[PATCH] D54516: [AMDGPU] Do not mark llvm.amdgcn.set.inactive as IntrNoMem

[Tim Renouf via Phabricator via llvm-commits]

tpr created this revision.
Herald added subscribers: llvm-commits, t-tye, dstuttard, yaxunl, nhaehnle, wdng, jvesely, kzhuravl.

The result of this intrinsic depends on what control flow it is inside,
so I am removing the IntrNoMem mark to stop it being CSEd.

Change-Id: Iaf95db68f30c1b07e695d050a03c81f12ca7629d


Repository:
  rL LLVM

https://reviews.llvm.org/D54516

Files:
  include/llvm/IR/IntrinsicsAMDGPU.td


Index: include/llvm/IR/IntrinsicsAMDGPU.td
===================================================================
--- include/llvm/IR/IntrinsicsAMDGPU.td
+++ include/llvm/IR/IntrinsicsAMDGPU.td
@@ -1362,7 +1362,7 @@
   Intrinsic<[llvm_anyint_ty],
             [LLVMMatchType<0>, // value to be copied
              LLVMMatchType<0>], // value for the inactive lanes to take
-            [IntrNoMem, IntrConvergent]>;
+            [IntrConvergent]>;
 
 // Waterfall intrinics used to tag a region as requiring waterfall loops to
 // transform VGPR values into SGPR values in a loop


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D54516.173994.patch
Type: text/x-patch
Size: 578 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20181114/f622b113/attachment.bin>