[ARM] Expand vXi1 VSELECT's

[dmgreen]

Summary

We have no lowering for VSELECT vXi1, vXi1, vXi1, so mark them as expanded to turn them into a series of logical operations.

[SjoerdMeijer]

llvm/test/Analysis/CostModel/ARM/arith-overflow.ll

387

More of a curiosity, how was this lowered before? Given the added codegen tests, I guess this is not only a cost-model tweak? And some of these costs are extremely high. Do they make sense?

[dmgreen]

llvm/test/Analysis/CostModel/ARM/arith-overflow.ll

387

Yep. The costs are based on whether the operation is legal and add.with.overflow checks the cost of selects of predicates (v4i1). There is no legal instruction for them for MVE, so a high cost is probably fine.

D94958 updates the similar costs for sadd.sat and friends, which also have costs based in sadd.with.overflow by default.

[SjoerdMeijer]

llvm/test/Analysis/CostModel/ARM/arith-overflow.ll

387

Yep. The costs are based on whether the operation is legal and add.with.overflow checks the cost of selects of predicates (v4i1). There is no legal instruction for them for MVE, so a high cost is probably fine.

D94958 updates the similar costs for sadd.sat and friends, which also have costs based in sadd.with.overflow by default.

Yeah, thanks, now this is more consistent with each other; I was asking because of this difference.

LGTM

llvm/lib/Target/ARM/ARMISelLowering.cpp

for (auto VT : pTypes) {
     setOperationAction(ISD::CONCAT_VECTORS, VT, Custom);
     setOperationAction(ISD::INSERT_VECTOR_ELT, VT, Custom);
     setOperationAction(ISD::EXTRACT_VECTOR_ELT, VT, Custom);
     setOperationAction(ISD::SETCC, VT, Custom);
     setOperationAction(ISD::SCALAR_TO_VECTOR, VT, Expand);
     setOperationAction(ISD::LOAD, VT, Custom);
     setOperationAction(ISD::STORE, VT, Custom);
     setOperationAction(ISD::TRUNCATE, VT, Custom);
+    setOperationAction(ISD::VSELECT, VT, Expand);
+    setOperationAction(ISD::SELECT, VT, Expand);
   }
 }
 
 ARMTargetLowering::ARMTargetLowering(const TargetMachine &TM,
                                      const ARMSubtarget &STI)
     : TargetLowering(TM), Subtarget(&STI) {
   RegInfo = Subtarget->getRegisterInfo();
   Itins = Subtarget->getInstrItineraryData();

llvm/test/Analysis/CostModel/ARM/arith-overflow.ll

 ; NEON-RECIP-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16I8 = call { <16 x i8>, <16 x i1> } @llvm.sadd.with.overflow.v16i8(<16 x i8> undef, <16 x i8> undef)
 ; NEON-RECIP-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V32I8 = call { <32 x i8>, <32 x i1> } @llvm.sadd.with.overflow.v32i8(<32 x i8> undef, <32 x i8> undef)
 ; NEON-RECIP-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V64I8 = call { <64 x i8>, <64 x i1> } @llvm.sadd.with.overflow.v64i8(<64 x i8> undef, <64 x i8> undef)
 ; NEON-RECIP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; MVE-RECIP-LABEL: 'sadd'
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %I64 = call { i64, i1 } @llvm.sadd.with.overflow.i64(i64 undef, i64 undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 74 for instruction: %V2I64 = call { <2 x i64>, <2 x i1> } @llvm.sadd.with.overflow.v2i64(<2 x i64> undef, <2 x i64> undef)
-; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 166 for instruction: %V4I64 = call { <4 x i64>, <4 x i1> } @llvm.sadd.with.overflow.v4i64(<4 x i64> undef, <4 x i64> undef)
-; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 582 for instruction: %V8I64 = call { <8 x i64>, <8 x i1> } @llvm.sadd.with.overflow.v8i64(<8 x i64> undef, <8 x i64> undef)
+; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 242 for instruction: %V4I64 = call { <4 x i64>, <4 x i1> } @llvm.sadd.with.overflow.v4i64(<4 x i64> undef, <4 x i64> undef)
+; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 866 for instruction: %V8I64 = call { <8 x i64>, <8 x i1> } @llvm.sadd.with.overflow.v8i64(<8 x i64> undef, <8 x i64> undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %I32 = call { i32, i1 } @llvm.sadd.with.overflow.i32(i32 undef, i32 undef)
-; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V4I32 = call { <4 x i32>, <4 x i1> } @llvm.sadd.with.overflow.v4i32(<4 x i32> undef, <4 x i32> undef)
-; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V8I32 = call { <8 x i32>, <8 x i1> } @llvm.sadd.with.overflow.v8i32(<8 x i32> undef, <8 x i32> undef)
-; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V16I32 = call { <16 x i32>, <16 x i1> } @llvm.sadd.with.overflow.v16i32(<16 x i32> undef, <16 x i32> undef)
+; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 90 for instruction: %V4I32 = call { <4 x i32>, <4 x i1> } @llvm.sadd.with.overflow.v4i32(<4 x i32> undef, <4 x i32> undef)
+; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 306 for instruction: %V8I32 = call { <8 x i32>, <8 x i1> } @llvm.sadd.with.overflow.v8i32(<8 x i32> undef, <8 x i32> undef)
+; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 1122 for instruction: %V16I32 = call { <16 x i32>, <16 x i1> } @llvm.sadd.with.overflow.v16i32(<16 x i32> undef, <16 x i32> undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %I16 = call { i16, i1 } @llvm.sadd.with.overflow.i16(i16 undef, i16 undef)
-; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8I16 = call { <8 x i16>, <8 x i1> } @llvm.sadd.with.overflow.v8i16(<8 x i16> undef, <8 x i16> undef)
-; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V16I16 = call { <16 x i16>, <16 x i1> } @llvm.sadd.with.overflow.v16i16(<16 x i16> undef, <16 x i16> undef)
-; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V32I16 = call { <32 x i16>, <32 x i1> } @llvm.sadd.with.overflow.v32i16(<32 x i16> undef, <32 x i16> undef)
+; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 298 for instruction: %V8I16 = call { <8 x i16>, <8 x i1> } @llvm.sadd.with.overflow.v8i16(<8 x i16> undef, <8 x i16> undef)
+; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 1106 for instruction: %V16I16 = call { <16 x i16>, <16 x i1> } @llvm.sadd.with.overflow.v16i16(<16 x i16> undef, <16 x i16> undef)
+; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 4260 for instruction: %V32I16 = call { <32 x i16>, <32 x i1> } @llvm.sadd.with.overflow.v32i16(<32 x i16> undef, <32 x i16> undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %I8 = call { i8, i1 } @llvm.sadd.with.overflow.i8(i8 undef, i8 undef)
-; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V16I8 = call { <16 x i8>, <16 x i1> } @llvm.sadd.with.overflow.v16i8(<16 x i8> undef, <16 x i8> undef)
-; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V32I8 = call { <32 x i8>, <32 x i1> } @llvm.sadd.with.overflow.v32i8(<32 x i8> undef, <32 x i8> undef)
-; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 56 for instruction: %V64I8 = call { <64 x i8>, <64 x i1> } @llvm.sadd.with.overflow.v64i8(<64 x i8> undef, <64 x i8> undef)
+; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 1098 for instruction: %V16I8 = call { <16 x i8>, <16 x i1> } @llvm.sadd.with.overflow.v16i8(<16 x i8> undef, <16 x i8> undef)
+; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 4244 for instruction: %V32I8 = call { <32 x i8>, <32 x i1> } @llvm.sadd.with.overflow.v32i8(<32 x i8> undef, <32 x i8> undef)
+; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 16680 for instruction: %V64I8 = call { <64 x i8>, <64 x i1> } @llvm.sadd.with.overflow.v64i8(<64 x i8> undef, <64 x i8> undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; V8M-SIZE-LABEL: 'sadd'
 ; V8M-SIZE-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %I64 = call { i64, i1 } @llvm.sadd.with.overflow.i64(i64 undef, i64 undef)
 ; V8M-SIZE-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V2I64 = call { <2 x i64>, <2 x i1> } @llvm.sadd.with.overflow.v2i64(<2 x i64> undef, <2 x i64> undef)
 ; V8M-SIZE-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V4I64 = call { <4 x i64>, <4 x i1> } @llvm.sadd.with.overflow.v4i64(<4 x i64> undef, <4 x i64> undef)
 ; V8M-SIZE-NEXT:  Cost Model: Found an estimated cost of 29 for instruction: %V8I64 = call { <8 x i64>, <8 x i1> } @llvm.sadd.with.overflow.v8i64(<8 x i64> undef, <8 x i64> undef)
 ; V8M-SIZE-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %I32 = call { i32, i1 } @llvm.sadd.with.overflow.i32(i32 undef, i32 undef)

