July 2012 - ffmpeg-cvslog - lists.ffmpeg.org

aac: reference supported sample rates in encoders.
by git＠videolan.org 06 Nov '12

06 Nov '12

ffmpeg | branch: master | Clément Bœsch <clement.boesch(a)smartjog.com> | Mon Mar 26 22:32:17 2012 +0200| [29679f4a372eb208af31c57177ab49f67c698362] | committer: Clément Bœsch aac: reference supported sample rates in encoders. > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=29679f4a372eb208a… --- libavcodec/aacenc.c | 1 + libavcodec/libvo-aacenc.c | 1 + 2 files changed, 2 insertions(+), 0 deletions(-) diff --git a/libavcodec/aacenc.c b/libavcodec/aacenc.c index cf4aef9..fd72daf 100644 --- a/libavcodec/aacenc.c +++ b/libavcodec/aacenc.c @@ -816,6 +816,7 @@ AVCodec ff_aac_encoder = { .init = aac_encode_init, .encode2 = aac_encode_frame, .close = aac_encode_end, + .supported_samplerates = avpriv_mpeg4audio_sample_rates, .capabilities = CODEC_CAP_SMALL_LAST_FRAME | CODEC_CAP_DELAY | CODEC_CAP_EXPERIMENTAL, .sample_fmts = (const enum AVSampleFormat[]){AV_SAMPLE_FMT_FLT,AV_SAMPLE_FMT_NONE}, .long_name = NULL_IF_CONFIG_SMALL("Advanced Audio Coding"), diff --git a/libavcodec/libvo-aacenc.c b/libavcodec/libvo-aacenc.c index 8f7c581..c5191ec 100644 --- a/libavcodec/libvo-aacenc.c +++ b/libavcodec/libvo-aacenc.c @@ -193,6 +193,7 @@ AVCodec ff_libvo_aacenc_encoder = { .init = aac_encode_init, .encode2 = aac_encode_frame, .close = aac_encode_close, + .supported_samplerates = avpriv_mpeg4audio_sample_rates, .capabilities = CODEC_CAP_SMALL_LAST_FRAME | CODEC_CAP_DELAY, .sample_fmts = (const enum AVSampleFormat[]){AV_SAMPLE_FMT_S16,AV_SAMPLE_FMT_NONE}, .long_name = NULL_IF_CONFIG_SMALL("Android VisualOn AAC"),

3 2

ffmpeg: avoid bogus error with "test.flac -c copy -f null -"
by git＠videolan.org 01 Aug '12

01 Aug '12

ffmpeg | branch: master | Michael Niedermayer <michaelni(a)gmx.at> | Wed Aug 1 01:47:51 2012 +0200| [d57a9928eac1fdc908954eb89dc3d2783e4d1996] | committer: Michael Niedermayer ffmpeg: avoid bogus error with "test.flac -c copy -f null -" Reported-by: durandal_1707 Signed-off-by: Michael Niedermayer <michaelni(a)gmx.at> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=d57a9928eac1fdc90… --- ffmpeg.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/ffmpeg.c b/ffmpeg.c index c2ea5bd..f85d8e0 100644 --- a/ffmpeg.c +++ b/ffmpeg.c @@ -2343,7 +2343,8 @@ static void do_streamcopy(InputStream *ist, OutputStream *ost, const AVPacket *p opkt.data = pkt->data; opkt.size = pkt->size; } - if (of->ctx->oformat->flags & AVFMT_RAWPICTURE) { + + if (ost->st->codec->codec_type == AVMEDIA_TYPE_VIDEO && (of->ctx->oformat->flags & AVFMT_RAWPICTURE)) { /* store AVPicture in AVPacket, as expected by the output format */ avpicture_fill(&pict, opkt.data, ost->st->codec->pix_fmt, ost->st->codec->width, ost->st->codec->height); opkt.data = (uint8_t *)&pict;

1 0

Merge remote-tracking branch 'qatar/master'
by git＠videolan.org 31 Jul '12

31 Jul '12

ffmpeg | branch: master | Michael Niedermayer <michaelni(a)gmx.at> | Tue Jul 31 22:41:00 2012 +0200| [d1dad7c824c14939a0e18fa262f3e51957de6e3e] | committer: Michael Niedermayer Merge remote-tracking branch 'qatar/master' * qatar/master: mpc8: return more meaningful error codes. mpc: return more meaningful error codes. wv,mpc8: don't return apetag data in packets. rtmp: do not warn about receiving metadata packets x86: h264dsp: Adjust YASM #ifdefs x86: yadif: Mark mmxext optimizations as such h264: convert loop filter strength dsp function to yasm. Improve descriptiveness of a number of codec and container long names Conflicts: libavcodec/flvdec.c libavcodec/libopenjpegdec.c libavformat/apetag.c libavformat/mp3dec.c Merged-by: Michael Niedermayer <michaelni(a)gmx.at> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=d1dad7c824c14939a… ---

