[PowerPC] Handle constant zero bits in BitPermutationSelector

We currently crash when analyzing an AssertZExt node that has some bits that are
constant zeros (i.e. as a result of an and with a constant).

This issue was reported in https://bugs.llvm.org/show_bug.cgi?id=41088 and this
patch fixes that.

Differential revision: https://reviews.llvm.org/D72038

1 files changed, 7 insertions, 4 deletions

diff --git a/llvm/lib/Target/PowerPC/PPCISelDAGToDAG.cpp b/llvm/lib/Target/PowerPC/PPCISelDAGToDAG.cpp
index 41042e26f3f..776ec52e260 100644
--- a/llvm/lib/Target/PowerPC/PPCISelDAGToDAG.cpp
+++ b/llvm/lib/Target/PowerPC/PPCISelDAGToDAG.cpp
@@ -1400,11 +1400,14 @@ class BitPermutationSelector {
       for (unsigned i = 0; i < NumValidBits; ++i)
         Bits[i] = (*LHSBits)[i];
 
-      // These bits are known to be zero.
+      // These bits are known to be zero but the AssertZext may be from a value
+      // that already has some constant zero bits (i.e. from a masking and).
       for (unsigned i = NumValidBits; i < NumBits; ++i)
-        Bits[i] = ValueBit((*LHSBits)[i].getValue(),
-                           (*LHSBits)[i].getValueBitIndex(),
-                           ValueBit::VariableKnownToBeZero);
+        Bits[i] = (*LHSBits)[i].hasValue()
+                      ? ValueBit((*LHSBits)[i].getValue(),
+                                 (*LHSBits)[i].getValueBitIndex(),
+                                 ValueBit::VariableKnownToBeZero)
+                      : ValueBit(ValueBit::ConstZero);
 
       return std::make_pair(Interesting, &Bits);
     }