Nagy-Egri Máté Ferenc

This issue was reported in 2022 September. I ran into it on 2023 August. I'd have thought that installing a defunct toolchain was something that would be prioritized higher. I...

Dear @ROCmSupport, > We can not continue supporting cards for more number of years as per business standards. Nvidia recently dropped Kepler support in CUDA in June 2021, when Kepler...

I know I'm late to the party, but, eh... @himanshugoel2797 The hurdles of GPGPU development on Windows using AMD is that aside from an extremely outdated OpenCL runtime, there's nothing...

@JonathanDotCel Nobody's expecting you or anyone to build from source. It doesn't come as a surprise that there are working compilers for Windows in the wild, otherwise Blender couldn't ship....

The 404 is fairly recent and is quite a large facepalm indeed.

I believe nohow. WSL2 will use virtuization under the hood, so ROCm will only work as much as it does today in virtual machines: nohow. Because you cannot have two...

[An Early Benchmark Of The NVIDIA CUDA GPU Performance On WSL2](https://www.phoronix.com/scan.php?page=news_item&px=WSL2-CUDA-Perf-Early-Look) from Phoronix shows that initial CUDA performance under WSL2 isn't stellar, but it will likely improve, plus it's still...

> Is it possible to run CUDA code with DirectML or ROCM on AMD GPU ? With DirectML:no. With ROCm: yes, but this issue is about ROCm working under WSL2....

FWIW AMD seems to have most things in place to get ROCm acceleration in WSL. Microsoft's Antares already uses the undocumented Windows HIP runtime shipping with drivers (amdhip64.dll or whatever),...

I just want to add that such project introspection would be most welcome by C++ developers using CMake and VSCode. Unfortunately, big brother VS has this feature and makes like...