[InstCombine] Drop range attributes in foldIsPowerOf2

[dtcxzyw]

Fixes #111934.

[llvmbot]

@llvm/pr-subscribers-llvm-transforms

Author: Yingwei Zheng (dtcxzyw)

Changes

Fixes #111934.

---

Full diff: https://github.com/llvm/llvm-project/pull/111946.diff

2 Files Affected:

• (modified) llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp (+13-5)
• (modified) llvm/test/Transforms/InstCombine/ispow2.ll (+32)

diff --git a/llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp b/llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp
index 688601a8ffa543..964616a4eb35e2 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp
@@ -955,9 +955,11 @@ static Value *foldIsPowerOf2OrZero(ICmpInst *Cmp0, ICmpInst *Cmp1, bool IsAnd,
 }
 
 /// Reduce a pair of compares that check if a value has exactly 1 bit set.
-/// Also used for logical and/or, must be poison safe.
+/// Also used for logical and/or, must be poison safe if range attributes are
+/// dropped.
 static Value *foldIsPowerOf2(ICmpInst *Cmp0, ICmpInst *Cmp1, bool JoinedByAnd,
-                             InstCombiner::BuilderTy &Builder) {
+                             InstCombiner::BuilderTy &Builder,
+                             InstCombinerImpl &IC) {
   // Handle 'and' / 'or' commutation: make the equality check the first operand.
   if (JoinedByAnd && Cmp1->getPredicate() == ICmpInst::ICMP_NE)
     std::swap(Cmp0, Cmp1);
@@ -971,7 +973,10 @@ static Value *foldIsPowerOf2(ICmpInst *Cmp0, ICmpInst *Cmp1, bool JoinedByAnd,
       match(Cmp1, m_SpecificICmp(ICmpInst::ICMP_ULT,
                                  m_Intrinsic<Intrinsic::ctpop>(m_Specific(X)),
                                  m_SpecificInt(2)))) {
-    Value *CtPop = Cmp1->getOperand(0);
+    auto *CtPop = cast<Instruction>(Cmp1->getOperand(0));
+    // Drop range attributes and re-infer them in the next iteration.
+    CtPop->dropPoisonGeneratingAnnotations();
+    IC.addToWorklist(CtPop);
     return Builder.CreateICmpEQ(CtPop, ConstantInt::get(CtPop->getType(), 1));
   }
   // (X == 0) || (ctpop(X) u> 1) --> ctpop(X) != 1
@@ -980,7 +985,10 @@ static Value *foldIsPowerOf2(ICmpInst *Cmp0, ICmpInst *Cmp1, bool JoinedByAnd,
       match(Cmp1, m_SpecificICmp(ICmpInst::ICMP_UGT,
                                  m_Intrinsic<Intrinsic::ctpop>(m_Specific(X)),
                                  m_SpecificInt(1)))) {
-    Value *CtPop = Cmp1->getOperand(0);
+    auto *CtPop = cast<Instruction>(Cmp1->getOperand(0));
+    // Drop range attributes and re-infer them in the next iteration.
+    CtPop->dropPoisonGeneratingAnnotations();
+    IC.addToWorklist(CtPop);
     return Builder.CreateICmpNE(CtPop, ConstantInt::get(CtPop->getType(), 1));
   }
   return nullptr;
@@ -3375,7 +3383,7 @@ Value *InstCombinerImpl::foldAndOrOfICmps(ICmpInst *LHS, ICmpInst *RHS,
     if (Value *V = foldSignedTruncationCheck(LHS, RHS, I, Builder))
       return V;
 
-  if (Value *V = foldIsPowerOf2(LHS, RHS, IsAnd, Builder))
+  if (Value *V = foldIsPowerOf2(LHS, RHS, IsAnd, Builder, *this))
     return V;
 
   if (Value *V = foldPowerOf2AndShiftedMask(LHS, RHS, IsAnd, Builder))
diff --git a/llvm/test/Transforms/InstCombine/ispow2.ll b/llvm/test/Transforms/InstCombine/ispow2.ll
index c21ad95f83a1c4..832c066370b0f8 100644
--- a/llvm/test/Transforms/InstCombine/ispow2.ll
+++ b/llvm/test/Transforms/InstCombine/ispow2.ll
@@ -1522,3 +1522,35 @@ define <2 x i1> @not_pow2_or_z_known_bits_fail_wrong_cmp(<2 x i32> %xin) {
   %r = icmp ugt <2 x i32> %cnt, <i32 2, i32 2>
   ret <2 x i1> %r
 }
+
+; Make sure that range attributes on return values are dropped after merging these two icmps
+
+define i1 @has_single_bit(i32 %x) {
+; CHECK-LABEL: @has_single_bit(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[POPCNT:%.*]] = call range(i32 0, 33) i32 @llvm.ctpop.i32(i32 [[X:%.*]])
+; CHECK-NEXT:    [[SEL:%.*]] = icmp eq i32 [[POPCNT]], 1
+; CHECK-NEXT:    ret i1 [[SEL]]
+;
+entry:
+  %cmp1 = icmp ne i32 %x, 0
+  %popcnt = call range(i32 1, 33) i32 @llvm.ctpop.i32(i32 %x)
+  %cmp2 = icmp ult i32 %popcnt, 2
+  %sel = select i1 %cmp1, i1 %cmp2, i1 false
+  ret i1 %sel
+}
+
+define i1 @has_single_bit_inv(i32 %x) {
+; CHECK-LABEL: @has_single_bit_inv(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[POPCNT:%.*]] = call range(i32 0, 33) i32 @llvm.ctpop.i32(i32 [[X:%.*]])
+; CHECK-NEXT:    [[SEL:%.*]] = icmp ne i32 [[POPCNT]], 1
+; CHECK-NEXT:    ret i1 [[SEL]]
+;
+entry:
+  %cmp1 = icmp eq i32 %x, 0
+  %popcnt = call range(i32 1, 33) i32 @llvm.ctpop.i32(i32 %x)
+  %cmp2 = icmp ugt i32 %popcnt, 1
+  %sel = select i1 %cmp1, i1 true, i1 %cmp2
+  ret i1 %sel
+}

[dtcxzyw]

Add ctpop to worklist to pass the fix-point verification.

[nikic]

LGTM

[dtcxzyw]

/cherry-pick 6a65e98

[llvmbot]

Failed to cherry-pick: 6a65e98

https://github.com/llvm/llvm-project/actions/runs/11290628445

Please manually backport the fix and push it to your github fork. Once this is done, please create a pull request