[InstCombine] Fold x & (-1 >> y) s>= x to x s<= (-1 >> y)

https://bugs.llvm.org/show_bug.cgi?id=38123
https://rise4fun.com/Alive/I3O

This pattern is not commutative!
We must make sure not to fold the commuted version!

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@337109 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Transforms/InstCombine/InstCombineCompares.cpp b/lib/Transforms/InstCombine/InstCombineCompares.cpp
index 9bd5d53382e528a3569ed37f44650a965f95b0fb..b43520d9e9223ec47202c9a5fd51ffceb8b97556 100644 (file)
--- a/lib/Transforms/InstCombine/InstCombineCompares.cpp
+++ b/lib/Transforms/InstCombine/InstCombineCompares.cpp
@@ -2918,6 +2918,12 @@ static Value *foldICmpWithLowBitMaskedVal(ICmpInst &I,
       return nullptr;         // Ignore the other case.
     DstPred = ICmpInst::Predicate::ICMP_SGT;
     break;
+  case ICmpInst::Predicate::ICMP_SGE:
+    //  x & (-1 >> y) s>= x    ->    x s<= (-1 >> y)
+    if (X != I.getOperand(1)) // X must be on RHS of comparison!
+      return nullptr;         // Ignore the other case.
+    DstPred = ICmpInst::Predicate::ICMP_SLE;
+    break;
   case ICmpInst::Predicate::ICMP_SLE:
     //  x s<= x & (-1 >> y)    ->    x s<= (-1 >> y)
     if (X != I.getOperand(0)) // X must be on LHS of comparison!

diff --git a/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-sge-to-icmp-sle.ll b/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-sge-to-icmp-sle.ll
index 802488d16f29ca9ad9e772550ae1a6383edc24a6..7be784a452fd1f2454d85e3ffee355f51cd7920a 100644 (file)
--- a/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-sge-to-icmp-sle.ll
+++ b/test/Transforms/InstCombine/canonicalize-constant-low-bit-mask-and-icmp-sge-to-icmp-sle.ll
@@ -15,9 +15,8 @@
 
 define i1 @p0(i8 %x) {
 ; CHECK-LABEL: @p0(
-; CHECK-NEXT:    [[TMP0:%.*]] = and i8 [[X:%.*]], 3
-; CHECK-NEXT:    [[RET:%.*]] = icmp sge i8 [[TMP0]], [[X]]
-; CHECK-NEXT:    ret i1 [[RET]]
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp slt i8 [[X:%.*]], 4
+; CHECK-NEXT:    ret i1 [[TMP1]]
 ;
   %tmp0 = and i8 %x, 3
   %ret = icmp sge i8 %tmp0, %x
@@ -27,9 +26,8 @@ define i1 @p0(i8 %x) {
 define i1 @pv(i8 %x, i8 %y) {
 ; CHECK-LABEL: @pv(
 ; CHECK-NEXT:    [[TMP0:%.*]] = lshr i8 -1, [[Y:%.*]]
-; CHECK-NEXT:    [[TMP1:%.*]] = and i8 [[TMP0]], [[X:%.*]]
-; CHECK-NEXT:    [[RET:%.*]] = icmp sge i8 [[TMP1]], [[X]]
-; CHECK-NEXT:    ret i1 [[RET]]
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp sge i8 [[TMP0]], [[X:%.*]]
+; CHECK-NEXT:    ret i1 [[TMP1]]
 ;
   %tmp0 = lshr i8 -1, %y
   %tmp1 = and i8 %tmp0, %x
@@ -43,9 +41,8 @@ define i1 @pv(i8 %x, i8 %y) {
 
