483628 100644 --- a/libavcodec/ratecontrol.c +++ b/libavcodec/ratecontrol.c @@ -103,12 +103,6 @@ av_cold int ff_rate_control_init(MpegEncContext *s) "isB", "avgQP", "qComp", -#if 0 - "lastIQP", - "lastPQP", - "lastBQP", - "nextNonBQP", -#endif "avgIITex", "avgPITex", "avgPPTex", @@ -380,12 +374,6 @@ static double get_qscale(MpegEncContext *s, RateControlEntry *rce, rce->pict_type == AV_PICTURE_TYPE_B, rcc->qscale_sum[pict_type] / (double)rcc->frame_count[pict_type], a->qcompress, -#if 0 - rcc->last_qscale_for[AV_PICTURE_TYPE_I], - rcc->last_qscale_for[AV_PICTURE_TYPE_P], - rcc->last_qscale_for[AV_PICTURE_TYPE_B], - rcc->next_non_b_qscale, -#endif rcc->i_cplx_sum[AV_PICTURE_TYPE_I] / (double)rcc->frame_count[AV_PICTURE_TYPE_I], rcc->i_cplx_sum[AV_PICTURE_TYPE_P] / (double)rcc->frame_count[AV_PICTURE_TYPE_P], rcc->p_cplx_sum[AV_PICTURE_TYPE_P] / (double)rcc->frame_count[AV_PICTURE_TYPE_P],

1 0

Merge commit 'd06dfaa5cbdd20acfd2364b16c0f4ae4ddb30a65'
by git＠videolan.org 31 Jan '17

