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

Merge commit '1f7b4f9abc6bae94e576e710b8d10117ca3c8238'
by git＠videolan.org 31 Jan '17

31 Jan '17

ffmpeg | branch: master | Clément Bœsch <cboesch(a)gopro.com> | Tue Jan 31 17:11:24 2017 +0100| [5f633c39cd3ec13ea97e0a7474d452fd5cff8243] | committer: Clément Bœsch Merge commit '1f7b4f9abc6bae94e576e710b8d10117ca3c8238' * commit '1f7b4f9abc6bae94e576e710b8d10117ca3c8238': h264dec: make sure not to call finish_setup() more than once per frame This commit is a noop, see bdbbb8f11edbf10add874508c5125c174d8939be Merged-by: Clément Bœsch <cboesch(a)gopro.com> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=5f633c39cd3ec13ea… ---

1 0

h264dec: make sure not to call finish_setup() more than once per frame
by git＠videolan.org 31 Jan '17

31 Jan '17

ffmpeg | branch: master | Anton Khirnov <anton(a)khirnov.net> | Wed Jul 20 09:59:45 2016 +0200| [1f7b4f9abc6bae94e576e710b8d10117ca3c8238] | committer: Anton Khirnov h264dec: make sure not to call finish_setup() more than once per frame It does not break anything currently, but should not be done anyway. > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=1f7b4f9abc6bae94e… --- libavcodec/h264dec.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/libavcodec/h264dec.c b/libavcodec/h264dec.c index 1086eab..faa502e 100644 --- a/libavcodec/h264dec.c +++ b/libavcodec/h264dec.c @@ -550,7 +550,7 @@ static int decode_nal_units(H264Context *h, const uint8_t *buf, int buf_size) break; if (avctx->active_thread_type & FF_THREAD_FRAME && !h->avctx->hwaccel && - i >= nals_needed) { + i >= nals_needed && !h->setup_finished) { ff_thread_finish_setup(avctx); h->setup_finished = 1; }

1 0

Merge commit 'da917fcf5183ed249ad1285b8edd330f421376c4'
by git＠videolan.org 31 Jan '17

31 Jan '17

ffmpeg | branch: master | Clément Bœsch <cboesch(a)gopro.com> | Tue Jan 31 17:06:49 2017 +0100| [dc0ca508ea2521010ac6168e330b0339edfc518c] | committer: Clément Bœsch Merge commit 'da917fcf5183ed249ad1285b8edd330f421376c4' * commit 'da917fcf5183ed249ad1285b8edd330f421376c4': avconv_dxva2: add a profile check for hevc This commit is a noop, see a655bc834479024ff4d60d0b8ecca290e73eb96c Merged-by: Clément Bœsch <cboesch(a)gopro.com> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=dc0ca508ea2521010… ---

1 0

avconv_dxva2: add a profile check for hevc
by git＠videolan.org 31 Jan '17

31 Jan '17

ffmpeg | branch: master | Hendrik Leppkes <h.leppkes(a)gmail.com> | Sat Jan 30 17:11:13 2016 +0100| [da917fcf5183ed249ad1285b8edd330f421376c4] | committer: Anton Khirnov avconv_dxva2: add a profile check for hevc Signed-off-by: Anton Khirnov <anton(a)khirnov.net> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=da917fcf5183ed249… --- avconv_dxva2.c | 6 ++++++ 1 file changed, 6 insertions(+) diff --git a/avconv_dxva2.c b/avconv_dxva2.c index 9cf09cd..7578c3f 100644 --- a/avconv_dxva2.c +++ b/avconv_dxva2.c @@ -422,6 +422,12 @@ int dxva2_init(AVCodecContext *s) return AVERROR(EINVAL); } + if (s->codec_id == AV_CODEC_ID_HEVC && + s->profile != FF_PROFILE_HEVC_MAIN && s->profile != FF_PROFILE_HEVC_MAIN_10) { + av_log(NULL, loglevel, "Unsupported HEVC profile for DXVA2 HWAccel: %d\n", s->profile); + return AVERROR(EINVAL); + } + av_buffer_unref(&ctx->hw_frames_ctx); ret = dxva2_create_decoder(s);

1 0

lavc/hevc: remove a few random spaces to reduce diff with libav
by git＠videolan.org 31 Jan '17

31 Jan '17

ffmpeg | branch: master | Clément Bœsch <cboesch(a)gopro.com> | Tue Jan 31 17:02:24 2017 +0100| [7c300a8ed4ef093a1dea40dd4d524cb32cef5c08] | committer: Clément Bœsch lavc/hevc: remove a few random spaces to reduce diff with libav > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=7c300a8ed4ef093a1… --- libavcodec/hevcdsp_template.c | 2 -- libavcodec/x86/hevcdsp_init.c | 2 +- 2 files changed, 1 insertion(+), 3 deletions(-) diff --git a/libavcodec/hevcdsp_template.c b/libavcodec/hevcdsp_template.c index 665436c..b95984f 100644 --- a/libavcodec/hevcdsp_template.c +++ b/libavcodec/hevcdsp_template.c @@ -26,7 +26,6 @@ #include "bit_depth_template.c" #include "hevcdsp.h" - static void FUNC(put_pcm)(uint8_t *_dst, ptrdiff_t stride, int width, int height, GetBitContext *gb, int pcm_bit_depth) { @@ -83,7 +82,6 @@ static void FUNC(add_residual32x32)(uint8_t *_dst, int16_t *res, FUNC(add_residual)(_dst, res, stride, 32); } - static void FUNC(transform_rdpcm)(int16_t *_coeffs, int16_t log2_size, int mode) { int16_t *coeffs = (int16_t *) _coeffs; diff --git a/libavcodec/x86/hevcdsp_init.c b/libavcodec/x86/hevcdsp_init.c index d16e59d..c4d9564 100644 --- a/libavcodec/x86/hevcdsp_init.c +++ b/libavcodec/x86/hevcdsp_init.c @@ -2,7 +2,6 @@ * Copyright (c) 2013 Seppo Tomperi * Copyright (c) 2013 - 2014 Pierre-Edouard Lepere * - * * This file is part of FFmpeg. * * FFmpeg is free software; you can redistribute it and/or @@ -21,6 +20,7 @@ */ #include "config.h" + #include "libavutil/cpu.h" #include "libavutil/x86/asm.h" #include "libavutil/x86/cpu.h"

1 0

Merge commit '1ecb63cd1c1a4ddc5efed4abbc3158b969d8c5e4'
by git＠videolan.org 31 Jan '17

31 Jan '17

ffmpeg | branch: master | Clément Bœsch <cboesch(a)gopro.com> | Tue Jan 31 16:57:18 2017 +0100| [bd370738baf8aea73ba3d6ecced967451c47ae88] | committer: Clément Bœsch Merge commit '1ecb63cd1c1a4ddc5efed4abbc3158b969d8c5e4' * commit '1ecb63cd1c1a4ddc5efed4abbc3158b969d8c5e4': hevc: set profile based on the profile compatibility flags if needed This commit is a noop, see f85cc3bf12236e974403667610b39b802b8651d6 Merged-by: Clément Bœsch <cboesch(a)gopro.com> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=bd370738baf8aea73… ---

1 0

hevc: set profile based on the profile compatibility flags if needed
by git＠videolan.org 31 Jan '17

31 Jan '17

ffmpeg | branch: master | Hendrik Leppkes <h.leppkes(a)gmail.com> | Sat Jan 30 17:44:34 2016 +0100| [1ecb63cd1c1a4ddc5efed4abbc3158b969d8c5e4] | committer: Anton Khirnov hevc: set profile based on the profile compatibility flags if needed This fixes retrieving a valid profile for many of the FATE conformance samples, allowing them to be properly decoded by the HWAccel after adding a profile check. Signed-off-by: Anton Khirnov <anton(a)khirnov.net> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=1ecb63cd1c1a4ddc5… --- libavcodec/hevc_ps.c | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) diff --git a/libavcodec/hevc_ps.c b/libavcodec/hevc_ps.c index 583fa06..5c249d8 100644 --- a/libavcodec/hevc_ps.c +++ b/libavcodec/hevc_ps.c @@ -244,8 +244,12 @@ static void decode_profile_tier_level(GetBitContext *gb, AVCodecContext *avctx, else av_log(avctx, AV_LOG_WARNING, "Unknown HEVC profile: %d\n", ptl->profile_idc); - for (i = 0; i < 32; i++) + for (i = 0; i < 32; i++) { ptl->profile_compatibility_flag[i] = get_bits1(gb); + + if (ptl->profile_idc == 0 && i > 0 && ptl->profile_compatibility_flag[i]) + ptl->profile_idc = i; + } ptl->progressive_source_flag = get_bits1(gb); ptl->interlaced_source_flag = get_bits1(gb); ptl->non_packed_constraint_flag = get_bits1(gb);

1 0

Merge commit 'fca3c3b61952aacc45e9ca54d86a762946c21942'
by git＠videolan.org 31 Jan '17

