LLVM ISel failure on CDNA 3 (gfx942): Cannot select: i64 = FrameIndex<4> in complex collision kernel

[ZJLi2013]

Root cause hypothesis: The AMDGPU LLVM backend's scratch memory lowering for CDNA targets uses flat addressing (i64), and the instruction selection for i64 = FrameIndex fails when the kernel has a large stack frame. This code path is likely not exercised by typical HPC/AI workloads. The same kernel may compile successfully on RDNA targets where scratch is accessed via buffer instructions with i32 offsets.

releated to genesis issue Genesis-Embodied-AI/genesis-world#2570

Thanks for checking
David

[hughperkins]

Thanks! Don't suppose... do you have a simple-ish Quadrants-only script (no Genesis) to reproduce the issue? As you say, you have already logged the issue as a Genesis issue, and I'm not sure it makes sense to log it also in Quadrants, if the reproduction script is the same one?

[v01dXYZ]

@ZJLi2013 The best thing actually would have been to have a reproducing script. The second best thing would be to have a dump of the LLVM module that is compiled with the full target and the compiling options.

[hughperkins]

In case it's useful to either of you, you can dump partially-compiled kernels using QD_DUMP_IR=1. They'll appear in your /tmp/ir folder, (although there is another env var to redirect to a different path).

[hughperkins]

@ZJLi2013 still waiting on your providing a simple reproduction script please.