 ; NEON-RECIP-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16I8 = call { <16 x i8>, <16 x i1> } @llvm.ssub.with.overflow.v16i8(<16 x i8> undef, <16 x i8> undef)
 ; NEON-RECIP-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V32I8 = call { <32 x i8>, <32 x i1> } @llvm.ssub.with.overflow.v32i8(<32 x i8> undef, <32 x i8> undef)
 ; NEON-RECIP-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V64I8 = call { <64 x i8>, <64 x i1> } @llvm.ssub.with.overflow.v64i8(<64 x i8> undef, <64 x i8> undef)
 ; NEON-RECIP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; MVE-RECIP-LABEL: 'ssub'
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %I64 = call { i64, i1 } @llvm.ssub.with.overflow.i64(i64 undef, i64 undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 74 for instruction: %V2I64 = call { <2 x i64>, <2 x i1> } @llvm.ssub.with.overflow.v2i64(<2 x i64> undef, <2 x i64> undef)
-; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 166 for instruction: %V4I64 = call { <4 x i64>, <4 x i1> } @llvm.ssub.with.overflow.v4i64(<4 x i64> undef, <4 x i64> undef)
-; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 582 for instruction: %V8I64 = call { <8 x i64>, <8 x i1> } @llvm.ssub.with.overflow.v8i64(<8 x i64> undef, <8 x i64> undef)
+; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 242 for instruction: %V4I64 = call { <4 x i64>, <4 x i1> } @llvm.ssub.with.overflow.v4i64(<4 x i64> undef, <4 x i64> undef)
+; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 866 for instruction: %V8I64 = call { <8 x i64>, <8 x i1> } @llvm.ssub.with.overflow.v8i64(<8 x i64> undef, <8 x i64> undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %I32 = call { i32, i1 } @llvm.ssub.with.overflow.i32(i32 undef, i32 undef)
-; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V4I32 = call { <4 x i32>, <4 x i1> } @llvm.ssub.with.overflow.v4i32(<4 x i32> undef, <4 x i32> undef)
-; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V8I32 = call { <8 x i32>, <8 x i1> } @llvm.ssub.with.overflow.v8i32(<8 x i32> undef, <8 x i32> undef)
-; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V16I32 = call { <16 x i32>, <16 x i1> } @llvm.ssub.with.overflow.v16i32(<16 x i32> undef, <16 x i32> undef)
+; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 90 for instruction: %V4I32 = call { <4 x i32>, <4 x i1> } @llvm.ssub.with.overflow.v4i32(<4 x i32> undef, <4 x i32> undef)
+; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 306 for instruction: %V8I32 = call { <8 x i32>, <8 x i1> } @llvm.ssub.with.overflow.v8i32(<8 x i32> undef, <8 x i32> undef)
+; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 1122 for instruction: %V16I32 = call { <16 x i32>, <16 x i1> } @llvm.ssub.with.overflow.v16i32(<16 x i32> undef, <16 x i32> undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %I16 = call { i16, i1 } @llvm.ssub.with.overflow.i16(i16 undef, i16 undef)
-; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8I16 = call { <8 x i16>, <8 x i1> } @llvm.ssub.with.overflow.v8i16(<8 x i16> undef, <8 x i16> undef)
-; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V16I16 = call { <16 x i16>, <16 x i1> } @llvm.ssub.with.overflow.v16i16(<16 x i16> undef, <16 x i16> undef)
-; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V32I16 = call { <32 x i16>, <32 x i1> } @llvm.ssub.with.overflow.v32i16(<32 x i16> undef, <32 x i16> undef)
+; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 298 for instruction: %V8I16 = call { <8 x i16>, <8 x i1> } @llvm.ssub.with.overflow.v8i16(<8 x i16> undef, <8 x i16> undef)
+; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 1106 for instruction: %V16I16 = call { <16 x i16>, <16 x i1> } @llvm.ssub.with.overflow.v16i16(<16 x i16> undef, <16 x i16> undef)
+; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 4260 for instruction: %V32I16 = call { <32 x i16>, <32 x i1> } @llvm.ssub.with.overflow.v32i16(<32 x i16> undef, <32 x i16> undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %I8 = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 undef, i8 undef)
-; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V16I8 = call { <16 x i8>, <16 x i1> } @llvm.ssub.with.overflow.v16i8(<16 x i8> undef, <16 x i8> undef)
-; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V32I8 = call { <32 x i8>, <32 x i1> } @llvm.ssub.with.overflow.v32i8(<32 x i8> undef, <32 x i8> undef)
-; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 56 for instruction: %V64I8 = call { <64 x i8>, <64 x i1> } @llvm.ssub.with.overflow.v64i8(<64 x i8> undef, <64 x i8> undef)
+; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 1098 for instruction: %V16I8 = call { <16 x i8>, <16 x i1> } @llvm.ssub.with.overflow.v16i8(<16 x i8> undef, <16 x i8> undef)
+; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 4244 for instruction: %V32I8 = call { <32 x i8>, <32 x i1> } @llvm.ssub.with.overflow.v32i8(<32 x i8> undef, <32 x i8> undef)
+; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 16680 for instruction: %V64I8 = call { <64 x i8>, <64 x i1> } @llvm.ssub.with.overflow.v64i8(<64 x i8> undef, <64 x i8> undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; V8M-SIZE-LABEL: 'ssub'
 ; V8M-SIZE-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %I64 = call { i64, i1 } @llvm.ssub.with.overflow.i64(i64 undef, i64 undef)
 ; V8M-SIZE-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V2I64 = call { <2 x i64>, <2 x i1> } @llvm.ssub.with.overflow.v2i64(<2 x i64> undef, <2 x i64> undef)
 ; V8M-SIZE-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V4I64 = call { <4 x i64>, <4 x i1> } @llvm.ssub.with.overflow.v4i64(<4 x i64> undef, <4 x i64> undef)
 ; V8M-SIZE-NEXT:  Cost Model: Found an estimated cost of 29 for instruction: %V8I64 = call { <8 x i64>, <8 x i1> } @llvm.ssub.with.overflow.v8i64(<8 x i64> undef, <8 x i64> undef)
 ; V8M-SIZE-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %I32 = call { i32, i1 } @llvm.ssub.with.overflow.i32(i32 undef, i32 undef)