31 Jan '17

ffmpeg | branch: master | James Almer <jamrial(a)gmail.com> | Tue Jan 31 15:36:49 2017 -0300| [ba5d08938130b2545f8b3e73a701989563316e49] | committer: James Almer Merge commit 'd06dfaa5cbdd20acfd2364b16c0f4ae4ddb30a65' * commit 'd06dfaa5cbdd20acfd2364b16c0f4ae4ddb30a65': x86: huffyuv: Use EXTERNAL_SSSE3_FAST convenience macro where appropriate Merged-by: James Almer <jamrial(a)gmail.com> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=ba5d08938130b2545… --- libavcodec/x86/lossless_videodsp.asm | 4 ++-- libavcodec/x86/lossless_videodsp_init.c | 11 ++++++----- 2 files changed, 8 insertions(+), 7 deletions(-) diff --git a/libavcodec/x86/lossless_videodsp.asm b/libavcodec/x86/lossless_videodsp.asm index 7fcae89..443fe02 100644 --- a/libavcodec/x86/lossless_videodsp.asm +++ b/libavcodec/x86/lossless_videodsp.asm @@ -160,8 +160,8 @@ cglobal add_left_pred, 3,3,7, dst, src, w, left psllq m0, 56 ADD_LEFT_LOOP 1, 1 -INIT_XMM sse4 -cglobal add_left_pred, 3,3,7, dst, src, w, left +INIT_XMM ssse3 +cglobal add_left_pred_unaligned, 3,3,7, dst, src, w, left mova m5, [pb_15] mova m6, [pb_zzzzzzzz77777777] mova m4, [pb_zzzz3333zzzzbbbb] diff --git a/libavcodec/x86/lossless_videodsp_init.c b/libavcodec/x86/lossless_videodsp_init.c index 58715e2..8d58344 100644 --- a/libavcodec/x86/lossless_videodsp_init.c +++ b/libavcodec/x86/lossless_videodsp_init.c @@ -35,8 +35,8 @@ void ff_add_median_pred_sse2(uint8_t *dst, const uint8_t *top, int ff_add_left_pred_ssse3(uint8_t *dst, const uint8_t *src, intptr_t w, int left); -int ff_add_left_pred_sse4(uint8_t *dst, const uint8_t *src, - intptr_t w, int left); +int ff_add_left_pred_unaligned_ssse3(uint8_t *dst, const uint8_t *src, + intptr_t w, int left); int ff_add_left_pred_int16_ssse3(uint16_t *dst, const uint16_t *src, unsigned mask, int w, unsigned acc); int ff_add_left_pred_int16_sse4(uint16_t *dst, const uint16_t *src, unsigned mask, int w, unsigned acc); @@ -105,12 +105,13 @@ void ff_llviddsp_init_x86(LLVidDSPContext *c) if (EXTERNAL_SSSE3(cpu_flags)) { c->add_left_pred = ff_add_left_pred_ssse3; - if (cpu_flags & AV_CPU_FLAG_SSE4) // not really SSE4, just slow on Conroe - c->add_left_pred = ff_add_left_pred_sse4; - c->add_left_pred_int16 = ff_add_left_pred_int16_ssse3; } + if (EXTERNAL_SSSE3_FAST(cpu_flags)) { + c->add_left_pred = ff_add_left_pred_unaligned_ssse3; + } + if (EXTERNAL_SSE4(cpu_flags)) { c->add_left_pred_int16 = ff_add_left_pred_int16_sse4; } ====================================================================== diff --cc libavcodec/x86/lossless_videodsp.asm index 7fcae89,0000000..443fe02 mode 100644,000000..100644 --- a/libavcodec/x86/lossless_videodsp.asm +++ b/libavcodec/x86/lossless_videodsp.asm @@@ -1,290 -1,0 +1,290 @@@ +;****************************************************************************** +;* SIMD lossless video DSP utils +;* Copyright (c) 2008 Loren Merritt +;* Copyright (c) 2014 Michael Niedermayer +;* +;* 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" + +SECTION_RODATA + +cextern pb_15 +pb_zzzzzzzz77777777: times 8 db -1 +pb_7: times 8 db 7 +pb_ef: times 8 db 14,15 +pb_67: times 8 db 6, 7 +pb_zzzz3333zzzzbbbb: db -1,-1,-1,-1,3,3,3,3,-1,-1,-1,-1,11,11,11,11 +pb_zz11zz55zz99zzdd: db -1,-1,1,1,-1,-1,5,5,-1,-1,9,9,-1,-1,13,13 +pb_zzzz2323zzzzabab: db -1,-1,-1,-1, 2, 3, 2, 3,-1,-1,-1,-1,10,11,10,11 +pb_zzzzzzzz67676767: db -1,-1,-1,-1,-1,-1,-1,-1, 6, 7, 6, 7, 6, 7, 6, 7 + +SECTION .text + +; void ff_add_median_pred_mmxext(uint8_t *dst, const uint8_t *top, +; const uint8_t *diff, int w, +; int *left, int *left_top) +%macro MEDIAN_PRED 0 +cglobal add_median_pred, 6,6,8, dst, top, diff, w, left, left_top + movu m0, [topq] + mova m2, m0 + movd m4, [left_topq] + LSHIFT m2, 1 + mova m1, m0 + por m4, m2 + movd m3, [leftq] + psubb