Andrzej Janik
Andrzej Janik
Should be fixed with #41
Excellent question. Firstly, if you simply want those frameworks to run on an Intel GPUs, then ZLUDA might not be the best way. Deep Learning is quite high on the...
Wait, oneDNN supports Intel GPUs with full performance. AFAIK the problem is with actually making TF, pyTorch, etc. use oneDNN-GPU efficiently, expose API for custom primitives, etc. If you use...
#178 should fix this issue properly. Unfortunately whitespace in build path will still fail as it is a deeper issue with interaction between LLVM build system and Rust C/C++ compiler...
I don't plan to update HIP-RT. It's a lot of work for no benefit. Newer HIP-RT will not make ZLUDA-OptiX any better. Biggest problems are either with ZLUDA project itself...
No idea, but I still have the package. I've uploaded it here: https://files.catbox.moe/ud40bj.zip
Hmm, this bit of PTX looks extremally normal. The start & end offsets reported in UnrecognizedDirective are in bytes. Yes, feel free to open a separate issue, I'm not usually...
I've looked into ptx, the problem is unimplemented `ldu` instruction. In general, the UnrecognizedDirective error you are seeing is an error that comes from recovering from a parser error. The...
The warning about version information is benign, but there might be a bigger problem. I don't know Davinci Resolve , but is it possible to coerce it to not use...
Should be fixed with the newest commit, please try again