[PATCH] D81093: [CodeGen] Fix ComputeNumSignBits for scalable vectors

[David Sherwood via Phabricator via llvm-commits]

david-arm updated this revision to Diff 269456.
david-arm added reviewers: efriedma, ctetreau.

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D81093/new/

https://reviews.llvm.org/D81093

Files:
  llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp


Index: llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
===================================================================
--- llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
+++ llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
@@ -3530,6 +3530,11 @@
 
 unsigned SelectionDAG::ComputeNumSignBits(SDValue Op, unsigned Depth) const {
   EVT VT = Op.getValueType();
+
+  // TODO: Assume we don't know anything for now.
+  if (VT.isScalableVector())
+    return 1;
+
   APInt DemandedElts = VT.isVector()
                            ? APInt::getAllOnesValue(VT.getVectorNumElements())
                            : APInt(1, 1);
@@ -3553,7 +3558,7 @@
   if (Depth >= MaxRecursionDepth)
     return 1;  // Limit search depth.
 
-  if (!DemandedElts)
+  if (!DemandedElts || VT.isScalableVector())
     return 1;  // No demanded elts, better to assume we don't know anything.
 
   unsigned Opcode = Op.getOpcode();


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D81093.269456.patch
Type: text/x-patch
Size: 904 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20200609/7da89008/attachment-0001.bin>