September 2015 - ffmpeg-cvslog

avcodec/pngdec: mark previous_picture as done on end of decode_frame_common()
by git＠videolan.org 30 Sep '15

30 Sep '15

ffmpeg | branch: master | Michael Niedermayer <michael(a)niedermayer.cc> | Mon Sep 28 02:42:39 2015 +0200| [9e9d731b510666edac3a2b450dfc331279db818b] | committer: Michael Niedermayer avcodec/pngdec: mark previous_picture as done on end of decode_frame_common() Fixes deadlock with threads Found-by: Paul B Mahol Reviewed-by: Paul B Mahol <onemda(a)gmail.com> Signed-off-by: Michael Niedermayer <michael(a)niedermayer.cc> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=9e9d731b510666eda… --- libavcodec/pngdec.c | 2 ++ 1 file changed, 2 insertions(+) diff --git a/libavcodec/pngdec.c b/libavcodec/pngdec.c index 1b47b18..bd84ee9 100644 --- a/libavcodec/pngdec.c +++ b/libavcodec/pngdec.c @@ -1217,6 +1217,7 @@ exit_loop: } } ff_thread_report_progress(&s->picture, INT_MAX, 0); + ff_thread_report_progress(&s->previous_picture, INT_MAX, 0); av_frame_set_metadata(p, metadata); metadata = NULL; @@ -1225,6 +1226,7 @@ exit_loop: fail: av_dict_free(&metadata); ff_thread_report_progress(&s->picture, INT_MAX, 0); + ff_thread_report_progress(&s->previous_picture, INT_MAX, 0); return ret; }

1 0

avcodec/vp8: Do not use num_coeff_partitions in thread/buffer setup
by git＠videolan.org 30 Sep '15

30 Sep '15

ffmpeg | branch: master | Michael Niedermayer <michael(a)niedermayer.cc> | Wed Sep 30 13:10:48 2015 +0200| [dabea74d0e82ea80cd344f630497cafcb3ef872c] | committer: Michael Niedermayer avcodec/vp8: Do not use num_coeff_partitions in thread/buffer setup The variable is not a constant and can lead to race conditions Fixes: repro.webm (not reproducable with FFmpeg alone) Found-by: Dale Curtis <dalecurtis(a)google.com> Tested-by: Dale Curtis <dalecurtis(a)google.com> Signed-off-by: Michael Niedermayer <michael(a)niedermayer.cc> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=dabea74d0e82ea80c… --- libavcodec/vp8.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/libavcodec/vp8.c b/libavcodec/vp8.c index 730871a..cb0c7cd 100644 --- a/libavcodec/vp8.c +++ b/libavcodec/vp8.c @@ -164,7 +164,7 @@ int update_dimensions(VP8Context *s, int width, int height, int is_vp7) s->mb_height = (s->avctx->coded_height + 15) / 16; s->mb_layout = is_vp7 || avctx->active_thread_type == FF_THREAD_SLICE && - FFMIN(s->num_coeff_partitions, avctx->thread_count) > 1; + avctx->thread_count > 1; if (!s->mb_layout) { // Frame threading and one thread s->macroblocks_base = av_mallocz((s->mb_width + s->mb_height * 2 + 1) * sizeof(*s->macroblocks));

1 0

avfilter/vf_maskedmerge: get rid of MaskedMergeContext from functions that do actual work
by git＠videolan.org 30 Sep '15

30 Sep '15

ffmpeg | branch: master | Paul B Mahol <onemda(a)gmail.com> | Wed Sep 30 22:35:31 2015 +0200| [30ce6fd1069e6e65edae03da162871df896e5283] | committer: Paul B Mahol avfilter/vf_maskedmerge: get rid of MaskedMergeContext from functions that do actual work Signed-off-by: Paul B Mahol <onemda(a)gmail.com> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=30ce6fd1069e6e65e… --- libavfilter/vf_maskedmerge.c | 26 ++++++++++++-------------- 1 file changed, 12 insertions(+), 14 deletions(-) diff --git a/libavfilter/vf_maskedmerge.c b/libavfilter/vf_maskedmerge.c index 67f74be..41bd039 100644 --- a/libavfilter/vf_maskedmerge.c +++ b/libavfilter/vf_maskedmerge.c @@ -35,11 +35,11 @@ typedef struct MaskedMergeContext { int max, half, depth; FFFrameSync fs; - void (*maskedmerge)(struct MaskedMergeContext *s, - const uint8_t *bsrc, int blinesize, + void (*maskedmerge)(const uint8_t *bsrc, int blinesize, const uint8_t *osrc, int olinesize, const uint8_t *msrc, int mlinesize, - uint8_t *dst, int dlinesize, int w, int h); + uint8_t *dst, int dlinesize, int w, int h, + int max, int half, int shift); } MaskedMergeContext; #define OFFSET(x) offsetof(MaskedMergeContext, x) @@ -109,11 +109,12 @@ static int process_frame(FFFrameSync *fs) continue; } - s->maskedmerge(s, base->data[p], base->linesize[p], + s->maskedmerge(base->data[p], base->linesize[p], overlay->data[p], overlay->linesize[p], mask->data[p], mask->linesize[p], out->data[p], out->linesize[p], - s->width[p], s->height[p]); + s->width[p], s->height[p], + s->max, s->half, s->depth); } } out->pts = av_rescale_q(base->pts, s->fs.time_base, outlink->time_base); @@ -121,11 +122,11 @@ static int process_frame(FFFrameSync *fs) return ff_filter_frame(outlink, out); } -static void maskedmerge8(MaskedMergeContext *s, - const uint8_t *bsrc, int blinesize, +static void maskedmerge8(const uint8_t *bsrc, int blinesize, const uint8_t *osrc, int olinesize, const uint8_t *msrc, int mlinesize, - uint8_t *dst, int dlinesize, int w, int h) + uint8_t *dst, int dlinesize, int w, int h, + int max, int half, int shift) { int x, y; @@ -141,15 +142,12 @@ static void maskedmerge8(MaskedMergeContext *s, } } -static void maskedmerge16(MaskedMergeContext *s, - const uint8_t *bbsrc, int blinesize, +static void maskedmerge16(const uint8_t *bbsrc, int blinesize, const uint8_t *oosrc, int olinesize, const uint8_t *mmsrc, int mlinesize, - uint8_t *ddst, int dlinesize, int w, int h) + uint8_t *ddst, int dlinesize, int w, int h, + int max, int half, int shift) { - const int max = s->max; - const int half = s->half; - const int shift = s->depth; const uint16_t *bsrc = (const uint16_t *)bbsrc; const uint16_t *osrc = (const uint16_t *)oosrc; const uint16_t *msrc = (const uint16_t *)mmsrc;

1 0

avfilter/vf_maskedmerge: rewrite and remove some duplicated code
by git＠videolan.org 30 Sep '15

30 Sep '15

ffmpeg | branch: master | Paul B Mahol <onemda(a)gmail.com> | Wed Sep 30 22:23:26 2015 +0200| [1da1574002e31f95543bae1b103ed007485d9517] | committer: Paul B Mahol avfilter/vf_maskedmerge: rewrite and remove some duplicated code Signed-off-by: Paul B Mahol <onemda(a)gmail.com> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=1da1574002e31f955… --- libavfilter/vf_maskedmerge.c | 103 +++++++++++++++++++++--------------------- 1 file changed, 52 insertions(+), 51 deletions(-) diff --git a/libavfilter/vf_maskedmerge.c b/libavfilter/vf_maskedmerge.c index 72f9fb3..67f74be 100644 --- a/libavfilter/vf_maskedmerge.c +++ b/libavfilter/vf_maskedmerge.c @@ -35,8 +35,11 @@ typedef struct MaskedMergeContext { int max, half, depth; FFFrameSync fs; - void (*maskedmerge)(struct MaskedMergeContext *s, const AVFrame *base, - const AVFrame *overlay, const AVFrame *mask, AVFrame *out); + void (*maskedmerge)(struct MaskedMergeContext *s, + const uint8_t *bsrc, int blinesize, + const uint8_t *osrc, int olinesize, + const uint8_t *msrc, int mlinesize, + uint8_t *dst, int dlinesize, int w, int h); } MaskedMergeContext; #define OFFSET(x) offsetof(MaskedMergeContext, x) @@ -92,78 +95,76 @@ static int process_frame(FFFrameSync *fs) if (!out) return AVERROR(ENOMEM); } else { + int p; + out = ff_get_video_buffer(outlink, outlink->w, outlink->h); if (!out) return AVERROR(ENOMEM); av_frame_copy_props(out, base); - s->maskedmerge(s, base, overlay, mask, out); + for (p = 0; p < s->nb_planes; p++) { + if (!((1 << p) & s->planes)) { + av_image_copy_plane(out->data[p], out->linesize[p], base->data[p], base->linesize[p], + s->width[p], s->height[p]); + continue; + } + + s->maskedmerge(s, base->data[p], base->linesize[p], + overlay->data[p], overlay->linesize[p], + mask->data[p], mask->linesize[p], + out->data[p], out->linesize[p], + s->width[p], s->height[p]); + } } out->pts = av_rescale_q(base->pts, s->fs.time_base, outlink->time_base); return ff_filter_frame(outlink, out); } -static void maskedmerge8(MaskedMergeContext *s, const AVFrame *base, - const AVFrame *overlay, const AVFrame *mask, AVFrame *out) +static void maskedmerge8(MaskedMergeContext *s, + const uint8_t *bsrc, int blinesize, + const uint8_t *osrc, int olinesize, + const uint8_t *msrc, int mlinesize, + uint8_t *dst, int dlinesize, int w, int h) { - int p, x, y; - - for (p = 0; p < s->nb_planes; p++) { - const uint8_t *bsrc = base->data[p]; - const uint8_t *osrc = overlay->data[p]; - const uint8_t *msrc = mask->data[p]; - uint8_t *dst = out->data[p]; - - if (!((1 << p) & s->planes)) { - av_image_copy_plane(dst, out->linesize[p], bsrc, base->linesize[p], - s->width[p], s->height[p]); - continue; - } + int x, y; - for (y = 0; y < s->height[p]; y++) { - for (x = 0; x < s->width[p]; x++) { - dst[x] = ((256 - msrc[x]) * bsrc[x] + msrc[x] * osrc[x] + 128) >> 8; - } - - dst += out->linesize[p]; - bsrc += base->linesize[p]; - osrc += overlay->linesize[p]; - msrc += mask->linesize[p]; + for (y = 0; y < h; y++) { + for (x = 0; x < w; x++) { + dst[x] = ((256 - msrc[x]) * bsrc[x] + msrc[x] * osrc[x] + 128) >> 8; } + + dst += dlinesize; + bsrc += blinesize; + osrc += olinesize; + msrc += mlinesize; } } -static void maskedmerge16(MaskedMergeContext *s, const AVFrame *base, - const AVFrame *overlay, const AVFrame *mask, AVFrame *out) +static void maskedmerge16(MaskedMergeContext *s, + const uint8_t *bbsrc, int blinesize, + const uint8_t *oosrc, int olinesize, + const uint8_t *mmsrc, int mlinesize, + uint8_t *ddst, int dlinesize, int w, int h) { const int max = s->max; const int half = s->half; const int shift = s->depth; - int p, x, y; - - for (p = 0; p < s->nb_planes; p++) { - const uint16_t *bsrc = (const uint16_t *)base->data[p]; - const uint16_t *osrc = (const uint16_t *)overlay->data[p]; - const uint16_t *msrc = (const uint16_t *)mask->data[p]; - uint16_t *dst = (uint16_t *)out->data[p]; - - if (!((1 << p) & s->planes)) { - av_image_copy_plane(out->data[p], out->linesize[p], base->data[p], base->linesize[p], - s->width[p], s->height[p]); - continue; + const uint16_t *bsrc = (const uint16_t *)bbsrc; + const uint16_t *osrc = (const uint16_t *)oosrc; + const uint16_t *msrc = (const uint16_t *)mmsrc; + uint16_t *dst = (uint16_t *)ddst; + int x, y; + + for (y = 0; y < h; y++) { + for (x = 0; x < w; x++) { + dst[x] = ((max - msrc[x]) * bsrc[x] + msrc[x] * osrc[x] + half) >> shift; } - for (y = 0; y < s->height[p]; y++) { - for (x = 0; x < s->width[p]; x++) { - dst[x] = ((max - msrc[x]) * bsrc[x] + msrc[x] * osrc[x] + half) >> shift; - } - - dst += out->linesize[p] / 2; - bsrc += base->linesize[p] / 2; - osrc += overlay->linesize[p] / 2; - msrc += mask->linesize[p] / 2; - } + dst += dlinesize / 2; + bsrc += blinesize / 2; + osrc += olinesize / 2; + msrc += mlinesize / 2; } }

1 0

avcodec/mpeg12dec: Initialize chroma_format to 1
by git＠videolan.org 30 Sep '15

30 Sep '15

ffmpeg | branch: master | Michael Niedermayer <michael(a)niedermayer.cc> | Wed Sep 30 21:16:29 2015 +0200| [9b4dd0f87696353f10778d4c0ba80d9949c6fbca] | committer: Michael Niedermayer avcodec/mpeg12dec: Initialize chroma_format to 1 This ensures that chroma format is never set to an invalid value Signed-off-by: Michael Niedermayer <michael(a)niedermayer.cc> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=9b4dd0f87696353f1… --- libavcodec/mpeg12dec.c | 1 + 1 file changed, 1 insertion(+) diff --git a/libavcodec/mpeg12dec.c b/libavcodec/mpeg12dec.c index 5d916d1..21ccf6c 100644 --- a/libavcodec/mpeg12dec.c +++ b/libavcodec/mpeg12dec.c @@ -1130,6 +1130,7 @@ static av_cold int mpeg_decode_init(AVCodecContext *avctx) ff_mpeg12_common_init(&s->mpeg_enc_ctx); ff_mpeg12_init_vlcs(); + s2->chroma_format = 1; s->mpeg_enc_ctx_allocated = 0; s->mpeg_enc_ctx.picture_number = 0; s->repeat_field = 0;

1 0

avcodec/mp3: fix skipping zeros
by git＠videolan.org 30 Sep '15

30 Sep '15

ffmpeg | branch: master | wm4 <nfxjfg(a)googlemail.com> | Wed Sep 30 14:53:35 2015 +0200| [cb1da9fb8d71bb611a7b0028914c97afc3f5711d] | committer: wm4 avcodec/mp3: fix skipping zeros Commits 43bc5cf9 and c5371f77 add code for skipping initial zeros in mp3 packets. This code forgot to report to the user that data was skipped at all. Since audio codecs allow partial packet decoding, the user application has to rely on the return value. It will remove the data reported as consumed by the decoder, and feed it to the decoder again. This resulted in the mp3 frame after the zero region to be decoded over and over again, until the zero region was finally skipped by the application. Fix this by including the amount of skipped bytes to the number of consumed bytes returned by the decode call. Fixes trac ticket #4890. > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=cb1da9fb8d71bb611… --- libavcodec/mpegaudiodec_template.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) diff --git a/libavcodec/mpegaudiodec_template.c b/libavcodec/mpegaudiodec_template.c index b7b9dbd..0f32ac7 100644 --- a/libavcodec/mpegaudiodec_template.c +++ b/libavcodec/mpegaudiodec_template.c @@ -1657,9 +1657,11 @@ static int decode_frame(AVCodecContext * avctx, void *data, int *got_frame_ptr, uint32_t header; int ret; + int skipped = 0; while(buf_size && !*buf){ buf++; buf_size--; + skipped++; } if (buf_size < HEADER_SIZE) @@ -1714,7 +1716,7 @@ static int decode_frame(AVCodecContext * avctx, void *data, int *got_frame_ptr, return ret; } s->frame_size = 0; - return buf_size; + return buf_size + skipped; } static void mp_flush(MPADecodeContext *ctx)

1 0

avfilter/af_rubberband: flush only if there is something available
by git＠videolan.org 30 Sep '15

