[Ffmpeg-devel-irc] ffmpeg-devel.log.20140415

[burek]

[01:04] <cone-528> ffmpeg.git 03Michael Niedermayer 07master:3751e5a8214d: swscale/swscale: clear to/from convert pointers when they used.
[01:04] <cone-528> ffmpeg.git 03Michael Niedermayer 07master:a2f088c0f93c: swscale: factor ff_sws_init_range_convert() out
[01:04] <cone-528> ffmpeg.git 03Michael Niedermayer 07master:fa98885be4e7: swscale/utils: move handling of unscaled case to the end of init function
[01:04] <cone-528> ffmpeg.git 03Michael Niedermayer 07master:4959a4fcf76e: swscale/utils: fix changing src/dst range after initializing the context
[01:37] <cone-528> ffmpeg.git 03Martin Storsjö 07master:a6ac4fcce441: vp78: Align the intra4x4_pred_mode_top array within VP8Macroblock
[01:37] <cone-528> ffmpeg.git 03Michael Niedermayer 07master:2adcd15f2e02: Merge commit 'a6ac4fcce44108f6f87278b8d12e890576f46a62'
[02:44] <cone-528> ffmpeg.git 03Michael Niedermayer 07master:ef818d8bf07b: avformat: add av_format_inject_global_side_data(), and disable it by default
[04:38] <cone-528> ffmpeg.git 03Michael Niedermayer 07master:0bf25c7b08d4: configure: fix arm osx build
[06:08] <jamrial> morning
[06:11] <jamrial> michaelni, could you take a look at https://github.com/jamrial/FFmpeg/commit/0b2156663dfebc9b3203de17bee6e9d0d0ae3859 and check if the performance hit is smaller compared with the first version for you?
[06:12] <jamrial> it's a wip, don't try linear interpolation as i haven't yet ported it to yasm. it will use the c code instead
[06:47] <cone-528> ffmpeg.git 03Michael Niedermayer 07master:2b58c9c945f7: swresample/resample_template: try to consider src_size more exactly
[06:53] <michaelni> jamrial, before 100169 decicycles in multiple_resample, 16384 runs, 0 skips
[06:53] <michaelni> after 118160 decicycles in multiple_resample, 16384 runs, 0 skips
[06:55] <michaelni> jamrial, are you getting better values ?
[06:56] <michaelni> with cpuflags 0 i get 517579 decicycles in multiple_resample, 16384 runs, 0 skips
[06:56] <michaelni> so its using some asm
[06:57] Action: michaelni falls asleep
[06:57] <jamrial> The performace hit here was a tad lower, but at least now you're getting something much better than in the old version
[06:58] <jamrial> Do you consider that hit acceptable?
[06:58] <jamrial> As in, do you think the performance hit outweighs the benefits of the refactor?
[06:58] <michaelni> 18% speedloss, IMOno
[06:59] <michaelni> IMO no
[06:59] <michaelni> what was the benefit ?
[06:59] Action: michaelni is too sleeply to remember
[06:59] <jamrial> As things are right now, each asm version we add is like +5kb after compiling
[07:00] <jamrial> With this refactor is like 500 bytes or less
[07:00] <jamrial> Also this patch alone reduced the library size by 12kb i think
[07:01] <michaelni> i like 18% speedup for 12kb disk space, id love if we could do that with h264/h265
[07:01] <jamrial> Alright
[12:41] <BBB> michaelni: hevc needs asm :)
[13:30] <thardin> bah, mail server at work isn't workin
[13:34] <michaelni> use gmx, its not at work and probably isnt working either ;)
[13:38] <thardin> changing work mail is a bit of a hassle
[13:38] <thardin> I'll just whip the sysadmin into asking forgiveness from the other mail server admins for letting spam through
[13:57] <cone-155> ffmpeg.git 03Michael Niedermayer 07master:083b6e739bfd: avformat/utils: fix typo
[13:57] <cone-155> ffmpeg.git 03Michael Niedermayer 07master:337f74868c5c: avformat/mpegtsenc: use av_malloc_array()
[13:57] <cone-155> ffmpeg.git 03Michael Niedermayer 07master:eead2cddb773: avformat/xmv: use av_malloc_array()
[13:57] <cone-155> ffmpeg.git 03Michael Niedermayer 07master:4615ff0d61d5: avutil/frame: use av_malloc(z)_array()
[15:18] <cone-155> ffmpeg.git 03Michael Niedermayer 07master:c6902f96d75b: avcodec/mpegaudioenc_template: Fix integer overflow
[17:28] <cone-155> ffmpeg.git 03Michael Niedermayer 07master:37f3f32d51ee: fate: force 128kb/sec for mp2 test
[18:52] <cone-155> ffmpeg.git 03Lars Hammarstrand 07master:add5d7a8ce6f: configure: allow the gas-preprocessor to be called with an absolute path.
[22:05] <ubitux> aarh cpuflag() wasn't behaving as i though
[22:06] <ubitux> cpuflag(ssse3) is actually true with avx2
[22:06] <ubitux> i need to check for cpuflag(avx2) *before* cpuflag(ssse3), meh.
[22:08] <ubitux> so much time wasted, i'm stupid
[22:16] <jamrial> there's notcpuflag() as well
[22:20] <ubitux> there are some really nasty macros&
[22:30] <Compn> ubitux : maybe you could put a comment in the code for the next person who may run into the same problem as you :P
[22:30] <Compn> i always suggest that, leave a comment for the next guy :)
[22:30] <Compn> 'dont fall into the hole!' :)
[22:31] <cone-876> ffmpeg.git 03Peter Ross 07master:5f4f9ee99f4e: Direct Stream Digital (DSD) decoder
[22:54] <Spideru> Hi, were can I find sources for testing with fate ffprobe_compact?
[23:05] <ubitux> sources?
[23:11] <Daemon404> make fate-rsync?
[23:31] <Spideru> thank you Daemon404, make fate-rsync is runnig
[23:31] <Spideru> *running
[00:00] --- Wed Apr 16 2014