February 2010 - ffmpeg-devel - lists.ffmpeg.org

[PATCH] ARMv6 optimised VP3 loop filter
by lessen42＠gmail.com 30 Jun '10

30 Jun '10

10% faster overall on a 1g ipod touch --- libavcodec/Makefile | 3 + libavcodec/arm/dsputil_init_armv6.c | 8 ++ libavcodec/arm/vp3dsp_armv6.S | 126 +++++++++++++++++++++++++++++++++++ 3 files changed, 137 insertions(+), 0 deletions(-) create mode 100644 libavcodec/arm/vp3dsp_armv6.S diff --git a/libavcodec/Makefile b/libavcodec/Makefile index 0b58459..6108d30 100644 --- a/libavcodec/Makefile +++ b/libavcodec/Makefile @@ -610,9 +610,12 @@ OBJS-$(HAVE_ARMV5TE) += arm/dsputil_init_armv5te.o \ arm/mpegvideo_armv5te_s.o \ arm/simple_idct_armv5te.o \ +ARMV6-OBJS-$(CONFIG_VP3_DECODER) += arm/vp3dsp_armv6.o + OBJS-$(HAVE_ARMV6) += arm/dsputil_init_armv6.o \ arm/dsputil_armv6.o \ arm/simple_idct_armv6.o \ + $(ARMV6-OBJS-yes) OBJS-$(HAVE_ARMVFP) += arm/dsputil_vfp.o \ arm/dsputil_init_vfp.o \ diff --git a/libavcodec/arm/dsputil_init_armv6.c b/libavcodec/arm/dsputil_init_armv6.c index 3209062..2d7a354 100644 --- a/libavcodec/arm/dsputil_init_armv6.c +++ b/libavcodec/arm/dsputil_init_armv6.c @@ -70,6 +70,9 @@ int ff_sse16_armv6(void *s, uint8_t *blk1, uint8_t *blk2, int ff_pix_norm1_armv6(uint8_t *pix, int line_size); int ff_pix_sum_armv6(uint8_t *pix, int line_size); +void ff_vp3_v_loop_filter_armv6(uint8_t *, int, int *); +void ff_vp3_h_loop_filter_armv6(uint8_t *, int, int *); + void av_cold ff_dsputil_init_armv6(DSPContext* c, AVCodecContext *avctx) { if (!avctx->lowres && (avctx->idct_algo == FF_IDCT_AUTO || @@ -118,4 +121,9 @@ void av_cold ff_dsputil_init_armv6(DSPContext* c, AVCodecContext *avctx) c->pix_norm1 = ff_pix_norm1_armv6; c->pix_sum = ff_pix_sum_armv6; + + if (CONFIG_VP3_DECODER) { + c->vp3_v_loop_filter = ff_vp3_v_loop_filter_armv6; + c->vp3_h_loop_filter = ff_vp3_h_loop_filter_armv6; + } } diff --git a/libavcodec/arm/vp3dsp_armv6.S b/libavcodec/arm/vp3dsp_armv6.S new file mode 100644 index 0000000..3d751e6 --- /dev/null +++ b/libavcodec/arm/vp3dsp_armv6.S @@ -0,0 +1,126 @@ +/* + * Copyright (c) 2010 David Conrad + * + * 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 "asm.S" + +.text + +// ((p0 - p3) + 3*(p2 - p1) + 4) >> 3 has range -127, 128 +// so we instead calculate the negative, so it fits in signed 8 bits + +function ff_vp3_v_loop_filter_armv6, export=1 + push {r4-r10,lr} + sub r0, r0, r1 + ldr r2, [r2, #129*4] + ldr r3, =0x030003 + ldr ip, =0xff00ff + mov lr, #8 +1: + ldr r4, [r0,-r1] + ldr r5, [r0] + ldr r6, [r0, r1] + ldr r7, [r0, r1, lsl #1] + subs lr, lr, #4 + + uxtb16 r10, r4 + uxtb16 r8, r5 + uxtb16 r9, r6 + ssub16 r8, r8, r9 + sadd16 r9, r8, r8 + sadd16 r8, r8, r9 // 3*(p1-p2) + uxtb16 r9, r7 + ssub16 r9, r9, r10 + sadd16 r8, r8, r9 + sadd16 r8, r8, r3 // (p3 - p0) + 3*(p1 - p2) + 3 + uxtb16 r9, r5, ror #8 + and r8, ip, r8, lsr #3 + + uxtb16 r10, r6, ror #8 + ssub16 r9, r9, r10 + sadd16 r10, r9, r9 + sadd16 r9, r9, r10 + uxtb16 r4, r4, ror #8 + uxtb16 r7, r7, ror #8 + ssub16 r4, r7, r4 + sadd16 r4, r4, r9 + sadd16 r4, r4, r3 + mov r10, #0 + and r4, ip, r4, lsr #3 + orr r8, r8, r4, lsl #8 // 4 filter values + + ssub8 r9, r10, r8 + sel r9, r9, r8 // abs(filter_value) + uqsub8 r7, r2, r9 + uqsub8 r9, r7, r9 + usub8 r9, r7, r9 +// add and sub, then select the right one based on the original sign + uqsub8 r7, r6, r9 + uqadd8 r6, r6, r9 + uqadd8 r4, r5, r9 + uqsub8 r5, r5, r9 + ssub8 r10, r8, r10 + sel r6, r6, r7 + sel r5, r5, r4 + str r6, [r0, r1] + str r5, [r0], #4 + bgt 1b + pop {r4-r10,pc} +.endfunc + +function ff_vp3_h_loop_filter_armv6, export=1 + push {r4-r10,lr} + sub ip, r0, #1 + sub r0, r0, #2 + ldr r3, =0x010003 + mov r10, #4 + mov lr, #8 +1: + ldr r4, [r0], r1 + ldr r7, [r0], r1 + subs lr, lr, #2 + + uxtb16 r5, r4, ror #8 + uxtb16 r6, r4, ror #16 + ssub16 r6, r6, r5 // p0-p3 p2-p1 + smlad r6, r6, r3, r10 // (p0-p3) + 3*(p2-p1) + 4 + + uxtb16 r8, r7, ror #8 + uxtb16 r9, r7, ror #16 + ssub16 r9, r9, r8 + smlad r9, r9, r3, r10 + +.macro filter p1 p2 r + asr \r, \r, #3 + ldr \r, [r2, \r, lsl #2] // filter value + uxtb \p2, \p1, ror #16 // p2 + uxtb \p1, \p1, ror #8 // p1 + sub \p2, \p2, \r + add \p1, \p1, \r + usat \p2, #8, \p2 + usat \p1, #8, \p1 + strb \p2, [ip, #1] + strb \p1, [ip], r1 +.endm + + filter r4, r5, r6 + filter r7, r8, r9 + bgt 1b + pop {r4-r10,pc} +.endfunc -- 1.6.6