1 0

x86: yadif: Mark mmxext optimizations as such
by git＠videolan.org 31 Jul '12

31 Jul '12

ffmpeg | branch: master | Diego Biurrun <diego(a)biurrun.de> | Tue Jul 31 12:17:02 2012 +0200| [d1505db06799d625420ff75de40af106c54e8c66] | committer: Diego Biurrun x86: yadif: Mark mmxext optimizations as such The yadif mmx optimizations contain the pmaxsw and pmaxub mmxext instructions, causing sigills on CPUs that do not support mmxext. > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=d1505db06799d6254… --- libavfilter/x86/yadif.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-) diff --git a/libavfilter/x86/yadif.c b/libavfilter/x86/yadif.c index 72b06cc..beb0824 100644 --- a/libavfilter/x86/yadif.c +++ b/libavfilter/x86/yadif.c @@ -45,9 +45,9 @@ DECLARE_ASM_CONST(16, const xmm_reg, pw_1) = {0x0001000100010001ULL, 0x000100010 #undef COMPILE_TEMPLATE_SSE #endif -#if HAVE_MMX +#if HAVE_MMX2 #undef RENAME -#define RENAME(a) a ## _mmx +#define RENAME(a) a ## _mmx2 #include "yadif_template.c" #endif @@ -58,9 +58,9 @@ av_cold void ff_yadif_init_x86(YADIFContext *yadif) int cpu_flags = av_get_cpu_flags(); #if HAVE_INLINE_ASM -#if HAVE_MMX - if (cpu_flags & AV_CPU_FLAG_MMX) - yadif->filter_line = yadif_filter_line_mmx; +#if HAVE_MMX2 + if (cpu_flags & AV_CPU_FLAG_MMX2) + yadif->filter_line = yadif_filter_line_mmx2; #endif #if HAVE_SSE if (cpu_flags & AV_CPU_FLAG_SSE2)

1 0

mpc8: return more meaningful error codes.
by git＠videolan.org 31 Jul '12

31 Jul '12

ffmpeg | branch: master | Anton Khirnov <anton(a)khirnov.net> | Mon Jul 30 07:29:34 2012 +0200| [f2ed006c90ccb65cd143d8b0fcfc28ffb98c4289] | committer: Anton Khirnov mpc8: return more meaningful error codes. > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=f2ed006c90ccb65cd… --- libavformat/mpc8.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/libavformat/mpc8.c b/libavformat/mpc8.c index 6b9f77e..b6837c5 100644 --- a/libavformat/mpc8.c +++ b/libavformat/mpc8.c @@ -202,7 +202,7 @@ static int mpc8_read_header(AVFormatContext *s) c->header_pos = avio_tell(pb); if(avio_rl32(pb) != TAG_MPCK){ av_log(s, AV_LOG_ERROR, "Not a Musepack8 file\n"); - return -1; + return AVERROR_INVALIDDATA; } while(!pb->eof_reached){ @@ -214,14 +214,14 @@ static int mpc8_read_header(AVFormatContext *s) } if(tag != TAG_STREAMHDR){ av_log(s, AV_LOG_ERROR, "Stream header not found\n"); - return -1; + return AVERROR_INVALIDDATA; } pos = avio_tell(pb); avio_skip(pb, 4); //CRC c->ver = avio_r8(pb); if(c->ver != 8){ av_log(s, AV_LOG_ERROR, "Unknown stream version %d\n", c->ver); - return -1; + return AVERROR_PATCHWELCOME; } c->samples = ffio_read_varlen(pb); ffio_read_varlen(pb); //silence samples at the beginning