m0, m4 ; t-tl + add dstq, wq + add topq, wq + add diffq, wq + neg wq + jmp .skip +.loop: + movu m4, [topq+wq] + mova m0, m4 + LSHIFT m4, 1 + por m4, m1 + mova m1, m0 ; t + psubb m0, m4 ; t-tl +.skip: + movu m2, [diffq+wq] +%assign i 0 +%rep mmsize + mova m4, m0 + paddb m4, m3 ; t-tl+l + mova m5, m3 + pmaxub m3, m1 + pminub m5, m1 + pminub m3, m4 + pmaxub m3, m5 ; median + paddb m3, m2 ; +residual +%if i==0 + mova m7, m3 + LSHIFT m7, mmsize-1 +%else + mova m6, m3 + RSHIFT m7, 1 + LSHIFT m6, mmsize-1 + por m7, m6 +%endif +%if i<mmsize-1 + RSHIFT m0, 1 + RSHIFT m1, 1 + RSHIFT m2, 1 +%endif +%assign i i+1 +%endrep + movu [dstq+wq], m7 + add wq, mmsize + jl .loop + movzx r2d, byte [dstq-1] + mov [leftq], r2d + movzx r2d, byte [topq-1] + mov [left_topq], r2d + RET +%endmacro + +%if ARCH_X86_32 +INIT_MMX mmxext +MEDIAN_PRED +%endif +INIT_XMM sse2 +MEDIAN_PRED + + +%macro ADD_LEFT_LOOP 2 ; %1 = dst_is_aligned, %2 = src_is_aligned + add srcq, wq + add dstq, wq + neg wq +%%.loop: +%if %2 + mova m1, [srcq+wq] +%else + movu m1, [srcq+wq] +%endif + mova m2, m1 + psllw m1, 8 + paddb m1, m2 + mova m2, m1 + pshufb m1, m3 + paddb m1, m2 + pshufb m0, m5 + mova m2, m1 + pshufb m1, m4 + paddb m1, m2 +%if mmsize == 16 + mova m2, m1 + pshufb m1, m6 + paddb m1, m2 +%endif + paddb m0, m1 +%if %1 + mova [dstq+wq], m0 +%else + movq [dstq+wq], m0 + movhps [dstq+wq+8], m0 +%endif + add wq, mmsize + jl %%.loop + mov eax, mmsize-1 + sub eax, wd + movd m1, eax + pshufb m0, m1 + movd eax, m0 + RET +%endmacro + +; int ff_add_left_pred(uint8_t *dst, const uint8_t *src, int w, int left) +INIT_MMX ssse3 +cglobal add_left_pred, 3,3,7, dst, src, w, left +.skip_prologue: + mova m5, [pb_7] + mova m4, [pb_zzzz3333zzzzbbbb] + mova m3, [pb_zz11zz55zz99zzdd] + movd m0, leftm + psllq m0, 56 + ADD_LEFT_LOOP 1, 1 + - INIT_XMM sse4 - cglobal add_left_pred, 3,3,7, dst, src, w, left ++INIT_XMM ssse3 ++cglobal add_left_pred_unaligned, 3,3,7, dst, src, w, left + mova m5, [pb_15] + mova m6, [pb_zzzzzzzz77777777] + mova m4, [pb_zzzz3333zzzzbbbb] + mova m3, [pb_zz11zz55zz99zzdd] + movd m0, leftm + pslldq m0, 15 + test srcq, 15 + jnz .src_unaligned + test dstq, 15 + jnz .dst_unaligned + ADD_LEFT_LOOP 1, 1 +.dst_unaligned: + ADD_LEFT_LOOP 0, 1 +.src_unaligned: + ADD_LEFT_LOOP 0, 0 + +%macro ADD_BYTES 0 +cglobal add_bytes, 3,4,2, dst, src, w, size + mov sizeq, wq + and sizeq, -2*mmsize + jz .2 + add dstq, sizeq + add srcq, sizeq + neg sizeq +.1: + mova m0, [srcq + sizeq] + mova m1, [srcq + sizeq + mmsize] + paddb m0, [dstq + sizeq] + paddb m1, [dstq + sizeq + mmsize] + mova [dstq + sizeq], m0 + mova [dstq + sizeq + mmsize], m1 + add sizeq, 2*mmsize + jl .1 +.2: + and wq, 2*mmsize-1 + jz .end + add dstq, wq + add srcq, wq + neg wq +.3: + mov sizeb, [srcq + wq] + add [dstq + wq], sizeb + inc wq + jl .3 +.end: + REP_RET +%endmacro + +%if ARCH_X86_32 +INIT_MMX mmx +ADD_BYTES +%endif +INIT_XMM sse2 +ADD_BYTES + +%macro ADD_HFYU_LEFT_LOOP_INT16 2 ; %1 = dst alignment (a/u), %2 = src alignment (a/u) + add wd, wd + add srcq, wq + add dstq, wq + neg wq +%%.loop: + mov%2 m1, [srcq+wq] + mova m2, m1 + pslld m1, 16 + paddw m1, m2 + mova m2, m1 + + pshufb m1, m3 + paddw m1, m2 + pshufb m0, m5 +%if mmsize == 16 + mova m2, m1 + pshufb m1, m4 + paddw m1, m2 +%endif + paddw m0, m1 + pand m0, m7 +%ifidn %1, a + mova [dstq+wq], m0 +%else + movq [dstq+wq], m0 + movhps [dstq+wq+8], m0 +%endif + add wq, mmsize + jl %%.loop + mov eax, mmsize-1 + sub eax, wd + mov wd, eax + shl wd, 8 + lea eax, [wd+eax-1] + movd m1, eax + pshufb m0, m1 + movd eax, m0 + RET +%endmacro + +; int add_left_pred_int16(uint16_t *dst, const uint16_t *src, unsigned mask, int w, int left) +INIT_MMX ssse3 +cglobal add_left_pred_int16, 4,4,8, dst, src, mask, w, left +.skip_prologue: + mova