31 Jan '17

ffmpeg | branch: master | Clément Bœsch <cboesch(a)gopro.com> | Tue Jan 31 16:50:21 2017 +0100| [78d16eb45217f7ce811d1b05afe56427dd40021b] | committer: Clément Bœsch Merge commit 'fca3c3b61952aacc45e9ca54d86a762946c21942' * commit 'fca3c3b61952aacc45e9ca54d86a762946c21942': hevc: Add AVX2 DC IDCT Mostly noop as we already have that code. In the ASM, code is merged with the exception of SECTION which is kept uppercase for consistency with the rest of the codebase. Still in the ASM, the prototype comment is fixed to honor the '_' added from the original commit. idct_dc_proto() is dropped as it's not used anymore here. Merged-by: Clément Bœsch <cboesch(a)gopro.com> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=78d16eb45217f7ce8… --- libavcodec/x86/hevc_idct.asm | 69 ++++++++++++++++++++++--------------------- libavcodec/x86/hevcdsp.h | 3 -- libavcodec/x86/hevcdsp_init.c | 48 +++++++++++++++--------------- 3 files changed, 59 insertions(+), 61 deletions(-) diff --git a/libavcodec/x86/hevc_idct.asm b/libavcodec/x86/hevc_idct.asm index 2edaf9a..33b437c 100644 --- a/libavcodec/x86/hevc_idct.asm +++ b/libavcodec/x86/hevc_idct.asm @@ -1,37 +1,38 @@ -; /* -; * SIMD optimized idct functions for HEVC decoding -; * Copyright (c) 2014 Pierre-Edouard LEPERE -; * Copyright (c) 2014 James Almer -; * -; * This file is part of FFmpeg. -; * -; * FFmpeg is free software; you can redistribute it and/or -; * modify it under the terms of the GNU Lesser General Public -; * License as published by the Free Software Foundation; either -; * version 2.1 of the License, or (at your option) any later version. -; * -; * FFmpeg is distributed in the hope that it will be useful, -; * but WITHOUT ANY WARRANTY; without even the implied warranty of -; * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU -; * Lesser General Public License for more details. -; * -; * You should have received a copy of the GNU Lesser General Public -; * License along with FFmpeg; if not, write to the Free Software -; * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA -; */ +;******************************************************************************* +;* SIMD-optimized IDCT functions for HEVC decoding +;* Copyright (c) 2014 Pierre-Edouard LEPERE +;* Copyright (c) 2014 James Almer +;* +;* This file is part of FFmpeg. +;* +;* FFmpeg is free software; you can redistribute it and/or +;* modify it under the terms of the GNU Lesser General Public +;* License as published by the Free Software Foundation; either +;* version 2.1 of the License, or (at your option) any later version. +;* +;* FFmpeg is distributed in the hope that it will be useful, +;* but WITHOUT ANY WARRANTY; without even the implied warranty of +;* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU +;* Lesser General Public License for more details. +;* +;* You should have received a copy of the GNU Lesser General Public +;* License along with FFmpeg; if not, write to the Free Software +;* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA +;****************************************************************************** + %include "libavutil/x86/x86util.asm" SECTION .text -; void ff_hevc_idctHxW_dc_{8,10}_<opt>(int16_t *coeffs) +; void ff_hevc_idct_HxW_dc_{8,10}_<opt>(int16_t *coeffs) ; %1 = HxW ; %2 = number of loops ; %3 = bitdepth %macro IDCT_DC 3 -cglobal hevc_idct%1x%1_dc_%3, 1, 2, 1, coeff, tmp - movsx tmpq, word [coeffq] - add tmpw, ((1 << 14-%3) + 1) - sar tmpw, (15-%3) +cglobal hevc_idct_%1x%1_dc_%3, 1, 2, 1, coeff, tmp + movsx tmpd, word [coeffq] + add tmpd, (1 << (14 - %3)) + 1 + sar tmpd, (15 - %3) movd xm0, tmpd SPLATW m0, xm0 DEFINE_ARGS coeff, cnt @@ -41,11 +42,11 @@ cglobal hevc_idct%1x%1_dc_%3, 1, 2, 1, coeff, tmp mova [coeffq+mmsize*1], m0 mova [coeffq+mmsize*2], m0 mova [coeffq+mmsize*3], m0 - mova [coeffq+mmsize*4], m0 - mova [coeffq+mmsize*5], m0 - mova [coeffq+mmsize*6], m0 - mova [coeffq+mmsize*7], m0 add coeffq, mmsize*8 + mova [coeffq+mmsize*-4], m0 + mova [coeffq+mmsize*-3], m0 + mova [coeffq+mmsize*-2], m0 + mova [coeffq+mmsize*-1], m0 dec cntd jg .loop RET @@ -54,10 +55,10 @@ cglobal hevc_idct%1x%1_dc_%3, 1, 2, 1, coeff, tmp ; %1 = HxW ; %2 = bitdepth %macro IDCT_DC_NL 2 ; No loop -cglobal hevc_idct%1x%1_dc_%2, 1, 2, 1, coeff, tmp - movsx tmpq, word [coeffq] - add tmpw, ((1 << 14-%2) + 1) - sar tmpw, (15-%2) +cglobal hevc_idct_%1x%1_dc_%2, 1, 2, 1, coeff, tmp + movsx tmpd, word [coeffq] + add tmpd, (1 << (14 - %2)) + 1 + sar tmpd, (15 - %2) movd m0, tmpd SPLATW m0, xm0 mova [coeffq+mmsize*0], m0 diff --git a/libavcodec/x86/hevcdsp.h b/libavcodec/x86/hevcdsp.h index 3cfdc27..63a148e 100644 --- a/libavcodec/x86/hevcdsp.h +++ b/libavcodec/x86/hevcdsp.h @@ -29,9 +29,6 @@ #include <stdint.h> -#define idct_dc_proto(size, bitd, opt) \ - void ff_hevc_idct##size##_dc_add_##bitd##_##opt(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride) - #define PEL_LINK(dst, idx1, idx2, idx3, name, D, opt) \ dst[idx1][idx2][idx3] = ff_hevc_put_hevc_ ## name ## _ ## D ## _##opt; \ dst ## _bi[idx1][idx2][idx3] = ff_hevc_put_hevc_bi_ ## name ## _ ## D ## _##opt; \ diff --git a/libavcodec/x86/hevcdsp_init.c b/libavcodec/x86/hevcdsp_init.c index da73d76..d16e59d 100644 --- a/libavcodec/x86/hevcdsp_init.c +++ b/libavcodec/x86/hevcdsp_init.c @@ -59,9 +59,9 @@ LFL_FUNCS(uint8_t, 10, avx) LFL_FUNCS(uint8_t, 12, avx) #define IDCT_FUNCS(W, opt) \ -void ff_hevc_idct##W##_dc_8_##opt(int16_t *coeffs); \ -void ff_hevc_idct##W##_dc_10_##opt(int16_t *coeffs); \ -void ff_hevc_idct##W##_dc_12_##opt(int16_t *coeffs) +void ff_hevc_idct_ ## W ## _dc_8_ ## opt(int16_t *coeffs); \ +void ff_hevc_idct_ ## W ## _dc_10_ ## opt(int16_t *coeffs); \ +void ff_hevc_idct_ ## W ## _dc_12_ ## opt(int16_t *coeffs) IDCT_FUNCS(4x4, mmxext); IDCT_FUNCS(8x8, mmxext); @@ -698,8 +698,8 @@ void ff_hevc_dsp_init_x86(HEVCDSPContext *c, const int bit_depth) if (bit_depth == 8) { if (EXTERNAL_MMXEXT(cpu_flags)) { - c->idct_dc[0] = ff_hevc_idct4x4_dc_8_mmxext; - c->idct_dc[1] = ff_hevc_idct8x8_dc_8_mmxext; + c->idct_dc[0] = ff_hevc_idct_4x4_dc_8_mmxext; + c->idct_dc[1] = ff_hevc_idct_8x8_dc_8_mmxext; c->add_residual[0] = ff_hevc_add_residual4_8_mmxext; } if (EXTERNAL_SSE2(cpu_flags)) { @@ -712,9 +712,9 @@ void ff_hevc_dsp_init_x86(HEVCDSPContext *c, const int bit_depth) } SAO_BAND_INIT(8, sse2); - c->idct_dc[1] = ff_hevc_idct8x8_dc_8_sse2; - c->idct_dc[2] = ff_hevc_idct16x16_dc_8_sse2; - c->idct_dc[3] = ff_hevc_idct32x32_dc_8_sse2; + c->idct_dc[1] = ff_hevc_idct_8x8_dc_8_sse2; + c->idct_dc[2] = ff_hevc_idct_16x16_dc_8_sse2; + c->idct_dc[3] = ff_hevc_idct_32x32_dc_8_sse2; c->add_residual[1] = ff_hevc_add_residual8_8_sse2; c->add_residual[2] = ff_hevc_add_residual16_8_sse2; @@ -757,8 +757,8 @@ void ff_hevc_dsp_init_x86(HEVCDSPContext *c, const int bit_depth) c->sao_band_filter[1] = ff_hevc_sao_band_filter_16_8_avx2; } if (EXTERNAL_AVX2_FAST(cpu_flags)) { - c->idct_dc[2] = ff_hevc_idct16x16_dc_8_avx2; - c->idct_dc[3] = ff_hevc_idct32x32_dc_8_avx2; + c->idct_dc[2] = ff_hevc_idct_16x16_dc_8_avx2; + c->idct_dc[3] = ff_hevc_idct_32x32_dc_8_avx2; if (ARCH_X86_64) { c->put_hevc_epel[7][0][0] = ff_hevc_put_hevc_pel_pixels32_8_avx2; c->put_hevc_epel[8][0][0] = ff_hevc_put_hevc_pel_pixels48_8_avx2; @@ -855,8 +855,8 @@ void ff_hevc_dsp_init_x86(HEVCDSPContext *c, const int bit_depth) } else if (bit_depth == 10) { if (EXTERNAL_MMXEXT(cpu_flags)) { c->add_residual[0] = ff_hevc_add_residual4_10_mmxext; - c->idct_dc[0] = ff_hevc_idct4x4_dc_10_mmxext; - c->idct_dc[1] = ff_hevc_idct8x8_dc_10_mmxext; + c->idct_dc[0] = ff_hevc_idct_4x4_dc_10_mmxext; + c->idct_dc[1] = ff_hevc_idct_8x8_dc_10_mmxext; } if (EXTERNAL_SSE2(cpu_flags)) { c->hevc_v_loop_filter_chroma = ff_hevc_v_loop_filter_chroma_10_sse2; @@ -868,9 +868,9 @@ void ff_hevc_dsp_init_x86(HEVCDSPContext *c, const int bit_depth) SAO_BAND_INIT(10, sse2); SAO_EDGE_INIT(10, sse2); - c->idct_dc[1] = ff_hevc_idct8x8_dc_10_sse2; - c->idct_dc[2] = ff_hevc_idct16x16_dc_10_sse2; - c->idct_dc[3] = ff_hevc_idct32x32_dc_10_sse2; + c->idct_dc[1] = ff_hevc_idct_8x8_dc_10_sse2; + c->idct_dc[2] = ff_hevc_idct_16x16_dc_10_sse2; + c->idct_dc[3] = ff_hevc_idct_32x32_dc_10_sse2; c->add_residual[1] = ff_hevc_add_residual8_10_sse2; c->add_residual[2] = ff_hevc_add_residual16_10_sse2; @@ -904,8 +904,8 @@ void ff_hevc_dsp_init_x86(HEVCDSPContext *c, const int bit_depth) c->sao_band_filter[0] = ff_hevc_sao_band_filter_8_10_avx2; } if (EXTERNAL_AVX2_FAST(cpu_flags)) { - c->idct_dc[2] = ff_hevc_idct16x16_dc_10_avx2; - c->idct_dc[3] = ff_hevc_idct32x32_dc_10_avx2; + c->idct_dc[2] = ff_hevc_idct_16x16_dc_10_avx2; + c->idct_dc[3] = ff_hevc_idct_32x32_dc_10_avx2; if (ARCH_X86_64) { c->put_hevc_epel[5][0][0] = ff_hevc_put_hevc_pel_pixels16_10_avx2; c->put_hevc_epel[6][0][0] = ff_hevc_put_hevc_pel_pixels24_10_avx2; @@ -1059,8 +1059,8 @@ void ff_hevc_dsp_init_x86(HEVCDSPContext *c, const int bit_depth) } } else if (bit_depth == 12) { if (EXTERNAL_MMXEXT(cpu_flags)) { - c->idct_dc[0] = ff_hevc_idct4x4_dc_12_mmxext; - c->idct_dc[1] = ff_hevc_idct8x8_dc_12_mmxext; + c->idct_dc[0] = ff_hevc_idct_4x4_dc_12_mmxext; + c->idct_dc[1] = ff_hevc_idct_8x8_dc_12_mmxext; } if (EXTERNAL_SSE2(cpu_flags)) { c->hevc_v_loop_filter_chroma = ff_hevc_v_loop_filter_chroma_12_sse2; @@ -1072,9 +1072,9 @@ void ff_hevc_dsp_init_x86(HEVCDSPContext *c, const int bit_depth) SAO_BAND_INIT(12, sse2); SAO_EDGE_INIT(12, sse2); - c->idct_dc[1] = ff_hevc_idct8x8_dc_12_sse2; - c->idct_dc[2] = ff_hevc_idct16x16_dc_12_sse2; - c->idct_dc[3] = ff_hevc_idct32x32_dc_12_sse2; + c->idct_dc[1] = ff_hevc_idct_8x8_dc_12_sse2; + c->idct_dc[2] = ff_hevc_idct_16x16_dc_12_sse2; + c->idct_dc[3] = ff_hevc_idct_32x32_dc_12_sse2; } if (EXTERNAL_SSSE3(cpu_flags) && ARCH_X86_64) { c->hevc_v_loop_filter_luma = ff_hevc_v_loop_filter_luma_12_ssse3; @@ -1104,8 +1104,8 @@ void ff_hevc_dsp_init_x86(HEVCDSPContext *c, const int bit_depth) c->sao_band_filter[0] = ff_hevc_sao_band_filter_8_12_avx2; } if (EXTERNAL_AVX2_FAST(cpu_flags)) { - c->idct_dc[2] = ff_hevc_idct16x16_dc_12_avx2; - c->idct_dc[3] = ff_hevc_idct32x32_dc_12_avx2; + c->idct_dc[2] = ff_hevc_idct_16x16_dc_12_avx2; + c->idct_dc[3] = ff_hevc_idct_32x32_dc_12_avx2; SAO_BAND_INIT(12, avx2); SAO_EDGE_INIT(12, avx2); ====================================================================== diff --cc libavcodec/x86/hevc_idct.asm index 2edaf9a,d662aa9..33b437c --- a/libavcodec/x86/hevc_idct.asm +++ b/libavcodec/x86/hevc_idct.asm @@@ -1,29 -1,30 +1,30 @@@ - ; /* - ; * SIMD optimized idct functions for HEVC decoding - ; * Copyright (c) 2014 Pierre-Edouard LEPERE - ; * Copyright (c) 2014 James Almer - ; * - ; * This file is part of FFmpeg. - ; * - ; * FFmpeg is free software; you can redistribute it and/or - ; * modify it under the terms of the GNU Lesser General Public - ; * License as published by the Free Software Foundation; either - ; * version 2.1 of the License, or (at your option) any later version. - ; * - ; * FFmpeg is distributed in the hope that it will be useful, - ; * but WITHOUT ANY WARRANTY; without even the implied warranty of - ; * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - ; * Lesser General Public License for more details. - ; * - ; * You should have received a copy of the GNU Lesser General Public - ; * License along with FFmpeg; if not, write to the Free Software - ; * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA - ; */ + ;******************************************************************************* + ;* SIMD-optimized IDCT functions for HEVC decoding + ;* Copyright (c) 2014 Pierre-Edouard LEPERE + ;* Copyright (c) 2014 James Almer + ;* -;* 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 +SECTION .text --; void ff_hevc_idctHxW_dc_{8,10}_<opt>(int16_t *coeffs) ++; void ff_hevc_idct_HxW_dc_{8,10}_<opt>(int16_t *coeffs) ; %1 = HxW ; %2 = number of loops ; %3 = bitdepth diff --cc libavcodec/x86/hevcdsp.h index 3cfdc27,0000000..63a148e mode 100644,000000..100644 --- a/libavcodec/x86/hevcdsp.h +++ b/libavcodec/x86/hevcdsp.h @@@ -1,261 -1,0 +1,258 @@@ +/* + * HEVC video decoder + * + * Copyright (C) 2012 - 2013 Guillaume Martres + * Copyright (C) 2013 - 2014 Pierre-Edouard Lepere + * + * + * 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 + */ + +#ifndef AVCODEC_X86_HEVCDSP_H +#define AVCODEC_X86_HEVCDSP_H + +#include <stddef.h> +#include <stdint.h> + + - #define idct_dc_proto(size, bitd, opt) \ - void ff_hevc_idct##size##_dc_add_##bitd##_##opt(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride) - +#define PEL_LINK(dst, idx1, idx2, idx3, name, D, opt) \ +dst[idx1][idx2][idx3] = ff_hevc_put_hevc_ ## name ## _ ## D ## _##opt; \ +dst ## _bi[idx1][idx2][idx3] = ff_hevc_put_hevc_bi_ ## name ## _ ## D ## _##opt; \ +dst ## _uni[idx1][idx2][idx3] = ff_hevc_put_hevc_uni_ ## name ## _ ## D ## _##opt; \ +dst ## _uni_w[idx1][idx2][idx3] = ff_hevc_put_hevc_uni_w_ ## name ## _ ## D ## _##opt; \ +dst ## _bi_w[idx1][idx2][idx3] = ff_hevc_put_hevc_bi_w_ ## name ## _ ## D ## _##opt + + +#define PEL_PROTOTYPE(name, D, opt) \ +void ff_hevc_put_hevc_ ## name ## _ ## D ## _##opt(int16_t *dst, uint8_t *_src, ptrdiff_t _srcstride, int height, intptr_t mx, intptr_t my,int width); \ +void ff_hevc_put_hevc_bi_ ## name ## _ ## D ## _##opt(uint8_t *_dst, ptrdiff_t _dststride, uint8_t *_src, ptrdiff_t _srcstride, int16_t *src2, int height, intptr_t mx, intptr_t my, int width); \ +void ff_hevc_put_hevc_uni_ ## name ## _ ## D ## _##opt(uint8_t *_dst, ptrdiff_t _dststride, uint8_t *_src, ptrdiff_t _srcstride, int height, intptr_t mx, intptr_t my, int width); \ +void ff_hevc_put_hevc_uni_w_ ## name ## _ ## D ## _##opt(uint8_t *_dst, ptrdiff_t _dststride, uint8_t *_src, ptrdiff_t _srcstride, int height, int denom, int wx, int ox, intptr_t mx, intptr_t my, int width); \ +void ff_hevc_put_hevc_bi_w_ ## name ## _ ## D ## _##opt(uint8_t *_dst, ptrdiff_t _dststride, uint8_t *_src, ptrdiff_t _srcstride, int16_t *src2, int height, int denom, int wx0, int wx1, int ox0, int ox1, intptr_t mx, intptr_t my, int width) + + +/////////////////////////////////////////////////////////////////////////////// +// MC functions +/////////////////////////////////////////////////////////////////////////////// + +#define EPEL_PROTOTYPES(fname, bitd, opt) \ + PEL_PROTOTYPE(fname##4, bitd, opt); \ + PEL_PROTOTYPE(fname##6, bitd, opt); \ + PEL_PROTOTYPE(fname##8, bitd, opt); \ + PEL_PROTOTYPE(fname##12, bitd, opt); \ + PEL_PROTOTYPE(fname##16, bitd, opt); \ + PEL_PROTOTYPE(fname##24, bitd, opt); \ + PEL_PROTOTYPE(fname##32, bitd, opt); \ + PEL_PROTOTYPE(fname##48, bitd, opt); \ + PEL_PROTOTYPE(fname##64, bitd, opt) + +#define QPEL_PROTOTYPES(fname, bitd, opt) \ + PEL_PROTOTYPE(fname##4, bitd, opt); \ + PEL_PROTOTYPE(fname##8, bitd, opt); \ + PEL_PROTOTYPE(fname##12, bitd, opt); \ + PEL_PROTOTYPE(fname##16, bitd, opt); \ + PEL_PROTOTYPE(fname##24, bitd, opt); \ + PEL_PROTOTYPE(fname##32, bitd, opt); \ + PEL_PROTOTYPE(fname##48, bitd, opt); \ + PEL_PROTOTYPE(fname##64, bitd, opt) + +#define WEIGHTING_PROTOTYPE(width, bitd, opt) \ +void ff_hevc_put_hevc_uni_w##width##_##bitd##_##opt(uint8_t *dst, ptrdiff_t dststride, int16_t *_src, int height, int denom, int _wx, int _ox); \ +void ff_hevc_put_hevc_bi_w##width##_##bitd##_##opt(uint8_t *dst, ptrdiff_t dststride, int16_t *_src, int16_t *_src2, int height, int denom, int _wx0, int _wx1, int _ox0, int _ox1) + +#define WEIGHTING_PROTOTYPES(bitd, opt) \ + WEIGHTING_PROTOTYPE(2, bitd, opt); \ + WEIGHTING_PROTOTYPE(4, bitd, opt); \ + WEIGHTING_PROTOTYPE(6, bitd, opt); \ + WEIGHTING_PROTOTYPE(8, bitd, opt); \ + WEIGHTING_PROTOTYPE(12, bitd, opt); \ + WEIGHTING_PROTOTYPE(16, bitd, opt); \ + WEIGHTING_PROTOTYPE(24, bitd, opt); \ + WEIGHTING_PROTOTYPE(32, bitd, opt); \ + WEIGHTING_PROTOTYPE(48, bitd, opt); \ + WEIGHTING_PROTOTYPE(64, bitd, opt) + + +/////////////////////////////////////////////////////////////////////////////// +// QPEL_PIXELS EPEL_PIXELS +/////////////////////////////////////////////////////////////////////////////// +EPEL_PROTOTYPES(pel_pixels , 8, sse4); +EPEL_PROTOTYPES(pel_pixels , 10, sse4); +EPEL_PROTOTYPES(pel_pixels , 12, sse4); + +void ff_hevc_put_hevc_pel_pixels16_8_avx2(int16_t *dst, uint8_t *_src, ptrdiff_t _srcstride, int height, intptr_t mx, intptr_t my,int width); +void ff_hevc_put_hevc_pel_pixels24_8_avx2(int16_t *dst, uint8_t *_src, ptrdiff_t _srcstride, int height, intptr_t mx, intptr_t my,int width); +void ff_hevc_put_hevc_pel_pixels32_8_avx2(int16_t *dst, uint8_t *_src, ptrdiff_t _srcstride, int height, intptr_t mx, intptr_t my,int width); +void ff_hevc_put_hevc_pel_pixels48_8_avx2(int16_t *dst, uint8_t *_src, ptrdiff_t _srcstride, int height, intptr_t mx, intptr_t my,int width); +void ff_hevc_put_hevc_pel_pixels64_8_avx2(int16_t *dst, uint8_t *_src, ptrdiff_t _srcstride, int height, intptr_t mx, intptr_t my,int width); + +void ff_hevc_put_hevc_pel_pixels16_10_avx2(int16_t *dst, uint8_t *_src, ptrdiff_t _srcstride, int height, intptr_t mx, intptr_t my,int width); +void ff_hevc_put_hevc_pel_pixels24_10_avx2(int16_t *dst, uint8_t *_src, ptrdiff_t _srcstride, int height, intptr_t mx, intptr_t my,int width); +void ff_hevc_put_hevc_pel_pixels32_10_avx2(int16_t *dst, uint8_t *_src, ptrdiff_t _srcstride, int height, intptr_t mx, intptr_t my,int width); +void ff_hevc_put_hevc_pel_pixels48_10_avx2(int16_t *dst, uint8_t *_src, ptrdiff_t _srcstride, int height, intptr_t mx, intptr_t my,int width); +void ff_hevc_put_hevc_pel_pixels64_10_avx2(int16_t *dst, uint8_t *_src, ptrdiff_t _srcstride, int height, intptr_t mx, intptr_t my,int width); + + + +void ff_hevc_put_hevc_uni_pel_pixels32_8_avx2(uint8_t *dst, ptrdiff_t dststride,uint8_t *_src, ptrdiff_t _srcstride, int height, intptr_t mx, intptr_t my,int width); +void ff_hevc_put_hevc_uni_pel_pixels48_8_avx2(uint8_t *dst, ptrdiff_t dststride,uint8_t *_src, ptrdiff_t _srcstride, int height, intptr_t mx, intptr_t my,int width); +void ff_hevc_put_hevc_uni_pel_pixels64_8_avx2(uint8_t *dst, ptrdiff_t dststride,uint8_t *_src, ptrdiff_t _srcstride, int height, intptr_t mx, intptr_t my,int width); +void ff_hevc_put_hevc_uni_pel_pixels96_8_avx2(uint8_t *dst, ptrdiff_t dststride,uint8_t *_src, ptrdiff_t _srcstride, int height, intptr_t mx, intptr_t my,int width); //used for 10bit +void ff_hevc_put_hevc_uni_pel_pixels128_8_avx2(uint8_t *dst, ptrdiff_t dststride,uint8_t *_src, ptrdiff_t _srcstride, int height, intptr_t mx, intptr_t my,int width);//used for 10bit + + +void ff_hevc_put_hevc_bi_pel_pixels16_8_avx2(uint8_t *_dst, ptrdiff_t _dststride, uint8_t *_src, ptrdiff_t _srcstride, int16_t *src2, int height, intptr_t mx, intptr_t my, int width); +void ff_hevc_put_hevc_bi_pel_pixels24_8_avx2(uint8_t *_dst, ptrdiff_t _dststride, uint8_t *_src, ptrdiff_t _srcstride, int16_t *src2, int height, intptr_t mx, intptr_t my, int width); +void ff_hevc_put_hevc_bi_pel_pixels32_8_avx2(uint8_t *_dst, ptrdiff_t _dststride, uint8_t *_src, ptrdiff_t _srcstride, int16_t *src2, int height, intptr_t mx, intptr_t my, int width); +void ff_hevc_put_hevc_bi_pel_pixels48_8_avx2(uint8_t *_dst, ptrdiff_t _dststride, uint8_t *_src, ptrdiff_t _srcstride, int16_t *src2, int height, intptr_t mx, intptr_t my, int width); +void ff_hevc_put_hevc_bi_pel_pixels64_8_avx2(uint8_t *_dst, ptrdiff_t _dststride, uint8_t *_src, ptrdiff_t _srcstride, int16_t *src2, int height, intptr_t mx, intptr_t my, int width); + +void ff_hevc_put_hevc_bi_pel_pixels16_10_avx2(uint8_t *_dst, ptrdiff_t _dststride, uint8_t *_src, ptrdiff_t _srcstride, int16_t *src2, int height, intptr_t mx, intptr_t my, int width); +void ff_hevc_put_hevc_bi_pel_pixels24_10_avx2(uint8_t *_dst, ptrdiff_t _dststride, uint8_t *_src, ptrdiff_t _srcstride, int16_t *src2, int height, intptr_t mx, intptr_t my, int width); +void ff_hevc_put_hevc_bi_pel_pixels32_10_avx2(uint8_t *_dst, ptrdiff_t _dststride, uint8_t *_src, ptrdiff_t _srcstride, int16_t *src2, int height, intptr_t mx, intptr_t my, int width); +void ff_hevc_put_hevc_bi_pel_pixels48_10_avx2(uint8_t *_dst, ptrdiff_t _dststride, uint8_t *_src, ptrdiff_t _srcstride, int16_t *src2, int height, intptr_t mx, intptr_t my, int width); +void ff_hevc_put_hevc_bi_pel_pixels64_10_avx2(uint8_t *_dst, ptrdiff_t _dststride, uint8_t *_src, ptrdiff_t _srcstride, int16_t *src2, int height, intptr_t mx, intptr_t my, int width); + +/////////////////////////////////////////////////////////////////////////////// +// EPEL +/////////////////////////////////////////////////////////////////////////////// +EPEL_PROTOTYPES(epel_h , 8, sse4); +EPEL_PROTOTYPES(epel_h , 10, sse4); +EPEL_PROTOTYPES(epel_h , 12, sse4); + +EPEL_PROTOTYPES(epel_v , 8, sse4); +EPEL_PROTOTYPES(epel_v , 10, sse4); +EPEL_PROTOTYPES(epel_v , 12, sse4); + +EPEL_PROTOTYPES(epel_hv , 8, sse4); +EPEL_PROTOTYPES(epel_hv , 10, sse4); +EPEL_PROTOTYPES(epel_hv , 12, sse4); + +PEL_PROTOTYPE(epel_h16, 8, avx2); +PEL_PROTOTYPE(epel_h24, 8, avx2); +PEL_PROTOTYPE(epel_h32, 8, avx2); +PEL_PROTOTYPE(epel_h48, 8, avx2); +PEL_PROTOTYPE(epel_h64, 8, avx2); + +PEL_PROTOTYPE(epel_h16,10, avx2); +PEL_PROTOTYPE(epel_h24,10, avx2); +PEL_PROTOTYPE(epel_h32,10, avx2); +PEL_PROTOTYPE(epel_h48,10, avx2); +PEL_PROTOTYPE(epel_h64,10, avx2); + +PEL_PROTOTYPE(epel_v16, 