[AArch64] Expand vector SDIVREM/UDIVREM

[apazos]

Summary

While experimenting with vectorization I came across a backend failure in lowering vector SDIVREM/UDIVREM. So expanding these vector operations as we already do for the scalar ones.

[t.p.northover]

The change looks fine, but we should add a test too.

[apazos]

Hi Tim, I added a test case, let me know if you want any other change.

[t.p.northover]

Thanks for adding the tests. You can probably drop the CHECK-NOT lines (the test would fail via its exit code and they'd go to stderr anyway). Other than that this looks fine.

Tim.

[mcrosier]

Committed in r254773.

lib/Target/AArch64/AArch64ISelLowering.cpp

AArch64TargetLowering::AArch64TargetLowering(const TargetMachine &TM,
   setOperationAction(ISD::CTTZ_ZERO_UNDEF, MVT::i64, Expand);
   setOperationAction(ISD::CTLZ_ZERO_UNDEF, MVT::i64, Expand);
 
   setOperationAction(ISD::CTPOP, MVT::i32, Custom);
   setOperationAction(ISD::CTPOP, MVT::i64, Custom);
 
   setOperationAction(ISD::SDIVREM, MVT::i32, Expand);
   setOperationAction(ISD::SDIVREM, MVT::i64, Expand);
+  for (MVT VT : MVT::vector_valuetypes()) {
+    setOperationAction(ISD::SDIVREM, VT, Expand);
+    setOperationAction(ISD::UDIVREM, VT, Expand);
+  }
   setOperationAction(ISD::SREM, MVT::i32, Expand);
   setOperationAction(ISD::SREM, MVT::i64, Expand);
   setOperationAction(ISD::UDIVREM, MVT::i32, Expand);
   setOperationAction(ISD::UDIVREM, MVT::i64, Expand);
   setOperationAction(ISD::UREM, MVT::i32, Expand);
   setOperationAction(ISD::UREM, MVT::i64, Expand);
 
   // Custom lower Add/Sub/Mul with overflow.