azonenberg changed the topic of #scopehal to: libscopehal, libscopeprotocols, and glscopeclient development and testing | https://github.com/glscopeclient/scopehal-apps | Logs: https://libera.irclog.whitequark.org/scopehal
Degi_ has joined #scopehal
Degi has quit [Ping timeout: 250 seconds]
Degi_ is now known as Degi
<_whitenotifier> [scopehal] azonenberg pushed 1 commit to master [+0/-0/±3] https://github.com/glscopeclient/scopehal/compare/a7725f27b88b...e37146f8f130
<_whitenotifier> [scopehal] azonenberg e37146f - Added API to query refclk output frequency on a BERT
<_whitenotifier> [scopehal-apps] azonenberg pushed 2 commits to master [+3/-0/±6] https://github.com/glscopeclient/scopehal-apps/compare/453aab19ceb2...b9125fbd359a
<_whitenotifier> [scopehal-apps] azonenberg fb2a097 - Updated threshold icon with latest version from artist, added upsample icon
<_whitenotifier> [scopehal-apps] azonenberg b9125fb - BERTDialog: added "calculated refclk out" box
bvernoux has joined #scopehal
<d1b2> <246tnt> Just tried ngscopehal on my new laptop. Although the "present" bug (causing weird refreshes) isn't present on this as opposed to my old Haswell one, the GPU hang is still there 😢
<d1b2> <246tnt> Also, I have a weird bug with tooltips and my window manager.
<d1b2> <246tnt> When a tooltip would "go out" of the window area, it creates a new X window but that window isn't marked as being a popup style thing and so the tiling manager tries to tile it but then that in turn cause the current scopehal window to move (because of retiling), which causes the tooltip to disapear and then the ngscope window comes back to the original place, causing the tooltip to show ... and it oscillates.
<d1b2> <azonenberg> That's an imgui issue, not a scopehal issue
<d1b2> <azonenberg> In general, windows within the main app window are a single X window
<d1b2> <azonenberg> but as soon as they stick out they become a new top level window
<d1b2> <azonenberg> So if they're not given the right flags, the bug is on their side and should be filed against imgui upstream
<d1b2> <azonenberg> wrt gpu hang, hmmmmm. definitely need to look into things, i thought we had fixed all of them
<d1b2> <azonenberg> is it a true hang or is it just something takes too long and it times out before rendering finishes?
<d1b2> <azonenberg> Adding some early outs for reduced intensity grading on low end GPUs is definitely on the todo list
<d1b2> <246tnt> [851102.375286] i915 0000:00:02.0: [drm] GPU HANG: ecode 12:1:85dcfdfb, in ngscopeclient [9436] [851102.375291] i915 0000:00:02.0: [drm] ngscopeclient[9436] context reset due to GPU hang
<d1b2> <246tnt> But it does "hang" for a couple of seconds.
<d1b2> <246tnt> And I'm not 100% convinced it's "just" something taking too long because the zoom level before the one that crashes is perfectly fluid.
<d1b2> <246tnt> I would have expected some FPS degradation before getting a few sec of complete hang.
<d1b2> <azonenberg> Yes that's what i would expect too
<d1b2> <azonenberg> Very interesting and the more information you can collect the better
<d1b2> <azonenberg> It has not happened on any system i have access to
<d1b2> <azonenberg> and right now i think me and lain are the only ones who have ever worked on the renderer code so we know it best
<d1b2> <david.rysk> @246tnt what specific GPU?
<d1b2> <246tnt> 12th Gen intel iGPU.
<d1b2> <david.rysk> @azonenberg have you tested with your laptop with the nvidia dGPU disabled?
