[Ffmpeg-devel-irc] ffmpeg-devel.log.20140416

[burek]

[00:31] <BBB> ubitux: well yeah that's obvious right?
[00:31] <BBB> ubitux: the intended use is to go from high to low
[00:31] <BBB> e.g. if avx2 ... else if ssse3 ... else if mmx ... else ..
[00:31] <BBB> (the else is useless, but you get the point)
[00:31] <ubitux> yeah sure it makes sense
[00:32] <ubitux> i was just& braindead again
[00:32] <ubitux> i have a lot of trouble with my macro and the (x)mN variants all over for the mov*
[00:32] <ubitux> macros*
[00:33] <ubitux> like... error: undefined symbol `xymm6' (first use) :))
[00:34] <cone-876> ffmpeg.git 03Michael Niedermayer 07master:6d8ccc7ac7e6: avformat/flacdec/flac_read_timestamp: dont modify AVStream state
[00:34] <cone-876> ffmpeg.git 03Michael Niedermayer 07master:bdc812ab9edb: avformat/flacdec/flac_read_timestamp: free the correct packet
[00:40] <ubitux> BBB: http://pastie.org/9083303
[00:40] <ubitux> first line works, following raises error: undefined symbol `xxmm0' (first use)
[00:41] <ubitux> i have no idea why :(
[00:43] <Daemon404> looks like x86inc magic might be adding the x?
[00:43] <Daemon404> (the first)
[00:45] <ubitux> why in one case but not the other?
[00:45] <jamrial> What about http://pastie.org/9083312? Some x86util macros do this
[00:48] <ubitux> jamrial: it seems to work but i have absolutely no idea why
[00:48] <jamrial> Neither do i :P
[00:49] <ubitux> i guess it's time for me to sleep
[00:49] <ubitux> 'night, and thx
[00:49] <jamrial> night
[01:20] <BBB> ubitux: m0 is xmm0 or mm0
[01:20] <BBB> what are you trying to do?
[01:21] <BBB> ubitux: also, you might want to ask in #x264dev, there's a lot of brilliant x86inc.asm knowledge in there
[01:23] <jamrial> m0 is ymm0 if you use INIT_YMM, which i assume is why he's doing that
[01:24] <jamrial> He wants that mova inside the macro to use xmm0
[01:26] <jamrial> I suppose the easiest way is to use the reg number as argument for the macro and use m%1, xm%1 and ym%1 as required
[01:26] <jamrial> like the SBUTTERFLY macros
[01:38] <J_Darnley> Sometimes I wish I had an account on trac just so I could post sarcastic comments
[01:40] <iive> 4.9 is not final release, is it?
[01:40] <J_Darnley> From the talk in that ticket, not yet
[01:42] <J_Darnley> And then navigating through to http://gcc.gnu.org/gcc-4.9/ gives me this...
[01:42] <J_Darnley> "As of this time no releases of GCC 4.9 have yet been made."
[01:44] <jamrial> There's an RC out http://gcc.gnu.org/ml/gcc/2014-04/msg00091.html
[01:44] <iive> yeh, but somebody should track and report the bug to gcc...
[01:54] <cone-876> ffmpeg.git 03Michael Niedermayer 07master:79d82a65744b: avformat/hnm: fix decoding of 256x256 video
[01:57] <BBB> jamrial: I think what you said is the most typical solution, i.e. use %d instead of m%d, and for specific register types, use xmm+%d instead of m+%d
[02:15] <cone-876> ffmpeg.git 03Michael Niedermayer 07master:14fa7fc6a81d: swscale/swscale: fix srcStride/srcSlice typo
[02:15] <cone-876> ffmpeg.git 03Michael Niedermayer 07master:256f530603ef: avcodec/mjpegen: Fix declared argument size
[03:40] <Compnn> > [...9
[03:40] <Compnn> michaelni left a bracket open! :)
[04:03] <cone-876> ffmpeg.git 03Michael Niedermayer 07master:f202af29c93f: avcodec/dcadec: fix error message suppression code
[04:03] <cone-876> ffmpeg.git 03Michael Niedermayer 07master:230db1426d12: avcodec/fic: Fix return value check
[04:03] <cone-876> ffmpeg.git 03Michael Niedermayer 07master:ff6fa0b4b980: avformat/mpegts: Remove redundant check
[04:03] <cone-876> ffmpeg.git 03Michael Niedermayer 07master:b8598f6ce61c: avcodec/diracdec: fix undefined behavior with shifts
[04:03] <cone-876> ffmpeg.git 03Michael Niedermayer 07master:1457f3fd90e1: avcodec/g723_1: add assert to help static code analyzers
[04:03] <cone-876> ffmpeg.git 03Michael Niedermayer 07master:5a8ef3c66b76: avfilter/f_select: fix loss of precission in SAD calculation
[04:03] <cone-876> ffmpeg.git 03Michael Niedermayer 07master:73734282e0e4: avfilter/vf_deshake: fix loss of precission with odd resolutions
[04:03] <cone-876> ffmpeg.git 03Michael Niedermayer 07master:cb53beb81a5b: iavcodec/vc1dec: Fix missing {}
[04:04] <michaelni> Compnn, ]
[07:11] <pzich> hmm, I'm guessing I'm SOL on this, but is there any way to get ffmpeg to allow me to set my own video_handler for .mp4 files, is there?
[07:11] <pzich> It looks like it's hard-coded into the movenc.c, and I have no idea where to start with making that optional.