2 1

[PATCH 1/6] flacdec: change variable-length array to fixed length
by mans＠mansr.com 30 Jun '10

30 Jun '10

pred_order can never exceed 64, so always allocating that amount is safe and not very wasteful. --- libavcodec/flacdec.c | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-) diff --git a/libavcodec/flacdec.c b/libavcodec/flacdec.c index 4ff7fae..643ed24 100644 --- a/libavcodec/flacdec.c +++ b/libavcodec/flacdec.c @@ -358,7 +358,7 @@ static int decode_subframe_lpc(FLACContext *s, int channel, int pred_order) { int i, j; int coeff_prec, qlevel; - int coeffs[pred_order]; + int coeffs[64]; int32_t *decoded = s->decoded[channel]; /* warm up samples */ -- 1.6.4

5 24

[PATCH] Default to using libraries when enabled
by lessen42＠gmail.com 25 Jun '10

25 Jun '10

Hi, This changes the order of initialization so that external libraries that are enabled are used by default over native codecs. My main motivation in moving these is to be able to have the ogg and mkv muxers default to encoding via libvorbis if it's enabled; currently, specifying CODEC_ID_VORBIS will use the native encoder regardless, which is much worse quality. However, this also affects faad and xvid. IMO, faad should be preferred anyway until the native decoder supports SBR, then dropped soon after. My only argument for xvid is that it's probably what users want to use when they --enable-libxvid.