30 Sep '15

ffmpeg | branch: master | Paul B Mahol <onemda(a)gmail.com> | Wed Sep 30 13:02:03 2015 +0200| [6ce02126cec5ceb4c0f0be8099aba89d61345c09] | committer: Paul B Mahol avfilter/af_rubberband: flush only if there is something available Signed-off-by: Paul B Mahol <onemda(a)gmail.com> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=6ce02126cec5ceb4c… --- libavfilter/af_rubberband.c | 40 +++++++++++++++++++++------------------- 1 file changed, 21 insertions(+), 19 deletions(-) diff --git a/libavfilter/af_rubberband.c b/libavfilter/af_rubberband.c index abd6483..0a15fdc 100644 --- a/libavfilter/af_rubberband.c +++ b/libavfilter/af_rubberband.c @@ -175,29 +175,31 @@ static int request_frame(AVFilterLink *outlink) ret = ff_request_frame(ctx->inputs[0]); if (ret == AVERROR_EOF && !s->flushed) { - AVFrame *out = ff_get_audio_buffer(inlink, 1); - int nb_samples; + if (rubberband_available(s->rbs) > 0) { + AVFrame *out = ff_get_audio_buffer(inlink, 1); + int nb_samples; - if (!out) - return AVERROR(ENOMEM); - - rubberband_process(s->rbs, (const float *const *)out->data, 1, 1); - av_frame_free(&out); - s->flushed = 1; - nb_samples = rubberband_available(s->rbs); - - if (nb_samples > 0) { - out = ff_get_audio_buffer(inlink, nb_samples); if (!out) return AVERROR(ENOMEM); - out->pts = av_rescale_q(s->nb_samples_out, - (AVRational){ 1, outlink->sample_rate }, - outlink->time_base); - nb_samples = rubberband_retrieve(s->rbs, (float *const *)out->data, nb_samples); - out->nb_samples = nb_samples; - ret = ff_filter_frame(outlink, out); - s->nb_samples_out += nb_samples; + + rubberband_process(s->rbs, (const float *const *)out->data, 1, 1); + av_frame_free(&out); + nb_samples = rubberband_available(s->rbs); + + if (nb_samples > 0) { + out = ff_get_audio_buffer(inlink, nb_samples); + if (!out) + return AVERROR(ENOMEM); + out->pts = av_rescale_q(s->nb_samples_out, + (AVRational){ 1, outlink->sample_rate }, + outlink->time_base); + nb_samples = rubberband_retrieve(s->rbs, (float *const *)out->data, nb_samples); + out->nb_samples = nb_samples; + ret = ff_filter_frame(outlink, out); + s->nb_samples_out += nb_samples; + } } + s->flushed = 1; av_log(ctx, AV_LOG_DEBUG, "nb_samples_in %"PRId64" nb_samples_out %"PRId64"\n", s->nb_samples_in, s->nb_samples_out); }

1 0

x86/hevc_sao: move 10/12bit functions into a separate file
by git＠videolan.org 30 Sep '15