m5, [pb_67] + mova m3, [pb_zzzz2323zzzzabab] + movd m0, leftm + psllq m0, 48 + movd m7, maskm + SPLATW m7 ,m7 + ADD_HFYU_LEFT_LOOP_INT16 a, a + +INIT_XMM sse4 +cglobal add_left_pred_int16, 4,4,8, dst, src, mask, w, left + mova m5, [pb_ef] + mova m4, [pb_zzzzzzzz67676767] + mova m3, [pb_zzzz2323zzzzabab] + movd m0, leftm + pslldq m0, 14 + movd m7, maskm + SPLATW m7 ,m7 + test srcq, 15 + jnz .src_unaligned + test dstq, 15 + jnz .dst_unaligned + ADD_HFYU_LEFT_LOOP_INT16 a, a +.dst_unaligned: + ADD_HFYU_LEFT_LOOP_INT16 u, a +.src_unaligned: + ADD_HFYU_LEFT_LOOP_INT16 u, u diff --cc libavcodec/x86/lossless_videodsp_init.c index 58715e2,0000000..8d58344 mode 100644,000000..100644 --- a/libavcodec/x86/lossless_videodsp_init.c +++ b/libavcodec/x86/lossless_videodsp_init.c @@@ -1,117 -1,0 +1,118 @@@ +/* + * Lossless video DSP utils + * + * 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/x86/asm.h" +#include "../lossless_videodsp.h" +#include "libavutil/x86/cpu.h" + +void ff_add_bytes_mmx(uint8_t *dst, uint8_t *src, intptr_t w); +void ff_add_bytes_sse2(uint8_t *dst, uint8_t *src, intptr_t w); + +void ff_add_median_pred_mmxext(uint8_t *dst, const uint8_t *top, + const uint8_t *diff, intptr_t w, + int *left, int *left_top); +void ff_add_median_pred_sse2(uint8_t *dst, const uint8_t *top, + const uint8_t *diff, intptr_t w, + int *left, int *left_top); + +int ff_add_left_pred_ssse3(uint8_t *dst, const uint8_t *src, + intptr_t w, int left); - int ff_add_left_pred_sse4(uint8_t *dst, const uint8_t *src, - intptr_t w, int left); ++int ff_add_left_pred_unaligned_ssse3(uint8_t *dst, const uint8_t *src, ++ intptr_t w, int left); + +int ff_add_left_pred_int16_ssse3(uint16_t *dst, const uint16_t *src, unsigned mask, int w, unsigned acc); +int ff_add_left_pred_int16_sse4(uint16_t *dst, const uint16_t *src, unsigned mask, int w, unsigned acc); + +#if HAVE_INLINE_ASM && HAVE_7REGS && ARCH_X86_32 +static void add_median_pred_cmov(uint8_t *dst, const uint8_t *top, + const uint8_t *diff, intptr_t w, + int *left, int *left_top) +{ + x86_reg w2 = -w; + x86_reg x; + int l = *left & 0xff; + int tl = *left_top & 0xff; + int t; + __asm__ volatile ( + "mov %7, %3 \n" + "1: \n" + "movzbl (%3, %4), %2 \n" + "mov %2, %k3 \n" + "sub %b1, %b3 \n" + "add %b0, %b3 \n" + "mov %2, %1 \n" + "cmp %0, %2 \n" + "cmovg %0, %2 \n" + "cmovg %1, %0 \n" + "cmp %k3, %0 \n" + "cmovg %k3, %0 \n" + "mov %7, %3 \n" + "cmp %2, %0 \n" + "cmovl %2, %0 \n" + "add (%6, %4), %b0 \n" + "mov %b0, (%5, %4) \n" + "inc %4 \n" + "jl 1b \n" + : "+&q"(l), "+&q"(tl), "=&r"(t), "=&q"(x), "+&r"(w2) + : "r"(dst + w), "r"(diff + w), "rm"(top + w) + ); + *left = l; + *left_top = tl; +} +#endif + +void ff_llviddsp_init_x86(LLVidDSPContext *c) +{ + int cpu_flags = av_get_cpu_flags(); + +#if HAVE_INLINE_ASM && HAVE_7REGS && ARCH_X86_32 + if (cpu_flags & AV_CPU_FLAG_CMOV) + c->add_median_pred = add_median_pred_cmov; +#endif + + if (ARCH_X86_32 && EXTERNAL_MMX(cpu_flags)) { + c->add_bytes = ff_add_bytes_mmx; + } + + if (ARCH_X86_32 && EXTERNAL_MMXEXT(cpu_flags)) { + /* slower than cmov version on AMD */ + if (!(cpu_flags & AV_CPU_FLAG_3DNOW)) + c->add_median_pred = ff_add_median_pred_mmxext; + } + + if (EXTERNAL_SSE2(cpu_flags)) { + c->add_bytes = ff_add_bytes_sse2; + c->add_median_pred = ff_add_median_pred_sse2; + } + + if (EXTERNAL_SSSE3(cpu_flags)) { + c->add_left_pred = ff_add_left_pred_ssse3; - if (cpu_flags & AV_CPU_FLAG_SSE4) // not really SSE4, just slow on Conroe - c->add_left_pred = ff_add_left_pred_sse4; - + c->add_left_pred_int16 = ff_add_left_pred_int16_ssse3; + } + ++ if (EXTERNAL_SSSE3_FAST(cpu_flags)) { ++ c->add_left_pred = ff_add_left_pred_unaligned_ssse3; ++ } ++ + if (EXTERNAL_SSE4(cpu_flags)) { + c->add_left_pred_int16 = ff_add_left_pred_int16_sse4; + } +}