11 37

IEC61937 compatible muxer
by bartek.wolowiec＠gmail.com 16 Jun '10

16 Jun '10

Hi, Here is IEC61937 compatible muxer. It can be used for sending compressed data over S/PDIF link. It's currently not very useful, but can be tested by: ./ffmpeg -y -i test_48khz.dts -acodec copy -f spdif test.spdif && ./ffmpeg -ar 48000 -f s16le -ac 2 -i test.spdif -f alsa plughw:0 -- Bartlomiej Wolowiec

7 23

[PATCH] -fs parameter human friendly
by stefano.sabatini-lala＠poste.it 09 Jun '10

09 Jun '10

Hi all, addresses issue 551. It exports the av_strtod() function to eval.h and uses it in ffmpeg.c (minor bump missing for easying the maintainance of the patch). Regards. -- FFmpeg = Funny and Fiendish Meaningless Patchable Embarassing Guide

3 18

a violation of license
by nail＠nodomain.sk 30 May '10

30 May '10

Hello, just for case: I downloaded software for uploading/converting videos/pictures to server rajce.net [1] (developer notes: [2] english translation [3]) which containing ffmpeg and ffplay binaries for win32, without licence and ffmpeg source code. This is testing release. ~/.wine/drive_c/Program Files/rajce_test$ ls ffmpeg.exe ffplay.exe FreeImage.dll FreeImagePlus.ver rajce-cs.lng rajce-en-us.lng rajce-it-it.lng rajce_proxy_setting.ini rajce-uk.lng unins000.dat updater.exe ffmpeg.ver ffplay.ver FreeImagePlus.dll FreeImage.ver rajce-de-de.lng rajce.exe rajce-pl.lng rajce-sk.lng SDL.dll unins000.exe updater.ver wine ffplay.exe FFplay version SVN-r18133, Copyright (c) 2003-2009 Fabrice Bellard, et al. configuration: --enable-memalign-hack --extra-cflags=-fno-common --enable-gpl --enable-libfaac --enable-libfaad --enable-libmp3lame --enable-libtheora --enable-libx264 --enable-libxvid --enable-avfilter --enable-libamr-nb --enable-libamr-wb --enable-nonfree libavutil 50. 1. 0 / 50. 1. 0 libavcodec 52.22. 2 / 52.22. 2 libavformat 52.32. 0 / 52.32. 0 libavdevice 52. 1. 0 / 52. 1. 0 libavfilter 0. 4. 0 / 0. 4. 0 libswscale 0. 7. 1 / 0. 7. 1 built on Mar 31 2009 22:08:09, gcc: 4.2.1-sjlj (mingw32-2) An input file must be specified wine ffmpeg.exe FFmpeg version SVN-r18133, Copyright (c) 2000-2009 Fabrice Bellard, et al. configuration: --enable-memalign-hack --extra-cflags=-fno-common --enable-gpl --enable-libfaac --enable-libfaad --enable-libmp3lame --enable-libtheora --enable-libx264 --enable-libxvid --enable-avfilter --enable-libamr-nb --enable-libamr-wb --enable-nonfree libavutil 50. 1. 0 / 50. 1. 0 libavcodec 52.22. 2 / 52.22. 2 libavformat 52.32. 0 / 52.32. 0 libavdevice 52. 1. 0 / 52. 1. 0 libavfilter 0. 4. 0 / 0. 4. 0 libswscale 0. 7. 1 / 0. 7. 1 built on Mar 31 2009 22:08:09, gcc: 4.2.1-sjlj (mingw32-2) At least one output file must be specified [1] http://www.rajce.idnes.cz/stahnout/test/rajce_test56v-129.exe [2] https://docs.google.com/Doc?docid=ajkd99k8zcw6_186vg5dccfp [3] http://translate.google.com/translate?hl=en&sl=cs&tl=en&u=http%3A%2F%2Fdocs…

3 4

[PATCH] add md5 muxer
by Reimar.Doeffinger＠gmx.de 25 May '10

25 May '10

Hello, we have -f crc, but IMO 32 bit crc are a bit little for a whole file and also we use md5sum in the regression tests, so I'd propose to add a -f md5 as in attached patch.

