[llvm] 2508ef0 - [SelectionDAG] Remove unused FP constant in getNegatedExpression
Qiu Chaofan via llvm-commits
llvm-commits at lists.llvm.org
Tue Sep 15 03:04:18 PDT 2020
Author: Qiu Chaofan
Date: 2020-09-15T17:59:10+08:00
New Revision: 2508ef014e8b01006de4e5ee6fd451d1f68d550f
URL: https://github.com/llvm/llvm-project/commit/2508ef014e8b01006de4e5ee6fd451d1f68d550f
DIFF: https://github.com/llvm/llvm-project/commit/2508ef014e8b01006de4e5ee6fd451d1f68d550f.diff
LOG: [SelectionDAG] Remove unused FP constant in getNegatedExpression
960cbc53 immediately removes nodes that won't be used to avoid
compilation time explosion. This patch adds the removal to constants to
fix PR47517.
Reviewed By: RKSimon, steven.zhang
Differential Revision: https://reviews.llvm.org/D87614
Added:
llvm/test/CodeGen/X86/pr47517.ll
Modified:
llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp
Removed:
################################################################################
diff --git a/llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp b/llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp
index 3446ee0efc45..749a5e83058e 100644
--- a/llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp
@@ -5773,8 +5773,10 @@ SDValue TargetLowering::getNegatedExpression(SDValue Op, SelectionDAG &DAG,
// If we already have the use of the negated floating constant, it is free
// to negate it even it has multiple uses.
- if (!Op.hasOneUse() && CFP.use_empty())
+ if (!Op.hasOneUse() && CFP.use_empty()) {
+ RemoveDeadNode(CFP);
break;
+ }
Cost = NegatibleCost::Neutral;
return CFP;
}
diff --git a/llvm/test/CodeGen/X86/pr47517.ll b/llvm/test/CodeGen/X86/pr47517.ll
new file mode 100644
index 000000000000..6b508acf15dd
--- /dev/null
+++ b/llvm/test/CodeGen/X86/pr47517.ll
@@ -0,0 +1,28 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple x86_64 < %s | FileCheck %s
+
+; To ensure unused floating point constant is removed in negation
+define float @test(float %src, float* %p) {
+; CHECK-LABEL: test:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: movq $0, (%rdi)
+; CHECK-NEXT: xorps %xmm0, %xmm0
+; CHECK-NEXT: retq
+entry:
+ %a0 = getelementptr inbounds float, float* %p, i32 0
+ %a1 = getelementptr inbounds float, float* %p, i32 1
+ store float 0.000000e+00, float* %a0
+ store float 0.000000e+00, float* %a1
+ %zero = load float, float* %a0
+ %fmul1 = fmul fast float %zero, %src
+ %fadd1 = fadd fast float %fmul1, %zero
+ %fmul2 = fmul fast float %fadd1, 2.000000e+00
+ %fmul3 = fmul fast float %fmul2, %fmul2
+ %fmul4 = fmul fast float %fmul2, 2.000000e+00
+ %fadd2 = fadd fast float %fmul4, -3.000000e+00
+ %fmul5 = fmul fast float %fadd2, %fmul2
+ %fadd3 = fadd fast float %fmul2, %src
+ %fadd4 = fadd fast float %fadd3, %fmul5
+ %fmul6 = fmul fast float %fmul3, %fadd4
+ ret float %fmul6
+}
More information about the llvm-commits
mailing list