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

Merge commit '0a39c9ac0bfd7345fe676b4e2707d9cec3cbb553'
by git＠videolan.org 31 Jan '17

31 Jan '17

ffmpeg | branch: master | James Almer <jamrial(a)gmail.com> | Tue Jan 31 14:59:29 2017 -0300| [a956164e1eb3418922cae949f02ad4035f013213] | committer: James Almer Merge commit '0a39c9ac0bfd7345fe676b4e2707d9cec3cbb553' * commit '0a39c9ac0bfd7345fe676b4e2707d9cec3cbb553': x86: hpeldsp: Don't check for bitexact flag when initializing VP3-specific code Merged-by: James Almer <jamrial(a)gmail.com> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=a956164e1eb341892… --- libavcodec/x86/hpeldsp.h | 2 +- libavcodec/x86/hpeldsp_init.c | 2 +- libavcodec/x86/hpeldsp_vp3_init.c | 14 +++++--------- 3 files changed, 7 insertions(+), 11 deletions(-) diff --git a/libavcodec/x86/hpeldsp.h b/libavcodec/x86/hpeldsp.h index bf97029..0ecc97a 100644 --- a/libavcodec/x86/hpeldsp.h +++ b/libavcodec/x86/hpeldsp.h @@ -52,6 +52,6 @@ void ff_put_pixels16_xy2_sse2(uint8_t *block, const uint8_t *pixels, void ff_put_pixels16_xy2_ssse3(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h); -void ff_hpeldsp_vp3_init_x86(HpelDSPContext *c, int cpu_flags, int flags); +void ff_hpeldsp_vp3_init_x86(HpelDSPContext *c, int cpu_flags); #endif /* AVCODEC_X86_HPELDSP_H */ diff --git a/libavcodec/x86/hpeldsp_init.c b/libavcodec/x86/hpeldsp_init.c index 58e27e3..e583bd9 100644 --- a/libavcodec/x86/hpeldsp_init.c +++ b/libavcodec/x86/hpeldsp_init.c @@ -309,5 +309,5 @@ av_cold void ff_hpeldsp_init_x86(HpelDSPContext *c, int flags) hpeldsp_init_ssse3(c, flags); if (CONFIG_VP3_DECODER) - ff_hpeldsp_vp3_init_x86(c, cpu_flags, flags); + ff_hpeldsp_vp3_init_x86(c, cpu_flags); } diff --git a/libavcodec/x86/hpeldsp_vp3_init.c b/libavcodec/x86/hpeldsp_vp3_init.c index 5979f41..17fdd08 100644 --- a/libavcodec/x86/hpeldsp_vp3_init.c +++ b/libavcodec/x86/hpeldsp_vp3_init.c @@ -38,19 +38,15 @@ void ff_put_no_rnd_pixels8_y2_exact_3dnow(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h); -av_cold void ff_hpeldsp_vp3_init_x86(HpelDSPContext *c, int cpu_flags, int flags) +av_cold void ff_hpeldsp_vp3_init_x86(HpelDSPContext *c, int cpu_flags) { if (EXTERNAL_AMD3DNOW(cpu_flags)) { - if (flags & AV_CODEC_FLAG_BITEXACT) { - c->put_no_rnd_pixels_tab[1][1] = ff_put_no_rnd_pixels8_x2_exact_3dnow; - c->put_no_rnd_pixels_tab[1][2] = ff_put_no_rnd_pixels8_y2_exact_3dnow; - } + c->put_no_rnd_pixels_tab[1][1] = ff_put_no_rnd_pixels8_x2_exact_3dnow; + c->put_no_rnd_pixels_tab[1][2] = ff_put_no_rnd_pixels8_y2_exact_3dnow; } if (EXTERNAL_MMXEXT(cpu_flags)) { - if (flags & AV_CODEC_FLAG_BITEXACT) { - c->put_no_rnd_pixels_tab[1][1] = ff_put_no_rnd_pixels8_x2_exact_mmxext; - c->put_no_rnd_pixels_tab[1][2] = ff_put_no_rnd_pixels8_y2_exact_mmxext; - } + c->put_no_rnd_pixels_tab[1][1] = ff_put_no_rnd_pixels8_x2_exact_mmxext; + c->put_no_rnd_pixels_tab[1][2] = ff_put_no_rnd_pixels8_y2_exact_mmxext; } } ====================================================================== diff --cc libavcodec/x86/hpeldsp.h index bf97029,566e518..0ecc97a --- a/libavcodec/x86/hpeldsp.h +++ b/libavcodec/x86/hpeldsp.h @@@ -43,15 -34,9 +43,15 @@@ void ff_avg_pixels16_xy2_ssse3(uint8_t void ff_put_pixels8_xy2_mmx(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h); +void ff_put_pixels8_xy2_ssse3(uint8_t *block, const uint8_t *pixels, + ptrdiff_t line_size, int h); void ff_put_pixels16_xy2_mmx(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h); +void ff_put_pixels16_xy2_sse2(uint8_t *block, const uint8_t *pixels, + ptrdiff_t line_size, int h); +void ff_put_pixels16_xy2_ssse3(uint8_t *block, const uint8_t *pixels, + ptrdiff_t line_size, int h); - void ff_hpeldsp_vp3_init_x86(HpelDSPContext *c, int cpu_flags, int flags); + void ff_hpeldsp_vp3_init_x86(HpelDSPContext *c, int cpu_flags); #endif /* AVCODEC_X86_HPELDSP_H */ diff --cc libavcodec/x86/hpeldsp_init.c index 58e27e3,6a8d420..e583bd9 --- a/libavcodec/x86/hpeldsp_init.c +++ b/libavcodec/x86/hpeldsp_init.c @@@ -305,9 -254,6 +305,9 @@@ av_cold void ff_hpeldsp_init_x86(HpelDS if (EXTERNAL_SSE2_FAST(cpu_flags)) hpeldsp_init_sse2_fast(c, flags); + if (EXTERNAL_SSSE3(cpu_flags)) + hpeldsp_init_ssse3(c, flags); + if (CONFIG_VP3_DECODER) - ff_hpeldsp_vp3_init_x86(c, cpu_flags, flags); + ff_hpeldsp_vp3_init_x86(c, cpu_flags); }

1 0

x86: Use *_FAST/*_SLOW CPU feature detection macros where appropriate
by git＠videolan.org 31 Jan '17

31 Jan '17

ffmpeg | branch: master | Diego Biurrun <diego(a)biurrun.de> | Thu Jan 14 10:57:41 2016 +0100| [4efab89332ea39a77145e8b15562b981d9dbde68] | committer: Diego Biurrun x86: Use *_FAST/*_SLOW CPU feature detection macros where appropriate > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=4efab89332ea39a77… --- 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 23f0162..89044f4 100644 --- a/libavcodec/x86/ac3dsp_init.c +++ b/libavcodec/x86/ac3dsp_init.c @@ -223,16 +223,19 @@ 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 541ebb2..43e150c 100644 --- a/libavcodec/x86/h264_qpel.c +++ b/libavcodec/x86/h264_qpel.c @@ -566,11 +566,6 @@ 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); @@ -597,6 +592,12 @@ 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 ea5d2ea..e8cce42 100644 --- a/libavcodec/x86/lpc.c +++ b/libavcodec/x86/lpc.c @@ -152,7 +152,7 @@ av_cold void ff_lpc_init_x86(LPCContext *c) #if HAVE_SSE2_INLINE int cpu_flags = av_get_cpu_flags(); - if (INLINE_SSE2(cpu_flags) && (cpu_flags & AV_CPU_FLAG_SSE2SLOW)) { + if (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 8f6d43c..3e84bed 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 (EXTERNAL_SSE2(cpu_flags) && (cpu_flags & AV_CPU_FLAG_SSE2SLOW)) { + if (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 (EXTERNAL_SSE2(cpu_flags) && (cpu_flags & AV_CPU_FLAG_SSE2SLOW)) { + if (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

x86: hpeldsp: Don' t check for bitexact flag when initializing VP3-specific code
by git＠videolan.org 31 Jan '17

31 Jan '17

ffmpeg | branch: master | Diego Biurrun <diego(a)biurrun.de> | Tue Jan 19 17:55:59 2016 +0100| [0a39c9ac0bfd7345fe676b4e2707d9cec3cbb553] | committer: Diego Biurrun x86: hpeldsp: Don't check for bitexact flag when initializing VP3-specific code That code is only ever initialized with that flag set. > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=0a39c9ac0bfd7345f… --- libavcodec/x86/hpeldsp.h | 2 +- libavcodec/x86/hpeldsp_init.c | 2 +- libavcodec/x86/hpeldsp_vp3_init.c | 14 +++++--------- 3 files changed, 7 insertions(+), 11 deletions(-) diff --git a/libavcodec/x86/hpeldsp.h b/libavcodec/x86/hpeldsp.h index d624ed9..566e518 100644 --- a/libavcodec/x86/hpeldsp.h +++ b/libavcodec/x86/hpeldsp.h @@ -37,6 +37,6 @@ void ff_put_pixels8_xy2_mmx(uint8_t *block, const uint8_t *pixels, void ff_put_pixels16_xy2_mmx(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h); -void ff_hpeldsp_vp3_init_x86(HpelDSPContext *c, int cpu_flags, int flags); +void ff_hpeldsp_vp3_init_x86(HpelDSPContext *c, int cpu_flags); #endif /* AVCODEC_X86_HPELDSP_H */ diff --git a/libavcodec/x86/hpeldsp_init.c b/libavcodec/x86/hpeldsp_init.c index cd3999e..6a8d420 100644 --- a/libavcodec/x86/hpeldsp_init.c +++ b/libavcodec/x86/hpeldsp_init.c @@ -255,5 +255,5 @@ av_cold void ff_hpeldsp_init_x86(HpelDSPContext *c, int flags) hpeldsp_init_sse2_fast(c, flags); if (CONFIG_VP3_DECODER) - ff_hpeldsp_vp3_init_x86(c, cpu_flags, flags); + ff_hpeldsp_vp3_init_x86(c, cpu_flags); } diff --git a/libavcodec/x86/hpeldsp_vp3_init.c b/libavcodec/x86/hpeldsp_vp3_init.c index 2510c11..cc1f5e4 100644 --- a/libavcodec/x86/hpeldsp_vp3_init.c +++ b/libavcodec/x86/hpeldsp_vp3_init.c @@ -38,19 +38,15 @@ void ff_put_no_rnd_pixels8_y2_exact_3dnow(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h); -av_cold void ff_hpeldsp_vp3_init_x86(HpelDSPContext *c, int cpu_flags, int flags) +av_cold void ff_hpeldsp_vp3_init_x86(HpelDSPContext *c, int cpu_flags) { if (EXTERNAL_AMD3DNOW(cpu_flags)) { - if (flags & AV_CODEC_FLAG_BITEXACT) { - c->put_no_rnd_pixels_tab[1][1] = ff_put_no_rnd_pixels8_x2_exact_3dnow; - c->put_no_rnd_pixels_tab[1][2] = ff_put_no_rnd_pixels8_y2_exact_3dnow; - } + c->put_no_rnd_pixels_tab[1][1] = ff_put_no_rnd_pixels8_x2_exact_3dnow; + c->put_no_rnd_pixels_tab[1][2] = ff_put_no_rnd_pixels8_y2_exact_3dnow; } if (EXTERNAL_MMXEXT(cpu_flags)) { - if (flags & AV_CODEC_FLAG_BITEXACT) { - c->put_no_rnd_pixels_tab[1][1] = ff_put_no_rnd_pixels8_x2_exact_mmxext; - c->put_no_rnd_pixels_tab[1][2] = ff_put_no_rnd_pixels8_y2_exact_mmxext; - } + c->put_no_rnd_pixels_tab[1][1] = ff_put_no_rnd_pixels8_x2_exact_mmxext; + c->put_no_rnd_pixels_tab[1][2] = ff_put_no_rnd_pixels8_y2_exact_mmxext; } }

1 0

Merge commit '95c1df929b92d81454656c222a35ec5f7db576b4'
by git＠videolan.org 31 Jan '17

31 Jan '17

ffmpeg | branch: master | James Almer <jamrial(a)gmail.com> | Tue Jan 31 14:56:11 2017 -0300| [f6005907fdeb9e4de37568ed5c1a8e7b869126f6] | committer: James Almer Merge commit '95c1df929b92d81454656c222a35ec5f7db576b4' * commit '95c1df929b92d81454656c222a35ec5f7db576b4': x86: hpeldsp: Drop unused function parameters Merged-by: James Almer <jamrial(a)gmail.com> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=f6005907fdeb9e4de… --- libavcodec/x86/hpeldsp_init.c | 20 ++++++++++---------- 1 file changed, 10 insertions(+), 10 deletions(-) diff --git a/libavcodec/x86/hpeldsp_init.c b/libavcodec/x86/hpeldsp_init.c index a192924..58e27e3 100644 --- a/libavcodec/x86/hpeldsp_init.c +++ b/libavcodec/x86/hpeldsp_init.c @@ -184,7 +184,7 @@ HPELDSP_AVG_PIXELS16(_mmxext) } while (0) #endif -static void hpeldsp_init_mmx(HpelDSPContext *c, int flags, int cpu_flags) +static void hpeldsp_init_mmx(HpelDSPContext *c, int flags) { SET_HPEL_FUNCS(put, [0], 16, mmx); SET_HPEL_FUNCS(put_no_rnd, [0], 16, mmx); @@ -202,7 +202,7 @@ static void hpeldsp_init_mmx(HpelDSPContext *c, int flags, int cpu_flags) #endif } -static void hpeldsp_init_mmxext(HpelDSPContext *c, int flags, int cpu_flags) +static void hpeldsp_init_mmxext(HpelDSPContext *c, int flags) { #if HAVE_MMXEXT_EXTERNAL c->put_pixels_tab[0][1] = ff_put_pixels16_x2_mmxext; @@ -233,7 +233,7 @@ static void hpeldsp_init_mmxext(HpelDSPContext *c, int flags, int cpu_flags) #endif /* HAVE_MMXEXT_EXTERNAL */ } -static void hpeldsp_init_3dnow(HpelDSPContext *c, int flags, int cpu_flags) +static void hpeldsp_init_3dnow(HpelDSPContext *c, int flags) { #if HAVE_AMD3DNOW_EXTERNAL c->put_pixels_tab[0][1] = ff_put_pixels16_x2_3dnow; @@ -264,7 +264,7 @@ static void hpeldsp_init_3dnow(HpelDSPContext *c, int flags, int cpu_flags) #endif /* HAVE_AMD3DNOW_EXTERNAL */ } -static void hpeldsp_init_sse2_fast(HpelDSPContext *c, int flags, int cpu_flags) +static void hpeldsp_init_sse2_fast(HpelDSPContext *c, int flags) { #if HAVE_SSE2_EXTERNAL c->put_pixels_tab[0][0] = ff_put_pixels16_sse2; @@ -279,7 +279,7 @@ static void hpeldsp_init_sse2_fast(HpelDSPContext *c, int flags, int cpu_flags) #endif /* HAVE_SSE2_EXTERNAL */ } -static void hpeldsp_init_ssse3(HpelDSPContext *c, int flags, int cpu_flags) +static void hpeldsp_init_ssse3(HpelDSPContext *c, int flags) { #if HAVE_SSSE3_EXTERNAL c->put_pixels_tab[0][3] = ff_put_pixels16_xy2_ssse3; @@ -294,19 +294,19 @@ av_cold void ff_hpeldsp_init_x86(HpelDSPContext *c, int flags) int cpu_flags = av_get_cpu_flags(); if (INLINE_MMX(cpu_flags)) - hpeldsp_init_mmx(c, flags, cpu_flags); + hpeldsp_init_mmx(c, flags); if (EXTERNAL_AMD3DNOW(cpu_flags)) - hpeldsp_init_3dnow(c, flags, cpu_flags); + hpeldsp_init_3dnow(c, flags); if (EXTERNAL_MMXEXT(cpu_flags)) - hpeldsp_init_mmxext(c, flags, cpu_flags); + hpeldsp_init_mmxext(c, flags); if (EXTERNAL_SSE2_FAST(cpu_flags)) - hpeldsp_init_sse2_fast(c, flags, cpu_flags); + hpeldsp_init_sse2_fast(c, flags); if (EXTERNAL_SSSE3(cpu_flags)) - hpeldsp_init_ssse3(c, flags, cpu_flags); + hpeldsp_init_ssse3(c, flags); if (CONFIG_VP3_DECODER) ff_hpeldsp_vp3_init_x86(c, cpu_flags, flags); ====================================================================== diff --cc libavcodec/x86/hpeldsp_init.c index a192924,cd3999e..58e27e3 --- a/libavcodec/x86/hpeldsp_init.c +++ b/libavcodec/x86/hpeldsp_init.c @@@ -177,32 -157,21 +177,32 @@@ HPELDSP_AVG_PIXELS16(_mmxext c->PFX ## _pixels_tab IDX [2] = PFX ## _pixels ## SIZE ## _y2_ ## CPU; \ c->PFX ## _pixels_tab IDX [3] = PFX ## _pixels ## SIZE ## _xy2_ ## CPU; \ } while (0) +#else +#define SET_HPEL_FUNCS(PFX, IDX, SIZE, CPU) \ + do { \ + SET_HPEL_FUNCS_EXT(PFX, IDX, SIZE, CPU) \ + } while (0) +#endif - static void hpeldsp_init_mmx(HpelDSPContext *c, int flags, int cpu_flags) + static void hpeldsp_init_mmx(HpelDSPContext *c, int flags) { -#if HAVE_MMX_INLINE SET_HPEL_FUNCS(put, [0], 16, mmx); SET_HPEL_FUNCS(put_no_rnd, [0], 16, mmx); SET_HPEL_FUNCS(avg, [0], 16, mmx); SET_HPEL_FUNCS(avg_no_rnd, , 16, mmx); SET_HPEL_FUNCS(put, [1], 8, mmx); SET_HPEL_FUNCS(put_no_rnd, [1], 8, mmx); - SET_HPEL_FUNCS(avg, [1], 8, mmx); -#endif /* HAVE_MMX_INLINE */ + if (HAVE_MMX_EXTERNAL) { + c->avg_pixels_tab[1][0] = ff_avg_pixels8_mmx; + c->avg_pixels_tab[1][1] = ff_avg_pixels8_x2_mmx; + } +#if HAVE_MMX_INLINE + c->avg_pixels_tab[1][2] = avg_pixels8_y2_mmx; + c->avg_pixels_tab[1][3] = ff_avg_pixels8_xy2_mmx; +#endif } - static void hpeldsp_init_mmxext(HpelDSPContext *c, int flags, int cpu_flags) + static void hpeldsp_init_mmxext(HpelDSPContext *c, int flags) { #if HAVE_MMXEXT_EXTERNAL c->put_pixels_tab[0][1] = ff_put_pixels16_x2_mmxext; @@@ -279,35 -238,22 +279,35 @@@ static void hpeldsp_init_sse2_fast(Hpel #endif /* HAVE_SSE2_EXTERNAL */ } - static void hpeldsp_init_ssse3(HpelDSPContext *c, int flags, int cpu_flags) ++static void hpeldsp_init_ssse3(HpelDSPContext *c, int flags) +{ +#if HAVE_SSSE3_EXTERNAL + c->put_pixels_tab[0][3] = ff_put_pixels16_xy2_ssse3; + c->avg_pixels_tab[0][3] = ff_avg_pixels16_xy2_ssse3; + c->put_pixels_tab[1][3] = ff_put_pixels8_xy2_ssse3; + c->avg_pixels_tab[1][3] = ff_avg_pixels8_xy2_ssse3; +#endif +} + av_cold void ff_hpeldsp_init_x86(HpelDSPContext *c, int flags) { int cpu_flags = av_get_cpu_flags(); if (INLINE_MMX(cpu_flags)) - hpeldsp_init_mmx(c, flags, cpu_flags); + hpeldsp_init_mmx(c, flags); if (EXTERNAL_AMD3DNOW(cpu_flags)) - hpeldsp_init_3dnow(c, flags, cpu_flags); + hpeldsp_init_3dnow(c, flags); if (EXTERNAL_MMXEXT(cpu_flags)) - hpeldsp_init_mmxext(c, flags, cpu_flags); + hpeldsp_init_mmxext(c, flags); if (EXTERNAL_SSE2_FAST(cpu_flags)) - hpeldsp_init_sse2_fast(c, flags, cpu_flags); + hpeldsp_init_sse2_fast(c, flags); + if (EXTERNAL_SSSE3(cpu_flags)) - hpeldsp_init_ssse3(c, flags, cpu_flags); ++ hpeldsp_init_ssse3(c, flags); + if (CONFIG_VP3_DECODER) ff_hpeldsp_vp3_init_x86(c, cpu_flags, flags); }

1 0

x86: hpeldsp: Drop unused function parameters
by git＠videolan.org 31 Jan '17

31 Jan '17

ffmpeg | branch: master | Diego Biurrun <diego(a)biurrun.de> | Mon Jan 11 16:06:33 2016 +0100| [95c1df929b92d81454656c222a35ec5f7db576b4] | committer: Diego Biurrun x86: hpeldsp: Drop unused function parameters > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=95c1df929b92d8145… --- libavcodec/x86/hpeldsp_init.c | 16 ++++++++-------- 1 file changed, 8 insertions(+), 8 deletions(-) diff --git a/libavcodec/x86/hpeldsp_init.c b/libavcodec/x86/hpeldsp_init.c index a7a4515..cd3999e 100644 --- a/libavcodec/x86/hpeldsp_init.c +++ b/libavcodec/x86/hpeldsp_init.c @@ -158,7 +158,7 @@ HPELDSP_AVG_PIXELS16(_mmxext) c->PFX ## _pixels_tab IDX [3] = PFX ## _pixels ## SIZE ## _xy2_ ## CPU; \ } while (0) -static void hpeldsp_init_mmx(HpelDSPContext *c, int flags, int cpu_flags) +static void hpeldsp_init_mmx(HpelDSPContext *c, int flags) { #if HAVE_MMX_INLINE SET_HPEL_FUNCS(put, [0], 16, mmx); @@ -171,7 +171,7 @@ static void hpeldsp_init_mmx(HpelDSPContext *c, int flags, int cpu_flags) #endif /* HAVE_MMX_INLINE */ } -static void hpeldsp_init_mmxext(HpelDSPContext *c, int flags, int cpu_flags) +static void hpeldsp_init_mmxext(HpelDSPContext *c, int flags) { #if HAVE_MMXEXT_EXTERNAL c->put_pixels_tab[0][1] = ff_put_pixels16_x2_mmxext; @@ -200,7 +200,7 @@ static void hpeldsp_init_mmxext(HpelDSPContext *c, int flags, int cpu_flags) #endif /* HAVE_MMXEXT_EXTERNAL */ } -static void hpeldsp_init_3dnow(HpelDSPContext *c, int flags, int cpu_flags) +static void hpeldsp_init_3dnow(HpelDSPContext *c, int flags) { #if HAVE_AMD3DNOW_EXTERNAL c->put_pixels_tab[0][1] = ff_put_pixels16_x2_3dnow; @@ -229,7 +229,7 @@ static void hpeldsp_init_3dnow(HpelDSPContext *c, int flags, int cpu_flags) #endif /* HAVE_AMD3DNOW_EXTERNAL */ } -static void hpeldsp_init_sse2_fast(HpelDSPContext *c, int flags, int cpu_flags) +static void hpeldsp_init_sse2_fast(HpelDSPContext *c, int flags) { #if HAVE_SSE2_EXTERNAL c->put_pixels_tab[0][0] = ff_put_pixels16_sse2; @@ -243,16 +243,16 @@ av_cold void ff_hpeldsp_init_x86(HpelDSPContext *c, int flags) int cpu_flags = av_get_cpu_flags(); if (INLINE_MMX(cpu_flags)) - hpeldsp_init_mmx(c, flags, cpu_flags); + hpeldsp_init_mmx(c, flags); if (EXTERNAL_AMD3DNOW(cpu_flags)) - hpeldsp_init_3dnow(c, flags, cpu_flags); + hpeldsp_init_3dnow(c, flags); if (EXTERNAL_MMXEXT(cpu_flags)) - hpeldsp_init_mmxext(c, flags, cpu_flags); + hpeldsp_init_mmxext(c, flags); if (EXTERNAL_SSE2_FAST(cpu_flags)) - hpeldsp_init_sse2_fast(c, flags, cpu_flags); + hpeldsp_init_sse2_fast(c, flags); if (CONFIG_VP3_DECODER) ff_hpeldsp_vp3_init_x86(c, cpu_flags, flags);

1 0

Merge commit 'c3e83ad3b7d75f3597f47ada2616ba4479665009'
by git＠videolan.org 31 Jan '17

31 Jan '17

ffmpeg | branch: master | James Almer <jamrial(a)gmail.com> | Tue Jan 31 14:53:27 2017 -0300| [4d0e89ce27bc8d995bf9be831e62ea0b3893e375] | committer: James Almer Merge commit 'c3e83ad3b7d75f3597f47ada2616ba4479665009' * commit 'c3e83ad3b7d75f3597f47ada2616ba4479665009': x86: hpeldsp: Use EXTERNAL_SSE2_FAST where appropriate Merged-by: James Almer <jamrial(a)gmail.com> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=4d0e89ce27bc8d995… --- libavcodec/x86/hpeldsp_init.c | 27 ++++++++++++--------------- 1 file changed, 12 insertions(+), 15 deletions(-) diff --git a/libavcodec/x86/hpeldsp_init.c b/libavcodec/x86/hpeldsp_init.c index e8da184..a192924 100644 --- a/libavcodec/x86/hpeldsp_init.c +++ b/libavcodec/x86/hpeldsp_init.c @@ -264,21 +264,18 @@ static void hpeldsp_init_3dnow(HpelDSPContext *c, int flags, int cpu_flags) #endif /* HAVE_AMD3DNOW_EXTERNAL */ } -static void hpeldsp_init_sse2(HpelDSPContext *c, int flags, int cpu_flags) +static void hpeldsp_init_sse2_fast(HpelDSPContext *c, int flags, int cpu_flags) { #if HAVE_SSE2_EXTERNAL - if (!(cpu_flags & AV_CPU_FLAG_SSE2SLOW)) { - // these functions are slower than mmx on AMD, but faster on Intel - c->put_pixels_tab[0][0] = ff_put_pixels16_sse2; - c->put_no_rnd_pixels_tab[0][0] = ff_put_pixels16_sse2; - c->put_pixels_tab[0][1] = ff_put_pixels16_x2_sse2; - c->put_pixels_tab[0][2] = ff_put_pixels16_y2_sse2; - c->put_pixels_tab[0][3] = ff_put_pixels16_xy2_sse2; - c->avg_pixels_tab[0][0] = ff_avg_pixels16_sse2; - c->avg_pixels_tab[0][1] = ff_avg_pixels16_x2_sse2; - c->avg_pixels_tab[0][2] = ff_avg_pixels16_y2_sse2; - c->avg_pixels_tab[0][3] = ff_avg_pixels16_xy2_sse2; - } + c->put_pixels_tab[0][0] = ff_put_pixels16_sse2; + c->put_no_rnd_pixels_tab[0][0] = ff_put_pixels16_sse2; + c->put_pixels_tab[0][1] = ff_put_pixels16_x2_sse2; + c->put_pixels_tab[0][2] = ff_put_pixels16_y2_sse2; + c->put_pixels_tab[0][3] = ff_put_pixels16_xy2_sse2; + c->avg_pixels_tab[0][0] = ff_avg_pixels16_sse2; + c->avg_pixels_tab[0][1] = ff_avg_pixels16_x2_sse2; + c->avg_pixels_tab[0][2] = ff_avg_pixels16_y2_sse2; + c->avg_pixels_tab[0][3] = ff_avg_pixels16_xy2_sse2; #endif /* HAVE_SSE2_EXTERNAL */ } @@ -305,8 +302,8 @@ av_cold void ff_hpeldsp_init_x86(HpelDSPContext *c, int flags) if (EXTERNAL_MMXEXT(cpu_flags)) hpeldsp_init_mmxext(c, flags, cpu_flags); - if (EXTERNAL_SSE2(cpu_flags)) - hpeldsp_init_sse2(c, flags, cpu_flags); + if (EXTERNAL_SSE2_FAST(cpu_flags)) + hpeldsp_init_sse2_fast(c, flags, cpu_flags); if (EXTERNAL_SSSE3(cpu_flags)) hpeldsp_init_ssse3(c, flags, cpu_flags); ====================================================================== diff --cc libavcodec/x86/hpeldsp_init.c index e8da184,a7a4515..a192924 --- a/libavcodec/x86/hpeldsp_init.c +++ b/libavcodec/x86/hpeldsp_init.c @@@ -264,21 -229,12 +264,18 @@@ static void hpeldsp_init_3dnow(HpelDSPC #endif /* HAVE_AMD3DNOW_EXTERNAL */ } - static void hpeldsp_init_sse2(HpelDSPContext *c, int flags, int cpu_flags) + static void hpeldsp_init_sse2_fast(HpelDSPContext *c, int flags, int cpu_flags) { #if HAVE_SSE2_EXTERNAL - if (!(cpu_flags & AV_CPU_FLAG_SSE2SLOW)) { - // these functions are slower than mmx on AMD, but faster on Intel - c->put_pixels_tab[0][0] = ff_put_pixels16_sse2; - c->put_no_rnd_pixels_tab[0][0] = ff_put_pixels16_sse2; - c->put_pixels_tab[0][1] = ff_put_pixels16_x2_sse2; - c->put_pixels_tab[0][2] = ff_put_pixels16_y2_sse2; - c->put_pixels_tab[0][3] = ff_put_pixels16_xy2_sse2; - c->avg_pixels_tab[0][0] = ff_avg_pixels16_sse2; - c->avg_pixels_tab[0][1] = ff_avg_pixels16_x2_sse2; - c->avg_pixels_tab[0][2] = ff_avg_pixels16_y2_sse2; - c->avg_pixels_tab[0][3] = ff_avg_pixels16_xy2_sse2; - } + c->put_pixels_tab[0][0] = ff_put_pixels16_sse2; + c->put_no_rnd_pixels_tab[0][0] = ff_put_pixels16_sse2; ++ c->put_pixels_tab[0][1] = ff_put_pixels16_x2_sse2; ++ c->put_pixels_tab[0][2] = ff_put_pixels16_y2_sse2; ++ c->put_pixels_tab[0][3] = ff_put_pixels16_xy2_sse2; + c->avg_pixels_tab[0][0] = ff_avg_pixels16_sse2; ++ c->avg_pixels_tab[0][1] = ff_avg_pixels16_x2_sse2; ++ c->avg_pixels_tab[0][2] = ff_avg_pixels16_y2_sse2; ++ c->avg_pixels_tab[0][3] = ff_avg_pixels16_xy2_sse2; #endif /* HAVE_SSE2_EXTERNAL */ } @@@ -305,12 -251,9 +302,12 @@@ av_cold void ff_hpeldsp_init_x86(HpelDS if (EXTERNAL_MMXEXT(cpu_flags)) hpeldsp_init_mmxext(c, flags, cpu_flags); - if (EXTERNAL_SSE2(cpu_flags)) - hpeldsp_init_sse2(c, flags, cpu_flags); + if (EXTERNAL_SSE2_FAST(cpu_flags)) + hpeldsp_init_sse2_fast(c, flags, cpu_flags); + if (EXTERNAL_SSSE3(cpu_flags)) + hpeldsp_init_ssse3(c, flags, cpu_flags); + if (CONFIG_VP3_DECODER) ff_hpeldsp_vp3_init_x86(c, cpu_flags, flags); }

1 0

x86: hpeldsp: Use EXTERNAL_SSE2_FAST where appropriate
by git＠videolan.org 31 Jan '17