1 0

mpc: return more meaningful error codes.
by git＠videolan.org 31 Jul '12

31 Jul '12

ffmpeg | branch: master | Anton Khirnov <anton(a)khirnov.net> | Sun Jul 29 22:19:35 2012 +0200| [263dbe9d953bfa717cd46698511cdcdc3d5d74b1] | committer: Anton Khirnov mpc: return more meaningful error codes. > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=263dbe9d953bfa717… --- libavformat/mpc.c | 16 ++++++++-------- 1 file changed, 8 insertions(+), 8 deletions(-) diff --git a/libavformat/mpc.c b/libavformat/mpc.c index 03d3629..404950b 100644 --- a/libavformat/mpc.c +++ b/libavformat/mpc.c @@ -59,17 +59,17 @@ static int mpc_read_header(AVFormatContext *s) if(avio_rl24(s->pb) != MKTAG('M', 'P', '+', 0)){ av_log(s, AV_LOG_ERROR, "Not a Musepack file\n"); - return -1; + return AVERROR_INVALIDDATA; } c->ver = avio_r8(s->pb); if(c->ver != 0x07 && c->ver != 0x17){ av_log(s, AV_LOG_ERROR, "Can demux Musepack SV7, got version %02X\n", c->ver); - return -1; + return AVERROR_INVALIDDATA; } c->fcount = avio_rl32(s->pb); if((int64_t)c->fcount * sizeof(MPCFrame) >= UINT_MAX){ av_log(s, AV_LOG_ERROR, "Too many frames, seeking is not possible\n"); - return -1; + return AVERROR_INVALIDDATA; } if(c->fcount){ c->frames = av_malloc(c->fcount * sizeof(MPCFrame)); @@ -122,7 +122,7 @@ static int mpc_read_packet(AVFormatContext *s, AVPacket *pkt) int64_t pos; if (c->curframe >= c->fcount && c->fcount) - return -1; + return AVERROR_EOF; if(c->curframe != c->lastframe + 1){ avio_seek(s->pb, c->frames[c->curframe].pos, SEEK_SET); @@ -151,8 +151,8 @@ static int mpc_read_packet(AVFormatContext *s, AVPacket *pkt) } c->curbits = (curbits + size2) & 0x1F; - if (av_new_packet(pkt, size) < 0) - return AVERROR(EIO); + if ((ret = av_new_packet(pkt, size)) < 0) + return ret; pkt->data[0] = curbits; pkt->data[1] = (c->curframe > c->fcount) && c->fcount; @@ -166,7 +166,7 @@ static int mpc_read_packet(AVFormatContext *s, AVPacket *pkt) avio_seek(s->pb, -4, SEEK_CUR); if(ret < size){ av_free_packet(pkt); - return AVERROR(EIO); + return ret < 0 ? ret : AVERROR(EIO); } pkt->size = ret + 4; @@ -214,7 +214,7 @@ static int mpc_read_seek(AVFormatContext *s, int stream_index, int64_t timestamp ret = av_read_frame(s, pkt); if (ret < 0){ c->curframe = lastframe; - return -1; + return ret; } av_free_packet(pkt); }