1 0

x86: huffyuv: Use EXTERNAL_SSSE3_FAST convenience macro where appropriate
by git＠videolan.org 31 Jan '17

31 Jan '17

ffmpeg | branch: master | Diego Biurrun <diego(a)biurrun.de> | Mon Jan 18 10:52:21 2016 +0100| [d06dfaa5cbdd20acfd2364b16c0f4ae4ddb30a65] | committer: Diego Biurrun x86: huffyuv: Use EXTERNAL_SSSE3_FAST convenience macro where appropriate > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=d06dfaa5cbdd20acf… --- libavcodec/x86/huffyuvdsp.asm | 4 ++-- libavcodec/x86/huffyuvdsp_init.c | 10 ++++++---- 2 files changed, 8 insertions(+), 6 deletions(-) diff --git a/libavcodec/x86/huffyuvdsp.asm b/libavcodec/x86/huffyuvdsp.asm index e7536da..692162b 100644 --- a/libavcodec/x86/huffyuvdsp.asm +++ b/libavcodec/x86/huffyuvdsp.asm @@ -146,8 +146,8 @@ cglobal add_hfyu_left_pred, 3,3,7, dst, src, w, left psllq m0, 56 ADD_HFYU_LEFT_LOOP 1, 1 -INIT_XMM sse4 -cglobal add_hfyu_left_pred, 3,3,7, dst, src, w, left +INIT_XMM ssse3 +cglobal add_hfyu_left_pred_unaligned, 3,3,7, dst, src, w, left mova m5, [pb_f] mova m6, [pb_zzzzzzzz77777777] mova m4, [pb_zzzz3333zzzzbbbb] diff --git a/libavcodec/x86/huffyuvdsp_init.c b/libavcodec/x86/huffyuvdsp_init.c index 75537d7..80e6cfb 100644 --- a/libavcodec/x86/huffyuvdsp_init.c +++ b/libavcodec/x86/huffyuvdsp_init.c @@ -31,8 +31,8 @@ void ff_add_hfyu_median_pred_mmxext(uint8_t *dst, const uint8_t *top, int ff_add_hfyu_left_pred_ssse3(uint8_t *dst, const uint8_t *src, int w, int left); -int ff_add_hfyu_left_pred_sse4(uint8_t *dst, const uint8_t *src, - int w, int left); +int ff_add_hfyu_left_pred_unaligned_ssse3(uint8_t *dst, const uint8_t *src, + int w, int left); #if HAVE_INLINE_ASM @@ -124,7 +124,9 @@ av_cold void ff_huffyuvdsp_init_x86(HuffYUVDSPContext *c) if (EXTERNAL_SSSE3(cpu_flags)) { c->add_hfyu_left_pred = ff_add_hfyu_left_pred_ssse3; - if (cpu_flags & AV_CPU_FLAG_SSE4) // not really SSE4, just slow on Conroe - c->add_hfyu_left_pred = ff_add_hfyu_left_pred_sse4; + } + + if (EXTERNAL_SSSE3_FAST(cpu_flags)) { + c->add_hfyu_left_pred = ff_add_hfyu_left_pred_unaligned_ssse3; } }

1 0

x86: cpu: Detect Conroe CPUs and their slow shuffle unit
by git＠videolan.org 31 Jan '17

31 Jan '17

ffmpeg | branch: master | Fiona Glaser <fiona(a)x264.com> | Thu Jan 14 10:54:27 2016 +0100| [8e9cd81d291b1010c625b2766058aadf4affb537] | committer: Diego Biurrun x86: cpu: Detect Conroe CPUs and their slow shuffle unit > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=8e9cd81d291b1010c… --- libavutil/x86/cpu.c | 6 ++++++ 1 file changed, 6 insertions(+) diff --git a/libavutil/x86/cpu.c b/libavutil/x86/cpu.c index 328ce6d..3b085e7 100644 --- a/libavutil/x86/cpu.c +++ b/libavutil/x86/cpu.c @@ -221,6 +221,12 @@ int ff_get_cpu_flags_x86(void) * functions on the Atom. */ if (family == 6 && model == 28) rval |= AV_CPU_FLAG_ATOM; + + /* Conroe has a slow shuffle unit. Check the model number to ensure not + * to include crippled low-end Penryns and Nehalems that lack SSE4. */ + if ((rval & AV_CPU_FLAG_SSSE3) && !(rval & AV_CPU_FLAG_SSE4) && + family == 6 && model < 23) + rval |= AV_CPU_FLAG_SSSE3SLOW; } #endif /* cpuid */

1 0

Merge commit '8e9cd81d291b1010c625b2766058aadf4affb537'
by git＠videolan.org 31 Jan '17

31 Jan '17

ffmpeg | branch: master | James Almer <jamrial(a)gmail.com> | Tue Jan 31 15:20:54 2017 -0300| [8d5df204d002bef25aadc4d7667e32c4fecaa2b7] | committer: James Almer Merge commit '8e9cd81d291b1010c625b2766058aadf4affb537' * commit '8e9cd81d291b1010c625b2766058aadf4affb537': x86: cpu: Detect Conroe CPUs and their slow shuffle unit Merged-by: James Almer <jamrial(a)gmail.com> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=8d5df204d002bef25… --- libavutil/x86/cpu.c | 6 ++++++ 1 file changed, 6 insertions(+) diff --git a/libavutil/x86/cpu.c b/libavutil/x86/cpu.c index f3a49c6..1b80847 100644 --- a/libavutil/x86/cpu.c +++ b/libavutil/x86/cpu.c @@ -221,6 +221,12 @@ int ff_get_cpu_flags_x86(void) * functions on the Atom. */ if (family == 6 && model == 28) rval |= AV_CPU_FLAG_ATOM; + + /* Conroe has a slow shuffle unit. Check the model number to ensure not + * to include crippled low-end Penryns and Nehalems that lack SSE4. */ + if ((rval & AV_CPU_FLAG_SSSE3) && !(rval & AV_CPU_FLAG_SSE4) && + family == 6 && model < 23) + rval |= AV_CPU_FLAG_SSSE3SLOW; } #endif /* cpuid */ ======================================================================

