[PATCH] D141075: [SelectionDAG] Implicitly truncate known bits in SPLAT_VECTOR

[Luke Lau via Phabricator via llvm-commits]

luke updated this revision to Diff 486784.
luke added a comment.

Address feedback


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D141075/new/

https://reviews.llvm.org/D141075

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
@@ -2977,17 +2977,11 @@
                             Depth + 1);
   case ISD::SPLAT_VECTOR: {
     SDValue SrcOp = Op.getOperand(0);
-    if (SrcOp.getValueSizeInBits() != BitWidth) {
-      assert(SrcOp.getValueSizeInBits() > BitWidth &&
-             "Expected SPLAT_VECTOR implicit truncation");
-      // FIXME: We should be able to truncate the known bits here to match
-      // the official semantics of SPLAT_VECTOR, but doing so exposes a
-      // Hexagon target bug which results in an infinite loop during
-      // DAGCombine.  (See D137140 for repo).  Once that's fixed, we can
-      // strengthen this.
-      break;
-    }
-    Known = computeKnownBits(SrcOp, Depth + 1);
+    assert(SrcOp.getValueSizeInBits() >= BitWidth &&
+           "Expected SPLAT_VECTOR implicit truncation");
+    // Implicitly truncate the bits to match the official semantics of
+    // SPLAT_VECTOR.
+    Known = computeKnownBits(SrcOp, Depth + 1).trunc(BitWidth);
     break;
   }
   case ISD::BUILD_VECTOR:


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D141075.486784.patch
Type: text/x-patch
Size: 1228 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20230106/eab0c199/attachment.bin>