1 0

wv,mpc8: don't return apetag data in packets.
by git＠videolan.org 31 Jul '12

31 Jul '12

ffmpeg | branch: master | Anton Khirnov <anton(a)khirnov.net> | Mon Jul 30 07:28:35 2012 +0200| [782e64fbe1daa84c38594db037d0edfac7193c37] | committer: Anton Khirnov wv,mpc8: don't return apetag data in packets. > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=782e64fbe1daa84c3… --- libavformat/apetag.c | 21 +++++++++++++++------ libavformat/apetag.h | 4 +++- libavformat/mpc8.c | 9 ++++++++- libavformat/wv.c | 9 ++++++++- 4 files changed, 34 insertions(+), 9 deletions(-) diff --git a/libavformat/apetag.c b/libavformat/apetag.c index 062f746..dd74631 100644 --- a/libavformat/apetag.c +++ b/libavformat/apetag.c @@ -113,40 +113,47 @@ static int ape_tag_read_field(AVFormatContext *s) return 0; } -void ff_ape_parse_tag(AVFormatContext *s) +int64_t ff_ape_parse_tag(AVFormatContext *s) { AVIOContext *pb = s->pb; int file_size = avio_size(pb); uint32_t val, fields, tag_bytes; uint8_t buf[8]; + int64_t tag_start; int i; if (file_size < APE_TAG_FOOTER_BYTES) - return; + return 0; avio_seek(pb, file_size - APE_TAG_FOOTER_BYTES, SEEK_SET); avio_read(pb, buf, 8); /* APETAGEX */ if (strncmp(buf, "APETAGEX", 8)) { - return; + return 0; } val = avio_rl32(pb); /* APE tag version */ if (val > APE_TAG_VERSION) { av_log(s, AV_LOG_ERROR, "Unsupported tag version. (>=%d)\n", APE_TAG_VERSION); - return; + return 0; } tag_bytes = avio_rl32(pb); /* tag size */ if (tag_bytes - APE_TAG_FOOTER_BYTES > (1024 * 1024 * 16)) { av_log(s, AV_LOG_ERROR, "Tag size is way too big\n"); - return; + return 0; + } + + tag_start = file_size - tag_bytes - APE_TAG_FOOTER_BYTES; + if (tag_start < 0) { + av_log(s, AV_LOG_ERROR, "Invalid tag size %u.\n", tag_bytes); + return 0; } fields = avio_rl32(pb); /* number of fields */ if (fields > 65536) { av_log(s, AV_LOG_ERROR, "Too many tag fields (%d)\n", fields); - return; + return 0; } val = avio_rl32(pb); /* flags */ @@ -159,4 +166,6 @@ void ff_ape_parse_tag(AVFormatContext *s) for (i=0; i<fields; i++) if (ape_tag_read_field(s) < 0) break; + + return tag_start; } diff --git a/libavformat/apetag.h b/libavformat/apetag.h index a0e6ead..279972f 100644 --- a/libavformat/apetag.h +++ b/libavformat/apetag.h @@ -27,7 +27,9 @@ /** * Read and parse an APE tag + * + * @return offset of the tag start in the file */ -void ff_ape_parse_tag(AVFormatContext *s); +int64_t ff_ape_parse_tag(AVFormatContext *s); #endif /* AVFORMAT_APETAG_H */ diff --git a/libavformat/mpc8.c b/libavformat/mpc8.c index 93848f4..6b9f77e 100644 --- a/libavformat/mpc8.c +++ b/libavformat/mpc8.c @@ -52,6 +52,8 @@ typedef struct { int frame; int64_t header_pos; int64_t samples; + + int64_t apetag_start; } MPCContext; static inline int64_t bs_get_v(uint8_t **bs) @@ -243,7 +245,7 @@ static int mpc8_read_header(AVFormatContext *s) if (pb->seekable) { int64_t pos = avio_tell(s->pb); - ff_ape_parse_tag(s); + c->apetag_start = ff_ape_parse_tag(s); avio_seek(s->pb, pos, SEEK_SET); } @@ -258,6 +260,11 @@ static int mpc8_read_packet(AVFormatContext *s, AVPacket *pkt) while(!s->pb->eof_reached){ pos = avio_tell(s->pb); + + /* don't return bogus packets with the ape tag data */ + if (c->apetag_start && pos >= c->apetag_start) + return AVERROR_EOF; + mpc8_get_chunk_header(s->pb, &tag, &size); if (size < 0) return -1; diff --git a/libavformat/wv.c b/libavformat/wv.c index 3fd1abc..39bb614 100644 --- a/libavformat/wv.c +++ b/libavformat/wv.c @@ -64,6 +64,8 @@ typedef struct { int block_parsed; uint8_t extra[WV_EXTRA_SIZE]; int64_t pos; + + int64_t apetag_start; } WVContext; static int wv_probe(AVProbeData *p) @@ -88,6 +90,11 @@ static int wv_read_block_header(AVFormatContext *ctx, AVIOContext *pb, uint32_t chmask; wc->pos = avio_tell(pb); + + /* don't return bogus packets with the ape tag data */ + if (wc->apetag_start && wc->pos >= wc->apetag_start) + return AVERROR_EOF; + if (!append) { tag = avio_rl32(pb); if (tag != MKTAG('w', 'v', 'p', 'k')) @@ -252,7 +259,7 @@ static int wv_read_header(AVFormatContext *s) if (s->pb->seekable) { int64_t cur = avio_tell(s->pb); - ff_ape_parse_tag(s); + wc->apetag_start = ff_ape_parse_tag(s); if (!av_dict_get(s->metadata, "", NULL, AV_DICT_IGNORE_SUFFIX)) ff_id3v1_read(s); avio_seek(s->pb, cur, SEEK_SET);

1 0

rtmp: do not warn about receiving metadata packets
by git＠videolan.org 31 Jul '12

31 Jul '12

ffmpeg | branch: master | Samuel Pitoiset <samuel.pitoiset(a)gmail.com> | Thu Jul 26 21:10:09 2012 +0200| [9c9c21eaa1f0adf1c5cd17f9b31a18b979bffe37] | committer: Luca Barbato rtmp: do not warn about receiving metadata packets They are managed in get_packet() Signed-off-by: Luca Barbato <lu_zero(a)gentoo.org> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=9c9c21eaa1f0adf1c… --- libavformat/rtmpproto.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/libavformat/rtmpproto.c b/libavformat/rtmpproto.c index 501e0ed..4a35319 100644 --- a/libavformat/rtmpproto.c +++ b/libavformat/rtmpproto.c @@ -1114,7 +1114,8 @@ static int rtmp_parse_result(URLContext *s, RTMPContext *rt, RTMPPacket *pkt) break; case RTMP_PT_VIDEO: case RTMP_PT_AUDIO: - /* Audio and Video packets are parsed in get_packet() */ + case RTMP_PT_METADATA: + /* Audio, Video and Metadata packets are parsed in get_packet() */ break; default: av_log(s, AV_LOG_VERBOSE, "Unknown packet type received 0x%02X\n", pkt->type);

