[PATCH] D57008: [AMDGPU] With XNACK, cannot clause a load with result coalesced with operand

[Tim Renouf via Phabricator via llvm-commits]

tpr marked 4 inline comments as done.
tpr added inline comments.


================
Comment at: lib/Target/AMDGPU/SIFormMemoryClauses.cpp:126
+    unsigned ResReg = ResMO.getReg();
+    for (const MachineOperand &MO : MI.operands()) {
+      if (!MO.isReg() || MO.isDef())
----------------
rampitec wrote:
> MI.uses() maybe? Then you do not need to check for isDef() and have less to scan.
Good idea to use MI.uses(), but it looks like I still need to check isDef() according to the comment in MachineInstr.h:
  /// Returns a range that includes all operands that are register uses.
  /// This may include unrelated operands which are not register uses.
So I guess it might include implicit defs at the end or something.


Repository:
  rL LLVM

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D57008/new/

https://reviews.llvm.org/D57008