llvm/test/Analysis/CostModel/ARM/arith-ssat.ll

 ; NEON-RECIP-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16I8 = call <16 x i8> @llvm.sadd.sat.v16i8(<16 x i8> undef, <16 x i8> undef)
 ; NEON-RECIP-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V32I8 = call <32 x i8> @llvm.sadd.sat.v32i8(<32 x i8> undef, <32 x i8> undef)
 ; NEON-RECIP-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V64I8 = call <64 x i8> @llvm.sadd.sat.v64i8(<64 x i8> undef, <64 x i8> undef)
 ; NEON-RECIP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; MVE-RECIP-LABEL: 'add'
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %I64 = call i64 @llvm.sadd.sat.i64(i64 undef, i64 undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 118 for instruction: %V2I64 = call <2 x i64> @llvm.sadd.sat.v2i64(<2 x i64> undef, <2 x i64> undef)
-; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 302 for instruction: %V4I64 = call <4 x i64> @llvm.sadd.sat.v4i64(<4 x i64> undef, <4 x i64> undef)
-; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 1046 for instruction: %V8I64 = call <8 x i64> @llvm.sadd.sat.v8i64(<8 x i64> undef, <8 x i64> undef)
+; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 378 for instruction: %V4I64 = call <4 x i64> @llvm.sadd.sat.v4i64(<4 x i64> undef, <4 x i64> undef)
+; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 1330 for instruction: %V8I64 = call <8 x i64> @llvm.sadd.sat.v8i64(<8 x i64> undef, <8 x i64> undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %I32 = call i32 @llvm.sadd.sat.i32(i32 undef, i32 undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 108 for instruction: %V2I32 = call <2 x i32> @llvm.sadd.sat.v2i32(<2 x i32> undef, <2 x i32> undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4I32 = call <4 x i32> @llvm.sadd.sat.v4i32(<4 x i32> undef, <4 x i32> undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8I32 = call <8 x i32> @llvm.sadd.sat.v8i32(<8 x i32> undef, <8 x i32> undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16I32 = call <16 x i32> @llvm.sadd.sat.v16i32(<16 x i32> undef, <16 x i32> undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %I16 = call i16 @llvm.sadd.sat.i16(i16 undef, i16 undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 108 for instruction: %V2I16 = call <2 x i16> @llvm.sadd.sat.v2i16(<2 x i16> undef, <2 x i16> undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4I16 = call <4 x i16> @llvm.sadd.sat.v4i16(<4 x i16> undef, <4 x i16> undef)

