Ilia Mirkin

No one is working on Fermi reclocking. If you're interested, join #nouveau on OFTC and get some advice there. Note that there's a big difference between getting reclocking working perfectly...

I think FIX would require you to manually set your session's sequence number to the "right thing". But if you do, and you've lost data, the other side won't know...

We're using this patch for resends specifically, which seems to work OK. Obviously a hack, esp since it will end up ignoring anything that ToApp does to the body, but...

@cbusbey would it be worthwhile to get the above into the quickfixgo codebase, or should we keep it in our vendor tree?

@cbusbey It's now been ~1.5 years since I posted the hack-fix. It's a hack, but an effective one. Nothing better has materialized since then. Making resends work seems like an...

> @imirkin did you create a branch with your change? No ... I thought @ackleymi said he'd do it properly... If someone says they'll apply it, then I can get...

There are two possible ways of looking at this... One is that gl_GlobalInvocationID is a function of gl_WorkGroupSize, and thus should not be defined until gl_WorkGroupSize is. Another is that...

OK, thanks for confirming. It's also the interpretation I went with for fixing mesa. It may be worth clarifying that these "derived" variables are meant to be there no matter...

In case this is of any interest, this was the bug that had originally triggered my question: https://bugs.freedesktop.org/show_bug.cgi?id=103393 It also includes a reference to the commit I pushed to resolve...

Same thing with KHR-GL45.enhanced_layouts.varying_location_limit, which sets an illegal location on an unused output: ``` Unexpected success: Shader source. Stage: vertex #version 430 core #extension GL_ARB_enhanced_layouts : require layout (location =...