31 Jan '17

ffmpeg | branch: master | Diego Biurrun <diego(a)biurrun.de> | Mon Jan 11 16:04:17 2016 +0100| [c3e83ad3b7d75f3597f47ada2616ba4479665009] | committer: Diego Biurrun x86: hpeldsp: Use EXTERNAL_SSE2_FAST where appropriate > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=c3e83ad3b7d75f359… --- libavcodec/x86/hpeldsp_init.c | 15 ++++++--------- 1 file changed, 6 insertions(+), 9 deletions(-) diff --git a/libavcodec/x86/hpeldsp_init.c b/libavcodec/x86/hpeldsp_init.c index 6731428..a7a4515 100644 --- a/libavcodec/x86/hpeldsp_init.c +++ b/libavcodec/x86/hpeldsp_init.c @@ -229,15 +229,12 @@ static void hpeldsp_init_3dnow(HpelDSPContext *c, int flags, int cpu_flags) #endif /* HAVE_AMD3DNOW_EXTERNAL */ } -static void hpeldsp_init_sse2(HpelDSPContext *c, int flags, int cpu_flags) +static void hpeldsp_init_sse2_fast(HpelDSPContext *c, int flags, int cpu_flags) { #if HAVE_SSE2_EXTERNAL - if (!(cpu_flags & AV_CPU_FLAG_SSE2SLOW)) { - // these functions are slower than mmx on AMD, but faster on Intel - c->put_pixels_tab[0][0] = ff_put_pixels16_sse2; - c->put_no_rnd_pixels_tab[0][0] = ff_put_pixels16_sse2; - c->avg_pixels_tab[0][0] = ff_avg_pixels16_sse2; - } + c->put_pixels_tab[0][0] = ff_put_pixels16_sse2; + c->put_no_rnd_pixels_tab[0][0] = ff_put_pixels16_sse2; + c->avg_pixels_tab[0][0] = ff_avg_pixels16_sse2; #endif /* HAVE_SSE2_EXTERNAL */ } @@ -254,8 +251,8 @@ av_cold void ff_hpeldsp_init_x86(HpelDSPContext *c, int flags) if (EXTERNAL_MMXEXT(cpu_flags)) hpeldsp_init_mmxext(c, flags, cpu_flags); - if (EXTERNAL_SSE2(cpu_flags)) - hpeldsp_init_sse2(c, flags, cpu_flags); + if (EXTERNAL_SSE2_FAST(cpu_flags)) + hpeldsp_init_sse2_fast(c, flags, cpu_flags); if (CONFIG_VP3_DECODER) ff_hpeldsp_vp3_init_x86(c, cpu_flags, flags);

1 0

Merge commit '1dfc3cf89d0eb026af28be46294b85d79499ffb5'
by git＠videolan.org 31 Jan '17

