mozilla :: #gfx

14 Mar 2017
00:12daoshengmujgilbert, ping
00:15jgilbertdaoshengmu, pong
00:20daoshengmujgilbert, I am investigating the depth demo on Mac, https://bugzilla.mozilla.org/show_bug.cgi?id=1281181#c23, and I think it should be the driver issus on RGB5_A1 format framebuffer. I am doing some experiment at WebGLFramebuffer::CheckFramebufferStatus() to make it not return false but still can't see the correct result. Do you have any idea about this?
00:20firebotBug 1281181 NEW, dmu@mozilla.com WebGL depth texture sampling does not work
00:21jgilbertdaoshengmu, RGB5_A1 isn't guaranteed to be renderable. I need to look at the spec for a few minutes.
03:42Jerry_IRCCloudmattwoodrow: ping. what's the actual format for MacIOSurface when we use hardware decoding at mac? I got the YUV422 format from the MacIOSurfacetextureHost, but I don't find the shader in compositorOGL to handle the YUV422 format.
06:03pulsebotCheck-in: https://hg.mozilla.org/projects/graphics/rev/569c7a783e96 - Ryan Hunt - Bug 1346982 - Only build WebRender DisplayItems for BoxShadowOuter when supported r=mchang
06:35pulsebotCheck-in: https://hg.mozilla.org/projects/graphics/rev/eaa70038e3f0 - Ryan Hunt - Bug 1346982 - Fix static bustage and backwards condition r=me
08:01mattwoodrowJerry_IRCCloud: I think its NV12
09:17ethanlingw, ping
11:10katslooks like bugzilla is taking a snow day too
11:10katsethanlin: i see the PR you submitted, it got merged already. i'll do another WR update to pull it in to m-c
11:18katsanybody around to rubberstamp a WR update?
12:43pulsebotCheck-in: https://hg.mozilla.org/projects/graphics/rev/142aa7bf0a84 - Kartikaya Gupta - Bug 1346915 - Update webrender to 7da345684dc29ca60f29cb79acc8d99c474ec477. r=me
15:45lsalzmanaosmond: i can no longer see out my window
15:54gw280there's a cm or maybe 2 here
15:55gw280time to hibernate
15:56lsalzmanyou're a bit late for that
15:57lsalzmaneven the bears are probably confused
15:57lsalzman70F one day, 10F and a blizzard the next :/
16:01aosmondlsalzman: I want the powder that was promised, dammit.
16:01aosmondwe have a little bit more falling now, but it looks like it will miss most of the cross country ski trails
16:02lsalzmanthe snow is BYOS
16:02lsalzmanbring-your-own-shovel
16:02lsalzmanleave with whatever you can :P
18:28pulsebotCheck-in: https://hg.mozilla.org/projects/graphics/pushloghtml?startID=386&endID=387 - 207 changesets - Merge m-c to graphics
18:28pulsebotCheck-in: https://hg.mozilla.org/projects/graphics/rev/563b634aec44 - Martin Robinson - Bug 1347155 - WebRender clips should not share ServoScrollRootIds. r=jrmuizel
19:27dvandermstange: ping
19:27mstangedvander: pong
19:28dvandermstange: hey, as much as i regret starting a conversation about mix-blend modes, i was looking at them today. i noticed when I did the shaders, I added a bunch of code for handling all the different layer types
19:28dvanderbut when i run css-blending tests, the only layer type that ever has a blend mode is a container. which makes since, they need intermediate surfaces
19:29dvanders/since/sense
19:29dvanderis all that code just pointless then?
19:35mstangedvander: hmm
19:37mstangedvander: ContainerLayers are the only type of layer that content sets mix blend modes on, but I thought we had code that propagated them to leaf layers in cases where it's possible, like we do with opacity
19:41mstangedvander: http://searchfox.org/mozilla-central/source/gfx/layers/Layers.cpp#761
19:41mstangedvander: it seems that with this loop, a leaf layer could pick up a mix blend mode that was set on a ContainerLayer ancestor
19:43mstangedvander: I think the parens in this condition are unfortunate: http://searchfox.org/mozilla-central/source/gfx/layers/Layers.cpp#1308-1310
19:43mstangedvander: we should only force an intermediate surface for container layers with mix-blend-mode if it has multiple children
19:44mstangedvander: actually, never mind, the parens are correct, I think
19:44mstangeso yeah
19:45mstangedvander: The way I read the current code means that it should be possible to have leaf layers with a non-"over" GetEffectiveMixBlendMode().
19:47dvandermstange: huh, okay, so maybe we just don't have any tests exercising that
19:47dvanderi'll try and write one to verify... thanks for checking!
19:52gw280roc: I don't know why but I just read that thread on lwn about the 52 release
20:39rocheh
20:42jgilbertsnorp, r?you
20:43jgilbertsnorp, is there any reason we still have GL context sharing support in EGL?
20:43snorpjgilbert: r? for what
20:43snorpjgilbert: we don't need it for egl
20:43snorpwe do need it for iOS :(
20:44jgilbertsnorp, that's surely eagl though?
20:44snorpbut I mean if you killed it who cares
20:44snorpjgilbert: it is, yeah
20:44jgilbertsnorp, https://reviewboard.mozilla.org/r/120288
20:44jgilbertfor EGL context creation fallback, re:webgl on Pixel not working
20:44jgilbertok, I'll remove EGL context sharing then
20:45jgilbertI swear I've removed context sharing like three times already
20:57jgilbertsnorp, does ios not have something like iosurface to supplant context sharing?
20:57snorpjgilbert: it does not
20:57snorpjgilbert: that exists on ios but it's private
20:58jgilbert:rolleyes:
20:58jgilbertof course it is
21:02snorpyeah.
21:02snorpjgilbert: I'm building this patch just to make sure it works
21:02snorpsince your patch is too complex for my brain to run it
21:03snorpI mean lambdas holy hell
21:03jgilberthaha
21:04jgilbertlambdas are the tastiest syntactic sugar
21:04snorpI think this patch is fine though
21:05jgilbertsnorp, can I extract a review from you for axing context sharing support except for eagl? :)
21:05snorpjgilbert: sure?
21:05jgilbertsnorp, https://bugzilla.mozilla.org/show_bug.cgi?id=1347289
21:05firebotBug 1347289 NEW, jgilbert@mozilla.com Remove GLContext sharing support (except for EAGL)
21:35jgilbertsnorp, for the record, that EGL fallback fix works on pixel?
21:35jgilbertI don't have one to test on
21:36snorpjgilbert: it just finished building
21:36snorpworks fine
21:36snorpI'll r+
21:36jgilbertawesome
21:36snorpjgilbert: no idea why context creation failed
21:36snorpunsupported flags?
21:36snorpsays it supports the ext though
21:36jgilbertbefore? yeah
21:37jgilbertit probably supports context robustness (context loss) but not robust_buffer_access_behavior
21:37jgilbertwe need this anyways, since the samsung s6 (maybe 7?) doesn't even support context loss, iirc (via google guys)
21:39jgilbertI'm adding the same fallback to other backends, but figured we can get the EGL one in first :)
21:44snorpjgilbert: ffs
21:44snorpyeah
21:57mstangehttps://irccloud.mozilla.com/pastebin/C5TeeGve/jrmuizel_%3A%20
22:11mstangehttps://irccloud.mozilla.com/pastebin/KFwuwSkg/my%20imports
23:17ashughesdvander: ping
23:17dvanderashughes: pong
23:19ashughesdvander: hi, I wanted to give you a heads up that there at least a 10x spike in GPU Process crashes in Fx55 -- seems to be bug 1160157 and bug 1306168 which combined account for 96% of the GPU Process crash volume in 55.
23:19firebothttps://bugzil.la/1160157 ASSIGNED, kechen non device reset startup forced crash in mozilla::layers::SyncObjectD3D11::FinalizeFrame()
23:19firebothttps://bugzil.la/1306168 NEW Crash in mozilla::layers::CompositorD3D11::BeginFrame
23:24dvanderashughes: okay, thanks, will keep an on eye them
23:24dvandereye on them*
23:24ashughesthe crashes are in Aurora and Beta but nowhere near the same degree
23:24dvanderit's worth noting that we turned on multiple restarts in 55
23:25dvanderso in theory we could have 4x the crashes or something
15 Mar 2017
No messages
   
Last message: 192 days and 8 hours ago