<d1b2> <azonenberg> yes, my recollection was that it was a lot slower but worked just fine
<d1b2> <246tnt> CPU is i7-1260P. GPU identifies as 00:02.0 VGA compatible controller [0300]: Intel Corporation Alder Lake-P Integrated Graphics Controller [8086:46a6] (rev 0c)
<d1b2> <azonenberg> i actually did that a lot when demoing to people since i was usually on power and using less power
<d1b2> <david.rysk> the other bit of information would be what versions of mesa are installed on both systems. I think we should have a template for such reports including: CPU make/model GPU make/model OS type/kernel/architecture/output of uname -a Mesa version GPU driver version
<d1b2> <azonenberg> We should make a "bug report helper" tool accessible via some menu
<d1b2> <azonenberg> that spits out some template info you can paste into a ticket
<d1b2> <246tnt> Mesa is 23.2-rc2 FWIW.
<_whitenotifier> [scopehal] azonenberg opened issue #796: Add mock instrument support for BERT - https://github.com/glscopeclient/scopehal/issues/796
<_whitenotifier> [scopehal] azonenberg labeled issue #796: Add mock instrument support for BERT - https://github.com/glscopeclient/scopehal/issues/796
<_whitenotifier> [scopehal] azonenberg opened issue #797: Add serialization support for BERT - https://github.com/glscopeclient/scopehal/issues/797
<_whitenotifier> [scopehal] azonenberg labeled issue #797: Add serialization support for BERT - https://github.com/glscopeclient/scopehal/issues/797
<_whitenotifier> [scopehal-apps] azonenberg opened issue #605: Add "bug report helper" utility under help menu - https://github.com/glscopeclient/scopehal-apps/issues/605
<_whitenotifier> [scopehal-apps] azonenberg labeled issue #605: Add "bug report helper" utility under help menu - https://github.com/glscopeclient/scopehal-apps/issues/605
<d1b2> <david.rysk> is there a way to turn off previews for the bot?
<d1b2> <azonenberg> I was just about to say they're very annoying on the discord side lol
<d1b2> <azonenberg> I'm also going to change bot settings so "labeled issue" doesn't notify
<d1b2> <azonenberg> because right now creating + labeling an issue makes two messages
<d1b2> <246tnt> Don't think glfw supports setting the right bits on the window a all for that behavior not to occur.
<d1b2> <246tnt> I just disabled multi-viewport 🤷‍♂️
<d1b2> <246tnt> I thought the filter graph bug had been tracked down ?
<d1b2> <azonenberg> Tracked down, yes. Fixed, no
<d1b2> <azonenberg> there has been progress upstream at identifying the issue but no fix has been merged afaik
<d1b2> <azonenberg> there is a WIP fix somebody posted that i may try to merge into our fork of the graph editor library
<d1b2> <azonenberg> i havent tested it yet but other people said it worked
<d1b2> <246tnt> Just tested it ... doesn't work.
<d1b2> <246tnt> Somebody said it worked in ngscopehal ?
<d1b2> <azonenberg> no they said it worked in general
<d1b2> <azonenberg> I have not tested in the graph editor specifically yet
<d1b2> <azonenberg> Playing with it, and perhaps modifying the patch, is on the todo but i've been busy
<d1b2> <azonenberg> Anyway, there has been (slow) upstream progress and it works fine when docked, so i'm not in a huge rush to fix it
<_whitenotifier> [scopehal] azonenberg pushed 1 commit to master [+0/-0/±4] https://github.com/glscopeclient/scopehal/compare/e37146f8f130...ef8a304eed30
<_whitenotifier> [scopehal] azonenberg ef8a304 - MultiLaneBERT: added RX CTLE config
<_whitenotifier> [scopehal-apps] azonenberg pushed 1 commit to master [+0/-0/±3] https://github.com/glscopeclient/scopehal-apps/compare/b9125fbd359a...49b446fccec5
<_whitenotifier> [scopehal-apps] azonenberg 49b446f - Added CTLE gain step configuration to BERT inputs
electronic_eel has quit [Quit: https://quassel-irc.org - Chat comfortably. Anywhere.]
electronic_eel has joined #scopehal
bvernoux has quit [Quit: Leaving]