freenode :: #aomedia

15 May 2017
15:23wwcvrillian: example? it clears something, i don't think it sets a -1 or -2
15:23wwcvnot sure why
15:50rillianwwcv: yeah. it seems to set Code-Review+1 when the style check passes, unset Code-Review when the ABI check passes, and then set Verified+1 when the rest of the ci tests complete.
15:50rillianjust seemed weird.
15:51rillianwwcv: do you might reviewing the updated patch at https://aomedia-review.googlesource.com/c/11969/4 ?
15:53wwcvrillian: james left a comment on that, i think you can rebase the config removal now
15:54rillianyes. Do I need to rebase the armv6 asm removal patch first?
15:54rillianI can't find any edsp bits in libaom; I think they might have been specific to vp8
15:56* rillian lands the asm removal and rebases the configure changes on master
15:59wwcvthey might have been removed already
15:59wwcvthey refer to armv5te cpu features
16:19rillianhe was refering the accepting the arch extension in configure, which I'd missed
16:20rillianI updated the patches, but the armv8 jenkins config still needs updating
16:21wwcviirc someone from mozilla added edsp/armv5te support
16:21rillianour tent is large
16:22rillianfortunately, I think we can safely say there's no point in optimizing aom for armv5
16:24wwcvhttps://chromium.googlesource.com/webm/libvpx/+/b71962fdc98c3fb9fdf95d74191452cd51cbc2b5
16:26rilliannice. That might have been before derf joined mozilla, actually.
16:26derfNo, it was after.
16:27rillianah
16:29rillianhas it really been that long?
16:31derfEven longer if you count the time I was a contractor.
16:47rillianit's like you're on your fourth codec
18:07derfrillian: There were some audio codecs in there, too.
18:10rillianI think I ran vp8+vp9 together in my head, like celt and opus
18:10rillianwwcv: https://aomedia-review.googlesource.com/c/11971 is ready for review
19:01peloverdeCan anyone explain why the encoder calls av1_set_high_precision_mv in the places it does? It seems to conditionally set a it a bunch of places then clobber it unconditionally at the last minute
19:48rilliandoes the compiler take care of neon register preservation when using intrinsics?
19:53rilliancertainly looks like it does
20:19wwcvyes
20:23rillianthanks
20:26rillianderf: I'm amused that https://people.xiph.org/~tterribe/daala/neon_tutorial.pdf keeps coming up when I search for neon documenation.
22:21utacki assume you did see xda...or even write it? https://www.xda-developers.com/av1-future-video-codecs-google-hevc/
22:22TD-LinuxI didn't see that yet
22:23derfrillian: Well, I wrote the documentation I wished I'd had...
22:25utackthe deasaster generation qualcom made (808,810) does not yet supprot vp9 for some reason....but i guess we are getting there
22:25quikeewhy people think there was a public beta release of AV1
22:28utackprobably because working encoder and decoder for them means beta, not bitstream fixed
22:29utacki mean technically it is called av1 what you encode now?
22:30quikeebut that's not a release
22:40peloverdeperhaps: https://aomedia.googlesource.com/aom/+blame/master/CHANGELOG#6
22:48utackany ideas when you have the bitstream fixed? this fall?
16 May 2017
No messages
   
Last message: 40 days and 19 hours ago