Nathan V. Morrical
Nathan V. Morrical
I’ve discovered this seems to be some sort of driver bug on NVIDIA. On my Intel ARC card, everything works correctly, with the same SPIR-V. @Erfan-Ahmadi do you have some...
> > I’ve discovered this seems to be some sort of driver bug on NVIDIA. On my Intel ARC card, everything works correctly, with the same SPIR-V. > > @Erfan-Ahmadi...
Haven’t had the chance, but since this seems to be an NVIDIA driver bug, we might want to submit a bug report to their developer forms. NVIDIA devs seem to...
I'll see if i can push NVIDIA to get this fixed in their driver.
Finding a couple more use cases for this feature. With hardware accelerated ray tracing, ray visibility flags are 8 bit integers. The TraceRay call in HLSL has users pass a...
Closing all these issues with the argument they should go to the HLSL spec—without actually bringing them over yourself to the HLSL spec repo—honestly comes off as lazy and dishonest....
Somewhat related, more of a complaint about Microsoft’s DX platform… The number of times DXIL gets in the way of progress in consumer-facing high performance computing these days is insane....
If any future dxc users finding their way here, unlike DXC, Slang has full support for uint8_t types! So if 8 bit types are a feature you value, I highly...
Sure, @s-perron here's a minimal reproducer on godbolt: https://godbolt.org/z/6T4P391fj Notice that on line 6 of the emitted SPIR-V, we get the following: `OpEntryPoint GLCompute %computeShader "computeShader" %2 %g_renderTarget %payload` where...
Similar issues occur in more realistic use cases, where a ray generation program and a closest hit program might exist in the same library and the closest hit uses a...