[PATCH] D32831: [AMDGPU] In the new waitcnt insertion pass, use getHeader instead of getTopBlock to find the loop header.

[Matt Arsenault via Phabricator via llvm-commits]

arsenm added inline comments.


================
Comment at: test/CodeGen/AMDGPU/waitcnt-looptest.ll:18
+
+define amdgpu_kernel void @testKernel(i32 addrspace(1)* nocapture %arg) local_unnamed_addr #0 !kernel_arg_addr_space !0 !kernel_arg_access_qual !1 !kernel_arg_type !2 !kernel_arg_base_type !2 !kernel_arg_type_qual !3 !kernel_arg_name !4 {
+bb:
----------------
Can drop all of this metadata too


================
Comment at: test/CodeGen/AMDGPU/waitcnt-looptest.ll:122
+  %tmp94 = getelementptr inbounds [100 x float], [100 x float] addrspace(1)* @data_generic, i64 0, i64 %tmp93
+  %tmp95 = load float, float addrspace(1)* %tmp94, align 4, !tbaa !17
+  %tmp96 = getelementptr inbounds [100 x float], [100 x float] addrspace(1)* @data_reference, i64 0, i64 %tmp93
----------------
You can still drop the tbaa metadata


https://reviews.llvm.org/D32831