[PATCH] D18197: AMDGPU/SI: Do not generate s_waitcnt for ds_permute/ds_bpermute Instructions.

[Tom Stellard via llvm-commits]

tstellarAMD added a comment.

LGTM. Just a few suggestions to improve the tests before you commit it.


================
Comment at: test/CodeGen/AMDGPU/llvm.amdgcn.ds.bpermute.ll:15
@@ +14,3 @@
+; CHECK: ds_bpermute_b32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
+; CHECK-NOT: s_waitcnt
+define void @bpermute_no_waitcnt_test(i32 addrspace(1)* %out, i32 %cond) {
----------------
You might want to add a CHECK here for the branch instruction.  This will make it more robust.

================
Comment at: test/CodeGen/AMDGPU/llvm.amdgcn.ds.permute.ll:15
@@ +14,3 @@
+; CHECK: ds_permute_b32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
+; CHECK-NOT: s_waitcnt
+define void @permute_no_waitcnt_test(i32 addrspace(1)* %out, i32 %cond) {
----------------
Same here.


http://reviews.llvm.org/D18197