https://github.com/petar-avramovic commented:
Think this should use mi_match more. Also I am not really convinced if use of `STI.getGeneration() >= AMDGPUSubtarget::GFX10` is justified in some places. Think we should also include CondDef "cost" and cover cases when it is SGPR-32(SCC) vs G_AMDGPU_COPY_SCC_VCC in this patch. If it is too complicated, can leave it for later. For next step, can you rebase this and precommit patch (the new tests should go in precommit) and try to give some summary of heuristic cost calc. https://github.com/llvm/llvm-project/pull/179352 _______________________________________________ llvm-branch-commits mailing list [email protected] https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