 ; NEON-RECIP-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16I8 = call <16 x i8> @llvm.ssub.sat.v16i8(<16 x i8> undef, <16 x i8> undef)
 ; NEON-RECIP-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V32I8 = call <32 x i8> @llvm.ssub.sat.v32i8(<32 x i8> undef, <32 x i8> undef)
 ; NEON-RECIP-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V64I8 = call <64 x i8> @llvm.ssub.sat.v64i8(<64 x i8> undef, <64 x i8> undef)
 ; NEON-RECIP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; MVE-RECIP-LABEL: 'sub'
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %I64 = call i64 @llvm.ssub.sat.i64(i64 undef, i64 undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 118 for instruction: %V2I64 = call <2 x i64> @llvm.ssub.sat.v2i64(<2 x i64> undef, <2 x i64> undef)
-; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 302 for instruction: %V4I64 = call <4 x i64> @llvm.ssub.sat.v4i64(<4 x i64> undef, <4 x i64> undef)
-; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 1046 for instruction: %V8I64 = call <8 x i64> @llvm.ssub.sat.v8i64(<8 x i64> undef, <8 x i64> undef)
+; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 378 for instruction: %V4I64 = call <4 x i64> @llvm.ssub.sat.v4i64(<4 x i64> undef, <4 x i64> undef)
+; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 1330 for instruction: %V8I64 = call <8 x i64> @llvm.ssub.sat.v8i64(<8 x i64> undef, <8 x i64> undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %I32 = call i32 @llvm.ssub.sat.i32(i32 undef, i32 undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 108 for instruction: %V2I32 = call <2 x i32> @llvm.ssub.sat.v2i32(<2 x i32> undef, <2 x i32> undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4I32 = call <4 x i32> @llvm.ssub.sat.v4i32(<4 x i32> undef, <4 x i32> undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8I32 = call <8 x i32> @llvm.ssub.sat.v8i32(<8 x i32> undef, <8 x i32> undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16I32 = call <16 x i32> @llvm.ssub.sat.v16i32(<16 x i32> undef, <16 x i32> undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %I16 = call i16 @llvm.ssub.sat.i16(i16 undef, i16 undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 108 for instruction: %V2I16 = call <2 x i16> @llvm.ssub.sat.v2i16(<2 x i16> undef, <2 x i16> undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4I16 = call <4 x i16> @llvm.ssub.sat.v4i16(<4 x i16> undef, <4 x i16> undef)

