-
Notifications
You must be signed in to change notification settings - Fork 412
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
CUDA Runtime error #41
Comments
Hm, that's strange... It's not immediately clear to me what the problem might be. Would you mind trying a few things?
Thanks! |
Thank you for your prompt reply. Currently I compiles pbrt-v4 in Debug mode with commit
Thanks for your help |
It looks like either you're not actually running a debug build or (more likely..) the build rules on Windows aren't setting the preprocessor definitions correctly there. (One way or another, in debug builds, NDEBUG should not be #defined, and so some additional logging and synchronization around kernel launches should be happening.) In any case, could you try adding |
I'm trying to figure out where's the actual bug but failed. Really eagering and thanking you for your help |
@neverfelly yes, i'm running on windows with gpu rendering had error same as your . before early commit has running correct. |
Unfortunately I don't have the bandwidth to debug Windows GPU support at the moment and am unlikely to be able to for a few months. I have therefore updated the build so that building with GPU support on Windows is prohibited for now. Needless to say, PRs to fix this would be highly welcome. :-) |
Some additional information from running PBRT through the CUDA debugger on Windows:
|
Interesting... There haven't been many changes to the OptiX code between the last time that it worked on Windows and now, but one of the few relates to shadow rays and updating the SampledSpectrum for pixels when they are unoccluded: 82ace32. However, I can't see anything suspicious in there... |
Thanks for the pointer! I will try reverting that commit, or commenting out the reads&writes to |
Reverting that commit allows me to render killeroo-gold.pbrt without any issues (apart from #108) in Debug mode on Windows. When trying in Release mode instead, I still get a launch error (could be in a different place though, I have not checked yet). |
This fixes the debug build on Windows on the GPU. (Release crashes with OptiX complaining about malformed PTX.) The issue is essentially the same as why *this is copied in GPU lambdas rather than being passed as a pointer; we are accessing the GPUPathIntegrator in read-only fashion from the CPU during rendering and with unified memory on Windows, it isn't allowed to concurrently access it on the GPU. This also fits with the data point that 82ace32 is when things first started crashing. Issues #41, #48, #72, #89 , and #96.
While Windows still has issues, at least this particular one is fixed now! |
This fixes the debug build on Windows on the GPU. (Release crashes with OptiX complaining about malformed PTX.) The issue is essentially the same as why *this is copied in GPU lambdas rather than being passed as a pointer; we are accessing the GPUPathIntegrator in read-only fashion from the CPU during rendering and with unified memory on Windows, it isn't allowed to concurrently access it on the GPU. This also fits with the data point that 82ace32 is when things first started crashing. Issues mmp#41, mmp#48, mmp#72, mmp#89 , and mmp#96.
Environment: Windows10+2xRTX2080TI+CUDA 11.0+Optix 7.1
Testing scene: Each scene of pbrt-v4-scenes
Error Message:
The text was updated successfully, but these errors were encountered: