ComputeMaskedBits: Make knownzero computation more aggressive for ctlz with undef zero.

unsigned foo(unsigned x) { return 31 - __builtin_clz(x); }
now compiles into a single "bsrl" instruction on x86.

llvm-svn: 147255

1 files changed, 8 insertions, 1 deletions

diff --git a/llvm/lib/Analysis/ValueTracking.cpp b/llvm/lib/Analysis/ValueTracking.cpp
index 4739d1b881a..b9a48eaf571 100644
--- a/llvm/lib/Analysis/ValueTracking.cpp
+++ b/llvm/lib/Analysis/ValueTracking.cpp
@@ -714,10 +714,17 @@ void llvm::ComputeMaskedBits(Value *V, const APInt &Mask,
     if (IntrinsicInst *II = dyn_cast<IntrinsicInst>(I)) {
       switch (II->getIntrinsicID()) {
       default: break;
-      case Intrinsic::ctpop:
       case Intrinsic::ctlz:
       case Intrinsic::cttz: {
         unsigned LowBits = Log2_32(BitWidth)+1;
+        // If this call is undefined for 0, the result will be less than 2^n.
+        if (II->getArgOperand(1) == ConstantInt::getTrue(II->getContext()))
+          LowBits -= 1;
+        KnownZero = APInt::getHighBitsSet(BitWidth, BitWidth - LowBits);
+        break;
+      }
+      case Intrinsic::ctpop: {
+        unsigned LowBits = Log2_32(BitWidth)+1;
         KnownZero = APInt::getHighBitsSet(BitWidth, BitWidth - LowBits);
         break;
       }