11 32

Again problems with lavfi/swscale
by stefano.sabatini-lala＠poste.it 16 May '10

16 May '10

I'm experiencing problems with commands of the kind: ffplay in.avi -vfilters "slicify=20, format=FORMAT" yuv444p,yuv411p,yuv422p,monow -> crash in RENAME(hScale) rgba,argb,rgb24,bgr24,rgb555 -> crash in YUV2RGBFUNC()/LOADCHROMA yuv410p -> chroma artefacts rgb8 -> crash in PUTRGB8 Formats which work fine: yuv420p, yuvj420p, monob nv12, nv21 (with the patch by Ramiro) gray, gray16 The commands without the slicify filter work just fine, so the problem seems to depend on the slicification of the image (yes at this point is clear that we need some testing system for spotting all these problems). If someone with more experience with libswscale wants to give some hint it would be cool. Regards. -- FFmpeg = Fundamental and Frenzy Merciful Ponderous Extreme Guide

7 49

feeding FFmpeg with images
by dennis＠conus.info 07 May '10

07 May '10

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Hi. For image2 option: it seems, it is not possible to set timestamps for all input images? Only possible to set framerate? If I want to add this feature, what is the best place if FFmpeg to start hacking? I probably need to modify img_read_header() and img_read_packet(). What structures fields I should use? Thanks in advance! - -- My PGP public key: http://yurichev.com/dennis.yurichev.asc -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.10 (MingW32) Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/ iEYEARECAAYFAkt9VNUACgkQ1YPmFmJG++MdQwCgoBMT5TkpIOLRyXEsnZ4A4sk2 9SoAoNKXjUPCg/j/GA3Ufx0lsT9hcrEt =FLfx -----END PGP SIGNATURE-----

6 9

feeding FFmpeg with images
by dennis＠conus.info 05 May '10

05 May '10

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Hi. I tried to use image2 option to feed FFmpeg with series of PNG files. I got latest FFmpeg from http://www.videohelp.com/tools/ffmpeg and it works. I got latest source code from SVN, build it with MinGW and configure with "--enable-memalign-hack" option. It doesn't work: == C:\msys\1.0\local\bin>ffmpeg -r 1 -f image2 -i tmp_%04d.png out.avi FFmpeg version SVN-r21861, Copyright (c) 2000-2010 Fabrice Bellard, et al. built on Feb 17 2010 16:15:26 with gcc 3.4.5 (mingw-vista special r3) configuration: --enable-memalign-hack --enable-bzlib --enable-avisynth libavutil 50. 9. 0 / 50. 9. 0 libavcodec 52.54. 0 / 52.54. 0 libavformat 52.52. 0 / 52.52. 0 libavdevice 52. 2. 0 / 52. 2. 0 libswscale 0.10. 0 / 0.10. 0 [image2 @ 0036c630]Could not find codec parameters (Video: 0x0000) tmp_%04d.png: could not find codec parameters == OK, do I need another library, which one? Next question: it seems, it is not possible to set timestamps for all input images? Only possible to set framerate? If I want to add this feature, what is the best place if FFmpeg to start hacking? I probably need to modify img_read_header() and img_read_packet(). What structures fields I should use? Thanks in advance! - -- My PGP public key: http://yurichev.com/dennis.yurichev.asc -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.10 (MingW32) Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/ iEYEARECAAYFAkt7/7IACgkQ1YPmFmJG++MIcwCgy0q+9Bw+emiE7UKwMlxNYEiP n8kAoLAuq1HIoyqy18106h+As6sEWEck =GJNA -----END PGP SIGNATURE-----

4 4