8, avx2); +PEL_PROTOTYPE(epel_v24, 8, avx2); +PEL_PROTOTYPE(epel_v32, 8, avx2); +PEL_PROTOTYPE(epel_v48, 8, avx2); +PEL_PROTOTYPE(epel_v64, 8, avx2); + +PEL_PROTOTYPE(epel_v16,10, avx2); +PEL_PROTOTYPE(epel_v24,10, avx2); +PEL_PROTOTYPE(epel_v32,10, avx2); +PEL_PROTOTYPE(epel_v48,10, avx2); +PEL_PROTOTYPE(epel_v64,10, avx2); + +PEL_PROTOTYPE(epel_hv16, 8, avx2); +PEL_PROTOTYPE(epel_hv24, 8, avx2); +PEL_PROTOTYPE(epel_hv32, 8, avx2); +PEL_PROTOTYPE(epel_hv48, 8, avx2); +PEL_PROTOTYPE(epel_hv64, 8, avx2); + +PEL_PROTOTYPE(epel_hv16,10, avx2); +PEL_PROTOTYPE(epel_hv24,10, avx2); +PEL_PROTOTYPE(epel_hv32,10, avx2); +PEL_PROTOTYPE(epel_hv48,10, avx2); +PEL_PROTOTYPE(epel_hv64,10, avx2); + +/////////////////////////////////////////////////////////////////////////////// +// QPEL +/////////////////////////////////////////////////////////////////////////////// +QPEL_PROTOTYPES(qpel_h , 8, sse4); +QPEL_PROTOTYPES(qpel_h , 10, sse4); +QPEL_PROTOTYPES(qpel_h , 12, sse4); + +QPEL_PROTOTYPES(qpel_v, 8, sse4); +QPEL_PROTOTYPES(qpel_v, 10, sse4); +QPEL_PROTOTYPES(qpel_v, 12, sse4); + +QPEL_PROTOTYPES(qpel_hv, 8, sse4); +QPEL_PROTOTYPES(qpel_hv, 10, sse4); +QPEL_PROTOTYPES(qpel_hv, 12, sse4); + +PEL_PROTOTYPE(qpel_h16, 8, avx2); +PEL_PROTOTYPE(qpel_h24, 8, avx2); +PEL_PROTOTYPE(qpel_h32, 8, avx2); +PEL_PROTOTYPE(qpel_h48, 8, avx2); +PEL_PROTOTYPE(qpel_h64, 8, avx2); + +PEL_PROTOTYPE(qpel_h16,10, avx2); +PEL_PROTOTYPE(qpel_h24,10, avx2); +PEL_PROTOTYPE(qpel_h32,10, avx2); +PEL_PROTOTYPE(qpel_h48,10, avx2); +PEL_PROTOTYPE(qpel_h64,10, avx2); + +PEL_PROTOTYPE(qpel_v16, 8, avx2); +PEL_PROTOTYPE(qpel_v24, 8, avx2); +PEL_PROTOTYPE(qpel_v32, 8, avx2); +PEL_PROTOTYPE(qpel_v48, 8, avx2); +PEL_PROTOTYPE(qpel_v64, 8, avx2); + +PEL_PROTOTYPE(qpel_v16,10, avx2); +PEL_PROTOTYPE(qpel_v24,10, avx2); +PEL_PROTOTYPE(qpel_v32,10, avx2); +PEL_PROTOTYPE(qpel_v48,10, avx2); +PEL_PROTOTYPE(qpel_v64,10, avx2); + +PEL_PROTOTYPE(qpel_hv16, 8, avx2); +PEL_PROTOTYPE(qpel_hv24, 8, avx2); +PEL_PROTOTYPE(qpel_hv32, 8, avx2); +PEL_PROTOTYPE(qpel_hv48, 8, avx2); +PEL_PROTOTYPE(qpel_hv64, 8, avx2); + +PEL_PROTOTYPE(qpel_hv16,10, avx2); +PEL_PROTOTYPE(qpel_hv24,10, avx2); +PEL_PROTOTYPE(qpel_hv32,10, avx2); +PEL_PROTOTYPE(qpel_hv48,10, avx2); +PEL_PROTOTYPE(qpel_hv64,10, avx2); + +WEIGHTING_PROTOTYPES(8, sse4); +WEIGHTING_PROTOTYPES(10, sse4); +WEIGHTING_PROTOTYPES(12, sse4); + +/////////////////////////////////////////////////////////////////////////////// +// TRANSFORM_ADD +/////////////////////////////////////////////////////////////////////////////// +void ff_hevc_add_residual4_8_mmxext(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride); +void ff_hevc_add_residual8_8_sse2(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride); +void ff_hevc_add_residual16_8_sse2(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride); +void ff_hevc_add_residual32_8_sse2(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride); + +void ff_hevc_add_residual8_8_avx(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride); +void ff_hevc_add_residual16_8_avx(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride); +void ff_hevc_add_residual32_8_avx(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride); + +void ff_hevc_add_residual32_8_avx2(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride); + +void ff_hevc_add_residual4_10_mmxext(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride); +void ff_hevc_add_residual8_10_sse2(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride); +void ff_hevc_add_residual16_10_sse2(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride); +void ff_hevc_add_residual32_10_sse2(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride); + +void ff_hevc_add_residual16_10_avx2(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride); +void ff_hevc_add_residual32_10_avx2(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride); + +#endif // AVCODEC_X86_HEVCDSP_H diff --cc libavcodec/x86/hevcdsp_init.c index da73d76,1a675ab..d16e59d --- a/libavcodec/x86/hevcdsp_init.c +++ b/libavcodec/x86/hevcdsp_init.c @@@ -34,34 -32,43 +34,34 @@@ void ff_hevc_ ## DIR ## _loop_filter_ch #define LFL_FUNC(DIR, DEPTH, OPT) \ void ff_hevc_ ## DIR ## _loop_filter_luma_ ## DEPTH ## _ ## OPT(uint8_t *pix, ptrdiff_t stride, int beta, int *tc, uint8_t *no_p, uint8_t *no_q); -#define LFC_FUNCS(type, depth) \ - LFC_FUNC(h, depth, sse2) \ - LFC_FUNC(v, depth, sse2) - -#define LFL_FUNCS(type, depth) \ - LFL_FUNC(h, depth, ssse3) \ - LFL_FUNC(v, depth, ssse3) - -LFC_FUNCS(uint8_t, 8) -LFC_FUNCS(uint8_t, 10) -LFL_FUNCS(uint8_t, 8) -LFL_FUNCS(uint8_t, 10) - -#define idct_dc_proto(size, bitd, opt) \ - void ff_hevc_idct_ ## size ## _dc_add_ ## bitd ## _ ## opt(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride) +#define LFC_FUNCS(type, depth, opt) \ + LFC_FUNC(h, depth, opt) \ + LFC_FUNC(v, depth, opt) -idct_dc_proto(4, 8,mmxext); -idct_dc_proto(8, 8,mmxext); -idct_dc_proto(16,8, sse2); -idct_dc_proto(32,8, sse2); +#define LFL_FUNCS(type, depth, opt) \ + LFL_FUNC(h, depth, opt) \ + LFL_FUNC(v, depth, opt) -idct_dc_proto(32,8, avx2); - -idct_dc_proto(4, 10,mmxext); -idct_dc_proto(8, 10, sse2); -idct_dc_proto(16,10, sse2); -idct_dc_proto(32,10, sse2); -idct_dc_proto(8, 10, avx); -idct_dc_proto(16,10, avx); -idct_dc_proto(32,10, avx); - -idct_dc_proto(16,10, avx2); -idct_dc_proto(32,10, avx2); +LFC_FUNCS(uint8_t, 8, sse2) +LFC_FUNCS(uint8_t, 10, sse2) +LFC_FUNCS(uint8_t, 12, sse2) +LFC_FUNCS(uint8_t, 8, avx) +LFC_FUNCS(uint8_t, 10, avx) +LFC_FUNCS(uint8_t, 12, avx) +LFL_FUNCS(uint8_t, 8, sse2) +LFL_FUNCS(uint8_t, 10, sse2) +LFL_FUNCS(uint8_t, 12, sse2) +LFL_FUNCS(uint8_t, 8, ssse3) +LFL_FUNCS(uint8_t, 10, ssse3) +LFL_FUNCS(uint8_t, 12, ssse3) +LFL_FUNCS(uint8_t, 8, avx) +LFL_FUNCS(uint8_t, 10, avx) +LFL_FUNCS(uint8_t, 12, avx) #define IDCT_FUNCS(W, opt) \ - void ff_hevc_idct##W##_dc_8_##opt(int16_t *coeffs); \ - void ff_hevc_idct##W##_dc_10_##opt(int16_t *coeffs); \ - void ff_hevc_idct##W##_dc_12_##opt(int16_t *coeffs) + void ff_hevc_idct_ ## W ## _dc_8_ ## opt(int16_t *coeffs); \ -void ff_hevc_idct_ ## W ## _dc_10_ ## opt(int16_t *coeffs) ++void ff_hevc_idct_ ## W ## _dc_10_ ## opt(int16_t *coeffs); \ ++void ff_hevc_idct_ ## W ## _dc_12_ ## opt(int16_t *coeffs) IDCT_FUNCS(4x4, mmxext); IDCT_FUNCS(8x8, mmxext); @@@ -696,419 -240,126 +696,419 @@@ void ff_hevc_dsp_init_x86(HEVCDSPContex { int cpu_flags = av_get_cpu_flags(); -#define SET_LUMA_FUNCS(tabname, funcname, depth, cf) \ - c->tabname[0] = funcname ## _4_ ## depth ## _ ## cf; \ - c->tabname[1] = funcname ## _8_ ## depth ## _ ## cf; \ - c->tabname[2] = funcname ## _12_ ## depth ## _ ## cf; \ - c->tabname[3] = funcname ## _16_ ## depth ## _ ## cf; \ - c->tabname[4] = funcname ## _24_ ## depth ## _ ## cf; \ - c->tabname[5] = funcname ## _32_ ## depth ## _ ## cf; \ - c->tabname[6] = funcname ## _48_ ## depth ## _ ## cf; \ - c->tabname[7] = funcname ## _64_ ## depth ## _ ## cf; - -#define SET_CHROMA_FUNCS(tabname, funcname, depth, cf) \ - c->tabname[1] = funcname ## _4_ ## depth ## _ ## cf; \ - c->tabname[3] = funcname ## _8_ ## depth ## _ ## cf; \ - c->tabname[4] = funcname ## _12_ ## depth ## _ ## cf; \ - c->tabname[5] = funcname ## _16_ ## depth ## _ ## cf; \ - c->tabname[6] = funcname ## _24_ ## depth ## _ ## cf; \ - c->tabname[7] = funcname ## _32_ ## depth ## _ ## cf; - -#define SET_QPEL_FUNCS(v, h, depth, cf, name) SET_LUMA_FUNCS (put_hevc_qpel[v][h], name, depth, cf) -#define SET_EPEL_FUNCS(v, h, depth, cf, name) SET_CHROMA_FUNCS(put_hevc_epel[v][h], name, depth, cf) - if (bit_depth == 8) { if (EXTERNAL_MMXEXT(cpu_flags)) { - c->idct_dc[0] = ff_hevc_idct4x4_dc_8_mmxext; - c->idct_dc[1] = ff_hevc_idct8x8_dc_8_mmxext; + c->idct_dc[0] = ff_hevc_idct_4x4_dc_8_mmxext; + c->idct_dc[1] = ff_hevc_idct_8x8_dc_8_mmxext; + c->add_residual[0] = ff_hevc_add_residual4_8_mmxext; } if (EXTERNAL_SSE2(cpu_flags)) { c->hevc_v_loop_filter_chroma = ff_hevc_v_loop_filter_chroma_8_sse2; c->hevc_h_loop_filter_chroma = ff_hevc_h_loop_filter_chroma_8_sse2; + if (ARCH_X86_64) { + c->hevc_v_loop_filter_luma = ff_hevc_v_loop_filter_luma_8_sse2; + c->hevc_h_loop_filter_luma = ff_hevc_h_loop_filter_luma_8_sse2; + + } + SAO_BAND_INIT(8, sse2); - c->idct_dc[1] = ff_hevc_idct8x8_dc_8_sse2; - c->idct_dc[2] = ff_hevc_idct16x16_dc_8_sse2; - c->idct_dc[3] = ff_hevc_idct32x32_dc_8_sse2; + c->idct_dc[1] = ff_hevc_idct_8x8_dc_8_sse2; + c->idct_dc[2] = ff_hevc_idct_16x16_dc_8_sse2; + c->idct_dc[3] = ff_hevc_idct_32x32_dc_8_sse2; - SET_QPEL_FUNCS(0, 0, 8, sse2, ff_hevc_get_pixels); - SET_EPEL_FUNCS(0, 0, 8, sse2, ff_hevc_get_pixels); - SET_LUMA_FUNCS(put_unweighted_pred, ff_hevc_put_unweighted_pred, 8, sse2); - SET_LUMA_FUNCS(put_unweighted_pred_avg, ff_hevc_put_unweighted_pred_avg, 8, sse2); - SET_CHROMA_FUNCS(put_unweighted_pred_chroma, ff_hevc_put_unweighted_pred, 8, sse2); - SET_CHROMA_FUNCS(put_unweighted_pred_avg_chroma, ff_hevc_put_unweighted_pred_avg, 8, sse2); + c->add_residual[1] = ff_hevc_add_residual8_8_sse2; + c->add_residual[2] = ff_hevc_add_residual16_8_sse2; + c->add_residual[3] = ff_hevc_add_residual32_8_sse2; } if (EXTERNAL_SSSE3(cpu_flags)) { - SET_QPEL_FUNCS(0, 1, 8, ssse3, ff_hevc_qpel_h); - SET_QPEL_FUNCS(1, 0, 8, ssse3, ff_hevc_qpel_v); - SET_EPEL_FUNCS(0, 1, 8, ssse3, ff_hevc_epel_h); - SET_EPEL_FUNCS(1, 0, 8, ssse3, ff_hevc_epel_v); + if(ARCH_X86_64) { + c->hevc_v_loop_filter_luma = ff_hevc_v_loop_filter_luma_8_ssse3; + c->hevc_h_loop_filter_luma = ff_hevc_h_loop_filter_luma_8_ssse3; + } + SAO_EDGE_INIT(8, ssse3); + } + if (EXTERNAL_SSE4(cpu_flags) && ARCH_X86_64) { + + EPEL_LINKS(c->put_hevc_epel, 0, 0, pel_pixels, 8, sse4); + EPEL_LINKS(c->put_hevc_epel, 0, 1, epel_h, 8, sse4); + EPEL_LINKS(c->put_hevc_epel, 1, 0, epel_v, 8, sse4); + EPEL_LINKS(c->put_hevc_epel, 1, 1, epel_hv, 8, sse4); + + QPEL_LINKS(c->put_hevc_qpel, 0, 0, pel_pixels, 8, sse4); + QPEL_LINKS(c->put_hevc_qpel, 0, 1, qpel_h, 8, sse4); + QPEL_LINKS(c->put_hevc_qpel, 1, 0, qpel_v, 8, sse4); + QPEL_LINKS(c->put_hevc_qpel, 1, 1, qpel_hv, 8, sse4); + } + if (EXTERNAL_AVX(cpu_flags)) { + c->hevc_v_loop_filter_chroma = ff_hevc_v_loop_filter_chroma_8_avx; + c->hevc_h_loop_filter_chroma = ff_hevc_h_loop_filter_chroma_8_avx; + if (ARCH_X86_64) { + c->hevc_v_loop_filter_luma = ff_hevc_v_loop_filter_luma_8_avx; + c->hevc_h_loop_filter_luma = ff_hevc_h_loop_filter_luma_8_avx; + } + SAO_BAND_INIT(8, avx); + + c->add_residual[1] = ff_hevc_add_residual8_8_avx; + c->add_residual[2] = ff_hevc_add_residual16_8_avx; + c->add_residual[3] = ff_hevc_add_residual32_8_avx; + } + if (EXTERNAL_AVX2(cpu_flags)) { + c->sao_band_filter[0] = ff_hevc_sao_band_filter_8_8_avx2; + c->sao_band_filter[1] = ff_hevc_sao_band_filter_16_8_avx2; + } + if (EXTERNAL_AVX2_FAST(cpu_flags)) { - c->idct_dc[2] = ff_hevc_idct16x16_dc_8_avx2; - c->idct_dc[3] = ff_hevc_idct32x32_dc_8_avx2; ++ c->idct_dc[2] = ff_hevc_idct_16x16_dc_8_avx2; ++ c->idct_dc[3] = ff_hevc_idct_32x32_dc_8_avx2; + if (ARCH_X86_64) { + c->put_hevc_epel[7][0][0] = ff_hevc_put_hevc_pel_pixels32_8_avx2; + c->put_hevc_epel[8][0][0] = ff_hevc_put_hevc_pel_pixels48_8_avx2; + c->put_hevc_epel[9][0][0] = ff_hevc_put_hevc_pel_pixels64_8_avx2; + + c->put_hevc_qpel[7][0][0] = ff_hevc_put_hevc_pel_pixels32_8_avx2; + c->put_hevc_qpel[8][0][0] = ff_hevc_put_hevc_pel_pixels48_8_avx2; + c->put_hevc_qpel[9][0][0] = ff_hevc_put_hevc_pel_pixels64_8_avx2; + + c->put_hevc_epel_uni[7][0][0] = ff_hevc_put_hevc_uni_pel_pixels32_8_avx2; + c->put_hevc_epel_uni[8][0][0] = ff_hevc_put_hevc_uni_pel_pixels48_8_avx2; + c->put_hevc_epel_uni[9][0][0] = ff_hevc_put_hevc_uni_pel_pixels64_8_avx2; + + c->put_hevc_qpel_uni[7][0][0] = ff_hevc_put_hevc_uni_pel_pixels32_8_avx2; + c->put_hevc_qpel_uni[8][0][0] = ff_hevc_put_hevc_uni_pel_pixels48_8_avx2; + c->put_hevc_qpel_uni[9][0][0] = ff_hevc_put_hevc_uni_pel_pixels64_8_avx2; + + c->put_hevc_qpel_bi[7][0][0] = ff_hevc_put_hevc_bi_pel_pixels32_8_avx2; + c->put_hevc_qpel_bi[8][0][0] = ff_hevc_put_hevc_bi_pel_pixels48_8_avx2; + c->put_hevc_qpel_bi[9][0][0] = ff_hevc_put_hevc_bi_pel_pixels64_8_avx2; + + c->put_hevc_epel_bi[7][0][0] = ff_hevc_put_hevc_bi_pel_pixels32_8_avx2; + c->put_hevc_epel_bi[8][0][0] = ff_hevc_put_hevc_bi_pel_pixels48_8_avx2; + c->put_hevc_epel_bi[9][0][0] = ff_hevc_put_hevc_bi_pel_pixels64_8_avx2; + + c->put_hevc_epel[7][0][1] = ff_hevc_put_hevc_epel_h32_8_avx2; + c->put_hevc_epel[8][0][1] = ff_hevc_put_hevc_epel_h48_8_avx2; + c->put_hevc_epel[9][0][1] = ff_hevc_put_hevc_epel_h64_8_avx2; + + c->put_hevc_epel_uni[7][0][1] = ff_hevc_put_hevc_uni_epel_h32_8_avx2; + c->put_hevc_epel_uni[8][0][1] = ff_hevc_put_hevc_uni_epel_h48_8_avx2; + c->put_hevc_epel_uni[9][0][1] = ff_hevc_put_hevc_uni_epel_h64_8_avx2; + c->put_hevc_epel_bi[7][0][1] = ff_hevc_put_hevc_bi_epel_h32_8_avx2; + c->put_hevc_epel_bi[8][0][1] = ff_hevc_put_hevc_bi_epel_h48_8_avx2; + c->put_hevc_epel_bi[9][0][1] = ff_hevc_put_hevc_bi_epel_h64_8_avx2; + + c->put_hevc_epel[7][1][0] = ff_hevc_put_hevc_epel_v32_8_avx2; + c->put_hevc_epel[8][1][0] = ff_hevc_put_hevc_epel_v48_8_avx2; + c->put_hevc_epel[9][1][0] = ff_hevc_put_hevc_epel_v64_8_avx2; + + c->put_hevc_epel_uni[7][1][0] = ff_hevc_put_hevc_uni_epel_v32_8_avx2; + c->put_hevc_epel_uni[8][1][0] = ff_hevc_put_hevc_uni_epel_v48_8_avx2; + c->put_hevc_epel_uni[9][1][0] = ff_hevc_put_hevc_uni_epel_v64_8_avx2; + + c->put_hevc_epel_bi[7][1][0] = ff_hevc_put_hevc_bi_epel_v32_8_avx2; + c->put_hevc_epel_bi[8][1][0] = ff_hevc_put_hevc_bi_epel_v48_8_avx2; + c->put_hevc_epel_bi[9][1][0] = ff_hevc_put_hevc_bi_epel_v64_8_avx2; + + c->put_hevc_epel[7][1][1] = ff_hevc_put_hevc_epel_hv32_8_avx2; + c->put_hevc_epel[8][1][1] = ff_hevc_put_hevc_epel_hv48_8_avx2; + c->put_hevc_epel[9][1][1] = ff_hevc_put_hevc_epel_hv64_8_avx2; + + c->put_hevc_epel_uni[7][1][1] = ff_hevc_put_hevc_uni_epel_hv32_8_avx2; + c->put_hevc_epel_uni[8][1][1] = ff_hevc_put_hevc_uni_epel_hv48_8_avx2; + c->put_hevc_epel_uni[9][1][1] = ff_hevc_put_hevc_uni_epel_hv64_8_avx2; + + c->put_hevc_epel_bi[7][1][1] = ff_hevc_put_hevc_bi_epel_hv32_8_avx2; + c->put_hevc_epel_bi[8][1][1] = ff_hevc_put_hevc_bi_epel_hv48_8_avx2; + c->put_hevc_epel_bi[9][1][1] = ff_hevc_put_hevc_bi_epel_hv64_8_avx2; + + c->put_hevc_qpel[7][0][1] = ff_hevc_put_hevc_qpel_h32_8_avx2; + c->put_hevc_qpel[8][0][1] = ff_hevc_put_hevc_qpel_h48_8_avx2; + c->put_hevc_qpel[9][0][1] = ff_hevc_put_hevc_qpel_h64_8_avx2; + + c->put_hevc_qpel[7][1][0] = ff_hevc_put_hevc_qpel_v32_8_avx2; + c->put_hevc_qpel[8][1][0] = ff_hevc_put_hevc_qpel_v48_8_avx2; + c->put_hevc_qpel[9][1][0] = ff_hevc_put_hevc_qpel_v64_8_avx2; + + c->put_hevc_qpel_uni[7][0][1] = ff_hevc_put_hevc_uni_qpel_h32_8_avx2; + c->put_hevc_qpel_uni[8][0][1] = ff_hevc_put_hevc_uni_qpel_h48_8_avx2; + c->put_hevc_qpel_uni[9][0][1] = ff_hevc_put_hevc_uni_qpel_h64_8_avx2; + + c->put_hevc_qpel_uni[7][1][0] = ff_hevc_put_hevc_uni_qpel_v32_8_avx2; + c->put_hevc_qpel_uni[8][1][0] = ff_hevc_put_hevc_uni_qpel_v48_8_avx2; + c->put_hevc_qpel_uni[9][1][0] = ff_hevc_put_hevc_uni_qpel_v64_8_avx2; + + c->put_hevc_qpel_bi[7][0][1] = ff_hevc_put_hevc_bi_qpel_h32_8_avx2; + c->put_hevc_qpel_bi[8][0][1] = ff_hevc_put_hevc_bi_qpel_h48_8_avx2; + c->put_hevc_qpel_bi[9][0][1] = ff_hevc_put_hevc_bi_qpel_h64_8_avx2; + + c->put_hevc_qpel_bi[7][1][0] = ff_hevc_put_hevc_bi_qpel_v32_8_avx2; + c->put_hevc_qpel_bi[8][1][0] = ff_hevc_put_hevc_bi_qpel_v48_8_avx2; + c->put_hevc_qpel_bi[9][1][0] = ff_hevc_put_hevc_bi_qpel_v64_8_avx2; + } + SAO_BAND_INIT(8, avx2); + + c->sao_edge_filter[2] = ff_hevc_sao_edge_filter_32_8_avx2; + c->sao_edge_filter[3] = ff_hevc_sao_edge_filter_48_8_avx2; + c->sao_edge_filter[4] = ff_hevc_sao_edge_filter_64_8_avx2; + + c->add_residual[3] = ff_hevc_add_residual32_8_avx2; } } else if (bit_depth == 10) { if (EXTERNAL_MMXEXT(cpu_flags)) { + c->add_residual[0] = ff_hevc_add_residual4_10_mmxext; - c->idct_dc[0] = ff_hevc_idct4x4_dc_10_mmxext; - c->idct_dc[1] = ff_hevc_idct8x8_dc_10_mmxext; + c->idct_dc[0] = ff_hevc_idct_4x4_dc_10_mmxext; + c->idct_dc[1] = ff_hevc_idct_8x8_dc_10_mmxext; } if (EXTERNAL_SSE2(cpu_flags)) { c->hevc_v_loop_filter_chroma = ff_hevc_v_loop_filter_chroma_10_sse2; c->hevc_h_loop_filter_chroma = ff_hevc_h_loop_filter_chroma_10_sse2; + if (ARCH_X86_64) { + c->hevc_v_loop_filter_luma = ff_hevc_v_loop_filter_luma_10_sse2; + c->hevc_h_loop_filter_luma = ff_hevc_h_loop_filter_luma_10_sse2; + } + SAO_BAND_INIT(10, sse2); + SAO_EDGE_INIT(10, sse2); - c->idct_dc[1] = ff_hevc_idct8x8_dc_10_sse2; - c->idct_dc[2] = ff_hevc_idct16x16_dc_10_sse2; - c->idct_dc[3] = ff_hevc_idct32x32_dc_10_sse2; + c->idct_dc[1] = ff_hevc_idct_8x8_dc_10_sse2; + c->idct_dc[2] = ff_hevc_idct_16x16_dc_10_sse2; + c->idct_dc[3] = ff_hevc_idct_32x32_dc_10_sse2; - SET_QPEL_FUNCS(0, 0, 10, sse2, ff_hevc_get_pixels); - SET_EPEL_FUNCS(0, 0, 10, sse2, ff_hevc_get_pixels); - - SET_LUMA_FUNCS(put_unweighted_pred, ff_hevc_put_unweighted_pred, 10, sse2); - SET_LUMA_FUNCS(put_unweighted_pred_avg, ff_hevc_put_unweighted_pred_avg, 10, sse2); - SET_CHROMA_FUNCS(put_unweighted_pred_chroma, ff_hevc_put_unweighted_pred, 10, sse2); - SET_CHROMA_FUNCS(put_unweighted_pred_avg_chroma, ff_hevc_put_unweighted_pred_avg, 10, sse2); + c->add_residual[1] = ff_hevc_add_residual8_10_sse2; + c->add_residual[2] = ff_hevc_add_residual16_10_sse2; + c->add_residual[3] = ff_hevc_add_residual32_10_sse2; } - } - -#if ARCH_X86_64 - if (bit_depth == 8) { - if (EXTERNAL_SSSE3(cpu_flags)) { - c->hevc_v_loop_filter_luma = ff_hevc_v_loop_filter_luma_8_ssse3; - c->hevc_h_loop_filter_luma = ff_hevc_h_loop_filter_luma_8_ssse3; + if (EXTERNAL_SSSE3(cpu_flags) && ARCH_X86_64) { + c->hevc_v_loop_filter_luma = ff_hevc_v_loop_filter_luma_10_ssse3; + c->hevc_h_loop_filter_luma = ff_hevc_h_loop_filter_luma_10_ssse3; } + if (EXTERNAL_SSE4(cpu_flags) && ARCH_X86_64) { + EPEL_LINKS(c->put_hevc_epel, 0, 0, pel_pixels, 10, sse4); + EPEL_LINKS(c->put_hevc_epel, 0, 1, epel_h, 10, sse4); + EPEL_LINKS(c->put_hevc_epel, 1, 0, epel_v, 10, sse4); + EPEL_LINKS(c->put_hevc_epel, 1, 1, epel_hv, 10, sse4); - if (EXTERNAL_SSE4(cpu_flags)) { - SET_LUMA_FUNCS(weighted_pred, ff_hevc_put_weighted_pred, 8, sse4); - SET_CHROMA_FUNCS(weighted_pred_chroma, ff_hevc_put_weighted_pred, 8, sse4); - SET_LUMA_FUNCS(weighted_pred_avg, ff_hevc_put_weighted_pred_avg, 8, sse4); - SET_CHROMA_FUNCS(weighted_pred_avg_chroma, ff_hevc_put_weighted_pred_avg, 8, sse4); + QPEL_LINKS(c->put_hevc_qpel, 0, 0, pel_pixels, 10, sse4); + QPEL_LINKS(c->put_hevc_qpel, 0, 1, qpel_h, 10, sse4); + QPEL_LINKS(c->put_hevc_qpel, 1, 0, qpel_v, 10, sse4); + QPEL_LINKS(c->put_hevc_qpel, 1, 1, qpel_hv, 10, sse4); } - if (EXTERNAL_AVX(cpu_flags)) { -#if HAVE_AVX_EXTERNAL - SET_QPEL_FUNCS(1, 1, 8, avx, hevc_qpel_hv); - SET_EPEL_FUNCS(1, 1, 8, avx, hevc_epel_hv); -#endif /* HAVE_AVX_EXTERNAL */ + c->hevc_v_loop_filter_chroma = ff_hevc_v_loop_filter_chroma_10_avx; + c->hevc_h_loop_filter_chroma = ff_hevc_h_loop_filter_chroma_10_avx; + if (ARCH_X86_64) { + c->hevc_v_loop_filter_luma = ff_hevc_v_loop_filter_luma_10_avx; + c->hevc_h_loop_filter_luma = ff_hevc_h_loop_filter_luma_10_avx; + } + SAO_BAND_INIT(10, avx); } if (EXTERNAL_AVX2(cpu_flags)) { - c->idct_dc[2] = ff_hevc_idct_16x16_dc_8_avx2; - c->idct_dc[3] = ff_hevc_idct_32x32_dc_8_avx2; + c->sao_band_filter[0] = ff_hevc_sao_band_filter_8_10_avx2; } - } else if (bit_depth == 10) { - if (EXTERNAL_SSSE3(cpu_flags)) { - c->hevc_v_loop_filter_luma = ff_hevc_v_loop_filter_luma_10_ssse3; - c->hevc_h_loop_filter_luma = ff_hevc_h_loop_filter_luma_10_ssse3; + if (EXTERNAL_AVX2_FAST(cpu_flags)) { - c->idct_dc[2] = ff_hevc_idct16x16_dc_10_avx2; - c->idct_dc[3] = ff_hevc_idct32x32_dc_10_avx2; ++ c->idct_dc[2] = ff_hevc_idct_16x16_dc_10_avx2; ++ c->idct_dc[3] = ff_hevc_idct_32x32_dc_10_avx2; + if (ARCH_X86_64) { + c->put_hevc_epel[5][0][0] = ff_hevc_put_hevc_pel_pixels16_10_avx2; + c->put_hevc_epel[6][0][0] = ff_hevc_put_hevc_pel_pixels24_10_avx2; + c->put_hevc_epel[7][0][0] = ff_hevc_put_hevc_pel_pixels32_10_avx2; + c->put_hevc_epel[8][0][0] = ff_hevc_put_hevc_pel_pixels48_10_avx2; + c->put_hevc_epel[9][0][0] = ff_hevc_put_hevc_pel_pixels64_10_avx2; + + c->put_hevc_qpel[5][0][0] = ff_hevc_put_hevc_pel_pixels16_10_avx2; + c->put_hevc_qpel[6][0][0] = ff_hevc_put_hevc_pel_pixels24_10_avx2; + c->put_hevc_qpel[7][0][0] = ff_hevc_put_hevc_pel_pixels32_10_avx2; + c->put_hevc_qpel[8][0][0] = ff_hevc_put_hevc_pel_pixels48_10_avx2; + c->put_hevc_qpel[9][0][0] = ff_hevc_put_hevc_pel_pixels64_10_avx2; + + c->put_hevc_epel_uni[5][0][0] = ff_hevc_put_hevc_uni_pel_pixels32_8_avx2; + c->put_hevc_epel_uni[6][0][0] = ff_hevc_put_hevc_uni_pel_pixels48_8_avx2; + c->put_hevc_epel_uni[7][0][0] = ff_hevc_put_hevc_uni_pel_pixels64_8_avx2; + c->put_hevc_epel_uni[8][0][0] = ff_hevc_put_hevc_uni_pel_pixels96_8_avx2; + c->put_hevc_epel_uni[9][0][0] = ff_hevc_put_hevc_uni_pel_pixels128_8_avx2; + + c->put_hevc_qpel_uni[5][0][0] = ff_hevc_put_hevc_uni_pel_pixels32_8_avx2; + c->put_hevc_qpel_uni[6][0][0] = ff_hevc_put_hevc_uni_pel_pixels48_8_avx2; + c->put_hevc_qpel_uni[7][0][0] = ff_hevc_put_hevc_uni_pel_pixels64_8_avx2; + c->put_hevc_qpel_uni[8][0][0] = ff_hevc_put_hevc_uni_pel_pixels96_8_avx2; + c->put_hevc_qpel_uni[9][0][0] = ff_hevc_put_hevc_uni_pel_pixels128_8_avx2; + + c->put_hevc_epel_bi[5][0][0] = ff_hevc_put_hevc_bi_pel_pixels16_10_avx2; + c->put_hevc_epel_bi[6][0][0] = ff_hevc_put_hevc_bi_pel_pixels24_10_avx2; + c->put_hevc_epel_bi[7][0][0] = ff_hevc_put_hevc_bi_pel_pixels32_10_avx2; + c->put_hevc_epel_bi[8][0][0] = ff_hevc_put_hevc_bi_pel_pixels48_10_avx2; + c->put_hevc_epel_bi[9][0][0] = ff_hevc_put_hevc_bi_pel_pixels64_10_avx2; + c->put_hevc_qpel_bi[5][0][0] = ff_hevc_put_hevc_bi_pel_pixels16_10_avx2; + c->put_hevc_qpel_bi[6][0][0] = ff_hevc_put_hevc_bi_pel_pixels24_10_avx2; + c->put_hevc_qpel_bi[7][0][0] = ff_hevc_put_hevc_bi_pel_pixels32_10_avx2; + c->put_hevc_qpel_bi[8][0][0] = ff_hevc_put_hevc_bi_pel_pixels48_10_avx2; + c->put_hevc_qpel_bi[9][0][0] = ff_hevc_put_hevc_bi_pel_pixels64_10_avx2; + + c->put_hevc_epel[5][0][1] = ff_hevc_put_hevc_epel_h16_10_avx2; + c->put_hevc_epel[6][0][1] = ff_hevc_put_hevc_epel_h24_10_avx2; + c->put_hevc_epel[7][0][1] = ff_hevc_put_hevc_epel_h32_10_avx2; + c->put_hevc_epel[8][0][1] = ff_hevc_put_hevc_epel_h48_10_avx2; + c->put_hevc_epel[9][0][1] = ff_hevc_put_hevc_epel_h64_10_avx2; + + c->put_hevc_epel_uni[5][0][1] = ff_hevc_put_hevc_uni_epel_h16_10_avx2; + c->put_hevc_epel_uni[6][0][1] = ff_hevc_put_hevc_uni_epel_h24_10_avx2; + c->put_hevc_epel_uni[7][0][1] = ff_hevc_put_hevc_uni_epel_h32_10_avx2; + c->put_hevc_epel_uni[8][0][1] = ff_hevc_put_hevc_uni_epel_h48_10_avx2; + c->put_hevc_epel_uni[9][0][1] = ff_hevc_put_hevc_uni_epel_h64_10_avx2; + + c->put_hevc_epel_bi[5][0][1] = ff_hevc_put_hevc_bi_epel_h16_10_avx2; + c->put_hevc_epel_bi[6][0][1] = ff_hevc_put_hevc_bi_epel_h24_10_avx2; + c->put_hevc_epel_bi[7][0][1] = ff_hevc_put_hevc_bi_epel_h32_10_avx2; + c->put_hevc_epel_bi[8][0][1] = ff_hevc_put_hevc_bi_epel_h48_10_avx2; + c->put_hevc_epel_bi[9][0][1] = ff_hevc_put_hevc_bi_epel_h64_10_avx2; + + c->put_hevc_epel[5][1][0] = ff_hevc_put_hevc_epel_v16_10_avx2; + c->put_hevc_epel[6][1][0] = ff_hevc_put_hevc_epel_v24_10_avx2; + c->put_hevc_epel[7][1][0] = ff_hevc_put_hevc_epel_v32_10_avx2; + c->put_hevc_epel[8][1][0] = ff_hevc_put_hevc_epel_v48_10_avx2; + c->put_hevc_epel[9][1][0] = ff_hevc_put_hevc_epel_v64_10_avx2; + + c->put_hevc_epel_uni[5][1][0] = ff_hevc_put_hevc_uni_epel_v16_10_avx2; + c->put_hevc_epel_uni[6][1][0] = ff_hevc_put_hevc_uni_epel_v24_10_avx2; + c->put_hevc_epel_uni[7][1][0] = ff_hevc_put_hevc_uni_epel_v32_10_avx2; + c->put_hevc_epel_uni[8][1][0] = ff_hevc_put_hevc_uni_epel_v48_10_avx2; + c->put_hevc_epel_uni[9][1][0] = ff_hevc_put_hevc_uni_epel_v64_10_avx2; + + c->put_hevc_epel_bi[5][1][0] = ff_hevc_put_hevc_bi_epel_v16_10_avx2; + c->put_hevc_epel_bi[6][1][0] = ff_hevc_put_hevc_bi_epel_v24_10_avx2; + c->put_hevc_epel_bi[7][1][0] = ff_hevc_put_hevc_bi_epel_v32_10_avx2; + c->put_hevc_epel_bi[8][1][0] = ff_hevc_put_hevc_bi_epel_v48_10_avx2; + c->put_hevc_epel_bi[9][1][0] = ff_hevc_put_hevc_bi_epel_v64_10_avx2; + + c->put_hevc_epel[5][1][1] = ff_hevc_put_hevc_epel_hv16_10_avx2; + c->put_hevc_epel[6][1][1] = ff_hevc_put_hevc_epel_hv24_10_avx2; + c->put_hevc_epel[7][1][1] = ff_hevc_put_hevc_epel_hv32_10_avx2; + c->put_hevc_epel[8][1][1] = ff_hevc_put_hevc_epel_hv48_10_avx2; + c->put_hevc_epel[9][1][1] = ff_hevc_put_hevc_epel_hv64_10_avx2; + + c->put_hevc_epel_uni[5][1][1] = ff_hevc_put_hevc_uni_epel_hv16_10_avx2; + c->put_hevc_epel_uni[6][1][1] = ff_hevc_put_hevc_uni_epel_hv24_10_avx2; + c->put_hevc_epel_uni[7][1][1] = ff_hevc_put_hevc_uni_epel_hv32_10_avx2; + c->put_hevc_epel_uni[8][1][1] = ff_hevc_put_hevc_uni_epel_hv48_10_avx2; + c->put_hevc_epel_uni[9][1][1] = ff_hevc_put_hevc_uni_epel_hv64_10_avx2; + + c->put_hevc_epel_bi[5][1][1] = ff_hevc_put_hevc_bi_epel_hv16_10_avx2; + c->put_hevc_epel_bi[6][1][1] = ff_hevc_put_hevc_bi_epel_hv24_10_avx2; + c->put_hevc_epel_bi[7][1][1] = ff_hevc_put_hevc_bi_epel_hv32_10_avx2; + c->put_hevc_epel_bi[8][1][1] = ff_hevc_put_hevc_bi_epel_hv48_10_avx2; + c->put_hevc_epel_bi[9][1][1] = ff_hevc_put_hevc_bi_epel_hv64_10_avx2; + + c->put_hevc_qpel[5][0][1] = ff_hevc_put_hevc_qpel_h16_10_avx2; + c->put_hevc_qpel[6][0][1] = ff_hevc_put_hevc_qpel_h24_10_avx2; + c->put_hevc_qpel[7][0][1] = ff_hevc_put_hevc_qpel_h32_10_avx2; + c->put_hevc_qpel[8][0][1] = ff_hevc_put_hevc_qpel_h48_10_avx2; + c->put_hevc_qpel[9][0][1] = ff_hevc_put_hevc_qpel_h64_10_avx2; + + c->put_hevc_qpel_uni[5][0][1] = ff_hevc_put_hevc_uni_qpel_h16_10_avx2; + c->put_hevc_qpel_uni[6][0][1] = ff_hevc_put_hevc_uni_qpel_h24_10_avx2; + c->put_hevc_qpel_uni[7][0][1] = ff_hevc_put_hevc_uni_qpel_h32_10_avx2; + c->put_hevc_qpel_uni[8][0][1] = ff_hevc_put_hevc_uni_qpel_h48_10_avx2; + c->put_hevc_qpel_uni[9][0][1] = ff_hevc_put_hevc_uni_qpel_h64_10_avx2; + + c->put_hevc_qpel_bi[5][0][1] = ff_hevc_put_hevc_bi_qpel_h16_10_avx2; + c->put_hevc_qpel_bi[6][0][1] = ff_hevc_put_hevc_bi_qpel_h24_10_avx2; + c->put_hevc_qpel_bi[7][0][1] = ff_hevc_put_hevc_bi_qpel_h32_10_avx2; + c->put_hevc_qpel_bi[8][0][1] = ff_hevc_put_hevc_bi_qpel_h48_10_avx2; + c->put_hevc_qpel_bi[9][0][1] = ff_hevc_put_hevc_bi_qpel_h64_10_avx2; + + c->put_hevc_qpel[5][1][0] = ff_hevc_put_hevc_qpel_v16_10_avx2; + c->put_hevc_qpel[6][1][0] = ff_hevc_put_hevc_qpel_v24_10_avx2; + c->put_hevc_qpel[7][1][0] = ff_hevc_put_hevc_qpel_v32_10_avx2; + c->put_hevc_qpel[8][1][0] = ff_hevc_put_hevc_qpel_v48_10_avx2; + c->put_hevc_qpel[9][1][0] = ff_hevc_put_hevc_qpel_v64_10_avx2; + + c->put_hevc_qpel_uni[5][1][0] = ff_hevc_put_hevc_uni_qpel_v16_10_avx2; + c->put_hevc_qpel_uni[6][1][0] = ff_hevc_put_hevc_uni_qpel_v24_10_avx2; + c->put_hevc_qpel_uni[7][1][0] = ff_hevc_put_hevc_uni_qpel_v32_10_avx2; + c->put_hevc_qpel_uni[8][1][0] = ff_hevc_put_hevc_uni_qpel_v48_10_avx2; + c->put_hevc_qpel_uni[9][1][0] = ff_hevc_put_hevc_uni_qpel_v64_10_avx2; + + c->put_hevc_qpel_bi[5][1][0] = ff_hevc_put_hevc_bi_qpel_v16_10_avx2; + c->put_hevc_qpel_bi[6][1][0] = ff_hevc_put_hevc_bi_qpel_v24_10_avx2; + c->put_hevc_qpel_bi[7][1][0] = ff_hevc_put_hevc_bi_qpel_v32_10_avx2; + c->put_hevc_qpel_bi[8][1][0] = ff_hevc_put_hevc_bi_qpel_v48_10_avx2; + c->put_hevc_qpel_bi[9][1][0] = ff_hevc_put_hevc_bi_qpel_v64_10_avx2; + + c->put_hevc_qpel[5][1][1] = ff_hevc_put_hevc_qpel_hv16_10_avx2; + c->put_hevc_qpel[6][1][1] = ff_hevc_put_hevc_qpel_hv24_10_avx2; + c->put_hevc_qpel[7][1][1] = ff_hevc_put_hevc_qpel_hv32_10_avx2; + c->put_hevc_qpel[8][1][1] = ff_hevc_put_hevc_qpel_hv48_10_avx2; + c->put_hevc_qpel[9][1][1] = ff_hevc_put_hevc_qpel_hv64_10_avx2; + + c->put_hevc_qpel_uni[5][1][1] = ff_hevc_put_hevc_uni_qpel_hv16_10_avx2; + c->put_hevc_qpel_uni[6][1][1] = ff_hevc_put_hevc_uni_qpel_hv24_10_avx2; + c->put_hevc_qpel_uni[7][1][1] = ff_hevc_put_hevc_uni_qpel_hv32_10_avx2; + c->put_hevc_qpel_uni[8][1][1] = ff_hevc_put_hevc_uni_qpel_hv48_10_avx2; + c->put_hevc_qpel_uni[9][1][1] = ff_hevc_put_hevc_uni_qpel_hv64_10_avx2; + + c->put_hevc_qpel_bi[5][1][1] = ff_hevc_put_hevc_bi_qpel_hv16_10_avx2; + c->put_hevc_qpel_bi[6][1][1] = ff_hevc_put_hevc_bi_qpel_hv24_10_avx2; + c->put_hevc_qpel_bi[7][1][1] = ff_hevc_put_hevc_bi_qpel_hv32_10_avx2; + c->put_hevc_qpel_bi[8][1][1] = ff_hevc_put_hevc_bi_qpel_hv48_10_avx2; + c->put_hevc_qpel_bi[9][1][1] = ff_hevc_put_hevc_bi_qpel_hv64_10_avx2; + } + SAO_BAND_INIT(10, avx2); + SAO_EDGE_INIT(10, avx2); + + c->add_residual[2] = ff_hevc_add_residual16_10_avx2; + c->add_residual[3] = ff_hevc_add_residual32_10_avx2; + } - if (EXTERNAL_SSE4(cpu_flags)) { - SET_LUMA_FUNCS(weighted_pred, ff_hevc_put_weighted_pred, 10, sse4); - SET_CHROMA_FUNCS(weighted_pred_chroma, ff_hevc_put_weighted_pred, 10, sse4); - SET_LUMA_FUNCS(weighted_pred_avg, ff_hevc_put_weighted_pred_avg, 10, sse4); - SET_CHROMA_FUNCS(weighted_pred_avg_chroma, ff_hevc_put_weighted_pred_avg, 10, sse4); + } else if (bit_depth == 12) { + if (EXTERNAL_MMXEXT(cpu_flags)) { - c->idct_dc[0] = ff_hevc_idct4x4_dc_12_mmxext; - c->idct_dc[1] = ff_hevc_idct8x8_dc_12_mmxext; ++ c->idct_dc[0] = ff_hevc_idct_4x4_dc_12_mmxext; ++ c->idct_dc[1] = ff_hevc_idct_8x8_dc_12_mmxext; + } + if (EXTERNAL_SSE2(cpu_flags)) { + c->hevc_v_loop_filter_chroma = ff_hevc_v_loop_filter_chroma_12_sse2; + c->hevc_h_loop_filter_chroma = ff_hevc_h_loop_filter_chroma_12_sse2; + if (ARCH_X86_64) { + c->hevc_v_loop_filter_luma = ff_hevc_v_loop_filter_luma_12_sse2; + c->hevc_h_loop_filter_luma = ff_hevc_h_loop_filter_luma_12_sse2; + } + SAO_BAND_INIT(12, sse2); + SAO_EDGE_INIT(12, sse2); + - c->idct_dc[1] = ff_hevc_idct8x8_dc_12_sse2; - c->idct_dc[2] = ff_hevc_idct16x16_dc_12_sse2; - c->idct_dc[3] = ff_hevc_idct32x32_dc_12_sse2; ++ c->idct_dc[1] = ff_hevc_idct_8x8_dc_12_sse2; ++ c->idct_dc[2] = ff_hevc_idct_16x16_dc_12_sse2; ++ c->idct_dc[3] = ff_hevc_idct_32x32_dc_12_sse2; + } + if (EXTERNAL_SSSE3(cpu_flags) && ARCH_X86_64) { + c->hevc_v_loop_filter_luma = ff_hevc_v_loop_filter_luma_12_ssse3; + c->hevc_h_loop_filter_luma = ff_hevc_h_loop_filter_luma_12_ssse3; + } + if (EXTERNAL_SSE4(cpu_flags) && ARCH_X86_64) { + EPEL_LINKS(c->put_hevc_epel, 0, 0, pel_pixels, 12, sse4); + EPEL_LINKS(c->put_hevc_epel, 0, 1, epel_h, 12, sse4); + EPEL_LINKS(c->put_hevc_epel, 1, 0, epel_v, 12, sse4); + EPEL_LINKS(c->put_hevc_epel, 1, 1, epel_hv, 12, sse4); + + QPEL_LINKS(c->put_hevc_qpel, 0, 0, pel_pixels, 12, sse4); + QPEL_LINKS(c->put_hevc_qpel, 0, 1, qpel_h, 12, sse4); + QPEL_LINKS(c->put_hevc_qpel, 1, 0, qpel_v, 12, sse4); + QPEL_LINKS(c->put_hevc_qpel, 1, 1, qpel_hv, 12, sse4); } if (EXTERNAL_AVX(cpu_flags)) { -#if HAVE_AVX_EXTERNAL - SET_QPEL_FUNCS(0, 1, 10, avx, ff_hevc_qpel_h); - SET_QPEL_FUNCS(1, 0, 10, avx, ff_hevc_qpel_v); - SET_QPEL_FUNCS(1, 1, 10, avx, hevc_qpel_hv); - SET_EPEL_FUNCS(0, 1, 10, avx, ff_hevc_epel_h); - SET_EPEL_FUNCS(1, 0, 10, avx, ff_hevc_epel_v); - SET_EPEL_FUNCS(1, 1, 10, avx, hevc_epel_hv); -#endif /* HAVE_AVX_EXTERNAL */ + c->hevc_v_loop_filter_chroma = ff_hevc_v_loop_filter_chroma_12_avx; + c->hevc_h_loop_filter_chroma = ff_hevc_h_loop_filter_chroma_12_avx; + if (ARCH_X86_64) { + c->hevc_v_loop_filter_luma = ff_hevc_v_loop_filter_luma_12_avx; + c->hevc_h_loop_filter_luma = ff_hevc_h_loop_filter_luma_12_avx; + } + SAO_BAND_INIT(12, avx); } if (EXTERNAL_AVX2(cpu_flags)) { - c->idct_dc[2] = ff_hevc_idct_16x16_dc_10_avx2; - c->idct_dc[3] = ff_hevc_idct_32x32_dc_10_avx2; + c->sao_band_filter[0] = ff_hevc_sao_band_filter_8_12_avx2; + } + if (EXTERNAL_AVX2_FAST(cpu_flags)) { - c->idct_dc[2] = ff_hevc_idct16x16_dc_12_avx2; - c->idct_dc[3] = ff_hevc_idct32x32_dc_12_avx2; ++ c->idct_dc[2] = ff_hevc_idct_16x16_dc_12_avx2; ++ c->idct_dc[3] = ff_hevc_idct_32x32_dc_12_avx2; + + SAO_BAND_INIT(12, avx2); + SAO_EDGE_INIT(12, avx2); } } -#endif /* ARCH_X86_64 */ }