 define <2 x i1> @p1_vec_splat(<2 x i8> %x) {
 ; CHECK-LABEL: @p1_vec_splat(
-; CHECK-NEXT:    [[TMP0:%.*]] = and <2 x i8> [[X:%.*]], <i8 3, i8 3>
-; CHECK-NEXT:    [[RET:%.*]] = icmp sge <2 x i8> [[TMP0]], [[X]]
-; CHECK-NEXT:    ret <2 x i1> [[RET]]
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp slt <2 x i8> [[X:%.*]], <i8 4, i8 4>
+; CHECK-NEXT:    ret <2 x i1> [[TMP1]]
 ;
   %tmp0 = and <2 x i8> %x, <i8 3, i8 3>
   %ret = icmp sge <2 x i8> %tmp0, %x
@@ -54,9 +51,8 @@ define <2 x i1> @p1_vec_splat(<2 x i8> %x) {
 
 define <2 x i1> @p2_vec_nonsplat(<2 x i8> %x) {
 ; CHECK-LABEL: @p2_vec_nonsplat(
-; CHECK-NEXT:    [[TMP0:%.*]] = and <2 x i8> [[X:%.*]], <i8 3, i8 15>
-; CHECK-NEXT:    [[RET:%.*]] = icmp sge <2 x i8> [[TMP0]], [[X]]
-; CHECK-NEXT:    ret <2 x i1> [[RET]]
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp slt <2 x i8> [[X:%.*]], <i8 4, i8 16>
+; CHECK-NEXT:    ret <2 x i1> [[TMP1]]
 ;
   %tmp0 = and <2 x i8> %x, <i8 3, i8 15> ; doesn't have to be splat.
   %ret = icmp sge <2 x i8> %tmp0, %x
@@ -65,9 +61,8 @@ define <2 x i1> @p2_vec_nonsplat(<2 x i8> %x) {
 
 define <3 x i1> @p3_vec_splat_undef(<3 x i8> %x) {
 ; CHECK-LABEL: @p3_vec_splat_undef(
-; CHECK-NEXT:    [[TMP0:%.*]] = and <3 x i8> [[X:%.*]], <i8 3, i8 undef, i8 3>
-; CHECK-NEXT:    [[RET:%.*]] = icmp sge <3 x i8> [[TMP0]], [[X]]
-; CHECK-NEXT:    ret <3 x i1> [[RET]]
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp slt <3 x i8> [[X:%.*]], <i8 4, i8 undef, i8 4>
+; CHECK-NEXT:    ret <3 x i1> [[TMP1]]
 ;
   %tmp0 = and <3 x i8> %x, <i8 3, i8 undef, i8 3>
   %ret = icmp sge <3 x i8> %tmp0, %x
@@ -84,8 +79,8 @@ define i1 @oneuse0(i8 %x) {
 ; CHECK-LABEL: @oneuse0(
 ; CHECK-NEXT:    [[TMP0:%.*]] = and i8 [[X:%.*]], 3
 ; CHECK-NEXT:    call void @use8(i8 [[TMP0]])
-; CHECK-NEXT:    [[RET:%.*]] = icmp sge i8 [[TMP0]], [[X]]
-; CHECK-NEXT:    ret i1 [[RET]]
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp slt i8 [[X]], 4
+; CHECK-NEXT:    ret i1 [[TMP1]]
 ;
   %tmp0 = and i8 %x, 3
   call void @use8(i8 %tmp0)
@@ -125,9 +120,8 @@ define i1 @cv0(i8 %y) {
 ; CHECK-LABEL: @cv0(
 ; CHECK-NEXT:    [[X:%.*]] = call i8 @gen8()
 ; CHECK-NEXT:    [[TMP0:%.*]] = lshr i8 -1, [[Y:%.*]]
-; CHECK-NEXT:    [[TMP1:%.*]] = and i8 [[X]], [[TMP0]]
-; CHECK-NEXT:    [[RET:%.*]] = icmp sge i8 [[TMP1]], [[X]]
-; CHECK-NEXT:    ret i1 [[RET]]
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp sle i8 [[X]], [[TMP0]]
+; CHECK-NEXT:    ret i1 [[TMP1]]
 ;
   %x = call i8 @gen8()
   %tmp0 = lshr i8 -1, %y