[llvm] [X86][SelectionDAG] - Add support for llvm.canonicalize intrinsic (PR #106370)

Matt Arsenault via llvm-commits llvm-commits at lists.llvm.org
Fri Sep 13 11:15:17 PDT 2024


================
@@ -58159,6 +58160,25 @@ static SDValue combineINTRINSIC_VOID(SDNode *N, SelectionDAG &DAG,
   return SDValue();
 }
 
+static SDValue combineCanonicalize(SDNode *N, SelectionDAG &DAG) {
+  SDValue Operand = N->getOperand(0);
+  EVT VT = Operand.getValueType();
+  SDLoc dl(N);
+
+  // Canonicalize scalar variable FP Nodes.
+  SDValue One =
+      DAG.getNode(ISD::SINT_TO_FP, dl, VT, DAG.getConstant(1, dl, MVT::i32));
+  // TODO: Fix Crash for bf16 when generating strict_fmul as it
+  // leads to a error : SoftPromoteHalfResult #0: t11: bf16,ch = strict_fmul t0,
+  // ConstantFP:bf16<APFloat(16256)>, t5 LLVM ERROR: Do not know how to soft
+  // promote this operator's result!
+  SDValue Chain = DAG.getEntryNode();
+  SDValue StrictFmul = DAG.getNode(ISD::STRICT_FMUL, dl, {VT, MVT::Other},
+                                   {Chain, One, Operand});
----------------
arsenm wrote:

Constant operands canonically should be the RHS 

https://github.com/llvm/llvm-project/pull/106370


More information about the llvm-commits mailing list