llvm/test/CodeGen/Thumb2/mve-pred-vselect.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=thumbv8.1m.main-none-none-eabi -mattr=+mve -verify-machineinstrs %s -o - | FileCheck %s
+
+define arm_aapcs_vfpcc <4 x i32> @cmpeqz_v4i1(<4 x i32> %a, <4 x i32> %b, <4 x i32> %c) {
+; CHECK-LABEL: cmpeqz_v4i1:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vpt.i32 eq, q1, zr
+; CHECK-NEXT:    vcmpt.i32 ne, q2, zr
+; CHECK-NEXT:    vmrs r0, p0
+; CHECK-NEXT:    vpt.i32 eq, q0, zr
+; CHECK-NEXT:    vcmpt.i32 eq, q2, zr
+; CHECK-NEXT:    vmrs r1, p0
+; CHECK-NEXT:    orrs r0, r1
+; CHECK-NEXT:    vmsr p0, r0
+; CHECK-NEXT:    vpsel q0, q0, q1
+; CHECK-NEXT:    bx lr
+entry:
+  %c1 = icmp eq <4 x i32> %a, zeroinitializer
+  %c2 = icmp eq <4 x i32> %b, zeroinitializer
+  %c3 = icmp eq <4 x i32> %c, zeroinitializer
+  %c4 = select <4 x i1> %c3, <4 x i1> %c1, <4 x i1> %c2
+  %s = select <4 x i1> %c4, <4 x i32> %a, <4 x i32> %b
+  ret <4 x i32> %s
+}
+
+define arm_aapcs_vfpcc <8 x i16> @cmpeqz_v8i1(<8 x i16> %a, <8 x i16> %b, <8 x i16> %c) {
+; CHECK-LABEL: cmpeqz_v8i1:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vpt.i16 eq, q1, zr
+; CHECK-NEXT:    vcmpt.i16 ne, q2, zr
+; CHECK-NEXT:    vmrs r0, p0
+; CHECK-NEXT:    vpt.i16 eq, q0, zr
+; CHECK-NEXT:    vcmpt.i16 eq, q2, zr
+; CHECK-NEXT:    vmrs r1, p0
+; CHECK-NEXT:    orrs r0, r1
+; CHECK-NEXT:    vmsr p0, r0
+; CHECK-NEXT:    vpsel q0, q0, q1
+; CHECK-NEXT:    bx lr
+entry:
+  %c1 = icmp eq <8 x i16> %a, zeroinitializer
+  %c2 = icmp eq <8 x i16> %b, zeroinitializer
+  %c3 = icmp eq <8 x i16> %c, zeroinitializer
+  %c4 = select <8 x i1> %c3, <8 x i1> %c1, <8 x i1> %c2
+  %s = select <8 x i1> %c4, <8 x i16> %a, <8 x i16> %b
+  ret <8 x i16> %s
+}
+
+define arm_aapcs_vfpcc <16 x i8> @cmpeqz_v16i1(<16 x i8> %a, <16 x i8> %b, <16 x i8> %c) {
+; CHECK-LABEL: cmpeqz_v16i1:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vpt.i8 eq, q1, zr
+; CHECK-NEXT:    vcmpt.i8 ne, q2, zr
+; CHECK-NEXT:    vmrs r0, p0
+; CHECK-NEXT:    vpt.i8 eq, q0, zr
+; CHECK-NEXT:    vcmpt.i8 eq, q2, zr
+; CHECK-NEXT:    vmrs r1, p0
+; CHECK-NEXT:    orrs r0, r1
+; CHECK-NEXT:    vmsr p0, r0
+; CHECK-NEXT:    vpsel q0, q0, q1
+; CHECK-NEXT:    bx lr
+entry:
+  %c1 = icmp eq <16 x i8> %a, zeroinitializer
+  %c2 = icmp eq <16 x i8> %b, zeroinitializer
+  %c3 = icmp eq <16 x i8> %c, zeroinitializer
+  %c4 = select <16 x i1> %c3, <16 x i1> %c1, <16 x i1> %c2
+  %s = select <16 x i1> %c4, <16 x i8> %a, <16 x i8> %b
+  ret <16 x i8> %s
+}
+
+define arm_aapcs_vfpcc <2 x i64> @cmpeqz_v2i1(<2 x i64> %a, <2 x i64> %b, <2 x i64> %c) {
+; CHECK-LABEL: cmpeqz_v2i1:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    .vsave {d8, d9}
+; CHECK-NEXT:    vpush {d8, d9}
+; CHECK-NEXT:    vmov r0, s11
+; CHECK-NEXT:    vmov r1, s10
+; CHECK-NEXT:    vmov r2, s8
+; CHECK-NEXT:    orrs r0, r1
+; CHECK-NEXT:    vmov r1, s9
+; CHECK-NEXT:    cset r0, eq
+; CHECK-NEXT:    tst.w r0, #1
+; CHECK-NEXT:    csetm r0, ne
+; CHECK-NEXT:    orrs r1, r2
+; CHECK-NEXT:    vmov r2, s4
+; CHECK-NEXT:    cset r1, eq
+; CHECK-NEXT:    tst.w r1, #1
+; CHECK-NEXT:    csetm r1, ne
+; CHECK-NEXT:    vmov q2[2], q2[0], r1, r0
+; CHECK-NEXT:    vmov q2[3], q2[1], r1, r0
+; CHECK-NEXT:    vmov r0, s7
+; CHECK-NEXT:    vmov r1, s6
+; CHECK-NEXT:    orrs r0, r1
+; CHECK-NEXT:    vmov r1, s5
+; CHECK-NEXT:    cset r0, eq
+; CHECK-NEXT:    tst.w r0, #1
+; CHECK-NEXT:    csetm r0, ne
+; CHECK-NEXT:    orrs r1, r2
+; CHECK-NEXT:    vmov r2, s0
+; CHECK-NEXT:    cset r1, eq
+; CHECK-NEXT:    tst.w r1, #1
+; CHECK-NEXT:    csetm r1, ne
+; CHECK-NEXT:    vmov q3[2], q3[0], r1, r0
+; CHECK-NEXT:    vmov q3[3], q3[1], r1, r0
+; CHECK-NEXT:    vmov r0, s3
+; CHECK-NEXT:    vmov r1, s2
+; CHECK-NEXT:    vbic q3, q3, q2
+; CHECK-NEXT:    orrs r0, r1
+; CHECK-NEXT:    vmov r1, s1
+; CHECK-NEXT:    cset r0, eq
+; CHECK-NEXT:    tst.w r0, #1
+; CHECK-NEXT:    csetm r0, ne
+; CHECK-NEXT:    orrs r1, r2
+; CHECK-NEXT:    cset r1, eq
+; CHECK-NEXT:    tst.w r1, #1
+; CHECK-NEXT:    csetm r1, ne
+; CHECK-NEXT:    vmov q4[2], q4[0], r1, r0
+; CHECK-NEXT:    vmov q4[3], q4[1], r1, r0
+; CHECK-NEXT:    vand q2, q4, q2
+; CHECK-NEXT:    vorr q2, q2, q3
+; CHECK-NEXT:    vbic q1, q1, q2
+; CHECK-NEXT:    vand q0, q0, q2
