[llvm] [IR][REVEC] Define llvm.vector.broadcast intrinsic (PR #208212)

[Benjamin Maxwell via llvm-commits]

=?utf-8?q?Gaëtan?= Bossu <gaetan.bossu at arm.com>,
=?utf-8?q?Gaëtan?= Bossu <gaetan.bossu at arm.com>,
=?utf-8?q?Gaëtan?= Bossu <gaetan.bossu at arm.com>,
=?utf-8?q?Gaëtan?= Bossu <gaetan.bossu at arm.com>,
=?utf-8?q?Gaëtan?= Bossu <gaetan.bossu at arm.com>,
=?utf-8?q?Gaëtan?= Bossu <gaetan.bossu at arm.com>,
=?utf-8?q?Gaëtan?= Bossu <gaetan.bossu at arm.com>,
=?utf-8?q?Gaëtan?= Bossu <gaetan.bossu at arm.com>
Message-ID:
In-Reply-To: <llvm.org/llvm/llvm-project/pull/208212 at github.com>


================
@@ -4019,6 +4072,13 @@ bool DAGTypeLegalizer::SplitVectorOperand(SDNode *N, unsigned OpNo) {
   return false;
 }
 
+SDValue DAGTypeLegalizer::SplitVecOp_VECTOR_BROADCAST(SDNode *N) {
+  SDLoc DL(N);
+  SDValue SrcLo, SrcHi;
+  GetSplitVector(N->getOperand(0), SrcLo, SrcHi);
+  return buildSplitVectorBroadcast(DAG, DL, N->getValueType(0), SrcLo, SrcHi);
+}
----------------
MacDue wrote:

Thinking about this, I think the current lowering could be fine, but it's worth a comment explaining why it's okay to ignore the second result of the interleave. 

https://github.com/llvm/llvm-project/pull/208212