[llvm] [AMDGPU][SIPreEmitPeephole] mustRetainExeczBranch: estimate ThenBlock cost using MachineTraceInfo (PR #111117)

[Juan Manuel Martinez Caamaño via llvm-commits]

jmmartinez wrote:

I'm not 100% confident about this patch at the moment. The transformation gets very aggressive.

I've run a benchmark pipeline (this patch vs amd-staging; I haven't checked the parent change alone) and I've got mixed results. Several big speedups but also several big slowdowns.

I didn't have time to investigate further.

https://github.com/llvm/llvm-project/pull/111117