1 0

Merge commit '7d7355aa92bb36ca0765c49a569a999bcb96f332'
by git＠videolan.org 31 Jan '17

31 Jan '17

ffmpeg | branch: master | James Almer <jamrial(a)gmail.com> | Tue Jan 31 15:15:47 2017 -0300| [2eab48177d74782cf948e7ae13aef2ee5d7ab1c1] | committer: James Almer Merge commit '7d7355aa92bb36ca0765c49a569a999bcb96f332' * commit '7d7355aa92bb36ca0765c49a569a999bcb96f332': x86: Add SSSE3_SLOW CPU flag and related convenience macros Merged-by: James Almer <jamrial(a)gmail.com> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=2eab48177d74782cf… --- doc/APIchanges | 3 +++ libavutil/cpu.h | 1 + libavutil/version.h | 2 +- libavutil/x86/cpu.h | 6 ++++++ 4 files changed, 11 insertions(+), 1 deletion(-) diff --git a/doc/APIchanges b/doc/APIchanges index 109d550..8bca71e 100644 --- a/doc/APIchanges +++ b/doc/APIchanges @@ -15,6 +15,9 @@ libavutil: 2015-08-28 API changes, most recent first: +2017-01-31 - xxxxxxx - lavu 55.46.100 / 55.20.0 - cpu.h + Add AV_CPU_FLAG_SSSE3SLOW. + 2017-01-24 - xxxxxxx - lavu 55.45.100 - channel_layout.h Add av_get_extended_channel_layout() diff --git a/libavutil/cpu.h b/libavutil/cpu.h index 8499f0e..de05593 100644 --- a/libavutil/cpu.h +++ b/libavutil/cpu.h @@ -39,6 +39,7 @@ #define AV_CPU_FLAG_SSE3SLOW 0x20000000 ///< SSE3 supported, but usually not faster ///< than regular MMX/SSE (e.g. Core1) #define AV_CPU_FLAG_SSSE3 0x0080 ///< Conroe SSSE3 functions +#define AV_CPU_FLAG_SSSE3SLOW 0x4000000 ///< SSSE3 supported, but usually not faster #define AV_CPU_FLAG_ATOM 0x10000000 ///< Atom processor, some SSSE3 instructions are slower #define AV_CPU_FLAG_SSE4 0x0100 ///< Penryn SSE4.1 functions #define AV_CPU_FLAG_SSE42 0x0200 ///< Nehalem SSE4.2 functions diff --git a/libavutil/version.h b/libavutil/version.h index 9d95086..8866064 100644 --- a/libavutil/version.h +++ b/libavutil/version.h @@ -79,7 +79,7 @@ */ #define LIBAVUTIL_VERSION_MAJOR 55 -#define LIBAVUTIL_VERSION_MINOR 45 +#define LIBAVUTIL_VERSION_MINOR 46 #define LIBAVUTIL_VERSION_MICRO 100 #define LIBAVUTIL_VERSION_INT AV_VERSION_INT(LIBAVUTIL_VERSION_MAJOR, \ diff --git a/libavutil/x86/cpu.h b/libavutil/x86/cpu.h index f171037..309b8e7 100644 --- a/libavutil/x86/cpu.h +++ b/libavutil/x86/cpu.h @@ -38,6 +38,8 @@ #define X86_SSE3_FAST(flags) CPUEXT_FAST(flags, SSE3) #define X86_SSE3_SLOW(flags) CPUEXT_SLOW(flags, SSE3) #define X86_SSSE3(flags) CPUEXT(flags, SSSE3) +#define X86_SSSE3_FAST(flags) CPUEXT_FAST(flags, SSSE3) +#define X86_SSSE3_SLOW(flags) CPUEXT_SLOW(flags, SSSE3) #define X86_SSE4(flags) CPUEXT(flags, SSE4) #define X86_SSE42(flags) CPUEXT(flags, SSE42) #define X86_AVX(flags) CPUEXT(flags, AVX) @@ -61,6 +63,8 @@ #define EXTERNAL_SSE3_FAST(flags) CPUEXT_SUFFIX_FAST(flags, _EXTERNAL, SSE3) #define EXTERNAL_SSE3_SLOW(flags) CPUEXT_SUFFIX_SLOW(flags, _EXTERNAL, SSE3) #define EXTERNAL_SSSE3(flags) CPUEXT_SUFFIX(flags, _EXTERNAL, SSSE3) +#define EXTERNAL_SSSE3_FAST(flags) CPUEXT_SUFFIX_FAST(flags, _EXTERNAL, SSSE3) +#define EXTERNAL_SSSE3_SLOW(flags) CPUEXT_SUFFIX_SLOW(flags, _EXTERNAL, SSSE3) #define EXTERNAL_SSE4(flags) CPUEXT_SUFFIX(flags, _EXTERNAL, SSE4) #define EXTERNAL_SSE42(flags) CPUEXT_SUFFIX(flags, _EXTERNAL, SSE42) #define EXTERNAL_AVX(flags) CPUEXT_SUFFIX(flags, _EXTERNAL, AVX) @@ -88,6 +92,8 @@ #define INLINE_SSE3_FAST(flags) CPUEXT_SUFFIX_FAST(flags, _INLINE, SSE3) #define INLINE_SSE3_SLOW(flags) CPUEXT_SUFFIX_SLOW(flags, _INLINE, SSE3) #define INLINE_SSSE3(flags) CPUEXT_SUFFIX(flags, _INLINE, SSSE3) +#define INLINE_SSSE3_FAST(flags) CPUEXT_SUFFIX_FAST(flags, _INLINE, SSSE3) +#define INLINE_SSSE3_SLOW(flags) CPUEXT_SUFFIX_SLOW(flags, _INLINE, SSSE3) #define INLINE_SSE4(flags) CPUEXT_SUFFIX(flags, _INLINE, SSE4) #define INLINE_SSE42(flags) CPUEXT_SUFFIX(flags, _INLINE, SSE42) #define INLINE_AVX(flags) CPUEXT_SUFFIX(flags, _INLINE, AVX) ====================================================================== diff --cc doc/APIchanges index 109d550,1b65c4b..8bca71e --- a/doc/APIchanges +++ b/doc/APIchanges @@@ -15,40 -13,10 +15,43 @@@ libavutil: 2015-08-2 API changes, most recent first: -2016-07-20 - xxxxxxx - lavu 55.20.0 - cpu.h ++2017-01-31 - xxxxxxx - lavu 55.46.100 / 55.20.0 - cpu.h + Add AV_CPU_FLAG_SSSE3SLOW. + -2016-07-02 - 4926fa9 - lavu 55.19.0 - hwcontext_vaapi.h +2017-01-24 - xxxxxxx - lavu 55.45.100 - channel_layout.h + Add av_get_extended_channel_layout() + +2017-01-22 - xxxxxxx - lavu 55.44.100 - lfg.h + Add av_lfg_init_from_data(). + +2017-01-xx - xxxxxxx - lavc 57.74.100 - vaapi.h + Deprecate struct vaapi_context and the vaapi.h installed header. + Callers should set AVCodecContext.hw_frames_ctx instead. + +2017-01-12 - xxxxxxx - lavfi 6.69.100- buffersink.h + Add av_buffersink_get_*() functions. + +2017-01-06 - xxxxxxx - lavf 57.62.100- avio.h + Add avio_get_dyn_buf() + +2016-12-10 - xxxxxxx - lavu xx.xx.100- imgutils.h + Add av_image_check_size2() + +2016-xx-xx - xxxxxxx - lavc 57.67.100 / 57.29.0 - avcodec.h + Add AV_PKT_DATA_SPHERICAL packet side data to export AVSphericalMapping + information from containers. + +2016-xx-xx - xxxxxxx - lavu 55.42.100 / 55.30.0 - spherical.h + Add AV_FRAME_DATA_SPHERICAL value, av_spherical_alloc() API and + AVSphericalMapping type to export and describe spherical video properties. + +2016-11-18 - xxxxxxx - lavf 57.58.100 - avformat.h + Add av_stream_add_side_data(). + +2016-xx-xx - xxxxxxx - lavu 55.39.100 - hwcontext_vaapi.h + Add AV_VAAPI_DRIVER_QUIRK_ATTRIB_MEMTYPE. + +2016-xx-xx - xxxxxxx - lavu 55.38.100 - hwcontext_vaapi.h Add driver quirks field to VAAPI-specific hwdevice and enum with members AV_VAAPI_DRIVER_QUIRK_* to represent its values. diff --cc libavutil/version.h index 9d95086,5cbc0e8..8866064 --- a/libavutil/version.h +++ b/libavutil/version.h @@@ -78,9 -53,9 +78,9 @@@ * @{ */ -#define LIBAVUTIL_VERSION_MAJOR 55 -#define LIBAVUTIL_VERSION_MINOR 20 -#define LIBAVUTIL_VERSION_MICRO 0 +#define LIBAVUTIL_VERSION_MAJOR 55 - #define LIBAVUTIL_VERSION_MINOR 45 ++#define LIBAVUTIL_VERSION_MINOR 46 +#define LIBAVUTIL_VERSION_MICRO 100 #define LIBAVUTIL_VERSION_INT AV_VERSION_INT(LIBAVUTIL_VERSION_MAJOR, \ LIBAVUTIL_VERSION_MINOR, \