1 0

x86: h264dsp: Adjust YASM #ifdefs
by git＠videolan.org 31 Jul '12

31 Jul '12

ffmpeg | branch: master | Diego Biurrun <diego(a)biurrun.de> | Tue Jul 31 12:32:29 2012 +0200| [8728b381cbacbdff23ba9a85c2fcb78e4fd84966] | committer: Diego Biurrun x86: h264dsp: Adjust YASM #ifdefs This fixes compilation with YASM disabled. > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=8728b381cbacbdff2… --- libavcodec/x86/h264dsp_mmx.c | 4 +--- 1 file changed, 1 insertion(+), 3 deletions(-) diff --git a/libavcodec/x86/h264dsp_mmx.c b/libavcodec/x86/h264dsp_mmx.c index 5d9da99..4bea788 100644 --- a/libavcodec/x86/h264dsp_mmx.c +++ b/libavcodec/x86/h264dsp_mmx.c @@ -194,9 +194,9 @@ H264_BIWEIGHT_10_SSE( 4, 10) void ff_h264dsp_init_x86(H264DSPContext *c, const int bit_depth, const int chroma_format_idc) { +#if HAVE_YASM int mm_flags = av_get_cpu_flags(); -#if HAVE_YASM if (chroma_format_idc == 1 && mm_flags & AV_CPU_FLAG_MMX2) { c->h264_loop_filter_strength = ff_h264_loop_filter_strength_mmx2; } @@ -282,9 +282,7 @@ void ff_h264dsp_init_x86(H264DSPContext *c, const int bit_depth, const int chrom } } } -#endif } else if (bit_depth == 10) { -#if HAVE_YASM if (mm_flags & AV_CPU_FLAG_MMX) { if (mm_flags & AV_CPU_FLAG_MMX2) { #if ARCH_X86_32

1 0

h264: convert loop filter strength dsp function to yasm.
by git＠videolan.org 31 Jul '12

