Cory Bloor

Interesting. Thanks for the insight. I'm curious to know more about the context for that. Which guardonce tool do you use with the sandboxed files and why? I'm not sure...

[It seems that C gives you a surprising amount of flexibility in placing comments](http://stackoverflow.com/a/789236/331041). e.g. ``` C /* comment */ # /* comment */ include /* comment */ // comment...

It can be even worse... ``` C /* comment */ # /* comment */ include /* comment */ /* comment */ ``` In the smart parser, I've been stripping comments...

I should probably mention that the 'smart parser' was dropped in 3af39a7107029265d2ab2887f47823af41f3c15d. Not sure if or when I'll work on it again. It would certainly handle a few more weird...

gfx1010 has been enabled by default in rocBLAS builds since ROCm 4.3.0. https://github.com/ROCmSoftwarePlatform/rocBLAS/blob/rocm-4.3.0/CMakeLists.txt#L169

@benjaminulmer, what is this waiting for?

Taking off my AMD hat for a moment and speaking just for myself, I'm thrilled to see a community contribution like this one. Please do bear with the rocBLAS team...

> I also wonder, if binary distribution can also avoid such problem. The code object files (`TensileLibrary_*.co`, `Kernels.so-000-*.hsaco`) are separated to different GPU archs, so maybe user can choose which...

Thanks for the information! I did a bit more research as well, and it appears that many chips that are ISA-compatible with each other in theory are being treated as...

It would be nice if clang could extract the version information directly from the library binary, but it's not obvious how that should be done. This is simpler and seems...