1 0

x86: Add SSSE3_SLOW CPU flag and related convenience macros
by git＠videolan.org 31 Jan '17

31 Jan '17

ffmpeg | branch: master | Diego Biurrun <diego(a)biurrun.de> | Thu Jan 14 11:37:34 2016 +0100| [7d7355aa92bb36ca0765c49a569a999bcb96f332] | committer: Diego Biurrun x86: Add SSSE3_SLOW CPU flag and related convenience macros > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=7d7355aa92bb36ca0… --- doc/APIchanges | 3 +++ libavutil/cpu.h | 1 + libavutil/version.h | 2 +- libavutil/x86/cpu.h | 6 ++++++ 4 files changed, 11 insertions(+), 1 deletion(-) diff --git a/doc/APIchanges b/doc/APIchanges index c554602..1b65c4b 100644 --- a/doc/APIchanges +++ b/doc/APIchanges @@ -13,6 +13,9 @@ libavutil: 2015-08-28 API changes, most recent first: +2016-07-20 - xxxxxxx - lavu 55.20.0 - cpu.h + Add AV_CPU_FLAG_SSSE3SLOW. + 2016-07-02 - 4926fa9 - lavu 55.19.0 - hwcontext_vaapi.h Add driver quirks field to VAAPI-specific hwdevice and enum with members AV_VAAPI_DRIVER_QUIRK_* to represent its values. diff --git a/libavutil/cpu.h b/libavutil/cpu.h index d640e79..deff4cc 100644 --- a/libavutil/cpu.h +++ b/libavutil/cpu.h @@ -38,6 +38,7 @@ #define AV_CPU_FLAG_SSE3SLOW 0x20000000 ///< SSE3 supported, but usually not faster ///< than regular MMX/SSE (e.g. Core1) #define AV_CPU_FLAG_SSSE3 0x0080 ///< Conroe SSSE3 functions +#define AV_CPU_FLAG_SSSE3SLOW 0x4000000 ///< SSSE3 supported, but usually not faster #define AV_CPU_FLAG_ATOM 0x10000000 ///< Atom processor, some SSSE3 instructions are slower #define AV_CPU_FLAG_SSE4 0x0100 ///< Penryn SSE4.1 functions #define AV_CPU_FLAG_SSE42 0x0200 ///< Nehalem SSE4.2 functions diff --git a/libavutil/version.h b/libavutil/version.h index 1f38c63..5cbc0e8 100644 --- a/libavutil/version.h +++ b/libavutil/version.h @@ -54,7 +54,7 @@ */ #define LIBAVUTIL_VERSION_MAJOR 55 -#define LIBAVUTIL_VERSION_MINOR 19 +#define LIBAVUTIL_VERSION_MINOR 20 #define LIBAVUTIL_VERSION_MICRO 0 #define LIBAVUTIL_VERSION_INT AV_VERSION_INT(LIBAVUTIL_VERSION_MAJOR, \ diff --git a/libavutil/x86/cpu.h b/libavutil/x86/cpu.h index 0695436..c0a525d 100644 --- a/libavutil/x86/cpu.h +++ b/libavutil/x86/cpu.h @@ -38,6 +38,8 @@ #define X86_SSE3_FAST(flags) CPUEXT_FAST(flags, SSE3) #define X86_SSE3_SLOW(flags) CPUEXT_SLOW(flags, SSE3) #define X86_SSSE3(flags) CPUEXT(flags, SSSE3) +#define X86_SSSE3_FAST(flags) CPUEXT_FAST(flags, SSSE3) +#define X86_SSSE3_SLOW(flags) CPUEXT_SLOW(flags, SSSE3) #define X86_SSE4(flags) CPUEXT(flags, SSE4) #define X86_SSE42(flags) CPUEXT(flags, SSE42) #define X86_AVX(flags) CPUEXT(flags, AVX) @@ -60,6 +62,8 @@ #define EXTERNAL_SSE3_FAST(flags) CPUEXT_SUFFIX_FAST(flags, _EXTERNAL, SSE3) #define EXTERNAL_SSE3_SLOW(flags) CPUEXT_SUFFIX_SLOW(flags, _EXTERNAL, SSE3) #define EXTERNAL_SSSE3(flags) CPUEXT_SUFFIX(flags, _EXTERNAL, SSSE3) +#define EXTERNAL_SSSE3_FAST(flags) CPUEXT_SUFFIX_FAST(flags, _EXTERNAL, SSSE3) +#define EXTERNAL_SSSE3_SLOW(flags) CPUEXT_SUFFIX_SLOW(flags, _EXTERNAL, SSSE3) #define EXTERNAL_SSE4(flags) CPUEXT_SUFFIX(flags, _EXTERNAL, SSE4) #define EXTERNAL_SSE42(flags) CPUEXT_SUFFIX(flags, _EXTERNAL, SSE42) #define EXTERNAL_AVX(flags) CPUEXT_SUFFIX(flags, _EXTERNAL, AVX) @@ -82,6 +86,8 @@ #define INLINE_SSE3_FAST(flags) CPUEXT_SUFFIX_FAST(flags, _INLINE, SSE3) #define INLINE_SSE3_SLOW(flags) CPUEXT_SUFFIX_SLOW(flags, _INLINE, SSE3) #define INLINE_SSSE3(flags) CPUEXT_SUFFIX(flags, _INLINE, SSSE3) +#define INLINE_SSSE3_FAST(flags) CPUEXT_SUFFIX_FAST(flags, _INLINE, SSSE3) +#define INLINE_SSSE3_SLOW(flags) CPUEXT_SUFFIX_SLOW(flags, _INLINE, SSSE3) #define INLINE_SSE4(flags) CPUEXT_SUFFIX(flags, _INLINE, SSE4) #define INLINE_SSE42(flags) CPUEXT_SUFFIX(flags, _INLINE, SSE42) #define INLINE_AVX(flags) CPUEXT_SUFFIX(flags, _INLINE, AVX)

1 0

Merge commit '4efab89332ea39a77145e8b15562b981d9dbde68'
by git＠videolan.org 31 Jan '17