30 Sep '15

ffmpeg | branch: master | James Almer <jamrial(a)gmail.com> | Mon Sep 28 00:58:01 2015 -0300| [3178931a14f5accf028495943f5bc3391f5043c8] | committer: James Almer x86/hevc_sao: move 10/12bit functions into a separate file Tested-by: Michael Niedermayer <michaelni(a)gmx.at> Signed-off-by: James Almer <jamrial(a)gmail.com> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=3178931a14f5accf0… --- libavcodec/x86/Makefile | 3 +- libavcodec/x86/hevc_sao.asm | 394 +++++---------------------------- libavcodec/x86/hevc_sao_10bit.asm | 433 +++++++++++++++++++++++++++++++++++++ 3 files changed, 490 insertions(+), 340 deletions(-) diff --git a/libavcodec/x86/Makefile b/libavcodec/x86/Makefile index b3cfb0b..febaccd 100644 --- a/libavcodec/x86/Makefile +++ b/libavcodec/x86/Makefile @@ -140,7 +140,8 @@ YASM-OBJS-$(CONFIG_HEVC_DECODER) += x86/hevc_mc.o \ x86/hevc_deblock.o \ x86/hevc_idct.o \ x86/hevc_res_add.o \ - x86/hevc_sao.o + x86/hevc_sao.o \ + x86/hevc_sao_10bit.o YASM-OBJS-$(CONFIG_JPEG2000_DECODER) += x86/jpeg2000dsp.o YASM-OBJS-$(CONFIG_MLP_DECODER) += x86/mlpdsp.o YASM-OBJS-$(CONFIG_MPEG4_DECODER) += x86/xvididct.o diff --git a/libavcodec/x86/hevc_sao.asm b/libavcodec/x86/hevc_sao.asm index fa45a24..888a28a 100644 --- a/libavcodec/x86/hevc_sao.asm +++ b/libavcodec/x86/hevc_sao.asm @@ -1,5 +1,5 @@ ;****************************************************************************** -;* SIMD optimized SAO functions for HEVC decoding +;* SIMD optimized SAO functions for HEVC 8bit decoding ;* ;* Copyright (c) 2013 Pierre-Edouard LEPERE ;* Copyright (c) 2014 James Almer @@ -25,27 +25,18 @@ SECTION_RODATA 32 -pw_mask10: times 16 dw 0x03FF -pw_mask12: times 16 dw 0x0FFF -pw_m2: times 16 dw -2 pb_edge_shuffle: times 2 db 1, 2, 0, 3, 4, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1 pb_eo: db -1, 0, 1, 0, 0, -1, 0, 1, -1, -1, 1, 1, 1, -1, -1, 1 -cextern pw_m1 -cextern pw_1 -cextern pw_2 cextern pb_1 cextern pb_2 SECTION .text -%define MAX_PB_SIZE 64 -%define PADDING_SIZE 32 ; AV_INPUT_BUFFER_PADDING_SIZE - ;****************************************************************************** ;SAO Band Filter ;****************************************************************************** -%macro HEVC_SAO_BAND_FILTER_INIT 1 +%macro HEVC_SAO_BAND_FILTER_INIT 0 and leftq, 31 movd xm0, leftd add leftq, 1 @@ -76,9 +67,6 @@ SECTION .text %endif %if ARCH_X86_64 -%if %1 > 8 - mova m13, [pw_mask %+ %1] -%endif pxor m14, m14 %else ; ARCH_X86_32 @@ -90,9 +78,6 @@ SECTION .text mova [rsp+mmsize*5], m5 mova [rsp+mmsize*6], m6 pxor m0, m0 -%if %1 > 8 - mova m1, [pw_mask %+ %1] -%endif %assign MMSIZE mmsize %define m14 m0 %define m13 m1 @@ -103,49 +88,49 @@ DEFINE_ARGS dst, src, dststride, srcstride, offset, height mov heightd, r7m %endmacro -%macro HEVC_SAO_BAND_FILTER_COMPUTE 3 - psraw %2, %3, %1-5 +%macro HEVC_SAO_BAND_FILTER_COMPUTE 2 + psraw %1, %2, 3 %if ARCH_X86_64 - pcmpeqw m10, %2, m0 - pcmpeqw m11, %2, m1 - pcmpeqw m12, %2, m2 - pcmpeqw %2, m3 + pcmpeqw m10, %1, m0 + pcmpeqw m11, %1, m1 + pcmpeqw m12, %1, m2 + pcmpeqw %1, m3 pand m10, m4 pand m11, m5 pand m12, m6 - pand %2, m7 + pand %1, m7 por m10, m11 - por m12, %2 + por m12, %1 por m10, m12 - paddw %3, m10 + paddw %2, m10 %else ; ARCH_X86_32 - pcmpeqw m4, %2, [rsp+MMSIZE*0] - pcmpeqw m5, %2, [rsp+MMSIZE*1] - pcmpeqw m6, %2, [rsp+MMSIZE*2] - pcmpeqw %2, [rsp+MMSIZE*3] + pcmpeqw m4, %1, [rsp+MMSIZE*0] + pcmpeqw m5, %1, [rsp+MMSIZE*1] + pcmpeqw m6, %1, [rsp+MMSIZE*2] + pcmpeqw %1, [rsp+MMSIZE*3] pand m4, [rsp+MMSIZE*4] pand m5, [rsp+MMSIZE*5] pand m6, [rsp+MMSIZE*6] - pand %2, m7 + pand %1, m7 por m4, m5 - por m6, %2 + por m6, %1 por m4, m6 - paddw %3, m4 + paddw %2, m4 %endif ; ARCH %endmacro ;void ff_hevc_sao_band_filter_<width>_8_<opt>(uint8_t *_dst, uint8_t *_src, ptrdiff_t _stride_dst, ptrdiff_t _stride_src, ; int16_t *sao_offset_val, int sao_left_class, int width, int height); -%macro HEVC_SAO_BAND_FILTER_8 2 +%macro HEVC_SAO_BAND_FILTER 2 cglobal hevc_sao_band_filter_%1_8, 6, 6, 15, 7*mmsize*ARCH_X86_32, dst, src, dststride, srcstride, offset, left - HEVC_SAO_BAND_FILTER_INIT 8 + HEVC_SAO_BAND_FILTER_INIT align 16 .loop: %if %1 == 8 movq m8, [srcq] punpcklbw m8, m14 - HEVC_SAO_BAND_FILTER_COMPUTE 8, m9, m8 + HEVC_SAO_BAND_FILTER_COMPUTE m9, m8 packuswb m8, m14 movq [dstq], m8 %endif ; %1 == 8 @@ -154,9 +139,9 @@ align 16 %rep %2 mova m13, [srcq + i] punpcklbw m8, m13, m14 - HEVC_SAO_BAND_FILTER_COMPUTE 8, m9, m8 + HEVC_SAO_BAND_FILTER_COMPUTE m9, m8 punpckhbw m13, m14 - HEVC_SAO_BAND_FILTER_COMPUTE 8, m9, m13 + HEVC_SAO_BAND_FILTER_COMPUTE m9, m13 packuswb m8, m13 mova [dstq + i], m8 %assign i i+mmsize @@ -167,9 +152,9 @@ INIT_XMM cpuname mova m13, [srcq + i] punpcklbw m8, m13, m14 - HEVC_SAO_BAND_FILTER_COMPUTE 8, m9, m8 + HEVC_SAO_BAND_FILTER_COMPUTE m9, m8 punpckhbw m13, m14 - HEVC_SAO_BAND_FILTER_COMPUTE 8, m9, m13 + HEVC_SAO_BAND_FILTER_COMPUTE m9, m13 packuswb m8, m13 mova [dstq + i], m8 %if cpuflag(avx2) @@ -184,76 +169,13 @@ INIT_YMM cpuname REP_RET %endmacro -;void ff_hevc_sao_band_filter_<width>_<depth>_<opt>(uint8_t *_dst, uint8_t *_src, ptrdiff_t _stride_dst, ptrdiff_t _stride_src, -; int16_t *sao_offset_val, int sao_left_class, int width, int height); -%macro HEVC_SAO_BAND_FILTER_16 3 -cglobal hevc_sao_band_filter_%2_%1, 6, 6, 15, 7*mmsize*ARCH_X86_32, dst, src, dststride, srcstride, offset, left - HEVC_SAO_BAND_FILTER_INIT %1 - -align 16 -.loop: -%if %2 == 8 - movu m8, [srcq] - HEVC_SAO_BAND_FILTER_COMPUTE %1, m9, m8 - CLIPW m8, m14, m13 - movu [dstq], m8 -%endif - -%assign i 0 -%rep %3 - mova m8, [srcq + i] - HEVC_SAO_BAND_FILTER_COMPUTE %1, m9, m8 - CLIPW m8, m14, m13 - mova [dstq + i], m8 - - mova m9, [srcq + i + mmsize] - HEVC_SAO_BAND_FILTER_COMPUTE %1, m8, m9 - CLIPW m9, m14, m13 - mova [dstq + i + mmsize], m9 -%assign i i+mmsize*2 -%endrep - -%if %2 == 48 -INIT_XMM cpuname - mova m8, [srcq + i] - HEVC_SAO_BAND_FILTER_COMPUTE %1, m9, m8 - CLIPW m8, m14, m13 - mova [dstq + i], m8 - - mova m9, [srcq + i + mmsize] - HEVC_SAO_BAND_FILTER_COMPUTE %1, m8, m9 - CLIPW m9, m14, m13 - mova [dstq + i + mmsize], m9 -%if cpuflag(avx2) -INIT_YMM cpuname -%endif -%endif ; %1 == 48 - - add dstq, dststrideq - add srcq, srcstrideq - dec heightd - jg .loop - REP_RET -%endmacro %macro HEVC_SAO_BAND_FILTER_FUNCS 0 -HEVC_SAO_BAND_FILTER_8 8, 0 -HEVC_SAO_BAND_FILTER_8 16, 1 -HEVC_SAO_BAND_FILTER_8 32, 2 -HEVC_SAO_BAND_FILTER_8 48, 2 -HEVC_SAO_BAND_FILTER_8 64, 4 - -HEVC_SAO_BAND_FILTER_16 10, 8, 0 -HEVC_SAO_BAND_FILTER_16 10, 16, 1 -HEVC_SAO_BAND_FILTER_16 10, 32, 2 -HEVC_SAO_BAND_FILTER_16 10, 48, 2 -HEVC_SAO_BAND_FILTER_16 10, 64, 4 - -HEVC_SAO_BAND_FILTER_16 12, 8, 0 -HEVC_SAO_BAND_FILTER_16 12, 16, 1 -HEVC_SAO_BAND_FILTER_16 12, 32, 2 -HEVC_SAO_BAND_FILTER_16 12, 48, 2 -HEVC_SAO_BAND_FILTER_16 12, 64, 4 +HEVC_SAO_BAND_FILTER 8, 0 +HEVC_SAO_BAND_FILTER 16, 1 +HEVC_SAO_BAND_FILTER 32, 2 +HEVC_SAO_BAND_FILTER 48, 2 +HEVC_SAO_BAND_FILTER 64, 4 %endmacro INIT_XMM sse2 @@ -263,37 +185,23 @@ HEVC_SAO_BAND_FILTER_FUNCS %if HAVE_AVX2_EXTERNAL INIT_XMM avx2 -HEVC_SAO_BAND_FILTER_8 8, 0 -HEVC_SAO_BAND_FILTER_8 16, 1 -INIT_YMM avx2 -HEVC_SAO_BAND_FILTER_8 32, 1 -HEVC_SAO_BAND_FILTER_8 48, 1 -HEVC_SAO_BAND_FILTER_8 64, 2 - -INIT_XMM avx2 -HEVC_SAO_BAND_FILTER_16 10, 8, 0 -HEVC_SAO_BAND_FILTER_16 10, 16, 1 +HEVC_SAO_BAND_FILTER 8, 0 +HEVC_SAO_BAND_FILTER 16, 1 INIT_YMM avx2 -HEVC_SAO_BAND_FILTER_16 10, 32, 1 -HEVC_SAO_BAND_FILTER_16 10, 48, 1 -HEVC_SAO_BAND_FILTER_16 10, 64, 2 - -INIT_XMM avx2 -HEVC_SAO_BAND_FILTER_16 12, 8, 0 -HEVC_SAO_BAND_FILTER_16 12, 16, 1 -INIT_YMM avx2 -HEVC_SAO_BAND_FILTER_16 12, 32, 1 -HEVC_SAO_BAND_FILTER_16 12, 48, 1 -HEVC_SAO_BAND_FILTER_16 12, 64, 2 +HEVC_SAO_BAND_FILTER 32, 1 +HEVC_SAO_BAND_FILTER 48, 1 +HEVC_SAO_BAND_FILTER 64, 2 %endif ;****************************************************************************** ;SAO Edge Filter ;****************************************************************************** +%define MAX_PB_SIZE 64 +%define PADDING_SIZE 32 ; AV_INPUT_BUFFER_PADDING_SIZE %define EDGE_SRCSTRIDE 2 * MAX_PB_SIZE + PADDING_SIZE -%macro HEVC_SAO_EDGE_FILTER_INIT 1 +%macro HEVC_SAO_EDGE_FILTER_INIT 0 %if WIN64 movsxd eoq, dword eom %elif ARCH_X86_64 @@ -304,15 +212,15 @@ HEVC_SAO_BAND_FILTER_16 12, 64, 2 lea tmp2q, [pb_eo] movsx a_strideq, byte [tmp2q+eoq*4+1] movsx b_strideq, byte [tmp2q+eoq*4+3] - imul a_strideq, EDGE_SRCSTRIDE>>%1 - imul b_strideq, EDGE_SRCSTRIDE>>%1 + imul a_strideq, EDGE_SRCSTRIDE + imul b_strideq, EDGE_SRCSTRIDE movsx tmpq, byte [tmp2q+eoq*4] add a_strideq, tmpq movsx tmpq, byte [tmp2q+eoq*4+2] add b_strideq, tmpq %endmacro -%macro HEVC_SAO_EDGE_FILTER_COMPUTE_8 1 +%macro HEVC_SAO_EDGE_FILTER_COMPUTE 1 pminub m4, m1, m2 pminub m5, m1, m3 pcmpeqb m2, m4 @@ -343,11 +251,11 @@ HEVC_SAO_BAND_FILTER_16 12, 64, 2 ;void ff_hevc_sao_edge_filter_<width>_8_<opt>(uint8_t *_dst, uint8_t *_src, ptrdiff_t stride_dst, int16_t *sao_offset_val, ; int eo, int width, int height); -%macro HEVC_SAO_EDGE_FILTER_8 2-3 +%macro HEVC_SAO_EDGE_FILTER 2-3 %if ARCH_X86_64 cglobal hevc_sao_edge_filter_%1_8, 4, 9, 8, dst, src, dststride, offset, eo, a_stride, b_stride, height, tmp %define tmp2q heightq - HEVC_SAO_EDGE_FILTER_INIT 0 + HEVC_SAO_EDGE_FILTER_INIT mov heightd, r6m %else ; ARCH_X86_32 @@ -356,7 +264,7 @@ cglobal hevc_sao_edge_filter_%1_8, 1, 6, 8, dst, src, dststride, a_stride, b_str %define tmpq heightq %define tmp2q dststrideq %define offsetq heightq - HEVC_SAO_EDGE_FILTER_INIT 0 + HEVC_SAO_EDGE_FILTER_INIT mov srcq, srcm mov offsetq, r3m mov dststrideq, dststridem @@ -383,7 +291,7 @@ align 16 movq m1, [srcq] movq m2, [srcq + a_strideq] movq m3, [srcq + b_strideq] - HEVC_SAO_EDGE_FILTER_COMPUTE_8 %1 + HEVC_SAO_EDGE_FILTER_COMPUTE %1 movq [dstq], m3 %endif @@ -392,7 +300,7 @@ align 16 mova m1, [srcq + i] movu m2, [srcq + a_strideq + i] movu m3, [srcq + b_strideq + i] - HEVC_SAO_EDGE_FILTER_COMPUTE_8 %1 + HEVC_SAO_EDGE_FILTER_COMPUTE %1 mov%3 [dstq + i], m3 %assign i i+mmsize %endrep @@ -403,7 +311,7 @@ INIT_XMM cpuname mova m1, [srcq + i] movu m2, [srcq + a_strideq + i] movu m3, [srcq + b_strideq + i] - HEVC_SAO_EDGE_FILTER_COMPUTE_8 %1 + HEVC_SAO_EDGE_FILTER_COMPUTE %1 mova [dstq + i], m3 %if cpuflag(avx2) INIT_YMM cpuname @@ -417,208 +325,16 @@ INIT_YMM cpuname RET %endmacro -%macro PMINUW 4 -%if cpuflag(sse4) - pminuw %1, %2, %3 -%else - psubusw %4, %2, %3 - psubw %1, %2, %4 -%endif -%endmacro - -%macro HEVC_SAO_EDGE_FILTER_COMPUTE_10 0 - PMINUW m4, m1, m2, m6 - PMINUW m5, m1, m3, m7 - pcmpeqw m2, m4 - pcmpeqw m3, m5 - pcmpeqw m4, m1 - pcmpeqw m5, m1 - psubw m4, m2 - psubw m5, m3 - - paddw m4, m5 - pcmpeqw m2, m4, [pw_m2] -%if ARCH_X86_64 - pcmpeqw m3, m4, m13 - pcmpeqw m5, m4, m0 - pcmpeqw m6, m4, m14 - pcmpeqw m7, m4, m15 - pand m2, m8 - pand m3, m9 - pand m5, m10 - pand m6, m11 - pand m7, m12 -%else - pcmpeqw m3, m4, [pw_m1] - pcmpeqw m5, m4, m0 - pcmpeqw m6, m4, [pw_1] - pcmpeqw m7, m4, [pw_2] - pand m2, [rsp+MMSIZE*0] - pand m3, [rsp+MMSIZE*1] - pand m5, [rsp+MMSIZE*2] - pand m6, [rsp+MMSIZE*3] - pand m7, [rsp+MMSIZE*4] -%endif - paddw m2, m3 - paddw m5, m6 - paddw m2, m7 - paddw m2, m1 - paddw m2, m5 -%endmacro - -;void ff_hevc_sao_edge_filter_<width>_<depth>_<opt>(uint8_t *_dst, uint8_t *_src, ptrdiff_t stride_dst, int16_t *sao_offset_val, -; int eo, int width, int height); -%macro HEVC_SAO_EDGE_FILTER_16 3 -%if ARCH_X86_64 -cglobal hevc_sao_edge_filter_%2_%1, 4, 9, 16, dst, src, dststride, offset, eo, a_stride, b_stride, height, tmp -%define tmp2q heightq - HEVC_SAO_EDGE_FILTER_INIT 1 - mov heightd, r6m - add a_strideq, a_strideq - add b_strideq, b_strideq - -%else ; ARCH_X86_32 -cglobal hevc_sao_edge_filter_%2_%1, 1, 6, 8, 5*mmsize, dst, src, dststride, a_stride, b_stride, height -%assign MMSIZE mmsize -%define eoq srcq -%define tmpq heightq -%define tmp2q dststrideq -%define offsetq heightq -%define m8 m1 -%define m9 m2 -%define m10 m3 -%define m11 m4 -%define m12 m5 - HEVC_SAO_EDGE_FILTER_INIT 1 - mov srcq, srcm - mov offsetq, r3m - mov dststrideq, dststridem - add a_strideq, a_strideq - add b_strideq, b_strideq - -%endif ; ARCH - -%if cpuflag(avx2) - SPLATW m8, [offsetq+2] - SPLATW m9, [offsetq+4] - SPLATW m10, [offsetq+0] - SPLATW m11, [offsetq+6] - SPLATW m12, [offsetq+8] -%else - movq m10, [offsetq+0] - movd m12, [offsetq+6] - SPLATW m8, xm10, 1 - SPLATW m9, xm10, 2 - SPLATW m10, xm10, 0 - SPLATW m11, xm12, 0 - SPLATW m12, xm12, 1 -%endif - pxor m0, m0 -%if ARCH_X86_64 - mova m13, [pw_m1] - mova m14, [pw_1] - mova m15, [pw_2] -%else - mov heightd, r6m - mova [rsp+mmsize*0], m8 - mova [rsp+mmsize*1], m9 - mova [rsp+mmsize*2], m10 - mova [rsp+mmsize*3], m11 - mova [rsp+mmsize*4], m12 -%endif - -align 16 -.loop: - -%if %2 == 8 - mova m1, [srcq] - movu