[PATCH] D35141: AMDGPU: Remove unnecessary check for constant operands

[Matt Arsenault via Phabricator via llvm-commits]

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

An instruction that has an immediate operand can't reach
this point. This is only called for a freshly shrunk instruction,
which prevously couldn't have had a literal constant operand.
This was also not conservative enough since it woudl also have
had to filter other constant-like inputs like frame indexes.


https://reviews.llvm.org/D35141

Files:
  lib/Target/AMDGPU/SIShrinkInstructions.cpp


Index: lib/Target/AMDGPU/SIShrinkInstructions.cpp
===================================================================
--- lib/Target/AMDGPU/SIShrinkInstructions.cpp
+++ lib/Target/AMDGPU/SIShrinkInstructions.cpp
@@ -138,11 +138,6 @@
 
   int Src0Idx = AMDGPU::getNamedOperandIdx(MI.getOpcode(), AMDGPU::OpName::src0);
 
-  // Only one literal constant is allowed per instruction, so if src0 is a
-  // literal constant then we can't do any folding.
-  if (TII->isLiteralConstant(MI, Src0Idx))
-    return;
-
   // Try to fold Src0
   MachineOperand &Src0 = MI.getOperand(Src0Idx);
   if (Src0.isReg() && MRI.hasOneUse(Src0.getReg())) {


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D35141.105686.patch
Type: text/x-patch
Size: 636 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20170709/bc2d995e/attachment.bin>