Jan Vesely
Jan Vesely
Hm, I think there's a better fix. I'll post it either tonight or tomorrow.
> Hey @jvesely and @kmantel, > > Sorry for the lack of input on this. I have been out on vacation and then sick for a bit. I tried the...
yeah, the problem is in converting ndarrays to LLVM IR: ``` elif isinstance(t, np.ndarray): return self.convert_python_struct_to_llvm_ir(t.tolist()) ``` This tries to reuse the code for converting lists, but it doesn't work...
Should be fixed on devel now—this time with a test.
atomic operations are system level (i.e. work between CPU and GPU). I haven't tested dGPUs which require special flag for the atomic instructions. the following tests work as expected on...
Sounds like either l2$ configuration bug, or wrong code generated by the compiler (missing SLC flag for atomic instructions). Either way you'll need someone from AMD to address it.
you can check the discussion here: https://reviews.llvm.org/D39350 looks like system scope atomics are not supported on dGPUs
The main problem with the workflow is that the GA security model doesn't allow jobs to depend on other jobs that include secrets. So the 'test-release' part cannot depend on...
> The logs from the previous run have expired, but I do see "Skip output 'sdist' since it may contain secret.". On my fork, it doesn't seem to have a...
@dependabot rebase