31 Jul '12

ffmpeg | branch: master | Ronald S. Bultje <rsbultje(a)gmail.com> | Sat Jul 28 10:11:00 2012 -0700| [b829b4ce29185625ab8cbcf0ce7a83cf8181ac3b] | committer: Ronald S. Bultje h264: convert loop filter strength dsp function to yasm. This completes the conversion of h264dsp to yasm; note that h264 also uses some dsputil functions, most notably qpel. Performance-wise, the yasm-version is ~10 cycles faster (182->172) on x86-64, and ~8 cycles faster (201->193) on x86-32. > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=b829b4ce29185625a… --- libavcodec/x86/h264_deblock.asm | 168 +++++++++++++++++++++++++++++++++++++++ libavcodec/x86/h264dsp_mmx.c | 162 ++----------------------------------- 2 files changed, 175 insertions(+), 155 deletions(-) diff --git a/libavcodec/x86/h264_deblock.asm b/libavcodec/x86/h264_deblock.asm index 0891ef3..940a8f7 100644 --- a/libavcodec/x86/h264_deblock.asm +++ b/libavcodec/x86/h264_deblock.asm @@ -27,6 +27,10 @@ %include "x86inc.asm" %include "x86util.asm" +SECTION_RODATA + +pb_3_1: times 4 db 3, 1 + SECTION .text cextern pb_0 @@ -911,3 +915,167 @@ ff_chroma_intra_body_mmx2: paddb m1, m5 paddb m2, m6 ret + +;----------------------------------------------------------------------------- +; void h264_loop_filter_strength(int16_t bs[2][4][4], uint8_t nnz[40], +; int8_t ref[2][40], int16_t mv[2][40][2], +; int bidir, int edges, int step, +; int mask_mv0, int mask_mv1, int field); +; +; bidir is 0 or 1 +; edges is 1 or 4 +; step is 1 or 2 +; mask_mv0 is 0 or 3 +; mask_mv1 is 0 or 1 +; field is 0 or 1 +;----------------------------------------------------------------------------- +%macro loop_filter_strength_iteration 7 ; edges, step, mask_mv, + ; dir, d_idx, mask_dir, bidir +%define edgesd %1 +%define stepd %2 +%define mask_mvd %3 +%define dir %4 +%define d_idx %5 +%define mask_dir %6 +%define bidir %7 + xor b_idxd, b_idxd ; for (b_idx = 0; b_idx < edges; b_idx += step) +%%.b_idx_loop: +%if mask_dir == 0 + pxor m0, m0 +%endif + test b_idxd, dword mask_mvd + jnz %%.skip_loop_iter ; if (!(b_idx & mask_mv)) +%if bidir == 1 + movd m2, [refq+b_idxq+d_idx+12] ; { ref0[bn] } + punpckldq m2, [refq+b_idxq+d_idx+52] ; { ref0[bn], ref1[bn] } + pshufw m0, [refq+b_idxq+12], 0x44 ; { ref0[b], ref0[b] } + pshufw m1, [refq+b_idxq+52], 0x44 ; { ref1[b], ref1[b] } + pshufw m3, m2, 0x4E ; { ref1[bn], ref0[bn] } + psubb m0, m2 ; { ref0[b] != ref0[bn], + ; ref0[b] != ref1[bn] } + psubb m1, m3 ; { ref1[b] != ref1[bn], + ; ref1[b] != ref0[bn] } + + por m0, m1 + mova m1, [mvq+b_idxq*4+(d_idx+12)*4] + mova m2, [mvq+b_idxq*4+(d_idx+12)*4+mmsize] + mova m3, m1 + mova m4, m2 + psubw m1, [mvq+b_idxq*4+12*4] + psubw m2, [mvq+b_idxq*4+12*4+mmsize] + psubw m3, [mvq+b_idxq*4+52*4] + psubw m4, [mvq+b_idxq*4+52*4+mmsize] + packsswb m1, m2 + packsswb m3, m4 + paddb m1, m6 + paddb m3, m6 + psubusb m1, m5 ; abs(mv[b] - mv[bn]) >= limit + psubusb m3, m5 + packsswb m1, m3 + + por m0, m1 + mova m1, [mvq+b_idxq*4+(d_idx+52)*4] + mova m2, [mvq+b_idxq*4+(d_idx+52)*4+mmsize] + mova m3, m1 + mova m4, m2 + psubw m1, [mvq+b_idxq*4+12*4] + psubw m2, [mvq+b_idxq*4+12*4+mmsize] + psubw m3, [mvq+b_idxq*4+52*4] + psubw m4, [mvq+b_idxq*4+52*4+mmsize] + packsswb m1, m2 + packsswb m3, m4 + paddb m1, m6 + paddb m3, m6 + psubusb m1, m5 ; abs(mv[b] - mv[bn]) >= limit + psubusb m3, m5 + packsswb m1, m3 + + pshufw m1, m1, 0x4E + por m0, m1 + pshufw m1, m0, 0x4E + pminub m0, m1 +%else ; bidir == 0 + movd m0, [refq+b_idxq+12] + psubb m0, [refq+b_idxq+d_idx+12] ; ref[b] != ref[bn] + + mova m1, [mvq+b_idxq*4+12*4] + mova m2, [mvq+b_idxq*4+12*4+mmsize] + psubw m1, [mvq+b_idxq*4+(d_idx+12)*4] + psubw m2, [mvq+b_idxq*4+(d_idx+12)*4+mmsize] + packsswb