+; CHECK-NEXT:    vorr q0, q0, q1
+; CHECK-NEXT:    vpop {d8, d9}
+; CHECK-NEXT:    bx lr
+entry:
+  %c1 = icmp eq <2 x i64> %a, zeroinitializer
+  %c2 = icmp eq <2 x i64> %b, zeroinitializer
+  %c3 = icmp eq <2 x i64> %c, zeroinitializer
+  %c4 = select <2 x i1> %c3, <2 x i1> %c1, <2 x i1> %c2
+  %s = select <2 x i1> %c4, <2 x i64> %a, <2 x i64> %b
+  ret <2 x i64> %s
+}
+
+define arm_aapcs_vfpcc <4 x i32> @cmpnez_v4i1(<4 x i32> %a, <4 x i32> %b, <4 x i32> %c) {
+; CHECK-LABEL: cmpnez_v4i1:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vpt.i32 ne, q1, zr
+; CHECK-NEXT:    vcmpt.i32 eq, q2, zr
+; CHECK-NEXT:    vmrs r0, p0
+; CHECK-NEXT:    vpt.i32 ne, q0, zr
+; CHECK-NEXT:    vcmpt.i32 ne, q2, zr
+; CHECK-NEXT:    vmrs r1, p0
+; CHECK-NEXT:    orrs r0, r1
+; CHECK-NEXT:    vmsr p0, r0
+; CHECK-NEXT:    vpsel q0, q0, q1
+; CHECK-NEXT:    bx lr
+entry:
+  %c1 = icmp ne <4 x i32> %a, zeroinitializer
+  %c2 = icmp ne <4 x i32> %b, zeroinitializer
+  %c3 = icmp ne <4 x i32> %c, zeroinitializer
+  %c4 = select <4 x i1> %c3, <4 x i1> %c1, <4 x i1> %c2
+  %s = select <4 x i1> %c4, <4 x i32> %a, <4 x i32> %b
+  ret <4 x i32> %s
+}
+
+define arm_aapcs_vfpcc <8 x i16> @cmpnez_v8i1(<8 x i16> %a, <8 x i16> %b, <8 x i16> %c) {
+; CHECK-LABEL: cmpnez_v8i1:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vpt.i16 ne, q1, zr
+; CHECK-NEXT:    vcmpt.i16 eq, q2, zr
+; CHECK-NEXT:    vmrs r0, p0
+; CHECK-NEXT:    vpt.i16 ne, q0, zr
+; CHECK-NEXT:    vcmpt.i16 ne, q2, zr
+; CHECK-NEXT:    vmrs r1, p0
+; CHECK-NEXT:    orrs r0, r1
+; CHECK-NEXT:    vmsr p0, r0
+; CHECK-NEXT:    vpsel q0, q0, q1
+; CHECK-NEXT:    bx lr
+entry:
+  %c1 = icmp ne <8 x i16> %a, zeroinitializer
+  %c2 = icmp ne <8 x i16> %b, zeroinitializer
+  %c3 = icmp ne <8 x i16> %c, zeroinitializer
+  %c4 = select <8 x i1> %c3, <8 x i1> %c1, <8 x i1> %c2
+  %s = select <8 x i1> %c4, <8 x i16> %a, <8 x i16> %b
+  ret <8 x i16> %s
+}
+
+define arm_aapcs_vfpcc <16 x i8> @cmpnez_v16i1(<16 x i8> %a, <16 x i8> %b, <16 x i8> %c) {
+; CHECK-LABEL: cmpnez_v16i1:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vpt.i8 ne, q1, zr
+; CHECK-NEXT:    vcmpt.i8 eq, q2, zr
+; CHECK-NEXT:    vmrs r0, p0
+; CHECK-NEXT:    vpt.i8 ne, q0, zr
+; CHECK-NEXT:    vcmpt.i8 ne, q2, zr
+; CHECK-NEXT:    vmrs r1, p0
+; CHECK-NEXT:    orrs r0, r1
+; CHECK-NEXT:    vmsr p0, r0
+; CHECK-NEXT:    vpsel q0, q0, q1
+; CHECK-NEXT:    bx lr
+entry:
+  %c1 = icmp ne <16 x i8> %a, zeroinitializer
+  %c2 = icmp ne <16 x i8> %b, zeroinitializer
+  %c3 = icmp ne <16 x i8> %c, zeroinitializer
+  %c4 = select <16 x i1> %c3, <16 x i1> %c1, <16 x i1> %c2
+  %s = select <16 x i1> %c4, <16 x i8> %a, <16 x i8> %b
+  ret <16 x i8> %s
+}
+
+define arm_aapcs_vfpcc <2 x i64> @cmpnez_v2i1(<2 x i64> %a, <2 x i64> %b, <2 x i64> %c) {
+; CHECK-LABEL: cmpnez_v2i1:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    .vsave {d8, d9}
+; CHECK-NEXT:    vpush {d8, d9}
+; CHECK-NEXT:    vmov r0, s11
+; CHECK-NEXT:    vmov r1, s10
+; CHECK-NEXT:    vmov r2, s8
+; CHECK-NEXT:    orrs r0, r1
+; CHECK-NEXT:    vmov r1, s9
+; CHECK-NEXT:    cset r0, ne
+; CHECK-NEXT:    tst.w r0, #1
+; CHECK-NEXT:    csetm r0, ne
+; CHECK-NEXT:    orrs r1, r2
+; CHECK-NEXT:    vmov r2, s4
+; CHECK-NEXT:    cset r1, ne
+; CHECK-NEXT:    tst.w r1, #1
+; CHECK-NEXT:    csetm r1, ne
+; CHECK-NEXT:    vmov q2[2], q2[0], r1, r0
+; CHECK-NEXT:    vmov q2[3], q2[1], r1, r0
+; CHECK-NEXT:    vmov r0, s7
+; CHECK-NEXT:    vmov r1, s6
+; CHECK-NEXT:    orrs r0, r1
+; CHECK-NEXT:    vmov r1, s5
+; CHECK-NEXT:    cset r0, ne
+; CHECK-NEXT:    tst.w r0, #1
+; CHECK-NEXT:    csetm r0, ne
+; CHECK-NEXT:    orrs r1, r2
+; CHECK-NEXT:    vmov r2, s0
+; CHECK-NEXT:    cset r1, ne
+; CHECK-NEXT:    tst.w r1, #1
+; CHECK-NEXT:    csetm r1, ne
+; CHECK-NEXT:    vmov q3[2], q3[0], r1, r0
+; CHECK-NEXT:    vmov q3[3], q3[1], r1, r0
+; CHECK-NEXT:    vmov r0, s3
+; CHECK-NEXT:    vmov r1, s2
+; CHECK-NEXT:    vbic q3, q3, q2
+; CHECK-NEXT:    orrs r0, r1
+; CHECK-NEXT:    vmov r1, s1
+; CHECK-NEXT:    cset r0, ne
+; CHECK-NEXT:    tst.w r0, #1
+; CHECK-NEXT:    csetm r0, ne
+; CHECK-NEXT:    orrs r1, r2
+; CHECK-NEXT:    cset r1, ne
+; CHECK-NEXT:    tst.w r1, #1
+; CHECK-NEXT:    csetm r1, ne
+; CHECK-NEXT:    vmov q4[2], q4[0], r1, r0
