[llvm] [AMDGPU] Compiler should synthesize private buffer resource descriptor from flat_scratch_init (PR #79586)

via llvm-commits llvm-commits at lists.llvm.org
Sat Feb 10 03:20:37 PST 2024


alex-t wrote:

> Hi Alex, do I understand correctly that the preferred reproducer is intermediate IR + final binary?

As far as I understand, the error happens in runtime while attempting to read memory beyond the aperture bounds.
So, what I would like ideally is guidance on how to get/configure/build the code that causes an error, or just the assembly itself to inspect the exact value of the V# (SRD) being used. Also, I would need an IR and compiler invocation command line to debug the reasons for setting SRD incorrectly.
Literally, I need to track down the compilation of the code that caused incorrect memory access.
Although, I know almost nothing about the openmp offload project and I would appreciate a reference on the docs that makes me understand better what went on.

https://github.com/llvm/llvm-project/pull/79586


More information about the llvm-commits mailing list