31 Jan '17

ffmpeg | branch: master | James Almer <jamrial(a)gmail.com> | Tue Jan 31 14:49:29 2017 -0300| [ca8a3978e57c7c8f6abab8547f47483e407469b7] | committer: James Almer Merge commit '1dfc3cf89d0eb026af28be46294b85d79499ffb5' * commit '1dfc3cf89d0eb026af28be46294b85d79499ffb5': x86: hpeldsp: Split off VP3-specific bits into a separate file Merged-by: James Almer <jamrial(a)gmail.com> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=ca8a3978e57c7c8f6… --- libavcodec/x86/Makefile | 2 + libavcodec/x86/hpeldsp.asm | 89 ------------------------------ libavcodec/x86/hpeldsp.h | 4 ++ libavcodec/x86/hpeldsp_init.c | 25 ++------- libavcodec/x86/hpeldsp_vp3.asm | 111 ++++++++++++++++++++++++++++++++++++++ libavcodec/x86/hpeldsp_vp3_init.c | 56 +++++++++++++++++++ 6 files changed, 176 insertions(+), 111 deletions(-) diff --git a/libavcodec/x86/Makefile b/libavcodec/x86/Makefile index 2f0354a..2864952 100644 --- a/libavcodec/x86/Makefile +++ b/libavcodec/x86/Makefile @@ -67,6 +67,7 @@ OBJS-$(CONFIG_TTA_ENCODER) += x86/ttaencdsp_init.o OBJS-$(CONFIG_V210_DECODER) += x86/v210-init.o OBJS-$(CONFIG_V210_ENCODER) += x86/v210enc_init.o OBJS-$(CONFIG_VORBIS_DECODER) += x86/vorbisdsp_init.o +OBJS-$(CONFIG_VP3_DECODER) += x86/hpeldsp_vp3_init.o OBJS-$(CONFIG_VP6_DECODER) += x86/vp6dsp_init.o OBJS-$(CONFIG_VP9_DECODER) += x86/vp9dsp_init.o \ x86/vp9dsp_init_10bpp.o \ @@ -169,6 +170,7 @@ YASM-OBJS-$(CONFIG_TTA_ENCODER) += x86/ttaencdsp.o YASM-OBJS-$(CONFIG_V210_ENCODER) += x86/v210enc.o YASM-OBJS-$(CONFIG_V210_DECODER) += x86/v210.o YASM-OBJS-$(CONFIG_VORBIS_DECODER) += x86/vorbisdsp.o +YASM-OBJS-$(CONFIG_VP3_DECODER) += x86/hpeldsp_vp3.o YASM-OBJS-$(CONFIG_VP6_DECODER) += x86/vp6dsp.o YASM-OBJS-$(CONFIG_VP9_DECODER) += x86/vp9intrapred.o \ x86/vp9intrapred_16bpp.o \ diff --git a/libavcodec/x86/hpeldsp.asm b/libavcodec/x86/hpeldsp.asm index 82fb893..ce5d7a4 100644 --- a/libavcodec/x86/hpeldsp.asm +++ b/libavcodec/x86/hpeldsp.asm @@ -175,53 +175,6 @@ INIT_MMX 3dnow PUT_NO_RND_PIXELS8_X2 -; void ff_put_no_rnd_pixels8_x2_exact(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h) -%macro PUT_NO_RND_PIXELS8_X2_EXACT 0 -cglobal put_no_rnd_pixels8_x2_exact, 4,5 - lea r4, [r2*3] - pcmpeqb m6, m6 -.loop: - mova m0, [r1] - mova m2, [r1+r2] - mova m1, [r1+1] - mova m3, [r1+r2+1] - pxor m0, m6 - pxor m2, m6 - pxor m1, m6 - pxor m3, m6 - PAVGB m0, m1 - PAVGB m2, m3 - pxor m0, m6 - pxor m2, m6 - mova [r0], m0 - mova [r0+r2], m2 - mova m0, [r1+r2*2] - mova m1, [r1+r2*2+1] - mova m2, [r1+r4] - mova m3, [r1+r4+1] - pxor m0, m6 - pxor m1, m6 - pxor m2, m6 - pxor m3, m6 - PAVGB m0, m1 - PAVGB m2, m3 - pxor m0, m6 - pxor m2, m6 - mova [r0+r2*2], m0 - mova [r0+r4], m2 - lea r1, [r1+r2*4] - lea r0, [r0+r2*4] - sub r3d, 4 - jg .loop - REP_RET -%endmacro - -INIT_MMX mmxext -PUT_NO_RND_PIXELS8_X2_EXACT -INIT_MMX 3dnow -PUT_NO_RND_PIXELS8_X2_EXACT - - ; void ff_put_pixels8_y2(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h) %macro PUT_PIXELS8_Y2 0 %if cpuflag(sse2) @@ -300,48 +253,6 @@ INIT_MMX 3dnow PUT_NO_RND_PIXELS8_Y2 -; void ff_put_no_rnd_pixels8_y2_exact(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h) -%macro PUT_NO_RND_PIXELS8_Y2_EXACT 0 -cglobal put_no_rnd_pixels8_y2_exact, 4,5 - lea r4, [r2*3] - mova m0, [r1] - pcmpeqb m6, m6 - add r1, r2 - pxor m0, m6 -.loop: - mova m1, [r1] - mova m2, [r1+r2] - pxor m1, m6 - pxor m2, m6 - PAVGB m0, m1 - PAVGB m1, m2 - pxor m0, m6 - pxor m1, m6 - mova [r0], m0 - mova [r0+r2], m1 - mova m1, [r1+r2*2] - mova m0, [r1+r4] - pxor m1, m6 - pxor m0, m6 - PAVGB m2, m1 - PAVGB m1, m0 - pxor m2, m6 - pxor m1, m6 - mova [r0+r2*2], m2 - mova [r0+r4], m1 - lea r1, [r1+r2*4] - lea r0, [r0+r2*4] - sub r3d, 4 - jg .loop - REP_RET -%endmacro - -INIT_MMX mmxext -PUT_NO_RND_PIXELS8_Y2_EXACT -INIT_MMX 3dnow -PUT_NO_RND_PIXELS8_Y2_EXACT - - ; void ff_avg_pixels8(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h) %macro AVG_PIXELS8 0 cglobal avg_pixels8, 4,5 diff --git a/libavcodec/x86/hpeldsp.h b/libavcodec/x86/hpeldsp.h index 5fae990..bf97029 100644 --- a/libavcodec/x86/hpeldsp.h +++ b/libavcodec/x86/hpeldsp.h @@ -22,6 +22,8 @@ #include <stddef.h> #include <stdint.h> +#include "libavcodec/hpeldsp.h" + void ff_avg_pixels8_x2_mmx(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h); @@ -50,4 +52,6 @@ void ff_put_pixels16_xy2_sse2(uint8_t *block, const uint8_t *pixels, void ff_put_pixels16_xy2_ssse3(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h); +void ff_hpeldsp_vp3_init_x86(HpelDSPContext *c, int cpu_flags, int flags); + #endif /* AVCODEC_X86_HPELDSP_H */ diff --git a/libavcodec/x86/hpeldsp_init.c b/libavcodec/x86/hpeldsp_init.c index f1ba4be..e8da184 100644 --- a/libavcodec/x86/hpeldsp_init.c +++ b/libavcodec/x86/hpeldsp_init.c @@ -51,12 +51,6 @@ void ff_put_no_rnd_pixels8_x2_mmxext(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h); void ff_put_no_rnd_pixels8_x2_3dnow(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h); -void ff_put_no_rnd_pixels8_x2_exact_mmxext(uint8_t *block, - const uint8_t *pixels, - ptrdiff_t line_size, int h); -void ff_put_no_rnd_pixels8_x2_exact_3dnow(uint8_t *block, - const uint8_t *pixels, - ptrdiff_t line_size, int h); void ff_put_pixels8_y2_mmxext(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h); void ff_put_pixels8_y2_3dnow(uint8_t *block, const uint8_t *pixels, @@ -65,12 +59,6 @@ void ff_put_no_rnd_pixels8_y2_mmxext(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h); void ff_put_no_rnd_pixels8_y2_3dnow(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h); -void ff_put_no_rnd_pixels8_y2_exact_mmxext(uint8_t *block, - const uint8_t *pixels, - ptrdiff_t line_size, int h); -void ff_put_no_rnd_pixels8_y2_exact_3dnow(uint8_t *block, - const uint8_t *pixels, - ptrdiff_t line_size, int h); void ff_avg_pixels8_3dnow(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h); void ff_avg_pixels8_x2_mmxext(uint8_t *block, const uint8_t *pixels, @@ -242,11 +230,6 @@ static void hpeldsp_init_mmxext(HpelDSPContext *c, int flags, int cpu_flags) c->avg_pixels_tab[0][3] = avg_approx_pixels16_xy2_mmxext; c->avg_pixels_tab[1][3] = ff_avg_approx_pixels8_xy2_mmxext; } - - if (CONFIG_VP3_DECODER && flags & AV_CODEC_FLAG_BITEXACT) { - c->put_no_rnd_pixels_tab[1][1] = ff_put_no_rnd_pixels8_x2_exact_mmxext; - c->put_no_rnd_pixels_tab[1][2] = ff_put_no_rnd_pixels8_y2_exact_mmxext; - } #endif /* HAVE_MMXEXT_EXTERNAL */ } @@ -278,11 +261,6 @@ static void hpeldsp_init_3dnow(HpelDSPContext *c, int flags, int cpu_flags) c->avg_pixels_tab[0][3] = avg_approx_pixels16_xy2_3dnow; c->avg_pixels_tab[1][3] = ff_avg_approx_pixels8_xy2_3dnow; } - - if (CONFIG_VP3_DECODER && flags & AV_CODEC_FLAG_BITEXACT) { - c->put_no_rnd_pixels_tab[1][1] = ff_put_no_rnd_pixels8_x2_exact_3dnow; - c->put_no_rnd_pixels_tab[1][2] = ff_put_no_rnd_pixels8_y2_exact_3dnow; - } #endif /* HAVE_AMD3DNOW_EXTERNAL */ } @@ -332,4 +310,7 @@ av_cold void ff_hpeldsp_init_x86(HpelDSPContext *c, int flags) if (EXTERNAL_SSSE3(cpu_flags)) hpeldsp_init_ssse3(c, flags, cpu_flags); + + if (CONFIG_VP3_DECODER) + ff_hpeldsp_vp3_init_x86(c, cpu_flags, flags); } diff --git a/libavcodec/x86/hpeldsp_vp3.asm b/libavcodec/x86/hpeldsp_vp3.asm new file mode 100644 index 0000000..cba96d0 --- /dev/null +++ b/libavcodec/x86/hpeldsp_vp3.asm @@ -0,0 +1,111 @@ +;****************************************************************************** +;* SIMD-optimized halfpel functions for VP3 +;* +;* 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 .text + +; void