[PATCH] D63269: AMDGPU: Fix input chain when gluing copies to m0

[Matt Arsenault via Phabricator via llvm-commits]

arsenm created this revision.
arsenm added reviewers: rampitec, nhaehnle.
Herald added subscribers: t-tye, tpr, dstuttard, yaxunl, wdng, jvesely, kzhuravl.

I don't think this was causing any observable issues, but was making
reading the DAG dump confusing.


https://reviews.llvm.org/D63269

Files:
  lib/Target/AMDGPU/AMDGPUISelDAGToDAG.cpp


Index: lib/Target/AMDGPU/AMDGPUISelDAGToDAG.cpp
===================================================================
--- lib/Target/AMDGPU/AMDGPUISelDAGToDAG.cpp
+++ lib/Target/AMDGPU/AMDGPUISelDAGToDAG.cpp
@@ -501,13 +501,16 @@
 
   // Write max value to m0 before each load operation
 
-  SDValue M0 = Lowering.copyToM0(*CurDAG, CurDAG->getEntryNode(), SDLoc(N),
+  assert(N->getOperand(0).getValueType() == MVT::Other && "Expected chain");
+
+  SDValue M0 = Lowering.copyToM0(*CurDAG, N->getOperand(0), SDLoc(N),
                                  Val);
 
   SDValue Glue = M0.getValue(1);
 
   SmallVector <SDValue, 8> Ops;
-  for (unsigned i = 0, e = N->getNumOperands(); i != e; ++i)
+  Ops.push_back(M0); // Replace the chain.
+  for (unsigned i = 1, e = N->getNumOperands(); i != e; ++i)
     Ops.push_back(N->getOperand(i));
 
   Ops.push_back(Glue);


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D63269.204536.patch
Type: text/x-patch
Size: 858 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20190613/58db789a/attachment.bin>