[InstCombine] Relax cttz/ctlz with select on zero

The cttz/ctlz intrinsics have a parameter specifying whether the
result is undefined for zero. cttz(x, false) can be relaxed to
cttz(x, true) if x is known non-zero, and in fact such an optimization
is already performed. However, this currently doesn't work if x is
non-zero as a result of a select rather than an explicit branch.
This patch adds handling for this case, thus allowing
x != 0 ? cttz(x, false) : y to simplify to x != 0 ? cttz(x, true) : y.

Differential Revision: https://reviews.llvm.org/D55786

llvm-svn: 350463

2 files changed, 19 insertions, 12 deletions

diff --git a/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp b/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp
index b1e0ffa6fa7..ebbe2afe3ec 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp
@@ -709,17 +709,18 @@ static Value *foldSelectCttzCtlz(ICmpInst *ICI, Value *TrueVal, Value *FalseVal,
       match(Count, m_Trunc(m_Value(V))))
     Count = V;
 
+  // Check that 'Count' is a call to intrinsic cttz/ctlz. Also check that the
+  // input to the cttz/ctlz is used as LHS for the compare instruction.
+  if (!match(Count, m_Intrinsic<Intrinsic::cttz>(m_Specific(CmpLHS))) &&
+      !match(Count, m_Intrinsic<Intrinsic::ctlz>(m_Specific(CmpLHS))))
+    return nullptr;
+
+  IntrinsicInst *II = cast<IntrinsicInst>(Count);
+
   // Check if the value propagated on zero is a constant number equal to the
   // sizeof in bits of 'Count'.
   unsigned SizeOfInBits = Count->getType()->getScalarSizeInBits();
-  if (!match(ValueOnZero, m_SpecificInt(SizeOfInBits)))
-    return nullptr;
-
-  // Check that 'Count' is a call to intrinsic cttz/ctlz. Also check that the
-  // input to the cttz/ctlz is used as LHS for the compare instruction.
-  if (match(Count, m_Intrinsic<Intrinsic::cttz>(m_Specific(CmpLHS))) ||
-      match(Count, m_Intrinsic<Intrinsic::ctlz>(m_Specific(CmpLHS)))) {
-    IntrinsicInst *II = cast<IntrinsicInst>(Count);
+  if (match(ValueOnZero, m_SpecificInt(SizeOfInBits))) {
     // Explicitly clear the 'undef_on_zero' flag.
     IntrinsicInst *NewI = cast<IntrinsicInst>(II->clone());
     NewI->setArgOperand(1, ConstantInt::getFalse(NewI->getContext()));
@@ -727,6 +728,12 @@ static Value *foldSelectCttzCtlz(ICmpInst *ICI, Value *TrueVal, Value *FalseVal,
     return Builder.CreateZExtOrTrunc(NewI, ValueOnZero->getType());
   }
 
+  // If the ValueOnZero is not the bitwidth, we can at least make use of the
+  // fact that the cttz/ctlz result will not be used if the input is zero, so
+  // it's okay to relax it to undef for that case.
+  if (II->hasOneUse() && !match(II->getArgOperand(1), m_One()))
+    II->setArgOperand(1, ConstantInt::getTrue(II->getContext()));
+
   return nullptr;
 }
 
diff --git a/llvm/test/Transforms/InstCombine/select-cmp-cttz-ctlz.ll b/llvm/test/Transforms/InstCombine/select-cmp-cttz-ctlz.ll
index f7690079b6f..606cdedca7a 100644
--- a/llvm/test/Transforms/InstCombine/select-cmp-cttz-ctlz.ll
+++ b/llvm/test/Transforms/InstCombine/select-cmp-cttz-ctlz.ll
@@ -345,7 +345,7 @@ define i128 @test8(i128 %x) {
 
 define i32 @test_ctlz_not_bw(i32 %x) {
 ; CHECK-LABEL: @test_ctlz_not_bw(
-; CHECK-NEXT:    [[CT:%.*]] = tail call i32 @llvm.ctlz.i32(i32 [[X:%.*]], i1 false), !range !1
+; CHECK-NEXT:    [[CT:%.*]] = tail call i32 @llvm.ctlz.i32(i32 [[X:%.*]], i1 true), !range !1
 ; CHECK-NEXT:    [[CMP:%.*]] = icmp eq i32 [[X]], 0
 ; CHECK-NEXT:    [[RES:%.*]] = select i1 [[CMP]], i32 123, i32 [[CT]]
 ; CHECK-NEXT:    ret i32 [[RES]]
@@ -373,7 +373,7 @@ define i32 @test_ctlz_not_bw_multiuse(i32 %x) {
 
 define i32 @test_cttz_not_bw(i32 %x) {
 ; CHECK-LABEL: @test_cttz_not_bw(
-; CHECK-NEXT:    [[CT:%.*]] = tail call i32 @llvm.cttz.i32(i32 [[X:%.*]], i1 false), !range !1
+; CHECK-NEXT:    [[CT:%.*]] = tail call i32 @llvm.cttz.i32(i32 [[X:%.*]], i1 true), !range !1
 ; CHECK-NEXT:    [[CMP:%.*]] = icmp eq i32 [[X]], 0
 ; CHECK-NEXT:    [[RES:%.*]] = select i1 [[CMP]], i32 123, i32 [[CT]]
 ; CHECK-NEXT:    ret i32 [[RES]]
@@ -412,7 +412,7 @@ define <2 x i32> @test_ctlz_bw_vec(<2 x i32> %x) {
 
 define <2 x i32> @test_ctlz_not_bw_vec(<2 x i32> %x) {
 ; CHECK-LABEL: @test_ctlz_not_bw_vec(
-; CHECK-NEXT:    [[CT:%.*]] = tail call <2 x i32> @llvm.ctlz.v2i32(<2 x i32> [[X:%.*]], i1 false)
+; CHECK-NEXT:    [[CT:%.*]] = tail call <2 x i32> @llvm.ctlz.v2i32(<2 x i32> [[X:%.*]], i1 true)
 ; CHECK-NEXT:    [[CMP:%.*]] = icmp eq <2 x i32> [[X]], zeroinitializer
 ; CHECK-NEXT:    [[RES:%.*]] = select <2 x i1> [[CMP]], <2 x i32> zeroinitializer, <2 x i32> [[CT]]
 ; CHECK-NEXT:    ret <2 x i32> [[RES]]
@@ -436,7 +436,7 @@ define <2 x i32> @test_cttz_bw_vec(<2 x i32> %x) {
 
 define <2 x i32> @test_cttz_not_bw_vec(<2 x i32> %x) {
 ; CHECK-LABEL: @test_cttz_not_bw_vec(
-; CHECK-NEXT:    [[CT:%.*]] = tail call <2 x i32> @llvm.cttz.v2i32(<2 x i32> [[X:%.*]], i1 false)
+; CHECK-NEXT:    [[CT:%.*]] = tail call <2 x i32> @llvm.cttz.v2i32(<2 x i32> [[X:%.*]], i1 true)
 ; CHECK-NEXT:    [[CMP:%.*]] = icmp eq <2 x i32> [[X]], zeroinitializer
 ; CHECK-NEXT:    [[RES:%.*]] = select <2 x i1> [[CMP]], <2 x i32> zeroinitializer, <2 x i32> [[CT]]
 ; CHECK-NEXT:    ret <2 x i32> [[RES]]