[InstCombine] Add freeze for lshr (mul (X, 2^N + 1)), N -> add (X, ls…

…hr(X, N))
llvm · Jun 21, 2024 · 04db79a · 04db79a
1 parent 4e6835d
commit 04db79a
Show file tree

Hide file tree

Showing 3 changed files with 25 additions and 14 deletions.
diff --git a/llvm/lib/Transforms/InstCombine/InstCombineShifts.cpp b/llvm/lib/Transforms/InstCombine/InstCombineShifts.cpp
@@ -1500,6 +1500,8 @@ Instruction *InstCombinerImpl::visitLShr(BinaryOperator &I) {
 
         // lshr (mul nuw (X, 2^N + 1)), N -> add nuw (X, lshr(X, N))
         if (Op0->hasOneUse()) {
+          if (!isGuaranteedNotToBeUndef(X, &AC, &I, &DT))
+            X = Builder.CreateFreeze(X, ".fr");
           auto *NewAdd = BinaryOperator::CreateNUWAdd(
               X, Builder.CreateLShr(X, ConstantInt::get(Ty, ShAmtC), "",
                                     I.isExact()));
@@ -1531,6 +1533,8 @@ Instruction *InstCombinerImpl::visitLShr(BinaryOperator &I) {
     if (match(Op0, m_OneUse(m_NSWMul(m_Value(X), m_APInt(MulC))))) {
       if (BitWidth > 2 && (*MulC - 1).isPowerOf2() &&
           MulC->logBase2() == ShAmtC) {
+        if (!isGuaranteedNotToBeUndef(X, &AC, &I, &DT))
+          X = Builder.CreateFreeze(X, ".fr");
         return BinaryOperator::CreateNSWAdd(
             X, Builder.CreateLShr(X, ConstantInt::get(Ty, ShAmtC), "",
                                   I.isExact()));

diff --git a/llvm/test/Transforms/InstCombine/ashr-lshr.ll b/llvm/test/Transforms/InstCombine/ashr-lshr.ll
@@ -607,8 +607,9 @@ define <2 x i8> @ashr_known_pos_exact_vec(<2 x i8> %x, <2 x i8> %y) {
 
 define i32 @lshr_mul_times_3_div_2(i32 %0) {
 ; CHECK-LABEL: @lshr_mul_times_3_div_2(
-; CHECK-NEXT:    [[TMP2:%.*]] = lshr i32 [[TMP0:%.*]], 1
-; CHECK-NEXT:    [[LSHR:%.*]] = add nuw nsw i32 [[TMP2]], [[TMP0]]
+; CHECK-NEXT:    [[DOTFR:%.*]] = freeze i32 [[TMP0:%.*]]
+; CHECK-NEXT:    [[TMP2:%.*]] = lshr i32 [[DOTFR]], 1
+; CHECK-NEXT:    [[LSHR:%.*]] = add nuw nsw i32 [[DOTFR]], [[TMP2]]
 ; CHECK-NEXT:    ret i32 [[LSHR]]
 ;
   %mul = mul nsw nuw i32 %0, 3
@@ -618,8 +619,9 @@ define i32 @lshr_mul_times_3_div_2(i32 %0) {
 
 define i32 @lshr_mul_times_3_div_2_exact(i32 %x) {
 ; CHECK-LABEL: @lshr_mul_times_3_div_2_exact(
-; CHECK-NEXT:    [[TMP1:%.*]] = lshr exact i32 [[X:%.*]], 1
-; CHECK-NEXT:    [[LSHR:%.*]] = add nsw i32 [[TMP1]], [[X]]
+; CHECK-NEXT:    [[DOTFR:%.*]] = freeze i32 [[X:%.*]]
+; CHECK-NEXT:    [[TMP1:%.*]] = lshr exact i32 [[DOTFR]], 1
+; CHECK-NEXT:    [[LSHR:%.*]] = add nsw i32 [[DOTFR]], [[TMP1]]
 ; CHECK-NEXT:    ret i32 [[LSHR]]
 ;
   %mul = mul nsw i32 %x, 3
@@ -657,8 +659,9 @@ define i32 @mul_times_3_div_2_multiuse_lshr(i32 %x) {
 