1 0

hevc: Add AVX2 DC IDCT
by git＠videolan.org 31 Jan '17

31 Jan '17

ffmpeg | branch: master | James Almer <jamrial(a)gmail.com> | Wed Jun 29 11:56:42 2016 +0200| [fca3c3b61952aacc45e9ca54d86a762946c21942] | committer: Luca Barbato hevc: Add AVX2 DC IDCT Originally written by Pierre Edouard Lepere <pierre-edouard.lepere(a)insa-rennes.fr>. Integrated to Libav by Josh de Kock <josh(a)itanimul.li>. Signed-off-by: Alexandra Hájková <alexandra(a)khirnov.net> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=fca3c3b61952aacc4… --- libavcodec/hevc.c | 2 +- libavcodec/x86/Makefile | 3 +- libavcodec/x86/hevc_idct.asm | 107 ++++++++++++++++++++++++++++++++++++++++++ libavcodec/x86/hevcdsp_init.c | 57 ++++++++++++++++++++++ 4 files changed, 167 insertions(+), 2 deletions(-) diff --git a/libavcodec/hevc.c b/libavcodec/hevc.c index 5d58b52..961991f 100644 --- a/libavcodec/hevc.c +++ b/libavcodec/hevc.c @@ -900,7 +900,7 @@ static void hls_residual_coding(HEVCContext *s, int x0, int y0, int vshift = s->ps.sps->vshift[c_idx]; uint8_t *dst = &s->frame->data[c_idx][(y0 >> vshift) * stride + ((x0 >> hshift) << s->ps.sps->pixel_shift)]; - DECLARE_ALIGNED(16, int16_t, coeffs[MAX_TB_SIZE * MAX_TB_SIZE]) = { 0 }; + DECLARE_ALIGNED(32, int16_t, coeffs[MAX_TB_SIZE * MAX_TB_SIZE]) = { 0 }; DECLARE_ALIGNED(8, uint8_t, significant_coeff_group_flag[8][8]) = { { 0 } }; int trafo_size = 1 << log2_trafo_size; diff --git a/libavcodec/x86/Makefile b/libavcodec/x86/Makefile index cdf7758..1460197 100644 --- a/libavcodec/x86/Makefile +++ b/libavcodec/x86/Makefile @@ -115,7 +115,8 @@ YASM-OBJS-$(CONFIG_APE_DECODER) += x86/apedsp.o YASM-OBJS-$(CONFIG_DCA_DECODER) += x86/dcadsp.o YASM-OBJS-$(CONFIG_DNXHD_ENCODER) += x86/dnxhdenc.o YASM-OBJS-$(CONFIG_HEVC_DECODER) += x86/hevc_deblock.o \ - x86/hevc_mc.o + x86/hevc_mc.o \ + x86/hevc_idct.o YASM-OBJS-$(CONFIG_PNG_DECODER) += x86/pngdsp.o YASM-OBJS-$(CONFIG_PRORES_DECODER) += x86/proresdsp.o YASM-OBJS-$(CONFIG_RV40_DECODER) += x86/rv40dsp.o diff --git a/libavcodec/x86/hevc_idct.asm b/libavcodec/x86/hevc_idct.asm new file mode 100644 index 0000000..d662aa9 --- /dev/null +++ b/libavcodec/x86/hevc_idct.asm @@ -0,0 +1,107 @@ +;******************************************************************************* +;* SIMD-optimized IDCT functions for HEVC decoding +;* Copyright (c) 2014 Pierre-Edouard LEPERE +;* Copyright (c) 2014 James Almer +;* +;* 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_hevc_idctHxW_dc_{8,10}_<opt>(int16_t *coeffs) +; %1 = HxW +; %2 = number of loops +; %3 = bitdepth +%macro IDCT_DC 3 +cglobal hevc_idct_%1x%1_dc_%3, 1, 2, 1, coeff, tmp + movsx tmpd, word [coeffq] + add tmpd, (1 << (14 - %3)) + 1 + sar tmpd, (15 - %3) + movd xm0, tmpd + SPLATW m0, xm0 + DEFINE_ARGS coeff, cnt + mov cntd, %2 +.loop: + mova [coeffq+mmsize*0], m0 + mova [coeffq+mmsize*1], m0 + mova [coeffq+mmsize*2], m0 + mova [coeffq+mmsize*3], m0 + add coeffq, mmsize*8 + mova [coeffq+mmsize*-4], m0 + mova [coeffq+mmsize*-3], m0 + mova [coeffq+mmsize*-2], m0 + mova [coeffq+mmsize*-1], m0 + dec cntd + jg .loop + RET +%endmacro + +; %1 = HxW +; %2 = bitdepth +%macro IDCT_DC_NL 2 ; No loop +cglobal hevc_idct_%1x%1_dc_%2, 1, 2, 1, coeff, tmp + movsx tmpd, word [coeffq] + add tmpd, (1 << (14 - %2)) + 1 + sar tmpd, (15 - %2) + movd m0, tmpd + SPLATW m0, xm0 + mova [coeffq+mmsize*0], m0 + mova [coeffq+mmsize*1], m0 + mova [coeffq+mmsize*2], m0 + mova [coeffq+mmsize*3], m0 +%if mmsize == 16 + mova [coeffq+mmsize*4], m0 + mova [coeffq+mmsize*5], m0 + mova [coeffq+mmsize*6], m0 + mova [coeffq+mmsize*7], m0 +%endif + RET +%endmacro + +; 8-bit +INIT_MMX mmxext +IDCT_DC_NL 4, 8 +IDCT_DC 8, 2, 8 + +INIT_XMM sse2 +IDCT_DC_NL 8, 8 +IDCT_DC 16, 4, 8 +IDCT_DC 32, 16, 8 + +%if HAVE_AVX2_EXTERNAL +INIT_YMM avx2 +IDCT_DC 16, 2, 8 +IDCT_DC 32, 8, 8 +%endif ;HAVE_AVX2_EXTERNAL + +; 10-bit +INIT_MMX mmxext +IDCT_DC_NL 4, 10 +IDCT_DC 8, 2, 10 + +INIT_XMM sse2 +IDCT_DC_NL 8, 10 +IDCT_DC 16, 4, 10 +IDCT_DC 32, 16, 10 + +%if HAVE_AVX2_EXTERNAL +INIT_YMM avx2 +IDCT_DC 16, 2, 10 +IDCT_DC 32, 8, 10 +%endif ;HAVE_AVX2_EXTERNAL diff --git a/libavcodec/x86/hevcdsp_init.c b/libavcodec/x86/hevcdsp_init.c index fd22fc3..1a675ab 100644 --- a/libavcodec/x86/hevcdsp_init.c +++ b/libavcodec/x86/hevcdsp_init.c @@ -45,6 +45,39 @@ LFC_FUNCS(uint8_t, 10) LFL_FUNCS(uint8_t, 8) LFL_FUNCS(uint8_t, 10) +#define idct_dc_proto(size, bitd, opt) \ + void ff_hevc_idct_ ## size ## _dc_add_ ## bitd ## _ ## opt(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride) + +idct_dc_proto(4, 8,mmxext); +idct_dc_proto(8, 8,mmxext); +idct_dc_proto(16,8, sse2); +idct_dc_proto(32,8, sse2); + +idct_dc_proto(32,8, avx2); + +idct_dc_proto(4, 10,mmxext); +idct_dc_proto(8, 10, sse2); +idct_dc_proto(16,10, sse2); +idct_dc_proto(32,10, sse2); +idct_dc_proto(8, 10, avx); +idct_dc_proto(16,10, avx); +idct_dc_proto(32,10, avx); + +idct_dc_proto(16,10, avx2); +idct_dc_proto(32,10, avx2); + +#define IDCT_FUNCS(W, opt) \ +void ff_hevc_idct_ ## W ## _dc_8_ ## opt(int16_t *coeffs); \ +void ff_hevc_idct_ ## W ## _dc_10_ ## opt(int16_t *coeffs) + +IDCT_FUNCS(4x4, mmxext); +IDCT_FUNCS(8x8, mmxext); +IDCT_FUNCS(8x8, sse2); +IDCT_FUNCS(16x16, sse2); +IDCT_FUNCS(32x32, sse2); +IDCT_FUNCS(16x16, avx2); +IDCT_FUNCS(32x32, avx2); + #define GET_PIXELS(width, depth, cf) \ void ff_hevc_get_pixels_ ## width ## _ ## depth ## _ ## cf(int16_t *dst, ptrdiff_t dststride, \ uint8_t *src, ptrdiff_t srcstride, \ @@ -229,10 +262,17 @@ void ff_hevc_dsp_init_x86(HEVCDSPContext *c, const int bit_depth) #define SET_EPEL_FUNCS(v, h, depth, cf, name) SET_CHROMA_FUNCS(put_hevc_epel[v][h], name, depth, cf) if (bit_depth == 8) { + if (EXTERNAL_MMXEXT(cpu_flags)) { + c->idct_dc[0] = ff_hevc_idct_4x4_dc_8_mmxext; + c->idct_dc[1] = ff_hevc_idct_8x8_dc_8_mmxext; + } if (EXTERNAL_SSE2(cpu_flags)) { c->hevc_v_loop_filter_chroma = ff_hevc_v_loop_filter_chroma_8_sse2; c->hevc_h_loop_filter_chroma = ff_hevc_h_loop_filter_chroma_8_sse2; + c->idct_dc[1] = ff_hevc_idct_8x8_dc_8_sse2; + c->idct_dc[2] = ff_hevc_idct_16x16_dc_8_sse2; + c->idct_dc[3] = ff_hevc_idct_32x32_dc_8_sse2; SET_QPEL_FUNCS(0, 0, 8, sse2, ff_hevc_get_pixels); SET_EPEL_FUNCS(0, 0, 8, sse2, ff_hevc_get_pixels); @@ -246,12 +286,21 @@ void ff_hevc_dsp_init_x86(HEVCDSPContext *c, const int bit_depth) SET_QPEL_FUNCS(1, 0, 8, ssse3, ff_hevc_qpel_v); SET_EPEL_FUNCS(0, 1, 8, ssse3, ff_hevc_epel_h); SET_EPEL_FUNCS(1, 0, 8, ssse3, ff_hevc_epel_v); + } } else if (bit_depth == 10) { + if (EXTERNAL_MMXEXT(cpu_flags)) { + c->idct_dc[0] = ff_hevc_idct_4x4_dc_10_mmxext; + c->idct_dc[1] = ff_hevc_idct_8x8_dc_10_mmxext; + } if (EXTERNAL_SSE2(cpu_flags)) { c->hevc_v_loop_filter_chroma = ff_hevc_v_loop_filter_chroma_10_sse2; c->hevc_h_loop_filter_chroma = ff_hevc_h_loop_filter_chroma_10_sse2; + c->idct_dc[1] = ff_hevc_idct_8x8_dc_10_sse2; + c->idct_dc[2] = ff_hevc_idct_16x16_dc_10_sse2; + c->idct_dc[3] = ff_hevc_idct_32x32_dc_10_sse2; + SET_QPEL_FUNCS(0, 0, 10, sse2, ff_hevc_get_pixels); SET_EPEL_FUNCS(0, 0, 10, sse2, ff_hevc_get_pixels); @@ -282,6 +331,10 @@ void ff_hevc_dsp_init_x86(HEVCDSPContext *c, const int bit_depth) SET_EPEL_FUNCS(1, 1, 8, avx, hevc_epel_hv); #endif /* HAVE_AVX_EXTERNAL */ } + if (EXTERNAL_AVX2(cpu_flags)) { + c->idct_dc[2] = ff_hevc_idct_16x16_dc_8_avx2; + c->idct_dc[3] = ff_hevc_idct_32x32_dc_8_avx2; + } } else if (bit_depth == 10) { if (EXTERNAL_SSSE3(cpu_flags)) { c->hevc_v_loop_filter_luma = ff_hevc_v_loop_filter_luma_10_ssse3; @@ -303,6 +356,10 @@ void ff_hevc_dsp_init_x86(HEVCDSPContext *c, const int bit_depth) SET_EPEL_FUNCS(1, 1, 10, avx, hevc_epel_hv); #endif /* HAVE_AVX_EXTERNAL */ } + if (EXTERNAL_AVX2(cpu_flags)) { + c->idct_dc[2] = ff_hevc_idct_16x16_dc_10_avx2; + c->idct_dc[3] = ff_hevc_idct_32x32_dc_10_avx2; + } } #endif /* ARCH_X86_64 */ }