31 Jan '17

ffmpeg | branch: master | James Almer <jamrial(a)gmail.com> | Tue Jan 31 15:08:19 2017 -0300| [ac774cfa571734c49c26e2d3387adccff8957ff8] | committer: James Almer Merge commit '4efab89332ea39a77145e8b15562b981d9dbde68' * commit '4efab89332ea39a77145e8b15562b981d9dbde68': x86: Use *_FAST/*_SLOW CPU feature detection macros where appropriate Merged-by: James Almer <jamrial(a)gmail.com> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=ac774cfa571734c49… --- libavcodec/x86/ac3dsp_init.c | 13 ++++++++----- libavcodec/x86/h264_qpel.c | 11 ++++++----- libavcodec/x86/lpc.c | 2 +- libavcodec/x86/vp8dsp_init.c | 4 ++-- 4 files changed, 17 insertions(+), 13 deletions(-) diff --git a/libavcodec/x86/ac3dsp_init.c b/libavcodec/x86/ac3dsp_init.c index 07f0d25..9fd0aef 100644 --- a/libavcodec/x86/ac3dsp_init.c +++ b/libavcodec/x86/ac3dsp_init.c @@ -228,16 +228,19 @@ av_cold void ff_ac3dsp_init_x86(AC3DSPContext *c, int bit_exact) c->float_to_fixed24 = ff_float_to_fixed24_sse2; c->compute_mantissa_size = ff_ac3_compute_mantissa_size_sse2; c->extract_exponents = ff_ac3_extract_exponents_sse2; - if (!(cpu_flags & AV_CPU_FLAG_SSE2SLOW)) { - c->ac3_lshift_int16 = ff_ac3_lshift_int16_sse2; - c->ac3_rshift_int32 = ff_ac3_rshift_int32_sse2; - } if (bit_exact) { c->apply_window_int16 = ff_apply_window_int16_sse2; - } else if (!(cpu_flags & AV_CPU_FLAG_SSE2SLOW)) { + } + } + + if (EXTERNAL_SSE2_FAST(cpu_flags)) { + c->ac3_lshift_int16 = ff_ac3_lshift_int16_sse2; + c->ac3_rshift_int32 = ff_ac3_rshift_int32_sse2; + if (!bit_exact) { c->apply_window_int16 = ff_apply_window_int16_round_sse2; } } + if (EXTERNAL_SSSE3(cpu_flags)) { c->ac3_max_msb_abs_int16 = ff_ac3_max_msb_abs_int16_ssse3; if (cpu_flags & AV_CPU_FLAG_ATOM) { diff --git a/libavcodec/x86/h264_qpel.c b/libavcodec/x86/h264_qpel.c index 8c4b1e1..96fa4a7 100644 --- a/libavcodec/x86/h264_qpel.c +++ b/libavcodec/x86/h264_qpel.c @@ -562,11 +562,6 @@ av_cold void ff_h264qpel_init_x86(H264QpelContext *c, int bit_depth) } if (EXTERNAL_SSE2(cpu_flags)) { - if (!(cpu_flags & AV_CPU_FLAG_SSE2SLOW) && !high_bit_depth) { - // these functions are slower than mmx on AMD, but faster on Intel - H264_QPEL_FUNCS(0, 0, sse2); - } - if (!high_bit_depth) { H264_QPEL_FUNCS(0, 1, sse2); H264_QPEL_FUNCS(0, 2, sse2); @@ -593,6 +588,12 @@ av_cold void ff_h264qpel_init_x86(H264QpelContext *c, int bit_depth) } } + if (EXTERNAL_SSE2_FAST(cpu_flags)) { + if (!high_bit_depth) { + H264_QPEL_FUNCS(0, 0, sse2); + } + } + if (EXTERNAL_SSSE3(cpu_flags)) { if (!high_bit_depth) { H264_QPEL_FUNCS(1, 0, ssse3); diff --git a/libavcodec/x86/lpc.c b/libavcodec/x86/lpc.c index 3a9493f..6c72e21 100644 --- a/libavcodec/x86/lpc.c +++ b/libavcodec/x86/lpc.c @@ -154,7 +154,7 @@ av_cold void ff_lpc_init_x86(LPCContext *c) #if HAVE_SSE2_INLINE int cpu_flags = av_get_cpu_flags(); - if (HAVE_SSE2_INLINE && cpu_flags & (AV_CPU_FLAG_SSE2 | AV_CPU_FLAG_SSE2SLOW)) { + if (INLINE_SSE2(cpu_flags) || INLINE_SSE2_SLOW(cpu_flags)) { c->lpc_apply_welch_window = lpc_apply_welch_window_sse2; c->lpc_compute_autocorr = lpc_compute_autocorr_sse2; } diff --git a/libavcodec/x86/vp8dsp_init.c b/libavcodec/x86/vp8dsp_init.c index 897d5a0..20c5fac 100644 --- a/libavcodec/x86/vp8dsp_init.c +++ b/libavcodec/x86/vp8dsp_init.c @@ -346,7 +346,7 @@ av_cold void ff_vp78dsp_init_x86(VP8DSPContext *c) c->put_vp8_bilinear_pixels_tab[0][0][0] = ff_put_vp8_pixels16_sse; } - if (HAVE_SSE2_EXTERNAL && cpu_flags & (AV_CPU_FLAG_SSE2 | AV_CPU_FLAG_SSE2SLOW)) { + if (EXTERNAL_SSE2(cpu_flags) || EXTERNAL_SSE2_SLOW(cpu_flags)) { VP8_LUMA_MC_FUNC(0, 16, sse2); VP8_MC_FUNC(1, 8, sse2); VP8_BILINEAR_MC_FUNC(0, 16, sse2); @@ -416,7 +416,7 @@ av_cold void ff_vp8dsp_init_x86(VP8DSPContext *c) c->vp8_luma_dc_wht = ff_vp8_luma_dc_wht_sse; } - if (HAVE_SSE2_EXTERNAL && cpu_flags & (AV_CPU_FLAG_SSE2 | AV_CPU_FLAG_SSE2SLOW)) { + if (EXTERNAL_SSE2(cpu_flags) || EXTERNAL_SSE2_SLOW(cpu_flags)) { c->vp8_v_loop_filter_simple = ff_vp8_v_loop_filter_simple_sse2; c->vp8_v_loop_filter16y_inner = ff_vp8_v_loop_filter16y_inner_sse2; ====================================================================== diff --cc libavcodec/x86/lpc.c index 3a9493f,e8cce42..6c72e21 --- a/libavcodec/x86/lpc.c +++ b/libavcodec/x86/lpc.c @@@ -154,7 -152,7 +154,7 @@@ av_cold void ff_lpc_init_x86(LPCContex #if HAVE_SSE2_INLINE int cpu_flags = av_get_cpu_flags(); - if (HAVE_SSE2_INLINE && cpu_flags & (AV_CPU_FLAG_SSE2 | AV_CPU_FLAG_SSE2SLOW)) { - if (INLINE_SSE2_SLOW(cpu_flags)) { ++ if (INLINE_SSE2(cpu_flags) || INLINE_SSE2_SLOW(cpu_flags)) { c->lpc_apply_welch_window = lpc_apply_welch_window_sse2; c->lpc_compute_autocorr = lpc_compute_autocorr_sse2; } diff --cc libavcodec/x86/vp8dsp_init.c index 897d5a0,3e84bed..20c5fac --- a/libavcodec/x86/vp8dsp_init.c +++ b/libavcodec/x86/vp8dsp_init.c @@@ -346,7 -346,7 +346,7 @@@ av_cold void ff_vp78dsp_init_x86(VP8DSP c->put_vp8_bilinear_pixels_tab[0][0][0] = ff_put_vp8_pixels16_sse; } - if (HAVE_SSE2_EXTERNAL && cpu_flags & (AV_CPU_FLAG_SSE2 | AV_CPU_FLAG_SSE2SLOW)) { - if (EXTERNAL_SSE2_SLOW(cpu_flags)) { ++ if (EXTERNAL_SSE2(cpu_flags) || EXTERNAL_SSE2_SLOW(cpu_flags)) { VP8_LUMA_MC_FUNC(0, 16, sse2); VP8_MC_FUNC(1, 8, sse2); VP8_BILINEAR_MC_FUNC(0, 16, sse2); @@@ -416,7 -416,7 +416,7 @@@ av_cold void ff_vp8dsp_init_x86(VP8DSPC c->vp8_luma_dc_wht = ff_vp8_luma_dc_wht_sse; } - if (HAVE_SSE2_EXTERNAL && cpu_flags & (AV_CPU_FLAG_SSE2 | AV_CPU_FLAG_SSE2SLOW)) { - if (EXTERNAL_SSE2_SLOW(cpu_flags)) { ++ if (EXTERNAL_SSE2(cpu_flags) || EXTERNAL_SSE2_SLOW(cpu_flags)) { c->vp8_v_loop_filter_simple = ff_vp8_v_loop_filter_simple_sse2; c->vp8_v_loop_filter16y_inner = ff_vp8_v_loop_filter16y_inner_sse2;

1 0