[llvm] r287700 - Fixed the lost FastMathFlags in GVN(Global Value Numbering).
Vyacheslav Klochkov via llvm-commits
llvm-commits at lists.llvm.org
Tue Nov 22 12:52:54 PST 2016
Author: v_klochkov
Date: Tue Nov 22 14:52:53 2016
New Revision: 287700
URL: http://llvm.org/viewvc/llvm-project?rev=287700&view=rev
Log:
Fixed the lost FastMathFlags in GVN(Global Value Numbering).
Reviewer: Hal Finkel.
Differential Revision: https://reviews.llvm.org/D26952
Added:
llvm/trunk/test/Transforms/GVN/propagate-ir-flags.ll
Modified:
llvm/trunk/lib/Transforms/Scalar/GVN.cpp
Modified: llvm/trunk/lib/Transforms/Scalar/GVN.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Scalar/GVN.cpp?rev=287700&r1=287699&r2=287700&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/Scalar/GVN.cpp (original)
+++ llvm/trunk/lib/Transforms/Scalar/GVN.cpp Tue Nov 22 14:52:53 2016
@@ -1731,7 +1731,12 @@ static void patchReplacementInstruction(
// Patch the replacement so that it is not more restrictive than the value
// being replaced.
- ReplInst->andIRFlags(I);
+ // Note that if 'I' is a load being replaced by some operation,
+ // for example, by an arithmetic operation, then andIRFlags()
+ // would just erase all math flags from the original arithmetic
+ // operation, which is clearly not wanted and not needed.
+ if (!isa<LoadInst>(I))
+ ReplInst->andIRFlags(I);
// FIXME: If both the original and replacement value are part of the
// same control-flow region (meaning that the execution of one
Added: llvm/trunk/test/Transforms/GVN/propagate-ir-flags.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/GVN/propagate-ir-flags.ll?rev=287700&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/GVN/propagate-ir-flags.ll (added)
+++ llvm/trunk/test/Transforms/GVN/propagate-ir-flags.ll Tue Nov 22 14:52:53 2016
@@ -0,0 +1,29 @@
+
+; RUN: opt < %s -gvn -S | FileCheck %s
+
+; CHECK-LABEL: func_fast
+; CHECK: fadd fast double
+; CHECK-NEXT: store
+; CHECK-NEXT: ret
+define double @func_fast(double %a, double %b) {
+entry:
+ %a.addr = alloca double, align 8
+ %add = fadd fast double %b, 3.000000e+00
+ store double %add, double* %a.addr, align 8
+ %load_add = load double, double* %a.addr, align 8
+ ret double %load_add
+}
+
+; CHECK-LABEL: func_no_fast
+; CHECK: fadd double
+; CHECK-NEXT: store
+; CHECK-NEXT: ret
+define double @func_no_fast(double %a, double %b) {
+entry:
+ %a.addr = alloca double, align 8
+ %add = fadd fast double %b, 3.000000e+00
+ store double %add, double* %a.addr, align 8
+ %duplicated_add = fadd double %b, 3.000000e+00
+ ret double %duplicated_add
+}
+
More information about the llvm-commits
mailing list