[09:01] <ubitux> BBB: i'm expliciting xmN instead of mN in most of the lpf code where there are mov*
[09:01] <ubitux> basically, mova m0, [fooq] ’ mova xm0, [fooq]
[09:02] <ubitux> unfortunately when it's in a macro, it just doesn't work as shown in my pastie
[09:02] <ubitux> but i think i'll rewrite most of the code instead
[11:29] <Skyler_> ubitux: yeah, you can't do that because of the order of macro expansion in yasm
[11:29] <Skyler_> m0 gets expanded before the x gets concatenated
[11:29] <Skyler_> this isn't my forte though, I think only pengvado really understands that
[11:29] <ubitux> :(
[11:31] <nevcairiel> cant you just pass the number only
[11:31] <Skyler_> yes, that's the way that sbutterfly works for example
[11:31] <Skyler_> what exactly are you trying to do?
[11:32] <Skyler_> like.  why do you need to do this.
[11:32] <ubitux> 09:01:38 <@ubitux> BBB: i'm expliciting xmN instead of mN in most of the lpf code where there are mov*
[11:32] <ubitux> i'm probably doing it wrong anyway for other reasons
[11:33] <Skyler_> yes but why.  like.  why are you trying to do this.
[11:33] <ubitux> it's likely i'll rewrite all that code for avx2 anyway
[11:33] <ubitux> Skyler_: well, because the code works the same in avx for a large part of it, except i must make sure the load are 128 and not 256
[11:33] <ubitux> in avx2*
[11:34] <ubitux> but in fact, i can do better by rewriting most of the code around anyway, so... i'll probably do that
[12:46] <BBB> ubitux: can you pass a register number and do xm %+ %arg?
[12:46] <ubitux> probably
[12:46] <BBB> ubitux: so %macro 1 with implementation mova xm %+ %1, [mem] %endmacro
[12:46] <ubitux> i need to change the current existing macro first
[12:47] <ubitux> assuming i'm still going that way
[12:47] <BBB> well test it first :)
[12:47] <ubitux> yeah sure, it was suggested several times
[12:47] <ubitux> i'll do that tonight
[13:23] <michaelni> pzich, dont register the mov (de)muxer and register your own
[14:29] <crow> Hi, I posted it firstly in #ffmpeg but then i realized there is -devel , Hello, i am VDR with softhddevice outplut plugin, which use ffmpeg 2.2 and i am getting an segfault. Is it caused by ffmpeg or softhddevice output plugin from VDR? http://sprunge.us/PEjZ
[14:30] <crow> I am using VDR application which uses softhddevice as output plugin, which uses ffmpeg
[14:32] <crow> ffmpeg info : http://sprunge.us/cLdb
[15:53] <cone-972> ffmpeg.git 03James Almer 07master:3b808900af61: x86/float_dsp: use SWAP in vector_fmac_scalar Win64
[17:01] <plepere> w00t, finally done with the weighting.
[17:02] <plepere> got stuck on stupid bugs and all, but I think I'll be able to submit something soon
[17:21] <cone-972> ffmpeg.git 03Michael Niedermayer 07master:572f5eb1afc6: avformat/mpegenc: try to improve the VBV buffer warning
[18:36] <cone-972> ffmpeg.git 03Matt Oliver 07master:27f184ef40e8: Fix OpenGL device compilation with msvc gl.h
[19:13] <cone-972> ffmpeg.git 03James Almer 07master:11b36b1ee051: x86/float_dsp: unroll loop in vector_fmac_scalar
[19:20] <cone-972> ffmpeg.git 03Michael Niedermayer 07release/2.2:fe87a40de679: ffprobe: do not lose non ascii characters in non utf8 strings
[19:20] <cone-972> ffmpeg.git 03Michael Niedermayer 07release/2.2:56f44c26f06a: avutil/avstring: do not lose ascii characters when decoding non utf-8 with av_utf8_decode()
[19:52] <cone-972> ffmpeg.git 03Michael Niedermayer 07master:dd77f6469a89: avformat/ico: use avmalloc(z)_array()
[19:52] <cone-972> ffmpeg.git 03Michael Niedermayer 07master:7999b41d4808: avfilter/vf_pixdesctest: use av_malloc_array()
[19:52] <cone-972> ffmpeg.git 03Michael Niedermayer 07master:58a53123bae9: avcodec/vble: use av_malloc_array()
[19:57] <cone-972> ffmpeg.git 03Michael Niedermayer 07release/1.0:ea66bd0d24f1: avformat/matroskaenc: wrap V_QUICKTIME codec private in something that looks like its part of quicktime stsd
[19:57] <cone-972> ffmpeg.git 03Michael Niedermayer 07release/1.0:66c927d2172a: avformat/matroskadec: support SVQ3 as generated by mkvtoolnix-6.6.0
[19:57] <cone-972> ffmpeg.git 03Michael Niedermayer 07release/1.2:9060c10c659c: avformat/matroskaenc: wrap V_QUICKTIME codec private in something that looks like its part of quicktime stsd
[19:57] <cone-972> ffmpeg.git 03Michael Niedermayer 07release/1.2:8ee704c39297: avformat/matroskadec: support SVQ3 as generated by mkvtoolnix-6.6.0
[19:57] <cone-972> ffmpeg.git 03Michael Niedermayer 07release/2.1:92d336670f82: avformat/matroskaenc: wrap V_QUICKTIME codec private in something that looks like its part of quicktime stsd
[19:57] <cone-972> ffmpeg.git 03Michael Niedermayer 07release/2.1:353f79defc1d: avformat/matroskadec: support SVQ3 as generated by mkvtoolnix-6.6.0
[20:04] <cone-972> ffmpeg.git 03Anh 07master:c150e2cf324c: Fix compilation with --disable-everything --enable-muxer=avi --disable-network.
[20:08] <cone-972> ffmpeg.git 03Anh 07release/2.2:32919db4fb8a: Fix compilation with --disable-everything --enable-muxer=avi --disable-network.
[00:00] --- Thu Apr 17 2014