ff_put_no_rnd_pixels8_x2_exact(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h) +%macro PUT_NO_RND_PIXELS8_X2_EXACT 0 +cglobal put_no_rnd_pixels8_x2_exact, 4,5 + lea r4, [r2*3] + pcmpeqb m6, m6 +.loop: + mova m0, [r1] + mova m2, [r1+r2] + mova m1, [r1+1] + mova m3, [r1+r2+1] + pxor m0, m6 + pxor m2, m6 + pxor m1, m6 + pxor m3, m6 + PAVGB m0, m1 + PAVGB m2, m3 + pxor m0, m6 + pxor m2, m6 + mova [r0], m0 + mova [r0+r2], m2 + mova m0, [r1+r2*2] + mova m1, [r1+r2*2+1] + mova m2, [r1+r4] + mova m3, [r1+r4+1] + pxor m0, m6 + pxor m1, m6 + pxor m2, m6 + pxor m3, m6 + PAVGB m0, m1 + PAVGB m2, m3 + pxor m0, m6 + pxor m2, m6 + mova [r0+r2*2], m0 + mova [r0+r4], m2 + lea r1, [r1+r2*4] + lea r0, [r0+r2*4] + sub r3d, 4 + jg .loop + REP_RET +%endmacro + +INIT_MMX mmxext +PUT_NO_RND_PIXELS8_X2_EXACT +INIT_MMX 3dnow +PUT_NO_RND_PIXELS8_X2_EXACT + + +; void ff_put_no_rnd_pixels8_y2_exact(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h) +%macro PUT_NO_RND_PIXELS8_Y2_EXACT 0 +cglobal put_no_rnd_pixels8_y2_exact, 4,5 + lea r4, [r2*3] + mova m0, [r1] + pcmpeqb m6, m6 + add r1, r2 + pxor m0, m6 +.loop: + mova m1, [r1] + mova m2, [r1+r2] + pxor m1, m6 + pxor m2, m6 + PAVGB m0, m1 + PAVGB m1, m2 + pxor m0, m6 + pxor m1, m6 + mova [r0], m0 + mova [r0+r2], m1 + mova m1, [r1+r2*2] + mova m0, [r1+r4] + pxor m1, m6 + pxor m0, m6 + PAVGB m2, m1 + PAVGB m1, m0 + pxor m2, m6 + pxor m1, m6 + mova [r0+r2*2], m2 + mova [r0+r4], m1 + lea r1, [r1+r2*4] + lea r0, [r0+r2*4] + sub r3d, 4 + jg .loop + REP_RET +%endmacro + +INIT_MMX mmxext +PUT_NO_RND_PIXELS8_Y2_EXACT +INIT_MMX 3dnow +PUT_NO_RND_PIXELS8_Y2_EXACT diff --git a/libavcodec/x86/hpeldsp_vp3_init.c b/libavcodec/x86/hpeldsp_vp3_init.c new file mode 100644 index 0000000..5979f41 --- /dev/null +++ b/libavcodec/x86/hpeldsp_vp3_init.c @@ -0,0 +1,56 @@ +/* + * 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/attributes.h" +#include "libavutil/cpu.h" +#include "libavutil/x86/cpu.h" + +#include "libavcodec/avcodec.h" +#include "libavcodec/hpeldsp.h" + +#include "hpeldsp.h" + +void ff_put_no_rnd_pixels8_x2_exact_mmxext(uint8_t *block, + const uint8_t *pixels, + ptrdiff_t line_size, int h); +void ff_put_no_rnd_pixels8_x2_exact_3dnow(uint8_t *block, + const uint8_t *pixels, + ptrdiff_t line_size, int h); +void ff_put_no_rnd_pixels8_y2_exact_mmxext(uint8_t *block, + const uint8_t *pixels, + ptrdiff_t line_size, int h); +void ff_put_no_rnd_pixels8_y2_exact_3dnow(uint8_t *block, + const uint8_t *pixels, + ptrdiff_t line_size, int h); + +av_cold void ff_hpeldsp_vp3_init_x86(HpelDSPContext *c, int cpu_flags, int flags) +{ + if (EXTERNAL_AMD3DNOW(cpu_flags)) { + if (flags & AV_CODEC_FLAG_BITEXACT) { + c->put_no_rnd_pixels_tab[1][1] = ff_put_no_rnd_pixels8_x2_exact_3dnow; + c->put_no_rnd_pixels_tab[1][2] = ff_put_no_rnd_pixels8_y2_exact_3dnow; + } + } + + if (EXTERNAL_MMXEXT(cpu_flags)) { + if (flags & AV_CODEC_FLAG_BITEXACT) { + c->put_no_rnd_pixels_tab[1][1] = ff_put_no_rnd_pixels8_x2_exact_mmxext; + c->put_no_rnd_pixels_tab[1][2] = ff_put_no_rnd_pixels8_y2_exact_mmxext; + } + } +} ====================================================================== diff --cc libavcodec/x86/Makefile index 2f0354a,3208699..2864952 --- a/libavcodec/x86/Makefile +++ b/libavcodec/x86/Makefile @@@ -57,22 -46,14 +57,23 @@@ OBJS-$(CONFIG_MLP_DECODER) OBJS-$(CONFIG_MPEG4_DECODER) += x86/xvididct_init.o OBJS-$(CONFIG_PNG_DECODER) += x86/pngdsp_init.o OBJS-$(CONFIG_PRORES_DECODER) += x86/proresdsp_init.o +OBJS-$(CONFIG_PRORES_LGPL_DECODER) += x86/proresdsp_init.o OBJS-$(CONFIG_RV40_DECODER) += x86/rv40dsp_init.o -OBJS-$(CONFIG_SVQ1_ENCODER) += x86/svq1enc.o -OBJS-$(CONFIG_TRUEHD_DECODER) += x86/mlpdsp.o +OBJS-$(CONFIG_SVQ1_ENCODER) += x86/svq1enc_init.o +OBJS-$(CONFIG_TAK_DECODER) += x86/takdsp_init.o +OBJS-$(CONFIG_TRUEHD_DECODER) += x86/mlpdsp_init.o +OBJS-$(CONFIG_TTA_DECODER) += x86/ttadsp_init.o +OBJS-$(CONFIG_TTA_ENCODER) += x86/ttaencdsp_init.o +OBJS-$(CONFIG_V210_DECODER) += x86/v210-init.o OBJS-$(CONFIG_V210_ENCODER) += x86/v210enc_init.o OBJS-$(CONFIG_VORBIS_DECODER) += x86/vorbisdsp_init.o + OBJS-$(CONFIG_VP3_DECODER) += x86/hpeldsp_vp3_init.o OBJS-$(CONFIG_VP6_DECODER) += x86/vp6dsp_init.o -OBJS-$(CONFIG_VP9_DECODER) += x86/vp9dsp_init.o +OBJS-$(CONFIG_VP9_DECODER) += x86/vp9dsp_init.o \ + x86/vp9dsp_init_10bpp.o \ + x86/vp9dsp_init_12bpp.o \ + x86/vp9dsp_init_16bpp.o +OBJS-$(CONFIG_WEBP_DECODER) += x86/vp8dsp_init.o # GCC inline assembly optimizations @@@ -133,49 -111,18 +134,50 @@@ YASM-OBJS-$(CONFIG_VP8DSP) x86/vp8dsp_loopfilter.o # decoders/encoders -YASM-OBJS-$(CONFIG_AAC_DECODER) += x86/sbrdsp.o -YASM-OBJS-$(CONFIG_APE_DECODER) += x86/apedsp.o -YASM-OBJS-$(CONFIG_DCA_DECODER) += x86/dcadsp.o +YASM-OBJS-$(CONFIG_AAC_DECODER) += x86/aacpsdsp.o \ + x86/sbrdsp.o +YASM-OBJS-$(CONFIG_AAC_ENCODER) += x86/aacencdsp.o +YASM-OBJS-$(CONFIG_ADPCM_G722_DECODER) += x86/g722dsp.o +YASM-OBJS-$(CONFIG_ADPCM_G722_ENCODER) += x86/g722dsp.o +YASM-OBJS-$(CONFIG_ALAC_DECODER) += x86/alacdsp.o +YASM-OBJS-$(CONFIG_APNG_DECODER) += x86/pngdsp.o +YASM-OBJS-$(CONFIG_DCA_DECODER) += x86/dcadsp.o x86/synth_filter.o +YASM-OBJS-$(CONFIG_DIRAC_DECODER) += x86/diracdsp.o \ + x86/dirac_dwt.o YASM-OBJS-$(CONFIG_DNXHD_ENCODER) += x86/dnxhdenc.o -YASM-OBJS-$(CONFIG_HEVC_DECODER) += x86/hevc_deblock.o \ - x86/hevc_mc.o \ - x86/hevc_idct.o +YASM-OBJS-$(CONFIG_FLAC_DECODER) += x86/flacdsp.o +ifdef CONFIG_GPL +YASM-OBJS-$(CONFIG_FLAC_ENCODER) += x86/flac_dsp_gpl.o +endif +YASM-OBJS-$(CONFIG_HEVC_DECODER) += x86/hevc_mc.o \ + x86/hevc_deblock.o \ + x86/hevc_idct.o \ + x86/hevc_res_add.o \ + x86/hevc_sao.o \ + x86/hevc_sao_10bit.o +YASM-OBJS-$(CONFIG_JPEG2000_DECODER) += x86/jpeg2000dsp.o +YASM-OBJS-$(CONFIG_MLP_DECODER) += x86/mlpdsp.o +YASM-OBJS-$(CONFIG_MPEG4_DECODER) += x86/xvididct.o YASM-OBJS-$(CONFIG_PNG_DECODER) += x86/pngdsp.o YASM-OBJS-$(CONFIG_PRORES_DECODER) += x86/proresdsp.o +YASM-OBJS-$(CONFIG_PRORES_LGPL_DECODER) += x86/proresdsp.o YASM-OBJS-$(CONFIG_RV40_DECODER) += x86/rv40dsp.o +YASM-OBJS-$(CONFIG_SVQ1_ENCODER) += x86/svq1enc.o +YASM-OBJS-$(CONFIG_TAK_DECODER) += x86/takdsp.o +YASM-OBJS-$(CONFIG_TRUEHD_DECODER) += x86/mlpdsp.o +YASM-OBJS-$(CONFIG_TTA_DECODER) += x86/ttadsp.o +YASM-OBJS-$(CONFIG_TTA_ENCODER) += x86/ttaencdsp.o YASM-OBJS-$(CONFIG_V210_ENCODER) += x86/v210enc.o +YASM-OBJS-$(CONFIG_V210_DECODER) += x86/v210.o YASM-OBJS-$(CONFIG_VORBIS_DECODER) += x86/vorbisdsp.o + YASM-OBJS-$(CONFIG_VP3_DECODER) += x86/hpeldsp_vp3.o YASM-OBJS-$(CONFIG_VP6_DECODER) += x86/vp6dsp.o -YASM-OBJS-$(CONFIG_VP9_DECODER) += x86/vp9dsp.o +YASM-OBJS-$(CONFIG_VP9_DECODER) += x86/vp9intrapred.o \ + x86/vp9intrapred_16bpp.o \ + x86/vp9itxfm.o \ + x86/vp9itxfm_16bpp.o \ + x86/vp9lpf.o \ + x86/vp9lpf_16bpp.o \ + x86/vp9mc.o \ + x86/vp9mc_16bpp.o +YASM-OBJS-$(CONFIG_WEBP_DECODER) += x86/vp8dsp.o diff --cc libavcodec/x86/hpeldsp.asm index 82fb893,8e21114..ce5d7a4 --- a/libavcodec/x86/hpeldsp.asm +++ b/libavcodec/x86/hpeldsp.asm @@@ -175,66 -142,15 +175,19 @@@ INIT_MMX 3dno PUT_NO_RND_PIXELS8_X2 - ; void ff_put_no_rnd_pixels8_x2_exact(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h) - %macro PUT_NO_RND_PIXELS8_X2_EXACT 0 - cglobal put_no_rnd_pixels8_x2_exact, 4,5 - lea r4, [r2*3] - pcmpeqb m6, m6 - .loop: - mova m0, [r1] - mova m2, [r1+r2] - mova m1, [r1+1] - mova m3, [r1+r2+1] - pxor m0, m6 - pxor m2, m6 - pxor m1, m6 - pxor m3, m6 - PAVGB m0, m1 - PAVGB m2, m3 - pxor m0, m6 - pxor m2, m6 - mova [r0], m0 - mova [r0+r2], m2 - mova m0, [r1+r2*2] - mova m1, [r1+r2*2+1] - mova m2, [r1+r4] - mova m3, [r1+r4+1] - pxor m0, m6 - pxor m1, m6 - pxor m2, m6 - pxor m3, m6 - PAVGB m0, m1 - PAVGB m2, m3 - pxor m0, m6 - pxor m2, m6 - mova [r0+r2*2], m0 - mova [r0+r4], m2 - lea r1, [r1+r2*4] - lea r0, [r0+r2*4] - sub r3d, 4 - jg .loop - REP_RET - %endmacro - - INIT_MMX mmxext - PUT_NO_RND_PIXELS8_X2_EXACT - INIT_MMX 3dnow - PUT_NO_RND_PIXELS8_X2_EXACT - - ; void ff_put_pixels8_y2(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h) %macro PUT_PIXELS8_Y2 0 +%if cpuflag(sse2) +cglobal put_pixels16_y2, 4,5,3 +%else cglobal put_pixels8_y2, 4,5 +%endif lea r4, [r2*2] - mova m0, [r1] + movu m0, [r1] sub r0, r2 .loop: - mova m1, [r1+r2] - mova m2, [r1+r4] + movu m1, [r1+r2] + movu m2, [r1+r4] add r1, r4 PAVGB m0, m1 PAVGB m1, m2 diff --cc libavcodec/x86/hpeldsp.h index 5fae990,d624ed9..bf97029 --- a/libavcodec/x86/hpeldsp.h +++ b/libavcodec/x86/hpeldsp.h @@@ -41,13 -34,9 +43,15 @@@ void ff_avg_pixels16_xy2_ssse3(uint8_t void ff_put_pixels8_xy2_mmx(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h); +void ff_put_pixels8_xy2_ssse3(uint8_t *block, const uint8_t *pixels, + ptrdiff_t line_size, int h); void ff_put_pixels16_xy2_mmx(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h); +void ff_put_pixels16_xy2_sse2(uint8_t *block, const uint8_t *pixels, + ptrdiff_t line_size, int h); +void ff_put_pixels16_xy2_ssse3(uint8_t *block, const uint8_t *pixels, + ptrdiff_t line_size, int h); + void ff_hpeldsp_vp3_init_x86(HpelDSPContext *c, int cpu_flags, int flags); + #endif /* AVCODEC_X86_HPELDSP_H */ diff --cc libavcodec/x86/hpeldsp_init.c index f1ba4be,6731428..e8da184 --- a/libavcodec/x86/hpeldsp_init.c +++ b/libavcodec/x86/hpeldsp_init.c @@@ -239,14 -194,9 +227,9 @@@ static void hpeldsp_init_mmxext(HpelDSP c->put_no_rnd_pixels_tab[1][1] = ff_put_no_rnd_pixels8_x2_mmxext; c->put_no_rnd_pixels_tab[1][2] = ff_put_no_rnd_pixels8_y2_mmxext; - c->avg_pixels_tab[0][3] = avg_pixels16_xy2_mmxext; - c->avg_pixels_tab[1][3] = ff_avg_pixels8_xy2_mmxext; + c->avg_pixels_tab[0][3] = avg_approx_pixels16_xy2_mmxext; + c->avg_pixels_tab[1][3] = ff_avg_approx_pixels8_xy2_mmxext; } - - if (CONFIG_VP3_DECODER && flags & AV_CODEC_FLAG_BITEXACT) { - c->put_no_rnd_pixels_tab[1][1] = ff_put_no_rnd_pixels8_x2_exact_mmxext; - c->put_no_rnd_pixels_tab[1][2] = ff_put_no_rnd_pixels8_y2_exact_mmxext; - } #endif /* HAVE_MMXEXT_EXTERNAL */ } @@@ -275,14 -223,9 +258,9 @@@ static void hpeldsp_init_3dnow(HpelDSPC c->put_no_rnd_pixels_tab[1][1] = ff_put_no_rnd_pixels8_x2_3dnow; c->put_no_rnd_pixels_tab[1][2] = ff_put_no_rnd_pixels8_y2_3dnow; - c->avg_pixels_tab[0][3] = avg_pixels16_xy2_3dnow; - c->avg_pixels_tab[1][3] = ff_avg_pixels8_xy2_3dnow; + c->avg_pixels_tab[0][3] = avg_approx_pixels16_xy2_3dnow; + c->avg_pixels_tab[1][3] = ff_avg_approx_pixels8_xy2_3dnow; } - - if (CONFIG_VP3_DECODER && flags & AV_CODEC_FLAG_BITEXACT) { - c->put_no_rnd_pixels_tab[1][1] = ff_put_no_rnd_pixels8_x2_exact_3dnow; - c->put_no_rnd_pixels_tab[1][2] = ff_put_no_rnd_pixels8_y2_exact_3dnow; - } #endif /* HAVE_AMD3DNOW_EXTERNAL */ } @@@ -330,6 -257,6 +308,9 @@@ av_cold void ff_hpeldsp_init_x86(HpelDS if (EXTERNAL_SSE2(cpu_flags)) hpeldsp_init_sse2(c, flags, cpu_flags); + if (EXTERNAL_SSSE3(cpu_flags)) + hpeldsp_init_ssse3(c, flags, cpu_flags); ++ + if (CONFIG_VP3_DECODER) + ff_hpeldsp_vp3_init_x86(c, cpu_flags, flags); } diff --cc libavcodec/x86/hpeldsp_vp3.asm index 0000000,513f14e..cba96d0 mode 000000,100644..100644 --- a/libavcodec/x86/hpeldsp_vp3.asm +++ b/libavcodec/x86/hpeldsp_vp3.asm @@@ -1,0 -1,111 +1,111 @@@ + ;****************************************************************************** + ;* SIMD-optimized halfpel functions for VP3 + ;* -;* This file is part of Libav. ++;* This file is part of FFmpeg. + ;* -;* Libav is free software; you can redistribute it and/or ++;* 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. + ;* -;* Libav is distributed in the hope that it will be useful, ++;* 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 Libav; if not, write to the Free Software ++;* 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 .text + + ; void ff_put_no_rnd_pixels8_x2_exact(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h) + %macro PUT_NO_RND_PIXELS8_X2_EXACT 0 + cglobal put_no_rnd_pixels8_x2_exact, 4,5 + lea r4, [r2*3] + pcmpeqb m6, m6 + .loop: + mova m0, [r1] + mova m2, [r1+r2] + mova m1, [r1+1] + mova m3, [r1+r2+1] + pxor m0, m6 + pxor m2, m6 + pxor m1, m6 + pxor m3, m6 + PAVGB m0, m1 + PAVGB m2, m3 + pxor m0, m6 + pxor m2, m6 + mova [r0], m0 + mova [r0+r2], m2 + mova m0, [r1+r2*2] + mova m1, [r1+r2*2+1] + mova m2, [r1+r4] + mova m3, [r1+r4+1] + pxor m0, m6 + pxor m1, m6 + pxor m2, m6 + pxor m3, m6 + PAVGB m0, m1 + PAVGB m2, m3 + pxor m0, m6 + pxor m2, m6 + mova [r0+r2*2], m0 + mova [r0+r4], m2 + lea r1, [r1+r2*4] + lea r0, [r0+r2*4] + sub r3d, 4 + jg .loop + REP_RET + %endmacro + + INIT_MMX mmxext + PUT_NO_RND_PIXELS8_X2_EXACT + INIT_MMX 3dnow + PUT_NO_RND_PIXELS8_X2_EXACT + + + ; void ff_put_no_rnd_pixels8_y2_exact(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h) + %macro PUT_NO_RND_PIXELS8_Y2_EXACT 0 + cglobal put_no_rnd_pixels8_y2_exact, 4,5 + lea r4, [r2*3] + mova m0, [r1] + pcmpeqb m6, m6 + add r1, r2 + pxor m0, m6 + .loop: + mova m1, [r1] + mova m2, [r1+r2] + pxor m1, m6 + pxor m2, m6 + PAVGB m0, m1 + PAVGB m1, m2 + pxor m0, m6 + pxor m1, m6 + mova [r0], m0 + mova [r0+r2], m1 + mova m1, [r1+r2*2] + mova m0, [r1+r4] + pxor m1, m6 + pxor m0, m6 + PAVGB m2, m1 + PAVGB m1, m0 + pxor m2, m6 + pxor m1, m6 + mova [r0+r2*2], m2 + mova [r0+r4], m1 + lea r1, [r1+r2*4] + lea r0, [r0+r2*4] + sub r3d, 4 + jg .loop + REP_RET + %endmacro + + INIT_MMX mmxext + PUT_NO_RND_PIXELS8_Y2_EXACT + INIT_MMX 3dnow + PUT_NO_RND_PIXELS8_Y2_EXACT diff --cc libavcodec/x86/hpeldsp_vp3_init.c index 0000000,2510c11..5979f41 mode 000000,100644..100644 --- a/libavcodec/x86/hpeldsp_vp3_init.c +++ b/libavcodec/x86/hpeldsp_vp3_init.c @@@ -1,0 -1,56 +1,56 @@@ + /* - * This file is part of Libav. ++ * This file is part of FFmpeg. + * - * Libav is free software; you can redistribute it and/or ++ * 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. + * - * Libav is distributed in the hope that it will be useful, ++ * 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 Libav; if not, write to the Free Software ++ * 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/attributes.h" + #include "libavutil/cpu.h" + #include "libavutil/x86/cpu.h" + + #include "libavcodec/avcodec.h" + #include "libavcodec/hpeldsp.h" + + #include "hpeldsp.h" + + void ff_put_no_rnd_pixels8_x2_exact_mmxext(uint8_t *block, + const uint8_t *pixels, + ptrdiff_t line_size, int h); + void ff_put_no_rnd_pixels8_x2_exact_3dnow(uint8_t *block, + const uint8_t *pixels, + ptrdiff_t line_size, int h); + void ff_put_no_rnd_pixels8_y2_exact_mmxext(uint8_t *block, + const uint8_t *pixels, + ptrdiff_t line_size, int h); + void ff_put_no_rnd_pixels8_y2_exact_3dnow(uint8_t *block, + const uint8_t *pixels, + ptrdiff_t line_size, int h); + + av_cold void ff_hpeldsp_vp3_init_x86(HpelDSPContext *c, int cpu_flags, int flags) + { + if (EXTERNAL_AMD3DNOW(cpu_flags)) { + if (flags & AV_CODEC_FLAG_BITEXACT) { + c->put_no_rnd_pixels_tab[1][1] = ff_put_no_rnd_pixels8_x2_exact_3dnow; + c->put_no_rnd_pixels_tab[1][2] = ff_put_no_rnd_pixels8_y2_exact_3dnow; + } + } + + if (EXTERNAL_MMXEXT(cpu_flags)) { + if (flags & AV_CODEC_FLAG_BITEXACT) { + c->put_no_rnd_pixels_tab[1][1] = ff_put_no_rnd_pixels8_x2_exact_mmxext; + c->put_no_rnd_pixels_tab[1][2] = ff_put_no_rnd_pixels8_y2_exact_mmxext; + } + } + }

