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 quit [Ping timeout: 276 seconds]
Degi_ has joined #scopehal
Degi_ is now known as Degi
<azonenberg> (it's easy to do in ngscopeclient)
<tnt> azonenberg: yeah but you can't load/save in ngscopeclient :)
<d1b2> <tnt> @azonenberg Also glscopehal tends to lockup the intel gpu a whole lot less than ngscopehal does. (And this combined with lack of session save makes it even more annoying since you need to re-setup everything)
bvernoux has joined #scopehal
<azonenberg> glscopeclient hangs the gpu less??
<azonenberg> that's interesting, i've heard the opposite reliability wise
<azonenberg> more interestingly is the fact that both are using the same rendering shader
<azonenberg> the rendering and compute are vulkan in both
<azonenberg> the only difference is the final compositor and the gui chrome
<azonenberg> neither of which should be causing lockups
<tnt> azonenberg: maybe when something isn't included in a "display" type pipeline it is subject to less timeout checks by the driver ?
<tnt> (i.e. allowed to run longer)
<whitequark> the gpu hangs made the experience of using glscopeclient so poor a friend wrote her own software to talk to the scope instead
<whitequark> and protocol analysis too
<whitequark> like, it was faster than getting glscopeclient to work with all the crashes
<monochroma> what did they implement it in?
<d1b2> <david.rysk> gpu hangs -- what hardware? assuming intel integrated GPU?
<monochroma> (afaik intel is the one we have had most, if not all of the issues on?)
<tnt> monochroma: not really. Trying on my RX570 (and someone else tried alson on same gen amd card), it just hangs the whole system requiring a reboot.
<tnt> I mean I only tried twice on that system because I don't enjoy risking corrupting anything and having to restart the whole machine ...
<monochroma> oh fun. *baps the GPU firmware/driver devs*
kbeckmann has joined #scopehal
<azonenberg> whitequark: weird
<azonenberg> on my nvidia setups i have not experienced a single gpu hang i can remember. i've had a few slowdowns due to bugs that quickly got fixed
<azonenberg> but some of my dev machines have like 200+ day uptimes and i use glscopeclient and ngscopeclient on them on a daily basis
<azonenberg> i know we recently added a workaround for a mesa issue but that was a crash on startup not a gpu hang
<azonenberg> i'd be very interested if we can collect some data on which platforms are having these issues
<azonenberg> and see if we can find any commonalities
<azonenberg> also vulkan config / info dumps from them
bvernoux has quit [Read error: Connection reset by peer]
<d1b2> <edgetriggered> Same observation here. Also seeing what: vk::Device::waitForFences: ErrorDeviceLost when trying to zoom/align waveforms.
<d1b2> <edgetriggered> I was replying to tnt, but it looks like the context is lost on IRC.
<syscall> maybe I should chat here instead :)
<d1b2> <tnt> @edgetriggered It's fine, I'm on discord too 😅
<d1b2> <edgetriggered> Also unusual is that that my RTA4004 apparently only works with scopehal when the scope is already in high resolution mode.
<d1b2> <edgetriggered> It somewhat works, but takes a random number of seconds to acquire a new trace.
<whitequark> azonenberg: windows, intel