[CUDA/NVPTX] Generalize amdgpu.ignore.denormal.mode to NVPTX
Open
backend:NVPTX
cuda
- Dominant language
- LLVM
- Stars
- 40.5k
- Forks
- 18.7k
- PR merge metrics
- PR metrics pending
Description
We need to have per-instruction FTZ control for FP atomics in NVPTX.
Context:
- #200732
- #206154
Contributor guide
Assessment
This issue has not been assessed yet.