[llvm] d5db2cd - [LVI] Don't push both binop operands at once

Nikita Popov via llvm-commits llvm-commits at lists.llvm.org
Tue Jan 2 02:24:57 PST 2024


Author: Nikita Popov
Date: 2024-01-02T11:24:47+01:00
New Revision: d5db2cdb22ab302acbb6e1a066e791f25dc612de

URL: https://github.com/llvm/llvm-project/commit/d5db2cdb22ab302acbb6e1a066e791f25dc612de
DIFF: https://github.com/llvm/llvm-project/commit/d5db2cdb22ab302acbb6e1a066e791f25dc612de.diff

LOG: [LVI] Don't push both binop operands at once

If one of the binop operands depends on the other, this may end
up evaluating them in the wrong order, producing sub-optimal
results.

Make sure that only one unevaluated operand gets pushed per
iteration.

Fixes https://github.com/llvm/llvm-project/issues/76705.

Added: 
    

Modified: 
    llvm/lib/Analysis/LazyValueInfo.cpp
    llvm/test/Transforms/CorrelatedValuePropagation/basic.ll

Removed: 
    


################################################################################
diff  --git a/llvm/lib/Analysis/LazyValueInfo.cpp b/llvm/lib/Analysis/LazyValueInfo.cpp
index f7d87716482278..c94e29cabc3fe0 100644
--- a/llvm/lib/Analysis/LazyValueInfo.cpp
+++ b/llvm/lib/Analysis/LazyValueInfo.cpp
@@ -976,9 +976,11 @@ LazyValueInfoImpl::solveBlockValueBinaryOpImpl(
   // lets us pick up facts from expressions like "and i32 (call i32
   // @foo()), 32"
   std::optional<ConstantRange> LHSRes = getRangeFor(I->getOperand(0), I, BB);
+  if (!LHSRes)
+    return std::nullopt;
+
   std::optional<ConstantRange> RHSRes = getRangeFor(I->getOperand(1), I, BB);
-  if (!LHSRes || !RHSRes)
-    // More work to do before applying this transfer rule.
+  if (!RHSRes)
     return std::nullopt;
 
   const ConstantRange &LHSRange = *LHSRes;

diff  --git a/llvm/test/Transforms/CorrelatedValuePropagation/basic.ll b/llvm/test/Transforms/CorrelatedValuePropagation/basic.ll
index 7b9375784dae82..6227a5c822b107 100644
--- a/llvm/test/Transforms/CorrelatedValuePropagation/basic.ll
+++ b/llvm/test/Transforms/CorrelatedValuePropagation/basic.ll
@@ -1915,8 +1915,7 @@ define i1 @binop_eval_order(i32 %x) {
 ; CHECK-NEXT:    [[A:%.*]] = add nuw nsw i32 [[X:%.*]], 1
 ; CHECK-NEXT:    [[B:%.*]] = add nuw nsw i32 [[A]], 1
 ; CHECK-NEXT:    [[C:%.*]] = add nuw nsw i32 [[A]], [[B]]
-; CHECK-NEXT:    [[D:%.*]] = icmp ugt i32 [[C]], 2
-; CHECK-NEXT:    ret i1 [[D]]
+; CHECK-NEXT:    ret i1 true
 ;
   %a = add nuw nsw i32 %x, 1
   %b = add nuw nsw i32 %a, 1


        


More information about the llvm-commits mailing list