m1, m2 + paddb m1, m6 + psubusb m1, m5 ; abs(mv[b] - mv[bn]) >= limit + packsswb m1, m1 + por m0, m1 +%endif ; bidir == 1/0 + +%%.skip_loop_iter: + movd m1, [nnzq+b_idxq+12] + por m1, [nnzq+b_idxq+d_idx+12] ; nnz[b] || nnz[bn] + + pminub m1, m7 + pminub m0, m7 + psllw m1, 1 + pxor m2, m2 + pmaxub m1, m0 + punpcklbw m1, m2 + movq [bsq+b_idxq+32*dir], m1 + + add b_idxd, dword stepd + cmp b_idxd, dword edgesd + jl %%.b_idx_loop +%endmacro + +INIT_MMX mmx2 +cglobal h264_loop_filter_strength, 9, 9, 0, bs, nnz, ref, mv, bidir, edges, \ + step, mask_mv0, mask_mv1, field +%define b_idxq bidirq +%define b_idxd bidird + cmp dword fieldm, 0 + mova m7, [pb_1] + mova m5, [pb_3] + je .nofield + mova m5, [pb_3_1] +.nofield: + mova m6, m5 + paddb m5, m5 + + shl dword stepd, 3 + shl dword edgesd, 3 +%if ARCH_X86_32 +%define mask_mv0d mask_mv0m +%define mask_mv1d mask_mv1m +%endif + shl dword mask_mv1d, 3 + shl dword mask_mv0d, 3 + + cmp dword bidird, 0 + jne .bidir + loop_filter_strength_iteration edgesd, stepd, mask_mv1d, 1, -8, 0, 0 + loop_filter_strength_iteration 32, 8, mask_mv0d, 0, -1, -1, 0 + + mova m0, [bsq+mmsize*0] + mova m1, [bsq+mmsize*1] + mova m2, [bsq+mmsize*2] + mova m3, [bsq+mmsize*3] + TRANSPOSE4x4W 0, 1, 2, 3, 4 + mova [bsq+mmsize*0], m0 + mova [bsq+mmsize*1], m1 + mova [bsq+mmsize*2], m2 + mova [bsq+mmsize*3], m3 + RET + +.bidir: + loop_filter_strength_iteration edgesd, stepd, mask_mv1d, 1, -8, 0, 1 + loop_filter_strength_iteration 32, 8, mask_mv0d, 0, -1, -1, 1 + + mova m0, [bsq+mmsize*0] + mova m1, [bsq+mmsize*1] + mova m2, [bsq+mmsize*2] + mova m3, [bsq+mmsize*3] + TRANSPOSE4x4W 0, 1, 2, 3, 4 + mova [bsq+mmsize*0], m0 + mova [bsq+mmsize*1], m1 + mova [bsq+mmsize*2], m2 + mova [bsq+mmsize*3], m3 + RET diff --git a/libavcodec/x86/h264dsp_mmx.c b/libavcodec/x86/h264dsp_mmx.c index 3f18f64..5d9da99 100644 --- a/libavcodec/x86/h264dsp_mmx.c +++ b/libavcodec/x86/h264dsp_mmx.c @@ -88,158 +88,10 @@ void ff_h264_luma_dc_dequant_idct_sse2(DCTELEM *output, DCTELEM *input, int qmul /***********************************/ /* deblocking */ -#define h264_loop_filter_strength_iteration_mmx2(bS, nz, ref, mv, bidir, edges, step, mask_mv, dir, d_idx, mask_dir) \ - do { \ - x86_reg b_idx; \ - mask_mv <<= 3; \ - for( b_idx=0; b_idx<edges; b_idx+=step ) { \ - if (!mask_dir) \ - __asm__ volatile( \ - "pxor %%mm0, %%mm0 \n\t" \ - :: \ - ); \ - if(!(mask_mv & b_idx)) { \ - if(bidir) { \ - __asm__ volatile( \ - "movd %a3(%0,%2), %%mm2 \n" \ - "punpckldq %a4(%0,%2), %%mm2 \n" /* { ref0[bn], ref1[bn] } */ \ - "pshufw $0x44, 12(%0,%2), %%mm0 \n" /* { ref0[b], ref0[b] } */ \ - "pshufw $0x44, 52(%0,%2), %%mm1 \n" /* { ref1[b], ref1[b] } */ \ - "pshufw $0x4E, %%mm2, %%mm3 \n" \ - "psubb %%mm2, %%mm0 \n" /* { ref0[b]!=ref0[bn], ref0[b]!=ref1[bn] } */ \ - "psubb %%mm3, %%mm1 \n" /* { ref1[b]!=ref1[bn], ref1[b]!=ref0[bn] } */ \ - \ - "por %%mm1, %%mm0 \n" \ - "movq %a5(%1,%2,4), %%mm1 \n" \ - "movq %a6(%1,%2,4), %%mm2 \n" \ - "movq %%mm1, %%mm3 \n" \ - "movq %%mm2, %%mm4 \n" \ - "psubw 48(%1,%2,4), %%mm1 \n" \ - "psubw 56(%1,%2,4), %%mm2 \n" \ - "psubw 208(%1,%2,4), %%mm3 \n" \ - "psubw 216(%1,%2,4), %%mm4 \n" \ - "packsswb %%mm2, %%mm1 \n" \ - "packsswb %%mm4, %%mm3 \n" \ - "paddb %%mm6, %%mm1 \n" \ - "paddb %%mm6, %%mm3 \n" \ - "psubusb %%mm5, %%mm1 \n" /* abs(mv[b] - mv[bn]) >= limit */ \ - "psubusb %%mm5, %%mm3 \n" \ - "packsswb %%mm3, %%mm1 \n" \ - \ - "por %%mm1, %%mm0 \n" \ - "movq %a7(%1,%2,4), %%mm1 \n" \ - "movq %a8(%1,%2,4), %%mm2 \n" \ - "movq %%mm1, %%mm3 \n" \ - "movq %%mm2, %%mm4 \n" \ - "psubw 48(%1,%2,4), %%mm1 \n" \ - "psubw 56(%1,%2,4), %%mm2 \n" \ - "psubw 208(%1,%2,4), %%mm3 \n" \ - "psubw 216(%1,%2,4), %%mm4 \n" \ - "packsswb %%mm2, %%mm1 \n" \ - "packsswb %%mm4, %%mm3 \n" \ - "paddb %%mm6, %%mm1 \n" \ - "paddb %%mm6, %%mm3 \n" \ - "psubusb %%mm5, %%mm1 \n" /* abs(mv[b] - mv[bn]) >= limit */ \ - "psubusb %%mm5, %%mm3 \n" \ - "packsswb %%mm3, %%mm1 \n" \ - \ - "pshufw $0x4E, %%mm1, %%mm1 \n" \ - "por %%mm1, %%mm0 \n" \ - "pshufw $0x4E, %%mm0, %%mm1 \n" \ - "pminub %%mm1, %%mm0 \n" \ - ::"r"(ref), \ - "r"(mv), \ - "r"(b_idx), \ - "i"(d_idx+12), \ - "i"(d_idx+52), \ - "i"(d_idx*4+48), \ - "i"(d_idx*4+56), \ - "i"(d_idx*4+208), \ - "i"(d_idx*4+216) \ - ); \ - } else { \ - __asm__ volatile( \ - "movd 12(%0,%2), %%mm0 \n" \ - "psubb %a3(%0,%2), %%mm0 \n" /* ref[b] != ref[bn] */ \ - "movq 48(%1,%2,4), %%mm1 \n" \ - "movq 56(%1,%2,4), %%mm2 \n" \ - "psubw %a4(%1,%2,4), %%mm1 \n" \ - "psubw %a5(%1,%2,4), %%mm2 \n" \ - "packsswb %%mm2, %%mm1 \n" \ - "paddb %%mm6, %%mm1 \n" \ - "psubusb %%mm5, %%mm1 \n" /* abs(mv[b] - mv[bn]) >= limit */ \ - "packsswb %%mm1, %%mm1 \n" \ - "por %%mm1, %%mm0 \n" \ - ::"r"(ref), \ - "r"(mv), \ - "r"(b_idx), \ - "i"(d_idx+12), \ - "i"(d_idx*4+48), \ - "i"(d_idx*4+56) \ - ); \ - } \ - } \ - __asm__ volatile( \ - "movd 12(%0,%1), %%mm1 \n" \ - "por %a2(%0,%1), %%mm1 \n" /* nnz[b] || nnz[bn] */ \ - ::"r"(nnz), \ - "r"(b_idx), \ - "i"(d_idx+12) \ - ); \ - __asm__ volatile( \ - "pminub %%mm7, %%mm1 \n" \ - "pminub %%mm7, %%mm0 \n" \ - "psllw $1, %%mm1 \n" \ - "pxor %%mm2, %%mm2 \n" \ - "pmaxub %%mm0, %%mm1 \n" \ - "punpcklbw %%mm2, %%mm1 \n" \ - "movq %%mm1, %a1(%0,%2) \n" \ - ::"r"(bS), \ - "i"(32*dir), \ - "r"(b_idx) \ - :"memory" \ - ); \ - } \ - } while (0) - -static void h264_loop_filter_strength_mmx2( int16_t bS[2][4][4], uint8_t nnz[40], int8_t ref[2][40], int16_t mv[2][40][2], - int bidir, int edges, int step, int mask_mv0, int mask_mv1, int field ) { - __asm__ volatile( - "movq %0, %%mm7 \n" - "movq %1, %%mm6 \n" - ::"m"(ff_pb_1), "m"(ff_pb_3) - ); - if(field) - __asm__ volatile( - "movq %0, %%mm6 \n" - ::"m"(ff_pb_3_1) - ); - __asm__ volatile( - "movq %%mm6, %%mm5 \n" - "paddb %%mm5, %%mm5 \n" - :); - - // could do a special case for dir==0 && edges==1, but it only reduces the - // average filter time by 1.2% - step <<= 3; - edges <<= 3; - h264_loop_filter_strength_iteration_mmx2(bS, nnz, ref, mv, bidir, edges, step, mask_mv1, 1, -8, 0); - h264_loop_filter_strength_iteration_mmx2(bS, nnz, ref, mv, bidir, 32, 8, mask_mv0, 0, -1, -1); - - __asm__ volatile( - "movq (%0), %%mm0 \n\t" - "movq 8(%0), %%mm1 \n\t" - "movq 16(%0), %%mm2 \n\t" - "movq 24(%0), %%mm3 \n\t" - TRANSPOSE4(%%mm0, %%mm1, %%mm2, %%mm3, %%mm4) - "movq %%mm0, (%0) \n\t" - "movq %%mm3, 8(%0) \n\t" - "movq %%mm4, 16(%0) \n\t" - "movq %%mm2, 24(%0) \n\t" - ::"r"(bS[0]) - :"memory" - ); -} +void ff_h264_loop_filter_strength_mmx2(int16_t bS[2][4][4], uint8_t nnz[40], + int8_t ref[2][40], int16_t mv[2][40][2], + int bidir, int edges, int step, + int mask_mv0, int mask_mv1, int field); #define LF_FUNC(DIR, TYPE, DEPTH, OPT) \ void ff_deblock_ ## DIR ## _ ## TYPE ## _ ## DEPTH ## _ ## OPT (uint8_t *pix, int stride, \ @@ -344,12 +196,12 @@ void ff_h264dsp_init_x86(H264DSPContext *c, const int bit_depth, const int chrom { int mm_flags = av_get_cpu_flags(); +#if HAVE_YASM if (chroma_format_idc == 1 && mm_flags & AV_CPU_FLAG_MMX2) { - c->h264_loop_filter_strength= h264_loop_filter_strength_mmx2; + c->h264_loop_filter_strength = ff_h264_loop_filter_strength_mmx2; } if (bit_depth == 8) { -#if HAVE_YASM if (mm_flags & AV_CPU_FLAG_MMX) { c->h264_idct_dc_add = c->h264_idct_add = ff_h264_idct_add_8_mmx; @@ -510,6 +362,6 @@ void ff_h264dsp_init_x86(H264DSPContext *c, const int bit_depth, const int chrom #endif /* HAVE_AVX */ } } -#endif } +#endif }

1 0