[AMDGPU] Optimze old value of v_mov_b32_dpp
We can eliminate old value if bound_ctrl = 1 and row_mask = bank_mask = 0xf.
This is alternative implementation working with the intrinsic in InstCombine.
Original review for past-ISel optimization: D46570.
Differential Revision: https://reviews.llvm.org/D46596
llvm-svn: 332956
diff --git a/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp b/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
index 2516179..8b7c282 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
@@ -3407,6 +3407,23 @@
// amdgcn.kill(i1 1) is a no-op
return eraseInstFromFunction(CI);
}
+ case Intrinsic::amdgcn_update_dpp: {
+ Value *Old = II->getArgOperand(0);
+
+ auto BC = dyn_cast<ConstantInt>(II->getArgOperand(5));
+ auto RM = dyn_cast<ConstantInt>(II->getArgOperand(3));
+ auto BM = dyn_cast<ConstantInt>(II->getArgOperand(4));
+ if (!BC || !RM || !BM ||
+ BC->isZeroValue() ||
+ RM->getZExtValue() != 0xF ||
+ BM->getZExtValue() != 0xF ||
+ isa<UndefValue>(Old))
+ break;
+
+ // If bound_ctrl = 1, row mask = bank mask = 0xf we can omit old value.
+ II->setOperand(0, UndefValue::get(Old->getType()));
+ return II;
+ }
case Intrinsic::stackrestore: {
// If the save is right next to the restore, remove the restore. This can
// happen when variable allocas are DCE'd.