m2, [srcq+a_strideq] - movu m3, [srcq+b_strideq] - - HEVC_SAO_EDGE_FILTER_COMPUTE_10 - CLIPW m2, m0, [pw_mask %+ %1] - movu [dstq], m2 -%endif - -%assign i 0 -%rep %3 - mova m1, [srcq + i] - movu m2, [srcq+a_strideq + i] - movu m3, [srcq+b_strideq + i] - HEVC_SAO_EDGE_FILTER_COMPUTE_10 - CLIPW m2, m0, [pw_mask %+ %1] - mova [dstq + i], m2 - - mova m1, [srcq + i + mmsize] - movu m2, [srcq+a_strideq + i + mmsize] - movu m3, [srcq+b_strideq + i + mmsize] - HEVC_SAO_EDGE_FILTER_COMPUTE_10 - CLIPW m2, m0, [pw_mask %+ %1] - mova [dstq + i + mmsize], m2 -%assign i i+mmsize*2 -%endrep - -%if %2 == 48 -INIT_XMM cpuname - mova m1, [srcq + i] - movu m2, [srcq+a_strideq + i] - movu m3, [srcq+b_strideq + i] - HEVC_SAO_EDGE_FILTER_COMPUTE_10 - CLIPW m2, m0, [pw_mask %+ %1] - mova [dstq + i], m2 - - mova m1, [srcq + i + mmsize] - movu m2, [srcq+a_strideq + i + mmsize] - movu m3, [srcq+b_strideq + i + mmsize] - HEVC_SAO_EDGE_FILTER_COMPUTE_10 - CLIPW m2, m0, [pw_mask %+ %1] - mova [dstq + i + mmsize], m2 -%if cpuflag(avx2) -INIT_YMM cpuname -%endif -%endif - - add dstq, dststrideq - add srcq, EDGE_SRCSTRIDE - dec heightd - jg .loop - RET -%endmacro - INIT_XMM ssse3 -HEVC_SAO_EDGE_FILTER_8 8, 0 -HEVC_SAO_EDGE_FILTER_8 16, 1, a -HEVC_SAO_EDGE_FILTER_8 32, 2, a -HEVC_SAO_EDGE_FILTER_8 48, 2, a -HEVC_SAO_EDGE_FILTER_8 64, 4, a - -%if HAVE_AVX2_EXTERNAL -INIT_YMM avx2 -HEVC_SAO_EDGE_FILTER_8 32, 1, a -HEVC_SAO_EDGE_FILTER_8 48, 1, u -HEVC_SAO_EDGE_FILTER_8 64, 2, a -%endif - -INIT_XMM sse2 -HEVC_SAO_EDGE_FILTER_16 10, 8, 0 -HEVC_SAO_EDGE_FILTER_16 10, 16, 1 -HEVC_SAO_EDGE_FILTER_16 10, 32, 2 -HEVC_SAO_EDGE_FILTER_16 10, 48, 2 -HEVC_SAO_EDGE_FILTER_16 10, 64, 4 - -HEVC_SAO_EDGE_FILTER_16 12, 8, 0 -HEVC_SAO_EDGE_FILTER_16 12, 16, 1 -HEVC_SAO_EDGE_FILTER_16 12, 32, 2 -HEVC_SAO_EDGE_FILTER_16 12, 48, 2 -HEVC_SAO_EDGE_FILTER_16 12, 64, 4 +HEVC_SAO_EDGE_FILTER 8, 0 +HEVC_SAO_EDGE_FILTER 16, 1, a +HEVC_SAO_EDGE_FILTER 32, 2, a +HEVC_SAO_EDGE_FILTER 48, 2, a +HEVC_SAO_EDGE_FILTER 64, 4, a %if HAVE_AVX2_EXTERNAL INIT_YMM avx2 -HEVC_SAO_EDGE_FILTER_16 10, 32, 1 -HEVC_SAO_EDGE_FILTER_16 10, 48, 1 -HEVC_SAO_EDGE_FILTER_16 10, 64, 2 - -HEVC_SAO_EDGE_FILTER_16 12, 32, 1 -HEVC_SAO_EDGE_FILTER_16 12, 48, 1 -HEVC_SAO_EDGE_FILTER_16 12, 64, 2 +HEVC_SAO_EDGE_FILTER 32, 1, a +HEVC_SAO_EDGE_FILTER 48, 1, u +HEVC_SAO_EDGE_FILTER 64, 2, a %endif diff --git a/libavcodec/x86/hevc_sao_10bit.asm b/libavcodec/x86/hevc_sao_10bit.asm new file mode 100644 index 0000000..f45fc56 --- /dev/null +++ b/libavcodec/x86/hevc_sao_10bit.asm @@ -0,0 +1,433 @@ +;****************************************************************************** +;* SIMD optimized SAO functions for HEVC 10/12bit decoding +;* +;* Copyright (c) 2013 Pierre-Edouard LEPERE +;* Copyright (c) 2014 James Almer +;* +;* This file is part of FFmpeg. +;* +;* FFmpeg is free software; you can redistribute it and/or +;* modify it under the terms of the GNU Lesser General Public +;* License as published by the Free Software Foundation; either +;* version 2.1 of the License, or (at your option) any later version. +;* +;* FFmpeg is distributed in the hope that it will be useful, +;* but WITHOUT ANY WARRANTY; without even the implied warranty of +;* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU +;* Lesser General Public License for more details. +;* +;* You should have received a copy of the GNU Lesser General Public +;* License along with FFmpeg; if not, write to the Free Software +;* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA +;****************************************************************************** + +%include "libavutil/x86/x86util.asm" + +SECTION_RODATA 32 + +pw_m2: times 16 dw -2 +pw_mask10: times 16 dw 0x03FF +pw_mask12: times 16 dw 0x0FFF +pb_eo: db -1, 0, 1, 0, 0, -1, 0, 1, -1, -1, 1, 1, 1, -1, -1, 1 +cextern pw_m1 +cextern pw_1 +cextern pw_2 + +SECTION .text + +;****************************************************************************** +;SAO Band Filter +;****************************************************************************** + +%macro HEVC_SAO_BAND_FILTER_INIT 1 + and leftq, 31 + movd xm0, leftd + add leftq, 1 + and leftq, 31 + movd xm1, leftd + add leftq, 1 + and leftq, 31 + movd xm2, leftd + add leftq, 1 + and leftq, 31 + movd xm3, leftd + + SPLATW m0, xm0 + SPLATW m1, xm1 + SPLATW m2, xm2 + SPLATW m3, xm3 +%if mmsize > 16 + SPLATW m4, [offsetq + 2] + SPLATW m5, [offsetq + 4] + SPLATW m6, [offsetq + 6] + SPLATW m7, [offsetq + 8] +%else + movq m7, [offsetq + 2] + SPLATW m4, m7, 0 + SPLATW m5, m7, 1 + SPLATW m6, m7, 2 + SPLATW m7, m7, 3 +%endif + +%if ARCH_X86_64 + mova m13, [pw_mask %+ %1] + pxor m14, m14 + +%else ; ARCH_X86_32 + mova [rsp+mmsize*0], m0 + mova [rsp+mmsize*1], m1 + mova [rsp+mmsize*2], m2 + mova [rsp+mmsize*3], m3 + mova [rsp+mmsize*4], m4 + mova [rsp+mmsize*5], m5 + mova [rsp+mmsize*6], m6 + mova m1, [pw_mask %+ %1] + pxor m0, m0 + %assign MMSIZE mmsize + %define m14 m0 + %define m13 m1 + %define m9 m2 + %define m8 m3 +%endif ; ARCH +DEFINE_ARGS dst, src, dststride, srcstride, offset, height + mov heightd, r7m +%endmacro + +%macro HEVC_SAO_BAND_FILTER_COMPUTE 3 + psraw %2, %3, %1-5 +%if ARCH_X86_64 + pcmpeqw m10, %2, m0 + pcmpeqw m11, %2, m1 + pcmpeqw m12, %2, m2 + pcmpeqw %2, m3 + pand m10, m4 + pand m11, m5 + pand m12, m6 + pand %2, m7 + por m10, m11 + por m12, %2 + por m10, m12 + paddw %3, m10 +%else ; ARCH_X86_32 + pcmpeqw m4, %2, [rsp+MMSIZE*0] + pcmpeqw m5, %2, [rsp+MMSIZE*1] + pcmpeqw m6, %2, [rsp+MMSIZE*2] + pcmpeqw %2, [rsp+MMSIZE*3] + pand m4, [rsp+MMSIZE*4] + pand m5, [rsp+MMSIZE*5] + pand m6, [rsp+MMSIZE*6] + pand %2, m7 + por m4, m5 + por m6, %2 + por m4, m6 + paddw %3, m4 +%endif ; ARCH +%endmacro + +;void ff_hevc_sao_band_filter_<width>_<depth>_<opt>(uint8_t *_dst, uint8_t *_src, ptrdiff_t _stride_dst, ptrdiff_t _stride_src, +; int16_t *sao_offset_val, int sao_left_class, int width, int height); +%macro HEVC_SAO_BAND_FILTER 3 +cglobal hevc_sao_band_filter_%2_%1, 6, 6, 15, 7*mmsize*ARCH_X86_32, dst, src, dststride, srcstride, offset, left + HEVC_SAO_BAND_FILTER_INIT %1 + +align 16 +.loop: +%if %2 == 8 + movu m8, [srcq] + HEVC_SAO_BAND_FILTER_COMPUTE %1, m9, m8 + CLIPW m8, m14, m13 + movu [dstq], m8 +%endif + +%assign i 0 +%rep %3 + mova m8, [srcq + i] + HEVC_SAO_BAND_FILTER_COMPUTE %1, m9, m8 + CLIPW m8, m14, m13 + mova [dstq + i], m8 + + mova m9, [srcq + i + mmsize] + HEVC_SAO_BAND_FILTER_COMPUTE %1, m8, m9 + CLIPW m9, m14, m13 + mova [dstq + i + mmsize], m9 +%assign i i+mmsize*2 +%endrep + +%if %2 == 48 +INIT_XMM cpuname + mova m8, [srcq + i] + HEVC_SAO_BAND_FILTER_COMPUTE %1, m9, m8 + CLIPW m8, m14, m13 + mova [dstq + i], m8 + + mova m9, [srcq + i + mmsize] + HEVC_SAO_BAND_FILTER_COMPUTE %1, m8, m9 + CLIPW m9, m14, m13 + mova [dstq + i + mmsize], m9 +%if cpuflag(avx2) +INIT_YMM cpuname +%endif +%endif ; %1 == 48 + + add dstq, dststrideq + add srcq, srcstrideq + dec heightd + jg .loop + REP_RET +%endmacro + +%macro HEVC_SAO_BAND_FILTER_FUNCS 0 +HEVC_SAO_BAND_FILTER 10, 8, 0 +HEVC_SAO_BAND_FILTER 10, 16, 1 +HEVC_SAO_BAND_FILTER 10, 32, 2 +HEVC_SAO_BAND_FILTER 10, 48, 2 +HEVC_SAO_BAND_FILTER 10, 64, 4 + +HEVC_SAO_BAND_FILTER 12, 8, 0 +HEVC_SAO_BAND_FILTER 12, 16, 1 +HEVC_SAO_BAND_FILTER 12, 32, 2 +HEVC_SAO_BAND_FILTER 12, 48, 2 +HEVC_SAO_BAND_FILTER 12, 64, 4 +%endmacro + +INIT_XMM sse2 +HEVC_SAO_BAND_FILTER_FUNCS +INIT_XMM avx +HEVC_SAO_BAND_FILTER_FUNCS + +%if HAVE_AVX2_EXTERNAL +INIT_XMM avx2 +HEVC_SAO_BAND_FILTER 10, 8, 0 +HEVC_SAO_BAND_FILTER 10, 16, 1 +INIT_YMM avx2 +HEVC_SAO_BAND_FILTER 10, 32, 1 +HEVC_SAO_BAND_FILTER 10, 48, 1 +HEVC_SAO_BAND_FILTER 10, 64, 2 + +INIT_XMM avx2 +HEVC_SAO_BAND_FILTER 12, 8, 0 +HEVC_SAO_BAND_FILTER 12, 16, 1 +INIT_YMM avx2 +HEVC_SAO_BAND_FILTER 12, 32, 1 +HEVC_SAO_BAND_FILTER 12, 48, 1 +HEVC_SAO_BAND_FILTER 12, 64, 2 +%endif + +;****************************************************************************** +;SAO Edge Filter +;****************************************************************************** + +%define MAX_PB_SIZE 64 +%define PADDING_SIZE 32 ; AV_INPUT_BUFFER_PADDING_SIZE +%define EDGE_SRCSTRIDE 2 * MAX_PB_SIZE + PADDING_SIZE + +%macro PMINUW 4 +%if cpuflag(sse4) + pminuw %1, %2, %3 +%else + psubusw %4, %2, %3 + psubw %1, %2, %4 +%endif +%endmacro + +%macro HEVC_SAO_EDGE_FILTER_INIT 0 +%if WIN64 + movsxd eoq, dword eom +%elif ARCH_X86_64 + movsxd eoq, eod +%else + mov eoq, r4m +%endif + lea tmp2q, [pb_eo] + movsx a_strideq, byte [tmp2q+eoq*4+1] + movsx b_strideq, byte [tmp2q+eoq*4+3] + imul a_strideq, EDGE_SRCSTRIDE >> 1 + imul b_strideq, EDGE_SRCSTRIDE >> 1 + movsx tmpq, byte [tmp2q+eoq*4] + add a_strideq, tmpq + movsx tmpq, byte [tmp2q+eoq*4+2] + add b_strideq, tmpq +%endmacro + +%macro HEVC_SAO_EDGE_FILTER_COMPUTE 0 + PMINUW m4, m1, m2, m6 + PMINUW m5, m1, m3, m7 + pcmpeqw m2, m4 + pcmpeqw m3, m5 + pcmpeqw m4, m1 + pcmpeqw m5, m1 + psubw m4, m2 + psubw m5, m3 + + paddw m4, m5 + pcmpeqw m2, m4, [pw_m2] +%if ARCH_X86_64 + pcmpeqw m3, m4, m13 + pcmpeqw m5, m4, m0 + pcmpeqw m6, m4, m14 + pcmpeqw m7, m4, m15 + pand m2, m8 + pand m3, m9 + pand m5, m10 + pand m6, m11 + pand m7, m12 +%else + pcmpeqw m3, m4, [pw_m1] + pcmpeqw m5, m4, m0 + pcmpeqw m6, m4, [pw_1] + pcmpeqw m7, m4, [pw_2] + pand m2, [rsp+MMSIZE*0] + pand m3, [rsp+MMSIZE*1] + pand m5, [rsp+MMSIZE*2] + pand m6, [rsp+MMSIZE*3] + pand m7, [rsp+MMSIZE*4] +%endif + paddw m2, m3 + paddw m5, m6 + paddw m2, m7 + paddw m2, m1 + paddw m2, m5 +%endmacro + +;void ff_hevc_sao_edge_filter_<width>_<depth>_<opt>(uint8_t *_dst, uint8_t *_src, ptrdiff_t stride_dst, int16_t *sao_offset_val, +; int eo, int width, int height); +%macro HEVC_SAO_EDGE_FILTER 3 +%if ARCH_X86_64 +cglobal hevc_sao_edge_filter_%2_%1, 4, 9, 16, dst, src, dststride, offset, eo, a_stride, b_stride, height, tmp +%define tmp2q heightq + HEVC_SAO_EDGE_FILTER_INIT + mov heightd, r6m + add a_strideq, a_strideq + add b_strideq, b_strideq + +%else ; ARCH_X86_32 +cglobal hevc_sao_edge_filter_%2_%1, 1, 6, 8, 5*mmsize, dst, src, dststride, a_stride, b_stride, height +%assign MMSIZE mmsize +%define eoq srcq +%define tmpq heightq +%define tmp2q dststrideq +%define offsetq heightq +%define m8 m1 +%define m9 m2 +%define m10 m3 +%define m11 m4 +%define m12 m5 + HEVC_SAO_EDGE_FILTER_INIT + mov srcq, srcm + mov offsetq, r3m + mov dststrideq, dststridem + add a_strideq, a_strideq + add b_strideq, b_strideq + +%endif ; ARCH + +%if cpuflag(avx2) + SPLATW m8, [offsetq+2] + SPLATW m9, [offsetq+4] + SPLATW m10, [offsetq+0] + SPLATW m11, [offsetq+6] + SPLATW m12, [offsetq+8] +%else + movq m10, [offsetq+0] + movd m12, [offsetq+6] + SPLATW m8, xm10, 1 + SPLATW m9, xm10, 2 + SPLATW m10, xm10, 0 + SPLATW m11, xm12, 0 + SPLATW m12, xm12, 1 +%endif + pxor m0, m0 +%if ARCH_X86_64 + mova m13, [pw_m1] + mova m14, [pw_1] + mova m15, [pw_2] +%else + mov heightd, r6m + mova [rsp+mmsize*0], m8 + mova [rsp+mmsize*1], m9 + mova [rsp+mmsize*2], m10 + mova [rsp+mmsize*3], m11 + mova [rsp+mmsize*4], m12 +%endif + +align 16 +.loop: + +%if %2 == 8 + mova m1, [srcq] + movu m2, [srcq+a_strideq] + movu m3, [srcq+b_strideq] + + HEVC_SAO_EDGE_FILTER_COMPUTE + CLIPW m2, m0, [pw_mask %+ %1] + movu [dstq], m2 +%endif + +%assign i 0 +%rep %3 + mova m1, [srcq + i] + movu m2, [srcq+a_strideq + i] + movu m3, [srcq+b_strideq + i] + HEVC_SAO_EDGE_FILTER_COMPUTE + CLIPW m2, m0, [pw_mask %+ %1] + mova [dstq + i], m2 + + mova m1, [srcq + i + mmsize] + movu m2, [srcq+a_strideq + i + mmsize] + movu m3, [srcq+b_strideq + i + mmsize] + HEVC_SAO_EDGE_FILTER_COMPUTE + CLIPW m2, m0, [pw_mask %+ %1] + mova [dstq + i + mmsize], m2 +%assign i i+mmsize*2 +%endrep + +%if %2 == 48 +INIT_XMM cpuname + mova m1, [srcq + i] + movu m2, [srcq+a_strideq + i] + movu m3, [srcq+b_strideq + i] + HEVC_SAO_EDGE_FILTER_COMPUTE + CLIPW m2, m0, [pw_mask %+ %1] + mova [dstq + i], m2 + + mova m1, [srcq + i + mmsize] + movu m2, [srcq+a_strideq + i + mmsize] + movu m3, [srcq+b_strideq + i + mmsize] + HEVC_SAO_EDGE_FILTER_COMPUTE + CLIPW m2, m0, [pw_mask %+ %1] + mova [dstq + i + mmsize], m2 +%if cpuflag(avx2) +INIT_YMM cpuname +%endif +%endif + + add dstq, dststrideq + add srcq, EDGE_SRCSTRIDE + dec heightd + jg .loop + RET +%endmacro + +INIT_XMM sse2 +HEVC_SAO_EDGE_FILTER 10, 8, 0 +HEVC_SAO_EDGE_FILTER 10, 16, 1 +HEVC_SAO_EDGE_FILTER 10, 32, 2 +HEVC_SAO_EDGE_FILTER 10, 48, 2 +HEVC_SAO_EDGE_FILTER 10, 64, 4 + +HEVC_SAO_EDGE_FILTER 12, 8, 0 +HEVC_SAO_EDGE_FILTER 12, 16, 1 +HEVC_SAO_EDGE_FILTER 12, 32, 2 +HEVC_SAO_EDGE_FILTER 12, 48, 2 +HEVC_SAO_EDGE_FILTER 12, 64, 4 + +%if HAVE_AVX2_EXTERNAL +INIT_YMM avx2 +HEVC_SAO_EDGE_FILTER 10, 32, 1 +HEVC_SAO_EDGE_FILTER 10, 48, 1 +HEVC_SAO_EDGE_FILTER 10, 64, 2 + +HEVC_SAO_EDGE_FILTER 12, 32, 1 +HEVC_SAO_EDGE_FILTER 12, 48, 1 +HEVC_SAO_EDGE_FILTER 12, 64, 2 +%endif