1 0

Merge commit 'cc16da75c2f99d92f7a6461100f041352deb6d88'
by git＠videolan.org 31 Jan '17

31 Jan '17

ffmpeg | branch: master | Clément Bœsch <cboesch(a)gopro.com> | Tue Jan 31 16:02:35 2017 +0100| [05018c2cdaf3cbf5769bdf90951f707cf99edd52] | committer: Clément Bœsch Merge commit 'cc16da75c2f99d92f7a6461100f041352deb6d88' * commit 'cc16da75c2f99d92f7a6461100f041352deb6d88': hevc: Add coefficient limiting to speed up IDCT Noop again as we have these changes already, only random spacing changes. Merged-by: Clément Bœsch <cboesch(a)gopro.com> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=05018c2cdaf3cbf57… --- libavcodec/hevc_cabac.c | 4 ++-- libavcodec/hevcdsp_template.c | 13 +++++++------ 2 files changed, 9 insertions(+), 8 deletions(-) diff --git a/libavcodec/hevc_cabac.c b/libavcodec/hevc_cabac.c index 457e087..4f90dab 100644 --- a/libavcodec/hevc_cabac.c +++ b/libavcodec/hevc_cabac.c @@ -1490,7 +1490,7 @@ void ff_hevc_hls_residual_coding(HEVCContext *s, int x0, int y0, } else { int max_xy = FFMAX(last_significant_coeff_x, last_significant_coeff_y); if (max_xy == 0) - s->hevcdsp.idct_dc[log2_trafo_size-2](coeffs); + s->hevcdsp.idct_dc[log2_trafo_size - 2](coeffs); else { int col_limit = last_significant_coeff_x + last_significant_coeff_y + 4; if (max_xy < 4) @@ -1499,7 +1499,7 @@ void ff_hevc_hls_residual_coding(HEVCContext *s, int x0, int y0, col_limit = FFMIN(8, col_limit); else if (max_xy < 12) col_limit = FFMIN(24, col_limit); - s->hevcdsp.idct[log2_trafo_size-2](coeffs, col_limit); + s->hevcdsp.idct[log2_trafo_size - 2](coeffs, col_limit); } } } diff --git a/libavcodec/hevcdsp_template.c b/libavcodec/hevcdsp_template.c index 1dbd47a..665436c 100644 --- a/libavcodec/hevcdsp_template.c +++ b/libavcodec/hevcdsp_template.c @@ -223,7 +223,7 @@ static void FUNC(transform_4x4_luma)(int16_t *coeffs) for (i = 0; i < 16; i++) \ for (j = 1; j < end; j += 2) \ o_32[i] += transform[j][i] * src[j * sstep]; \ - TR_16(e_32, src, 1, 2 * sstep, SET, end/2); \ + TR_16(e_32, src, 1, 2 * sstep, SET, end / 2); \ \ for (i = 0; i < 16; i++) { \ assign(dst[i * dstep], e_32[i] + o_32[i]); \ @@ -232,21 +232,22 @@ static void FUNC(transform_4x4_luma)(int16_t *coeffs) } while (0) #define IDCT_VAR4(H) \ - int limit2 = FFMIN(col_limit + 4, H) + int limit2 = FFMIN(col_limit + 4, H) #define IDCT_VAR8(H) \ - int limit = FFMIN(col_limit, H); \ - int limit2 = FFMIN(col_limit + 4, H) + int limit = FFMIN(col_limit, H); \ + int limit2 = FFMIN(col_limit + 4, H) #define IDCT_VAR16(H) IDCT_VAR8(H) #define IDCT_VAR32(H) IDCT_VAR8(H) #define IDCT(H) \ static void FUNC(idct_ ## H ## x ## H )(int16_t *coeffs, \ - int col_limit) { \ + int col_limit) \ +{ \ int i; \ int shift = 7; \ int add = 1 << (shift - 1); \ int16_t *src = coeffs; \ - IDCT_VAR ##H(H); \ + IDCT_VAR ## H(H); \ \ for (i = 0; i < H; i++) { \ TR_ ## H(src, src, H, H, SCALE, limit2); \ ====================================================================== diff --cc libavcodec/hevc_cabac.c index 457e087,b01808f..4f90dab --- a/libavcodec/hevc_cabac.c +++ b/libavcodec/hevc_cabac.c @@@ -999,541 -867,6 +999,541 @@@ static av_always_inline int coeff_sign_ int ret = 0; for (i = 0; i < nb; i++) - ret = (ret << 1) | get_cabac_bypass(&s->HEVClc.cc); + ret = (ret << 1) | get_cabac_bypass(&s->HEVClc->cc); return ret; } + +void ff_hevc_hls_residual_coding(HEVCContext *s, int x0, int y0, + int log2_trafo_size, enum ScanType scan_idx, + int c_idx) +{ +#define GET_COORD(offset, n) \ + do { \ + x_c = (x_cg << 2) + scan_x_off[n]; \ + y_c = (y_cg << 2) + scan_y_off[n]; \ + } while (0) + HEVCLocalContext *lc = s->HEVClc; + int transform_skip_flag = 0; + + int last_significant_coeff_x, last_significant_coeff_y; + int last_scan_pos; + int n_end; + int num_coeff = 0; + int greater1_ctx = 1; + + int num_last_subset; + int x_cg_last_sig, y_cg_last_sig; + + const uint8_t *scan_x_cg, *scan_y_cg, *scan_x_off, *scan_y_off; + + ptrdiff_t stride = s->frame->linesize[c_idx]; + int hshift = s->ps.sps->hshift[c_idx]; + int vshift = s->ps.sps->vshift[c_idx]; + uint8_t *dst = &s->frame->data[c_idx][(y0 >> vshift) * stride + + ((x0 >> hshift) << s->ps.sps->pixel_shift)]; + int16_t *coeffs = (int16_t*)(c_idx ? lc->edge_emu_buffer2 : lc->edge_emu_buffer); + uint8_t significant_coeff_group_flag[8][8] = {{0}}; + int explicit_rdpcm_flag = 0; + int explicit_rdpcm_dir_flag; + + int trafo_size = 1 << log2_trafo_size; + int i; + int qp,shift,add,scale,scale_m; + static const uint8_t level_scale[] = { 40, 45, 51, 57, 64, 72 }; + const uint8_t *scale_matrix = NULL; + uint8_t dc_scale; + int pred_mode_intra = (c_idx == 0) ? lc->tu.intra_pred_mode : + lc->tu.intra_pred_mode_c; + + memset(coeffs, 0, trafo_size * trafo_size * sizeof(int16_t)); + + // Derive QP for dequant + if (!lc->cu.cu_transquant_bypass_flag) { + static const int qp_c[] = { 29, 30, 31, 32, 33, 33, 34, 34, 35, 35, 36, 36, 37, 37 }; + static const uint8_t rem6[51 + 4 * 6 + 1] = { + 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, 4, 5, 0, 1, 2, + 3, 4, 5, 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, 4, 5, + 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, + 4, 5, 0, 1, 2, 3, 4, 5, 0, 1 + }; + + static const uint8_t div6[51 + 4 * 6 + 1] = { + 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 2, 2, 2, 2, 2, 2, 3, 3, 3, + 3, 3, 3, 4, 4, 4, 4, 4, 4, 5, 5, 5, 5, 5, 5, 6, 6, 6, 6, 6, 6, + 7, 7, 7, 7, 7, 7, 8, 8, 8, 8, 8, 8, 9, 9, 9, 9, 9, 9, 10, 10, 10, 10, + 10, 10, 11, 11, 11, 11, 11, 11, 12, 12 + }; + int qp_y = lc->qp_y; + + if (s->ps.pps->transform_skip_enabled_flag && + log2_trafo_size <= s->ps.pps->log2_max_transform_skip_block_size) { + transform_skip_flag = hevc_transform_skip_flag_decode(s, c_idx); + } + + if (c_idx == 0) { + qp = qp_y + s->ps.sps->qp_bd_offset; + } else { + int qp_i, offset; + + if (c_idx == 1) + offset = s->ps.pps->cb_qp_offset + s->sh.slice_cb_qp_offset + + lc->tu.cu_qp_offset_cb; + else + offset = s->ps.pps->cr_qp_offset + s->sh.slice_cr_qp_offset + + lc->tu.cu_qp_offset_cr; + + qp_i = av_clip(qp_y + offset, - s->ps.sps->qp_bd_offset, 57); + if (s->ps.sps->chroma_format_idc == 1) { + if (qp_i < 30) + qp = qp_i; + else if (qp_i > 43) + qp = qp_i - 6; + else + qp = qp_c[qp_i - 30]; + } else { + if (qp_i > 51) + qp = 51; + else + qp = qp_i; + } + + qp += s->ps.sps->qp_bd_offset; + } + + shift = s->ps.sps->bit_depth + log2_trafo_size - 5; + add = 1 << (shift-1); + scale = level_scale[rem6[qp]] << (div6[qp]); + scale_m = 16; // default when no custom scaling lists. + dc_scale = 16; + + if (s->ps.sps->scaling_list_enable_flag && !(transform_skip_flag && log2_trafo_size > 2)) { + const ScalingList *sl = s->ps.pps->scaling_list_data_present_flag ? + &s->ps.pps->scaling_list : &s->ps.sps->scaling_list; + int matrix_id = lc->cu.pred_mode != MODE_INTRA; + + matrix_id = 3 * matrix_id + c_idx; + + scale_matrix = sl->sl[log2_trafo_size - 2][matrix_id]; + if (log2_trafo_size >= 4) + dc_scale = sl->sl_dc[log2_trafo_size - 4][matrix_id]; + } + } else { + shift = 0; + add = 0; + scale = 0; + dc_scale = 0; + } + + if (lc->cu.pred_mode == MODE_INTER && s->ps.sps->explicit_rdpcm_enabled_flag && + (transform_skip_flag || lc->cu.cu_transquant_bypass_flag)) { + explicit_rdpcm_flag = explicit_rdpcm_flag_decode(s, c_idx); + if (explicit_rdpcm_flag) { + explicit_rdpcm_dir_flag = explicit_rdpcm_dir_flag_decode(s, c_idx); + } + } + + last_significant_coeff_xy_prefix_decode(s, c_idx, log2_trafo_size, + &last_significant_coeff_x, &last_significant_coeff_y); + + if (last_significant_coeff_x > 3) { + int suffix = last_significant_coeff_suffix_decode(s, last_significant_coeff_x); + last_significant_coeff_x = (1 << ((last_significant_coeff_x >> 1) - 1)) * + (2 + (last_significant_coeff_x & 1)) + + suffix; + } + + if (last_significant_coeff_y > 3) { + int suffix = last_significant_coeff_suffix_decode(s, last_significant_coeff_y); + last_significant_coeff_y = (1 << ((last_significant_coeff_y >> 1) - 1)) * + (2 + (last_significant_coeff_y & 1)) + + suffix; + } + + if (scan_idx == SCAN_VERT) + FFSWAP(int, last_significant_coeff_x, last_significant_coeff_y); + + x_cg_last_sig = last_significant_coeff_x >> 2; + y_cg_last_sig = last_significant_coeff_y >> 2; + + switch (scan_idx) { + case SCAN_DIAG: { + int last_x_c = last_significant_coeff_x & 3; + int last_y_c = last_significant_coeff_y & 3; + + scan_x_off = ff_hevc_diag_scan4x4_x; + scan_y_off = ff_hevc_diag_scan4x4_y; + num_coeff = diag_scan4x4_inv[last_y_c][last_x_c]; + if (trafo_size == 4) { + scan_x_cg = scan_1x1; + scan_y_cg = scan_1x1; + } else if (trafo_size == 8) { + num_coeff += diag_scan2x2_inv[y_cg_last_sig][x_cg_last_sig] << 4; + scan_x_cg = diag_scan2x2_x; + scan_y_cg = diag_scan2x2_y; + } else if (trafo_size == 16) { + num_coeff += diag_scan4x4_inv[y_cg_last_sig][x_cg_last_sig] << 4; + scan_x_cg = ff_hevc_diag_scan4x4_x; + scan_y_cg = ff_hevc_diag_scan4x4_y; + } else { // trafo_size == 32 + num_coeff += diag_scan8x8_inv[y_cg_last_sig][x_cg_last_sig] << 4; + scan_x_cg = ff_hevc_diag_scan8x8_x; + scan_y_cg = ff_hevc_diag_scan8x8_y; + } + break; + } + case SCAN_HORIZ: + scan_x_cg = horiz_scan2x2_x; + scan_y_cg = horiz_scan2x2_y; + scan_x_off = horiz_scan4x4_x; + scan_y_off = horiz_scan4x4_y; + num_coeff = horiz_scan8x8_inv[last_significant_coeff_y][last_significant_coeff_x]; + break; + default: //SCAN_VERT + scan_x_cg = horiz_scan2x2_y; + scan_y_cg = horiz_scan2x2_x; + scan_x_off = horiz_scan4x4_y; + scan_y_off = horiz_scan4x4_x; + num_coeff = horiz_scan8x8_inv[last_significant_coeff_x][last_significant_coeff_y]; + break; + } + num_coeff++; + num_last_subset = (num_coeff - 1) >> 4; + + for (i = num_last_subset; i >= 0; i--) { + int n, m; + int x_cg, y_cg, x_c, y_c, pos; + int implicit_non_zero_coeff = 0; + int64_t trans_coeff_level; + int prev_sig = 0; + int offset = i << 4; + int rice_init = 0; + + uint8_t significant_coeff_flag_idx[16]; + uint8_t nb_significant_coeff_flag = 0; + + x_cg = scan_x_cg[i]; + y_cg = scan_y_cg[i]; + + if ((i < num_last_subset) && (i > 0)) { + int ctx_cg = 0; + if (x_cg < (1 << (log2_trafo_size - 2)) - 1) + ctx_cg += significant_coeff_group_flag[x_cg + 1][y_cg]; + if (y_cg < (1 << (log2_trafo_size - 2)) - 1) + ctx_cg += significant_coeff_group_flag[x_cg][y_cg + 1]; + + significant_coeff_group_flag[x_cg][y_cg] = + significant_coeff_group_flag_decode(s, c_idx, ctx_cg); + implicit_non_zero_coeff = 1; + } else { + significant_coeff_group_flag[x_cg][y_cg] = + ((x_cg == x_cg_last_sig && y_cg == y_cg_last_sig) || + (x_cg == 0 && y_cg == 0)); + } + + last_scan_pos = num_coeff - offset - 1; + + if (i == num_last_subset) { + n_end = last_scan_pos - 1; + significant_coeff_flag_idx[0] = last_scan_pos; + nb_significant_coeff_flag = 1; + } else { + n_end = 15; + } + + if (x_cg < ((1 << log2_trafo_size) - 1) >> 2) + prev_sig = !!significant_coeff_group_flag[x_cg + 1][y_cg]; + if (y_cg < ((1 << log2_trafo_size) - 1) >> 2) + prev_sig += (!!significant_coeff_group_flag[x_cg][y_cg + 1] << 1); + + if (significant_coeff_group_flag[x_cg][y_cg] && n_end >= 0) { + static const uint8_t ctx_idx_map[] = { + 0, 1, 4, 5, 2, 3, 4, 5, 6, 6, 8, 8, 7, 7, 8, 8, // log2_trafo_size == 2 + 1, 1, 1, 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, // prev_sig == 0 + 2, 2, 2, 2, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, // prev_sig == 1 + 2, 1, 0, 0, 2, 1, 0, 0, 2, 1, 0, 0, 2, 1, 0, 0, // prev_sig == 2 + 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2 // default + }; + const uint8_t *ctx_idx_map_p; + int scf_offset = 0; + if (s->ps.sps->transform_skip_context_enabled_flag && + (transform_skip_flag || lc->cu.cu_transquant_bypass_flag)) { + ctx_idx_map_p = (uint8_t*) &ctx_idx_map[4 * 16]; + if (c_idx == 0) { + scf_offset = 40; + } else { + scf_offset = 14 + 27; + } + } else { + if (c_idx != 0) + scf_offset = 27; + if (log2_trafo_size == 2) { + ctx_idx_map_p = (uint8_t*) &ctx_idx_map[0]; + } else { + ctx_idx_map_p = (uint8_t*) &ctx_idx_map[(prev_sig + 1) << 4]; + if (c_idx == 0) { + if ((x_cg > 0 || y_cg > 0)) + scf_offset += 3; + if (log2_trafo_size == 3) { + scf_offset += (scan_idx == SCAN_DIAG) ? 9 : 15; + } else { + scf_offset += 21; + } + } else { + if (log2_trafo_size == 3) + scf_offset += 9; + else + scf_offset += 12; + } + } + } + for (n = n_end; n > 0; n--) { + x_c = scan_x_off[n]; + y_c = scan_y_off[n]; + if (significant_coeff_flag_decode(s, x_c, y_c, scf_offset, ctx_idx_map_p)) { + significant_coeff_flag_idx[nb_significant_coeff_flag] = n; + nb_significant_coeff_flag++; + implicit_non_zero_coeff = 0; + } + } + if (implicit_non_zero_coeff == 0) { + if (s->ps.sps->transform_skip_context_enabled_flag && + (transform_skip_flag || lc->cu.cu_transquant_bypass_flag)) { + if (c_idx == 0) { + scf_offset = 42; + } else { + scf_offset = 16 + 27; + } + } else { + if (i == 0) { + if (c_idx == 0) + scf_offset = 0; + else + scf_offset = 27; + } else { + scf_offset = 2 + scf_offset; + } + } + if (significant_coeff_flag_decode_0(s, c_idx, scf_offset) == 1) { + significant_coeff_flag_idx[nb_significant_coeff_flag] = 0; + nb_significant_coeff_flag++; + } + } else { + significant_coeff_flag_idx[nb_significant_coeff_flag] = 0; + nb_significant_coeff_flag++; + } + } + + n_end = nb_significant_coeff_flag; + + + if (n_end) { + int first_nz_pos_in_cg; + int last_nz_pos_in_cg; + int c_rice_param = 0; + int first_greater1_coeff_idx = -1; + uint8_t coeff_abs_level_greater1_flag[8]; + uint16_t coeff_sign_flag; + int sum_abs = 0; + int sign_hidden; + int sb_type; + + + // initialize first elem of coeff_bas_level_greater1_flag + int ctx_set = (i > 0 && c_idx == 0) ? 2 : 0; + + if (s->ps.sps->persistent_rice_adaptation_enabled_flag) { + if (!transform_skip_flag && !lc->cu.cu_transquant_bypass_flag) + sb_type = 2 * (c_idx == 0 ? 1 : 0); + else + sb_type = 2 * (c_idx == 0 ? 1 : 0) + 1; + c_rice_param = lc->stat_coeff[sb_type] / 4; + } + + if (!(i == num_last_subset) && greater1_ctx == 0) + ctx_set++; + greater1_ctx = 1; + last_nz_pos_in_cg = significant_coeff_flag_idx[0]; + + for (m = 0; m < (n_end > 8 ? 8 : n_end); m++) { + int inc = (ctx_set << 2) + greater1_ctx; + coeff_abs_level_greater1_flag[m] = + coeff_abs_level_greater1_flag_decode(s, c_idx, inc); + if (coeff_abs_level_greater1_flag[m]) { + greater1_ctx = 0; + if (first_greater1_coeff_idx == -1) + first_greater1_coeff_idx = m; + } else if (greater1_ctx > 0 && greater1_ctx < 3) { + greater1_ctx++; + } + } + first_nz_pos_in_cg = significant_coeff_flag_idx[n_end - 1]; + + if (lc->cu.cu_transquant_bypass_flag || + (lc->cu.pred_mode == MODE_INTRA && + s->ps.sps->implicit_rdpcm_enabled_flag && transform_skip_flag && + (pred_mode_intra == 10 || pred_mode_intra == 26 )) || + explicit_rdpcm_flag) + sign_hidden = 0; + else + sign_hidden = (last_nz_pos_in_cg - first_nz_pos_in_cg >= 4); + + if (first_greater1_coeff_idx != -1) { + coeff_abs_level_greater1_flag[first_greater1_coeff_idx] += coeff_abs_level_greater2_flag_decode(s, c_idx, ctx_set); + } + if (!s->ps.pps->sign_data_hiding_flag || !sign_hidden ) { + coeff_sign_flag = coeff_sign_flag_decode(s, nb_significant_coeff_flag) << (16 - nb_significant_coeff_flag); + } else { + coeff_sign_flag = coeff_sign_flag_decode(s, nb_significant_coeff_flag - 1) << (16 - (nb_significant_coeff_flag - 1)); + } + + for (m = 0; m < n_end; m++) { + n = significant_coeff_flag_idx[m]; + GET_COORD(offset, n); + if (m < 8) { + trans_coeff_level = 1 + coeff_abs_level_greater1_flag[m]; + if (trans_coeff_level == ((m == first_greater1_coeff_idx) ? 3 : 2)) { + int last_coeff_abs_level_remaining = coeff_abs_level_remaining_decode(s, c_rice_param); + + trans_coeff_level += last_coeff_abs_level_remaining; + if (trans_coeff_level > (3 << c_rice_param)) + c_rice_param = s->ps.sps->persistent_rice_adaptation_enabled_flag ? c_rice_param + 1 : FFMIN(c_rice_param + 1, 4); + if (s->ps.sps->persistent_rice_adaptation_enabled_flag && !rice_init) { + int c_rice_p_init = lc->stat_coeff[sb_type] / 4; + if (last_coeff_abs_level_remaining >= (3 << c_rice_p_init)) + lc->stat_coeff[sb_type]++; + else if (2 * last_coeff_abs_level_remaining < (1 << c_rice_p_init)) + if (lc->stat_coeff[sb_type] > 0) + lc->stat_coeff[sb_type]--; + rice_init = 1; + } + } + } else { + int last_coeff_abs_level_remaining = coeff_abs_level_remaining_decode(s, c_rice_param); + + trans_coeff_level = 1 + last_coeff_abs_level_remaining; + if (trans_coeff_level > (3 << c_rice_param)) + c_rice_param = s->ps.sps->persistent_rice_adaptation_enabled_flag ? c_rice_param + 1 : FFMIN(c_rice_param + 1, 4); + if (s->ps.sps->persistent_rice_adaptation_enabled_flag && !rice_init) { + int c_rice_p_init = lc->stat_coeff[sb_type] / 4; + if (last_coeff_abs_level_remaining >= (3 << c_rice_p_init)) + lc->stat_coeff[sb_type]++; + else if (2 * last_coeff_abs_level_remaining < (1 << c_rice_p_init)) + if (lc->stat_coeff[sb_type] > 0) + lc->stat_coeff[sb_type]--; + rice_init = 1; + } + } + if (s->ps.pps->sign_data_hiding_flag && sign_hidden) { + sum_abs += trans_coeff_level; + if (n == first_nz_pos_in_cg && (sum_abs&1)) + trans_coeff_level = -trans_coeff_level; + } + if (coeff_sign_flag >> 15) + trans_coeff_level = -trans_coeff_level; + coeff_sign_flag <<= 1; + if(!lc->cu.cu_transquant_bypass_flag) { + if (s->ps.sps->scaling_list_enable_flag && !(transform_skip_flag && log2_trafo_size > 2)) { + if(y_c || x_c || log2_trafo_size < 4) { + switch(log2_trafo_size) { + case 3: pos = (y_c << 3) + x_c; break; + case 4: pos = ((y_c >> 1) << 3) + (x_c >> 1); break; + case 5: pos = ((y_c >> 2) << 3) + (x_c >> 2); break; + default: pos = (y_c << 2) + x_c; break; + } + scale_m = scale_matrix[pos]; + } else { + scale_m = dc_scale; + } + } + trans_coeff_level = (trans_coeff_level * (int64_t)scale * (int64_t)scale_m + add) >> shift; + if(trans_coeff_level < 0) { + if((~trans_coeff_level) & 0xFffffffffff8000) + trans_coeff_level = -32768; + } else { + if(trans_coeff_level & 0xffffffffffff8000) + trans_coeff_level = 32767; + } + } + coeffs[y_c * trafo_size + x_c] = trans_coeff_level; + } + } + } + + if (lc->cu.cu_transquant_bypass_flag) { + if (explicit_rdpcm_flag || (s->ps.sps->implicit_rdpcm_enabled_flag && + (pred_mode_intra == 10 || pred_mode_intra == 26))) { + int mode = s->ps.sps->implicit_rdpcm_enabled_flag ? (pred_mode_intra == 26) : explicit_rdpcm_dir_flag; + + s->hevcdsp.transform_rdpcm(coeffs, log2_trafo_size, mode); + } + } else { + if (transform_skip_flag) { + int rot = s->ps.sps->transform_skip_rotation_enabled_flag && + log2_trafo_size == 2 && + lc->cu.pred_mode == MODE_INTRA; + if (rot) { + for (i = 0; i < 8; i++) + FFSWAP(int16_t, coeffs[i], coeffs[16 - i - 1]); + } + + s->hevcdsp.dequant(coeffs, log2_trafo_size); + + if (explicit_rdpcm_flag || (s->ps.sps->implicit_rdpcm_enabled_flag && + lc->cu.pred_mode == MODE_INTRA && + (pred_mode_intra == 10 || pred_mode_intra == 26))) { + int mode = explicit_rdpcm_flag ? explicit_rdpcm_dir_flag : (pred_mode_intra == 26); + + s->hevcdsp.transform_rdpcm(coeffs, log2_trafo_size, mode); + } + } else if (lc->cu.pred_mode == MODE_INTRA && c_idx == 0 && log2_trafo_size == 2) { + s->hevcdsp.transform_4x4_luma(coeffs); + } else { + int max_xy = FFMAX(last_significant_coeff_x, last_significant_coeff_y); + if (max_xy == 0) - s->hevcdsp.idct_dc[log2_trafo_size-2](coeffs); ++ s->hevcdsp.idct_dc[log2_trafo_size - 2](coeffs); + else { + int col_limit = last_significant_coeff_x + last_significant_coeff_y + 4; + if (max_xy < 4) + col_limit = FFMIN(4, col_limit); + else if (max_xy < 8) + col_limit = FFMIN(8, col_limit); + else if (max_xy < 12) + col_limit = FFMIN(24, col_limit); - s->hevcdsp.idct[log2_trafo_size-2](coeffs, col_limit); ++ s->hevcdsp.idct[log2_trafo_size - 2](coeffs, col_limit); + } + } + } + if (lc->tu.cross_pf) { + int16_t *coeffs_y = (int16_t*)lc->edge_emu_buffer; + + for (i = 0; i < (trafo_size * trafo_size); i++) { + coeffs[i] = coeffs[i] + ((lc->tu.res_scale_val * coeffs_y[i]) >> 3); + } + } + s->hevcdsp.add_residual[log2_trafo_size-2](dst, coeffs, stride); +} + +void ff_hevc_hls_mvd_coding(HEVCContext *s, int x0, int y0, int log2_cb_size) +{ + HEVCLocalContext *lc = s->HEVClc; + int x = abs_mvd_greater0_flag_decode(s); + int y = abs_mvd_greater0_flag_decode(s); + + if (x) + x += abs_mvd_greater1_flag_decode(s); + if (y) + y += abs_mvd_greater1_flag_decode(s); + + switch (x) { + case 2: lc->pu.mvd.x = mvd_decode(s); break; + case 1: lc->pu.mvd.x = mvd_sign_flag_decode(s); break; + case 0: lc->pu.mvd.x = 0; break; + } + + switch (y) { + case 2: lc->pu.mvd.y = mvd_decode(s); break; + case 1: lc->pu.mvd.y = mvd_sign_flag_decode(s); break; + case 0: lc->pu.mvd.y = 0; break; + } +} +

1 0