+; CHECK-NEXT:    vmov q4[3], q4[1], r1, r0
+; CHECK-NEXT:    vand q2, q4, q2
+; CHECK-NEXT:    vorr q2, q2, q3
+; CHECK-NEXT:    vbic q1, q1, q2
+; CHECK-NEXT:    vand q0, q0, q2
+; CHECK-NEXT:    vorr q0, q0, q1
+; CHECK-NEXT:    vpop {d8, d9}
+; CHECK-NEXT:    bx lr
+entry:
+  %c1 = icmp ne <2 x i64> %a, zeroinitializer
+  %c2 = icmp ne <2 x i64> %b, zeroinitializer
+  %c3 = icmp ne <2 x i64> %c, zeroinitializer
+  %c4 = select <2 x i1> %c3, <2 x i1> %c1, <2 x i1> %c2
+  %s = select <2 x i1> %c4, <2 x i64> %a, <2 x i64> %b
+  ret <2 x i64> %s
+}
+
+
+
+define arm_aapcs_vfpcc <4 x i32> @cmpsltz_v4i1(<4 x i32> %a, <4 x i32> %b, <4 x i32> %c) {
+; CHECK-LABEL: cmpsltz_v4i1:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vpt.s32 lt, q1, zr
+; CHECK-NEXT:    vcmpt.s32 ge, q2, zr
+; CHECK-NEXT:    vmrs r0, p0
+; CHECK-NEXT:    vpt.s32 lt, q0, zr
+; CHECK-NEXT:    vcmpt.s32 lt, q2, zr
+; CHECK-NEXT:    vmrs r1, p0
+; CHECK-NEXT:    orrs r0, r1
+; CHECK-NEXT:    vmsr p0, r0
+; CHECK-NEXT:    vpsel q0, q0, q1
+; CHECK-NEXT:    bx lr
+entry:
+  %c1 = icmp slt <4 x i32> %a, zeroinitializer
+  %c2 = icmp slt <4 x i32> %b, zeroinitializer
+  %c3 = icmp slt <4 x i32> %c, zeroinitializer
+  %c4 = select <4 x i1> %c3, <4 x i1> %c1, <4 x i1> %c2
+  %s = select <4 x i1> %c4, <4 x i32> %a, <4 x i32> %b
+  ret <4 x i32> %s
+}
+
+define arm_aapcs_vfpcc <8 x i16> @cmpsltz_v8i1(<8 x i16> %a, <8 x i16> %b, <8 x i16> %c) {
+; CHECK-LABEL: cmpsltz_v8i1:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vpt.s16 lt, q1, zr
+; CHECK-NEXT:    vcmpt.s16 ge, q2, zr
+; CHECK-NEXT:    vmrs r0, p0
+; CHECK-NEXT:    vpt.s16 lt, q0, zr
+; CHECK-NEXT:    vcmpt.s16 lt, q2, zr
+; CHECK-NEXT:    vmrs r1, p0
+; CHECK-NEXT:    orrs r0, r1
+; CHECK-NEXT:    vmsr p0, r0
+; CHECK-NEXT:    vpsel q0, q0, q1
+; CHECK-NEXT:    bx lr
+entry:
+  %c1 = icmp slt <8 x i16> %a, zeroinitializer
+  %c2 = icmp slt <8 x i16> %b, zeroinitializer
+  %c3 = icmp slt <8 x i16> %c, zeroinitializer
+  %c4 = select <8 x i1> %c3, <8 x i1> %c1, <8 x i1> %c2
+  %s = select <8 x i1> %c4, <8 x i16> %a, <8 x i16> %b
+  ret <8 x i16> %s
+}
+
+define arm_aapcs_vfpcc <16 x i8> @cmpsltz_v16i1(<16 x i8> %a, <16 x i8> %b, <16 x i8> %c) {
+; CHECK-LABEL: cmpsltz_v16i1:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vpt.s8 lt, q1, zr
+; CHECK-NEXT:    vcmpt.s8 ge, q2, zr
+; CHECK-NEXT:    vmrs r0, p0
+; CHECK-NEXT:    vpt.s8 lt, q0, zr
+; CHECK-NEXT:    vcmpt.s8 lt, q2, zr
+; CHECK-NEXT:    vmrs r1, p0
+; CHECK-NEXT:    orrs r0, r1
+; CHECK-NEXT:    vmsr p0, r0
+; CHECK-NEXT:    vpsel q0, q0, q1
+; CHECK-NEXT:    bx lr
+entry:
+  %c1 = icmp slt <16 x i8> %a, zeroinitializer
+  %c2 = icmp slt <16 x i8> %b, zeroinitializer
+  %c3 = icmp slt <16 x i8> %c, zeroinitializer
+  %c4 = select <16 x i1> %c3, <16 x i1> %c1, <16 x i1> %c2
+  %s = select <16 x i1> %c4, <16 x i8> %a, <16 x i8> %b
+  ret <16 x i8> %s
+}
+
+define arm_aapcs_vfpcc <2 x i64> @cmpsltz_v2i1(<2 x i64> %a, <2 x i64> %b, <2 x i64> %c) {
+; CHECK-LABEL: cmpsltz_v2i1:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    .vsave {d8, d9}
+; CHECK-NEXT:    vpush {d8, d9}
+; CHECK-NEXT:    vmov r0, s11
+; CHECK-NEXT:    vmov r1, s9
+; CHECK-NEXT:    asrs r0, r0, #31
+; CHECK-NEXT:    asrs r1, r1, #31
+; CHECK-NEXT:    vmov q2[2], q2[0], r1, r0
+; CHECK-NEXT:    vmov q2[3], q2[1], r1, r0
+; CHECK-NEXT:    vmov r0, s7
+; CHECK-NEXT:    vmov r1, s5
+; CHECK-NEXT:    asrs r0, r0, #31
+; CHECK-NEXT:    asrs r1, r1, #31
+; CHECK-NEXT:    vmov q3[2], q3[0], r1, r0
+; CHECK-NEXT:    vmov q3[3], q3[1], r1, r0
+; CHECK-NEXT:    vmov r0, s3
+; CHECK-NEXT:    vmov r1, s1
+; CHECK-NEXT:    vbic q3, q3, q2
+; CHECK-NEXT:    asrs r0, r0, #31
+; CHECK-NEXT:    asrs r1, r1, #31
+; CHECK-NEXT:    vmov q4[2], q4[0], r1, r0
+; CHECK-NEXT:    vmov q4[3], q4[1], r1, r0
+; CHECK-NEXT:    vand q2, q4, q2
+; CHECK-NEXT:    vorr q2, q2, q3
+; CHECK-NEXT:    vbic q1, q1, q2
+; CHECK-NEXT:    vand q0, q0, q2
+; CHECK-NEXT:    vorr q0, q0, q1
+; CHECK-NEXT:    vpop {d8, d9}
+; CHECK-NEXT:    bx lr
+entry:
+  %c1 = icmp slt <2 x i64> %a, zeroinitializer
+  %c2 = icmp slt <2 x i64> %b, zeroinitializer
+  %c3 = icmp slt <2 x i64> %c, zeroinitializer
+  %c4 = select <2 x i1> %c3, <2 x i1> %c1, <2 x i1> %c2
+  %s = select <2 x i1> %c4, <2 x i64> %a, <2 x i64> %b
