jurahul wrote: This PR address a stack size regression that we see in some graphics workloads internally at NVIDIA which seems to be particularly sensitive to stack size. Thanks @Nadharm for triaging this. https://github.com/llvm/llvm-project/pull/152219