[ARM] Fix infinite loop when lowering STRICT_FP_EXTEND

If the target has FP64 but not FP16 then we have custom lowering for FP_EXTEND
and STRICT_FP_EXTEND with type f64. However if the extend is from f32 to f64 the
current implementation will cause in infinite loop for STRICT_FP_EXTEND due to
emitting a merge_values of the original node which after replacement becomes a
merge_values of itself.

Fix this by not doing anything for f32 to f64 extend when we have FP64, though
for STRICT_FP_EXTEND we have to do the strict-to-nonstrict mutation as that
doesn't happen automatically for opcodes with custom lowering.

Differential Revision: https://reviews.llvm.org/D74559

(cherry picked from commit 0ec57972967dfb43fc022c2e3788be041d1db730)

1 files changed, 9 insertions, 0 deletions

diff --git a/llvm/lib/Target/ARM/ARMISelLowering.cpp b/llvm/lib/Target/ARM/ARMISelLowering.cpp
index 8e4bb91d53b..559a4e8435c 100644
--- a/llvm/lib/Target/ARM/ARMISelLowering.cpp
+++ b/llvm/lib/Target/ARM/ARMISelLowering.cpp
@@ -16380,6 +16380,15 @@ SDValue ARMTargetLowering::LowerFP_EXTEND(SDValue Op, SelectionDAG &DAG) const {
   assert(!(DstSz == 32 && Subtarget->hasFP16()) &&
          "With FP16, 16 to 32 conversion is legal!");
 
+  // Converting from 32 -> 64 is valid if we have FP64.
+  if (SrcSz == 32 && DstSz == 64 && Subtarget->hasFP64()) {
+    // FIXME: Remove this when we have strict fp instruction selection patterns
+    if (IsStrict) {
+      DAG.mutateStrictFPToFP(Op.getNode());
+    }
+    return Op;
+  }
+
   // Either we are converting from 16 -> 64, without FP16 and/or
   // FP.double-precision or without Armv8-fp. So we must do it in two
   // steps.