 define i32 @lshr_mul_times_3_div_2_exact_2(i32 %x) {
 ; CHECK-LABEL: @lshr_mul_times_3_div_2_exact_2(
-; CHECK-NEXT:    [[TMP1:%.*]] = lshr exact i32 [[X:%.*]], 1
-; CHECK-NEXT:    [[LSHR:%.*]] = add nuw i32 [[TMP1]], [[X]]
+; CHECK-NEXT:    [[DOTFR:%.*]] = freeze i32 [[X:%.*]]
+; CHECK-NEXT:    [[TMP1:%.*]] = lshr exact i32 [[DOTFR]], 1
+; CHECK-NEXT:    [[LSHR:%.*]] = add nuw i32 [[DOTFR]], [[TMP1]]
 ; CHECK-NEXT:    ret i32 [[LSHR]]
 ;
   %mul = mul nuw i32 %x, 3
@@ -668,8 +671,9 @@ define i32 @lshr_mul_times_3_div_2_exact_2(i32 %x) {
 
 define i32 @lshr_mul_times_5_div_4(i32 %0) {
 ; CHECK-LABEL: @lshr_mul_times_5_div_4(
-; CHECK-NEXT:    [[TMP2:%.*]] = lshr i32 [[TMP0:%.*]], 2
-; CHECK-NEXT:    [[LSHR:%.*]] = add nuw nsw i32 [[TMP2]], [[TMP0]]
+; CHECK-NEXT:    [[DOTFR:%.*]] = freeze i32 [[TMP0:%.*]]
+; CHECK-NEXT:    [[TMP2:%.*]] = lshr i32 [[DOTFR]], 2
+; CHECK-NEXT:    [[LSHR:%.*]] = add nuw nsw i32 [[DOTFR]], [[TMP2]]
 ; CHECK-NEXT:    ret i32 [[LSHR]]
 ;
   %mul = mul nsw nuw i32 %0, 5
@@ -679,8 +683,9 @@ define i32 @lshr_mul_times_5_div_4(i32 %0) {
 
 define i32 @lshr_mul_times_5_div_4_exact(i32 %x) {
 ; CHECK-LABEL: @lshr_mul_times_5_div_4_exact(
-; CHECK-NEXT:    [[TMP1:%.*]] = lshr exact i32 [[X:%.*]], 2
-; CHECK-NEXT:    [[LSHR:%.*]] = add nsw i32 [[TMP1]], [[X]]
+; CHECK-NEXT:    [[DOTFR:%.*]] = freeze i32 [[X:%.*]]
+; CHECK-NEXT:    [[TMP1:%.*]] = lshr exact i32 [[DOTFR]], 2
+; CHECK-NEXT:    [[LSHR:%.*]] = add nsw i32 [[DOTFR]], [[TMP1]]
 ; CHECK-NEXT:    ret i32 [[LSHR]]
 ;
   %mul = mul nsw i32 %x, 5
@@ -718,8 +723,9 @@ define i32 @mul_times_5_div_4_multiuse_lshr(i32 %x) {
 
 define i32 @lshr_mul_times_5_div_4_exact_2(i32 %x) {
 ; CHECK-LABEL: @lshr_mul_times_5_div_4_exact_2(
-; CHECK-NEXT:    [[TMP1:%.*]] = lshr exact i32 [[X:%.*]], 2
-; CHECK-NEXT:    [[LSHR:%.*]] = add nuw i32 [[TMP1]], [[X]]
+; CHECK-NEXT:    [[DOTFR:%.*]] = freeze i32 [[X:%.*]]
+; CHECK-NEXT:    [[TMP1:%.*]] = lshr exact i32 [[DOTFR]], 2
+; CHECK-NEXT:    [[LSHR:%.*]] = add nuw i32 [[DOTFR]], [[TMP1]]
 ; CHECK-NEXT:    ret i32 [[LSHR]]
 ;
   %mul = mul nuw i32 %x, 5

diff --git a/llvm/test/Transforms/InstCombine/lshr.ll b/llvm/test/Transforms/InstCombine/lshr.ll
@@ -741,8 +741,9 @@ define i32 @mul_splat_fold_wrong_lshr_const(i32 %x) {
 
 define i32 @mul_splat_fold_no_nuw(i32 %x) {
 ; CHECK-LABEL: @mul_splat_fold_no_nuw(
-; CHECK-NEXT:    [[TMP1:%.*]] = lshr i32 [[X:%.*]], 16
-; CHECK-NEXT:    [[T:%.*]] = add nsw i32 [[TMP1]], [[X]]
+; CHECK-NEXT:    [[DOTFR:%.*]] = freeze i32 [[X:%.*]]
+; CHECK-NEXT:    [[TMP1:%.*]] = lshr i32 [[DOTFR]], 16
+; CHECK-NEXT:    [[T:%.*]] = add nsw i32 [[DOTFR]], [[TMP1]]
 ; CHECK-NEXT:    ret i32 [[T]]
 ;
   %m = mul nsw i32 %x, 65537