1 0

x86: hpeldsp: Split off VP3-specific bits into a separate file
by git＠videolan.org 31 Jan '17

31 Jan '17

ffmpeg | branch: master | Diego Biurrun <diego(a)biurrun.de> | Mon Jan 11 15:53:41 2016 +0100| [1dfc3cf89d0eb026af28be46294b85d79499ffb5] | committer: Diego Biurrun x86: hpeldsp: Split off VP3-specific bits into a separate file > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=1dfc3cf89d0eb026a… --- libavcodec/x86/Makefile | 2 + libavcodec/x86/hpeldsp.asm | 89 ------------------------------ libavcodec/x86/hpeldsp.h | 4 ++ libavcodec/x86/hpeldsp_init.c | 25 ++------- libavcodec/x86/hpeldsp_vp3.asm | 111 ++++++++++++++++++++++++++++++++++++++ libavcodec/x86/hpeldsp_vp3_init.c | 56 +++++++++++++++++++ 6 files changed, 176 insertions(+), 111 deletions(-) diff --git a/libavcodec/x86/Makefile b/libavcodec/x86/Makefile index 1460197..3208699 100644 --- a/libavcodec/x86/Makefile +++ b/libavcodec/x86/Makefile @@ -51,6 +51,7 @@ OBJS-$(CONFIG_SVQ1_ENCODER) += x86/svq1enc.o OBJS-$(CONFIG_TRUEHD_DECODER) += x86/mlpdsp.o OBJS-$(CONFIG_V210_ENCODER) += x86/v210enc_init.o OBJS-$(CONFIG_VORBIS_DECODER) += x86/vorbisdsp_init.o +OBJS-$(CONFIG_VP3_DECODER) += x86/hpeldsp_vp3_init.o OBJS-$(CONFIG_VP6_DECODER) += x86/vp6dsp_init.o OBJS-$(CONFIG_VP9_DECODER) += x86/vp9dsp_init.o @@ -122,5 +123,6 @@ YASM-OBJS-$(CONFIG_PRORES_DECODER) += x86/proresdsp.o YASM-OBJS-$(CONFIG_RV40_DECODER) += x86/rv40dsp.o YASM-OBJS-$(CONFIG_V210_ENCODER) += x86/v210enc.o YASM-OBJS-$(CONFIG_VORBIS_DECODER) += x86/vorbisdsp.o +YASM-OBJS-$(CONFIG_VP3_DECODER) += x86/hpeldsp_vp3.o YASM-OBJS-$(CONFIG_VP6_DECODER) += x86/vp6dsp.o YASM-OBJS-$(CONFIG_VP9_DECODER) += x86/vp9dsp.o diff --git a/libavcodec/x86/hpeldsp.asm b/libavcodec/x86/hpeldsp.asm index b8929b9..8e21114 100644 --- a/libavcodec/x86/hpeldsp.asm +++ b/libavcodec/x86/hpeldsp.asm @@ -142,53 +142,6 @@ INIT_MMX 3dnow PUT_NO_RND_PIXELS8_X2 -; void ff_put_no_rnd_pixels8_x2_exact(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h) -%macro PUT_NO_RND_PIXELS8_X2_EXACT 0 -cglobal put_no_rnd_pixels8_x2_exact, 4,5 - lea r4, [r2*3] - pcmpeqb m6, m6 -.loop: - mova m0, [r1] - mova m2, [r1+r2] - mova m1, [r1+1] - mova m3, [r1+r2+1] - pxor m0, m6 - pxor m2, m6 - pxor m1, m6 - pxor m3, m6 - PAVGB m0, m1 - PAVGB m2, m3 - pxor m0, m6 - pxor m2, m6 - mova [r0], m0 - mova [r0+r2], m2 - mova m0, [r1+r2*2] - mova m1, [r1+r2*2+1] - mova m2, [r1+r4] - mova m3, [r1+r4+1] - pxor m0, m6 - pxor m1, m6 - pxor m2, m6 - pxor m3, m6 - PAVGB m0, m1 - PAVGB m2, m3 - pxor m0, m6 - pxor m2, m6 - mova [r0+r2*2], m0 - mova [r0+r4], m2 - lea r1, [r1+r2*4] - lea r0, [r0+r2*4] - sub r3d, 4 - jg .loop - REP_RET -%endmacro - -INIT_MMX mmxext -PUT_NO_RND_PIXELS8_X2_EXACT -INIT_MMX 3dnow -PUT_NO_RND_PIXELS8_X2_EXACT - - ; void ff_put_pixels8_y2(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h) %macro PUT_PIXELS8_Y2 0 cglobal put_pixels8_y2, 4,5 @@ -260,48 +213,6 @@ INIT_MMX 3dnow PUT_NO_RND_PIXELS8_Y2 -; void ff_put_no_rnd_pixels8_y2_exact(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h) -%macro PUT_NO_RND_PIXELS8_Y2_EXACT 0 -cglobal put_no_rnd_pixels8_y2_exact, 4,5 - lea r4, [r2*3] - mova m0, [r1] - pcmpeqb m6, m6 - add r1, r2 - pxor m0, m6 -.loop: - mova m1, [r1] - mova m2, [r1+r2] - pxor m1, m6 - pxor m2, m6 - PAVGB m0, m1 - PAVGB m1, m2 - pxor m0, m6 - pxor m1, m6 - mova [r0], m0 - mova [r0+r2], m1 - mova m1, [r1+r2*2] - mova m0, [r1+r4] - pxor m1, m6 - pxor m0, m6 - PAVGB m2, m1 - PAVGB m1, m0 - pxor m2, m6 - pxor m1, m6 - mova [r0+r2*2], m2 - mova [r0+r4], m1 - lea r1, [r1+r2*4] - lea r0, [r0+r2*4] - sub r3d, 4 - jg .loop - REP_RET -%endmacro - -INIT_MMX mmxext -PUT_NO_RND_PIXELS8_Y2_EXACT -INIT_MMX 3dnow -PUT_NO_RND_PIXELS8_Y2_EXACT - - ; void ff_avg_pixels8(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h) %macro AVG_PIXELS8 0 cglobal avg_pixels8, 4,5 diff --git a/libavcodec/x86/hpeldsp.h b/libavcodec/x86/hpeldsp.h index 47b0b8b..d624ed9 100644 --- a/libavcodec/x86/hpeldsp.h +++ b/libavcodec/x86/hpeldsp.h @@ -22,6 +22,8 @@ #include <stddef.h> #include <stdint.h> +#include "libavcodec/hpeldsp.h" + void ff_avg_pixels8_x2_mmx(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h); @@ -35,4 +37,6 @@ void ff_put_pixels8_xy2_mmx(uint8_t *block, const uint8_t *pixels, void ff_put_pixels16_xy2_mmx(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h); +void ff_hpeldsp_vp3_init_x86(HpelDSPContext *c, int cpu_flags, int flags); + #endif /* AVCODEC_X86_HPELDSP_H */ diff --git a/libavcodec/x86/hpeldsp_init.c b/libavcodec/x86/hpeldsp_init.c index 65aac71..6731428 100644 --- a/libavcodec/x86/hpeldsp_init.c +++ b/libavcodec/x86/hpeldsp_init.c @@ -43,12 +43,6 @@ void ff_put_no_rnd_pixels8_x2_mmxext(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h); void ff_put_no_rnd_pixels8_x2_3dnow(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h); -void ff_put_no_rnd_pixels8_x2_exact_mmxext(uint8_t *block, - const uint8_t *pixels, - ptrdiff_t line_size, int h); -void ff_put_no_rnd_pixels8_x2_exact_3dnow(uint8_t *block, - const uint8_t *pixels, - ptrdiff_t line_size, int h); void ff_put_pixels8_y2_mmxext(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h); void ff_put_pixels8_y2_3dnow(uint8_t *block, const uint8_t *pixels, @@ -57,12 +51,6 @@ void ff_put_no_rnd_pixels8_y2_mmxext(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h); void ff_put_no_rnd_pixels8_y2_3dnow(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h); -void ff_put_no_rnd_pixels8_y2_exact_mmxext(uint8_t *block, - const uint8_t *pixels, - ptrdiff_t line_size, int h); -void ff_put_no_rnd_pixels8_y2_exact_3dnow(uint8_t *block, - const uint8_t *pixels, - ptrdiff_t line_size, int h); void ff_avg_pixels8_3dnow(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h); void ff_avg_pixels8_x2_mmxext(uint8_t *block, const uint8_t *pixels, @@ -209,11 +197,6 @@ static void hpeldsp_init_mmxext(HpelDSPContext *c, int flags, int cpu_flags) c->avg_pixels_tab[0][3] = avg_pixels16_xy2_mmxext; c->avg_pixels_tab[1][3] = ff_avg_pixels8_xy2_mmxext; } - - if (flags & AV_CODEC_FLAG_BITEXACT && CONFIG_VP3_DECODER) { - c->put_no_rnd_pixels_tab[1][1] = ff_put_no_rnd_pixels8_x2_exact_mmxext; - c->put_no_rnd_pixels_tab[1][2] = ff_put_no_rnd_pixels8_y2_exact_mmxext; - } #endif /* HAVE_MMXEXT_EXTERNAL */ } @@ -243,11 +226,6 @@ static void hpeldsp_init_3dnow(HpelDSPContext *c, int flags, int cpu_flags) c->avg_pixels_tab[0][3] = avg_pixels16_xy2_3dnow; c->avg_pixels_tab[1][3] = ff_avg_pixels8_xy2_3dnow; } - - if (flags & AV_CODEC_FLAG_BITEXACT && CONFIG_VP3_DECODER) { - c->put_no_rnd_pixels_tab[1][1] = ff_put_no_rnd_pixels8_x2_exact_3dnow; - c->put_no_rnd_pixels_tab[1][2] = ff_put_no_rnd_pixels8_y2_exact_3dnow; - } #endif /* HAVE_AMD3DNOW_EXTERNAL */ } @@ -278,4 +256,7 @@ av_cold void ff_hpeldsp_init_x86(HpelDSPContext *c, int flags) if (EXTERNAL_SSE2(cpu_flags)) hpeldsp_init_sse2(c, flags, cpu_flags); + + if (CONFIG_VP3_DECODER) + ff_hpeldsp_vp3_init_x86(c, cpu_flags, flags); } diff --git a/libavcodec/x86/hpeldsp_vp3.asm b/libavcodec/x86/hpeldsp_vp3.asm new file mode 100644 index 0000000..513f14e --- /dev/null +++ b/libavcodec/x86/hpeldsp_vp3.asm @@ -0,0 +1,111 @@ +;****************************************************************************** +;* SIMD-optimized halfpel functions for VP3 +;* +;* This file is part of Libav. +;* +;* Libav 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. +;* +;* Libav 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 Libav; 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 .text + +; void ff_put_no_rnd_pixels8_x2_exact(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h) +%macro PUT_NO_RND_PIXELS8_X2_EXACT 0 +cglobal put_no_rnd_pixels8_x2_exact, 4,5 + lea r4, [r2*3] + pcmpeqb m6, m6 +.loop: + mova m0, [r1] + mova m2, [r1+r2] + mova m1, [r1+1] + mova m3, [r1+r2+1] + pxor m0, m6 + pxor m2, m6 + pxor m1, m6 + pxor m3, m6 + PAVGB m0, m1 + PAVGB m2, m3 + pxor m0, m6 + pxor m2, m6 + mova [r0], m0 + mova [r0+r2], m2 + mova m0, [r1+r2*2] + mova m1, [r1+r2*2+1] + mova m2, [r1+r4] + mova m3, [r1+r4+1] + pxor m0, m6 + pxor m1, m6 + pxor m2, m6 + pxor m3, m6 + PAVGB m0, m1 + PAVGB m2, m3 + pxor m0, m6 + pxor m2, m6 + mova [r0+r2*2], m0 + mova [r0+r4], m2 + lea r1, [r1+r2*4] + lea r0, [r0+r2*4] + sub r3d, 4 + jg .loop + REP_RET +%endmacro + +INIT_MMX mmxext +PUT_NO_RND_PIXELS8_X2_EXACT +INIT_MMX 3dnow +PUT_NO_RND_PIXELS8_X2_EXACT + + +; void ff_put_no_rnd_pixels8_y2_exact(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h) +%macro PUT_NO_RND_PIXELS8_Y2_EXACT 0 +cglobal put_no_rnd_pixels8_y2_exact, 4,5 + lea r4, [r2*3] + mova m0, [r1] + pcmpeqb m6, m6 + add r1, r2 + pxor m0, m6 +.loop: + mova m1, [r1] + mova m2, [r1+r2] + pxor m1, m6 + pxor m2, m6 + PAVGB m0, m1 + PAVGB m1, m2 + pxor m0, m6 + pxor m1, m6 + mova [r0], m0 + mova [r0+r2], m1 + mova m1, [r1+r2*2] + mova m0, [r1+r4] + pxor m1, m6 + pxor m0, m6 + PAVGB m2, m1 + PAVGB m1, m0 + pxor m2, m6 + pxor m1, m6 + mova [r0+r2*2], m2 + mova [r0+r4], m1 + lea r1, [r1+r2*4] + lea r0, [r0+r2*4] + sub r3d, 4 + jg .loop + REP_RET +%endmacro + +INIT_MMX mmxext +PUT_NO_RND_PIXELS8_Y2_EXACT +INIT_MMX 3dnow +PUT_NO_RND_PIXELS8_Y2_EXACT diff --git a/libavcodec/x86/hpeldsp_vp3_init.c b/libavcodec/x86/hpeldsp_vp3_init.c new file mode 100644 index 0000000..2510c11 --- /dev/null +++ b/libavcodec/x86/hpeldsp_vp3_init.c @@ -0,0 +1,56 @@ +/* + * This file is part of Libav. + * + * Libav 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. + * + * Libav 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 Libav; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ + +#include "libavutil/attributes.h" +#include "libavutil/cpu.h" +#include "libavutil/x86/cpu.h" + +#include "libavcodec/avcodec.h" +#include "libavcodec/hpeldsp.h" + +#include "hpeldsp.h" + +void ff_put_no_rnd_pixels8_x2_exact_mmxext(uint8_t *block, + const uint8_t *pixels, + ptrdiff_t line_size, int h); +void ff_put_no_rnd_pixels8_x2_exact_3dnow(uint8_t *block, + const uint8_t *pixels, + ptrdiff_t line_size, int h); +void ff_put_no_rnd_pixels8_y2_exact_mmxext(uint8_t *block, + const uint8_t *pixels, + ptrdiff_t line_size, int h); +void ff_put_no_rnd_pixels8_y2_exact_3dnow(uint8_t *block, + const uint8_t *pixels, + ptrdiff_t line_size, int h); + +av_cold void ff_hpeldsp_vp3_init_x86(HpelDSPContext *c, int cpu_flags, int flags) +{ + if (EXTERNAL_AMD3DNOW(cpu_flags)) { + if (flags & AV_CODEC_FLAG_BITEXACT) { + c->put_no_rnd_pixels_tab[1][1] = ff_put_no_rnd_pixels8_x2_exact_3dnow; + c->put_no_rnd_pixels_tab[1][2] = ff_put_no_rnd_pixels8_y2_exact_3dnow; + } + } + + if (EXTERNAL_MMXEXT(cpu_flags)) { + if (flags & AV_CODEC_FLAG_BITEXACT) { + c->put_no_rnd_pixels_tab[1][1] = ff_put_no_rnd_pixels8_x2_exact_mmxext; + c->put_no_rnd_pixels_tab[1][2] = ff_put_no_rnd_pixels8_y2_exact_mmxext; + } + } +}

1 0

MAINTAINERS: Add myself for boadec.c
by git＠videolan.org 31 Jan '17

31 Jan '17

ffmpeg | branch: master | Michael Niedermayer <michael(a)niedermayer.cc> | Mon Jan 30 01:45:30 2017 +0100| [481884080977f15854fe06e1c742a7741e49555c] | committer: Michael Niedermayer MAINTAINERS: Add myself for boadec.c It seems ive written this thing though i cannot really remember Signed-off-by: Michael Niedermayer <michael(a)niedermayer.cc> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=481884080977f1585… --- MAINTAINERS | 1 + 1 file changed, 1 insertion(+) diff --git a/MAINTAINERS b/MAINTAINERS index 981d71b..80721e1 100644 --- a/MAINTAINERS +++ b/MAINTAINERS @@ -384,6 +384,7 @@ Muxers/Demuxers: avisynth.c Stephen Hutchinson avr.c Paul B Mahol bink.c Peter Ross + boadec.c Michael Niedermayer brstm.c Paul B Mahol caf* Peter Ross cdxl.c Paul B Mahol

1 0