+  ret <2 x i64> %s
+}
+
+
+
+define arm_aapcs_vfpcc <4 x i32> @cmpeqz_v4i1_i1(<4 x i32> %a, <4 x i32> %b, i32 %c) {
+; CHECK-LABEL: cmpeqz_v4i1_i1:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    cbz r0, .LBB12_2
+; CHECK-NEXT:  @ %bb.1: @ %select.false
+; CHECK-NEXT:    vcmp.i32 eq, q1, zr
+; CHECK-NEXT:    vpsel q0, q0, q1
+; CHECK-NEXT:    bx lr
+; CHECK-NEXT:  .LBB12_2:
+; CHECK-NEXT:    vcmp.i32 eq, q0, zr
+; CHECK-NEXT:    vpsel q0, q0, q1
+; CHECK-NEXT:    bx lr
+entry:
+  %c1 = icmp eq <4 x i32> %a, zeroinitializer
+  %c2 = icmp eq <4 x i32> %b, zeroinitializer
+  %c3 = icmp eq i32 %c, 0
+  %c4 = select i1 %c3, <4 x i1> %c1, <4 x i1> %c2
+  %s = select <4 x i1> %c4, <4 x i32> %a, <4 x i32> %b
+  ret <4 x i32> %s
+}
+
+define arm_aapcs_vfpcc <8 x i16> @cmpeqz_v8i1_i1(<8 x i16> %a, <8 x i16> %b, i16 %c) {
+; CHECK-LABEL: cmpeqz_v8i1_i1:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    lsls r0, r0, #16
+; CHECK-NEXT:    beq .LBB13_2
+; CHECK-NEXT:  @ %bb.1: @ %select.false
+; CHECK-NEXT:    vcmp.i16 eq, q1, zr
+; CHECK-NEXT:    vpsel q0, q0, q1
+; CHECK-NEXT:    bx lr
+; CHECK-NEXT:  .LBB13_2:
+; CHECK-NEXT:    vcmp.i16 eq, q0, zr
+; CHECK-NEXT:    vpsel q0, q0, q1
+; CHECK-NEXT:    bx lr
+entry:
+  %c1 = icmp eq <8 x i16> %a, zeroinitializer
+  %c2 = icmp eq <8 x i16> %b, zeroinitializer
+  %c3 = icmp eq i16 %c, 0
+  %c4 = select i1 %c3, <8 x i1> %c1, <8 x i1> %c2
+  %s = select <8 x i1> %c4, <8 x i16> %a, <8 x i16> %b
+  ret <8 x i16> %s
+}
+
+define arm_aapcs_vfpcc <16 x i8> @cmpeqz_v16i1_i1(<16 x i8> %a, <16 x i8> %b, i8 %c) {
+; CHECK-LABEL: cmpeqz_v16i1_i1:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    lsls r0, r0, #24
+; CHECK-NEXT:    beq .LBB14_2
+; CHECK-NEXT:  @ %bb.1: @ %select.false
+; CHECK-NEXT:    vcmp.i8 eq, q1, zr
+; CHECK-NEXT:    vpsel q0, q0, q1
+; CHECK-NEXT:    bx lr
+; CHECK-NEXT:  .LBB14_2:
+; CHECK-NEXT:    vcmp.i8 eq, q0, zr
+; CHECK-NEXT:    vpsel q0, q0, q1
+; CHECK-NEXT:    bx lr
+entry:
+  %c1 = icmp eq <16 x i8> %a, zeroinitializer
+  %c2 = icmp eq <16 x i8> %b, zeroinitializer
+  %c3 = icmp eq i8 %c, 0
+  %c4 = select i1 %c3, <16 x i1> %c1, <16 x i1> %c2
+  %s = select <16 x i1> %c4, <16 x i8> %a, <16 x i8> %b
+  ret <16 x i8> %s
+}
+
+define arm_aapcs_vfpcc <2 x i64> @cmpeqz_v2i1_i1(<2 x i64> %a, <2 x i64> %b, i64 %c) {
+; CHECK-LABEL: cmpeqz_v2i1_i1:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    .save {r4, lr}
+; CHECK-NEXT:    push {r4, lr}
+; CHECK-NEXT:    vmov r2, s7
+; CHECK-NEXT:    vmov r3, s6
+; CHECK-NEXT:    orrs r2, r3
+; CHECK-NEXT:    vmov r3, s4
+; CHECK-NEXT:    cset r2, eq
+; CHECK-NEXT:    tst.w r2, #1
+; CHECK-NEXT:    vmov r2, s5
+; CHECK-NEXT:    csetm r12, ne
+; CHECK-NEXT:    orrs r2, r3
+; CHECK-NEXT:    vmov r3, s2
+; CHECK-NEXT:    cset r2, eq
+; CHECK-NEXT:    tst.w r2, #1
+; CHECK-NEXT:    vmov r2, s3
+; CHECK-NEXT:    csetm r4, ne
+; CHECK-NEXT:    orrs r2, r3
+; CHECK-NEXT:    vmov r3, s0
+; CHECK-NEXT:    cset r2, eq
+; CHECK-NEXT:    tst.w r2, #1
+; CHECK-NEXT:    vmov r2, s1
+; CHECK-NEXT:    csetm lr, ne
+; CHECK-NEXT:    orrs r2, r3
+; CHECK-NEXT:    cset r2, eq
+; CHECK-NEXT:    tst.w r2, #1
+; CHECK-NEXT:    csetm r2, ne
+; CHECK-NEXT:    orrs r0, r1
+; CHECK-NEXT:    beq .LBB15_2
+; CHECK-NEXT:  @ %bb.1: @ %select.false
+; CHECK-NEXT:    vmov q2[2], q2[0], r4, r12
+; CHECK-NEXT:    vmov q2[3], q2[1], r4, r12
+; CHECK-NEXT:    b .LBB15_3
+; CHECK-NEXT:  .LBB15_2:
+; CHECK-NEXT:    vmov q2[2], q2[0], r2, lr
+; CHECK-NEXT:    vmov q2[3], q2[1], r2, lr
+; CHECK-NEXT:  .LBB15_3: @ %select.end
+; CHECK-NEXT:    vmov r0, s10
+; CHECK-NEXT:    vmov r1, s8
+; CHECK-NEXT:    and r0, r0, #1
+; CHECK-NEXT:    and r1, r1, #1
+; CHECK-NEXT:    rsbs r0, r0, #0
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    vmov q2[2], q2[0], r1, r0
+; CHECK-NEXT:    vmov q2[3], q2[1], r1, r0
+; CHECK-NEXT:    vbic q1, q1, q2
+; CHECK-NEXT:    vand q0, q0, q2
+; CHECK-NEXT:    vorr q0, q0, q1
+; CHECK-NEXT:    pop {r4, pc}
+entry:
+  %c1 = icmp eq <2 x i64> %a, zeroinitializer
+  %c2 = icmp eq <2 x i64> %b, zeroinitializer
+  %c3 = icmp eq i64 %c, zeroinitializer
+  %c4 = select i1 %c3, <2 x i1> %c1, <2 x i1> %c2
+  %s = select <2 x i1> %c4, <2 x i64> %a, <2 x i64> %b
+  ret <2 x i64> %s
+}