fix redundant umax

Author: el-ev

llvm/lib/Analysis/InstructionSimplify.cpp

@@ -2981,7 +2981,7 @@ static Value *simplifyICmpWithZero(CmpPredicate Pred, Value *LHS, Value *RHS,
 }
 
 static Value *simplifyICmpWithConstant(CmpPredicate Pred, Value *LHS,
-                                       Value *RHS, const InstrInfoQuery &IIQ) {
+                                       Value *RHS, const SimplifyQuery &Q) {
   Type *ITy = getCompareTy(RHS); // The return type.
 
   Value *X;
@@ -3007,7 +3007,7 @@ static Value *simplifyICmpWithConstant(CmpPredicate Pred, Value *LHS,
     return ConstantInt::getTrue(ITy);
 
   ConstantRange LHS_CR =
-      computeConstantRange(LHS, CmpInst::isSigned(Pred), IIQ.UseInstrInfo);
+      computeConstantRange(LHS, CmpInst::isSigned(Pred), Q.IIQ.UseInstrInfo);
   if (!LHS_CR.isFullSet()) {
     if (RHS_CR.contains(LHS_CR))
       return ConstantInt::getTrue(ITy);
@@ -3018,13 +3018,16 @@ static Value *simplifyICmpWithConstant(CmpPredicate Pred, Value *LHS,
   // (mul nuw/nsw X, MulC) != C --> true  (if C is not a multiple of MulC)
   // (mul nuw/nsw X, MulC) == C --> false (if C is not a multiple of MulC)
   const APInt *MulC;
-  if (IIQ.UseInstrInfo && ICmpInst::isEquality(Pred) &&
+  if (Q.IIQ.UseInstrInfo && ICmpInst::isEquality(Pred) &&
       ((match(LHS, m_NUWMul(m_Value(), m_APIntAllowPoison(MulC))) &&
         *MulC != 0 && C->urem(*MulC) != 0) ||
        (match(LHS, m_NSWMul(m_Value(), m_APIntAllowPoison(MulC))) &&
         *MulC != 0 && C->srem(*MulC) != 0)))
     return ConstantInt::get(ITy, Pred == ICmpInst::ICMP_NE);
 
+  if (ICmpInst::isGE(Pred) && C->isOne() && isKnownNonZero(LHS, Q))
+    return ConstantInt::getTrue(ITy);
+
   return nullptr;
 }
 
@@ -3776,7 +3779,7 @@ static Value *simplifyICmpInst(CmpPredicate Pred, Value *LHS, Value *RHS,
   if (Value *V = simplifyICmpWithZero(Pred, LHS, RHS, Q))
     return V;
 
-  if (Value *V = simplifyICmpWithConstant(Pred, LHS, RHS, Q.IIQ))
+  if (Value *V = simplifyICmpWithConstant(Pred, LHS, RHS, Q))
     return V;
 
   // If both operands have range metadata, use the metadata

llvm/test/Transforms/InstCombine/ceil-shift.ll

@@ -275,3 +275,34 @@ define i1 @ceil_shift_not_add_or(i32 %arg0) {
   %res = icmp eq i32 %quot_and_rem, 0
   ret i1 %res
 }
+
+define i32 @ceil_shift_should_infer_ge_zero(i32 %x) {
+; CHECK-LABEL: define i32 @ceil_shift_should_infer_ge_zero(
+; CHECK-SAME: i32 [[X:%.*]]) {
+; CHECK-NEXT:    [[COND_NOT:%.*]] = icmp eq i32 [[X]], 0
+; CHECK-NEXT:    br i1 [[COND_NOT]], label %[[IF_ELSE:.*]], label %[[IF_THEN:.*]]
+; CHECK:       [[IF_THEN]]:
+; CHECK-NEXT:    [[TMP1:%.*]] = lshr i32 [[X]], 20
+; CHECK-NEXT:    [[TMP2:%.*]] = and i32 [[X]], 1048575
+; CHECK-NEXT:    [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
+; CHECK-NEXT:    [[TMP4:%.*]] = zext i1 [[TMP3]] to i32
+; CHECK-NEXT:    [[TMP5:%.*]] = add nuw nsw i32 [[TMP1]], [[TMP4]]
+; CHECK-NEXT:    ret i32 [[TMP5]]
+; CHECK:       [[IF_ELSE]]:
+; CHECK-NEXT:    ret i32 0
+;
+  %cond = icmp ne i32 %x, 0
+  br i1 %cond, label %if.then, label %if.else
+
+if.then:
+  %12 = lshr i32 %x, 20
+  %13 = and i32 %x, 1048575
+  %14 = icmp ne i32 %13, 0
+  %15 = zext i1 %14 to i32
+  %16 = add nuw nsw i32 %12, %15
+  %max = call i32 @llvm.umax.i32(i32 %16, i32 1)
+  ret i32 %max
+
+if.else:
+  ret i32 0
+}