1 0

compat/w32pthreads: Add return values to match the simulated pthread functions.
by git＠videolan.org 30 Sep '15

30 Sep '15

ffmpeg | branch: master | Matt Oliver <protogonoi(a)gmail.com> | Wed Sep 30 13:41:33 2015 +1000| [ae58abeabbfc15eddff0e23cbf1a62c7bc50d3c8] | committer: Matt Oliver compat/w32pthreads: Add return values to match the simulated pthread functions. > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=ae58abeabbfc15edd… --- compat/w32pthreads.h | 36 +++++++++++++++++++++++------------- 1 file changed, 23 insertions(+), 13 deletions(-) diff --git a/compat/w32pthreads.h b/compat/w32pthreads.h index 87e816f..deb1c53 100644 --- a/compat/w32pthreads.h +++ b/compat/w32pthreads.h @@ -87,14 +87,19 @@ static av_unused int pthread_create(pthread_t *thread, const void *unused_attr, return !thread->handle; } -static av_unused void pthread_join(pthread_t thread, void **value_ptr) +static av_unused int pthread_join(pthread_t thread, void **value_ptr) { DWORD ret = WaitForSingleObject(thread.handle, INFINITE); - if (ret != WAIT_OBJECT_0) - return; + if (ret != WAIT_OBJECT_0) { + if (ret == WAIT_ABANDONED) + return EINVAL; + else + return EDEADLK; + } if (value_ptr) *value_ptr = thread.ret; CloseHandle(thread.handle); + return 0; } static inline int pthread_mutex_init(pthread_mutex_t *m, void* attr) @@ -126,14 +131,15 @@ static inline int pthread_cond_init(pthread_cond_t *cond, const void *unused_att } /* native condition variables do not destroy */ -static inline void pthread_cond_destroy(pthread_cond_t *cond) +static inline int pthread_cond_destroy(pthread_cond_t *cond) { - return; + return 0; } -static inline void pthread_cond_broadcast(pthread_cond_t *cond) +static inline int pthread_cond_broadcast(pthread_cond_t *cond) { WakeAllConditionVariable(cond); + return 0; } static inline int pthread_cond_wait(pthread_cond_t *cond, pthread_mutex_t *mutex) @@ -142,9 +148,10 @@ static inline int pthread_cond_wait(pthread_cond_t *cond, pthread_mutex_t *mutex return 0; } -static inline void pthread_cond_signal(pthread_cond_t *cond) +static inline int pthread_cond_signal(pthread_cond_t *cond) { WakeConditionVariable(cond); + return 0; } #else // _WIN32_WINNT < 0x0600 @@ -191,12 +198,12 @@ static av_unused int pthread_cond_init(pthread_cond_t *cond, const void *unused_ return 0; } -static av_unused void pthread_cond_destroy(pthread_cond_t *cond) +static av_unused int pthread_cond_destroy(pthread_cond_t *cond) { win32_cond_t *win32_cond = cond->Ptr; /* native condition variables do not destroy */ if (cond_init) - return; + return 0; /* non native condition variables */ CloseHandle(win32_cond->semaphore); @@ -205,16 +212,17 @@ static av_unused void pthread_cond_destroy(pthread_cond_t *cond) pthread_mutex_destroy(&win32_cond->mtx_broadcast); av_freep(&win32_cond); cond->Ptr = NULL; + return 0; } -static av_unused void pthread_cond_broadcast(pthread_cond_t *cond) +static av_unused int pthread_cond_broadcast(pthread_cond_t *cond) { win32_cond_t *win32_cond = cond->Ptr; int have_waiter; if (cond_broadcast) { cond_broadcast(cond); - return; + return 0; } /* non native condition variables */ @@ -236,6 +244,7 @@ static av_unused void pthread_cond_broadcast(pthread_cond_t *cond) } else pthread_mutex_unlock(&win32_cond->mtx_waiter_count); pthread_mutex_unlock(&win32_cond->mtx_broadcast); + return 0; } static av_unused int pthread_cond_wait(pthread_cond_t *cond, pthread_mutex_t *mutex) @@ -270,13 +279,13 @@ static av_unused int pthread_cond_wait(pthread_cond_t *cond, pthread_mutex_t *mu return pthread_mutex_lock(mutex); } -static av_unused void pthread_cond_signal(pthread_cond_t *cond) +static av_unused int pthread_cond_signal(pthread_cond_t *cond) { win32_cond_t *win32_cond = cond->Ptr; int have_waiter; if (cond_signal) { cond_signal(cond); - return; + return 0; } pthread_mutex_lock(&win32_cond->mtx_broadcast); @@ -293,6 +302,7 @@ static av_unused void pthread_cond_signal(pthread_cond_t *cond) } pthread_mutex_unlock(&win32_cond->mtx_broadcast); + return 0; } #endif

1 0

avformat/async: Allow compilation with native threads.
by git＠videolan.org 30 Sep '15

30 Sep '15

ffmpeg | branch: master | Matt Oliver <protogonoi(a)gmail.com> | Wed Sep 30 13:41:53 2015 +1000| [3b03bde46e331aa301bd16c6c6e94c11e073e8f8] | committer: Matt Oliver avformat/async: Allow compilation with native threads. > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=3b03bde46e331aa30… --- configure | 2 +- libavformat/async.c | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) diff --git a/configure b/configure index 01f1797..361c024 100755 --- a/configure +++ b/configure @@ -2693,7 +2693,7 @@ x11grab_indev_deps="x11grab" x11grab_xcb_indev_deps="libxcb" # protocols -async_protocol_deps="pthreads" +async_protocol_deps="threads" bluray_protocol_deps="libbluray" ffrtmpcrypt_protocol_deps="!librtmp_protocol" ffrtmpcrypt_protocol_deps_any="gcrypt gmp openssl" diff --git a/libavformat/async.c b/libavformat/async.c index 003212d..a01673d 100644 --- a/libavformat/async.c +++ b/libavformat/async.c @@ -34,9 +34,9 @@ #include "libavutil/fifo.h" #include "libavutil/log.h" #include "libavutil/opt.h" +#include "libavutil/thread.h" #include "url.h" #include <stdint.h> -#include <pthread.h> #if HAVE_UNISTD_H #include <unistd.h>

1 0