[SVE] Lower fixed length vector ISD::SPLAT_VECTOR operations.

[paulwalker-arm]

Summary

Also strengthens the CHECK lines for scalable vector splat tests.

llvm/lib/Target/AArch64/AArch64ISelLowering.cpp

void AArch64TargetLowering::addTypeForFixedLengthSVE(MVT VT) {
   setOperationAction(ISD::MUL, VT, Custom);
   setOperationAction(ISD::OR, VT, Custom);
   setOperationAction(ISD::SETCC, VT, Custom);
   setOperationAction(ISD::SHL, VT, Custom);
   setOperationAction(ISD::SIGN_EXTEND, VT, Custom);
   setOperationAction(ISD::SIGN_EXTEND_INREG, VT, Custom);
   setOperationAction(ISD::SMAX, VT, Custom);
   setOperationAction(ISD::SMIN, VT, Custom);
+  setOperationAction(ISD::SPLAT_VECTOR, VT, Custom);
   setOperationAction(ISD::SRA, VT, Custom);
   setOperationAction(ISD::SRL, VT, Custom);
   setOperationAction(ISD::STORE, VT, Custom);
   setOperationAction(ISD::SUB, VT, Custom);
   setOperationAction(ISD::TRUNCATE, VT, Custom);
   setOperationAction(ISD::UMAX, VT, Custom);
   setOperationAction(ISD::UMIN, VT, Custom);
   setOperationAction(ISD::XOR, VT, Custom);
SDValue AArch64TargetLowering::LowerVECTOR_SHUFFLE(SDValue Op,
   return GenerateTBL(Op, ShuffleMask, DAG);
 }
 
 SDValue AArch64TargetLowering::LowerSPLAT_VECTOR(SDValue Op,
                                                  SelectionDAG &DAG) const {
   SDLoc dl(Op);
   EVT VT = Op.getValueType();
   EVT ElemVT = VT.getScalarType();
-
   SDValue SplatVal = Op.getOperand(0);
 
+  if (useSVEForFixedLengthVectorVT(VT))
+    return LowerToScalableOp(Op, DAG);
+
   // Extend input splat value where needed to fit into a GPR (32b or 64b only)
   // FPRs don't have this restriction.
   switch (ElemVT.getSimpleVT().SimpleTy) {
   case MVT::i1: {
     // The only legal i1 vectors are SVE vectors, so we can use SVE-specific
     // lowering code.
     if (auto *ConstVal = dyn_cast<ConstantSDNode>(SplatVal)) {
       if (ConstVal->isOne())
SDValue AArch64TargetLowering::LowerToScalableOp(SDValue Op,
   EVT VT = Op.getValueType();
   assert(useSVEForFixedLengthVectorVT(VT) &&
          "Only expected to lower fixed length vector operation!");
   EVT ContainerVT = getContainerForFixedLengthVector(DAG, VT);
 
   // Create list of operands by converting existing ones to scalable types.
   SmallVector<SDValue, 4> Ops;
   for (const SDValue &V : Op->op_values()) {
+    assert(!isa<VTSDNode>(V) && "Unexpected VTSDNode node!");
+
+    // Pass through non-vector operands.
+    if (!V.getValueType().isVector()) {
+      Ops.push_back(V);
+      continue;
+    }
+
+    // "cast" fixed length vector to a scalable vector.
     assert(useSVEForFixedLengthVectorVT(V.getValueType()) &&
            "Only fixed length vectors are supported!");
     Ops.push_back(convertToScalableVector(DAG, ContainerVT, V));
   }
 
   auto ScalableRes = DAG.getNode(Op.getOpcode(), SDLoc(Op), ContainerVT, Ops);
   return convertFromScalableVector(DAG, VT, ScalableRes);
 }

llvm/test/CodeGen/AArch64/sve-fixed-length-splat-vector.ll

+; RUN: llc -aarch64-sve-vector-bits-min=128  < %s | FileCheck %s -check-prefix=NO_SVE
+; RUN: llc -aarch64-sve-vector-bits-min=256  < %s | FileCheck %s -check-prefixes=CHECK,VBITS_EQ_256
+; RUN: llc -aarch64-sve-vector-bits-min=384  < %s | FileCheck %s -check-prefixes=CHECK
+; RUN: llc -aarch64-sve-vector-bits-min=512  < %s | FileCheck %s -check-prefixes=CHECK,VBITS_GE_512
+; RUN: llc -aarch64-sve-vector-bits-min=640  < %s | FileCheck %s -check-prefixes=CHECK,VBITS_GE_512
+; RUN: llc -aarch64-sve-vector-bits-min=768  < %s | FileCheck %s -check-prefixes=CHECK,VBITS_GE_512
+; RUN: llc -aarch64-sve-vector-bits-min=896  < %s | FileCheck %s -check-prefixes=CHECK,VBITS_GE_512
+; RUN: llc -aarch64-sve-vector-bits-min=1024 < %s | FileCheck %s -check-prefixes=CHECK,VBITS_GE_512,VBITS_GE_1024
+; RUN: llc -aarch64-sve-vector-bits-min=1152 < %s | FileCheck %s -check-prefixes=CHECK,VBITS_GE_512,VBITS_GE_1024
+; RUN: llc -aarch64-sve-vector-bits-min=1280 < %s | FileCheck %s -check-prefixes=CHECK,VBITS_GE_512,VBITS_GE_1024
+; RUN: llc -aarch64-sve-vector-bits-min=1408 < %s | FileCheck %s -check-prefixes=CHECK,VBITS_GE_512,VBITS_GE_1024
+; RUN: llc -aarch64-sve-vector-bits-min=1536 < %s | FileCheck %s -check-prefixes=CHECK,VBITS_GE_512,VBITS_GE_1024
+; RUN: llc -aarch64-sve-vector-bits-min=1664 < %s | FileCheck %s -check-prefixes=CHECK,VBITS_GE_512,VBITS_GE_1024
+; RUN: llc -aarch64-sve-vector-bits-min=1792 < %s | FileCheck %s -check-prefixes=CHECK,VBITS_GE_512,VBITS_GE_1024
+; RUN: llc -aarch64-sve-vector-bits-min=1920 < %s | FileCheck %s -check-prefixes=CHECK,VBITS_GE_512,VBITS_GE_1024
+; RUN: llc -aarch64-sve-vector-bits-min=2048 < %s | FileCheck %s -check-prefixes=CHECK,VBITS_GE_512,VBITS_GE_1024,VBITS_GE_2048
+
+target triple = "aarch64-unknown-linux-gnu"
+
+; Don't use SVE when its registers are no bigger than NEON.
+; NO_SVE-NOT: ptrue
+
+;
+; DUP (integer)
+;
+
+; Don't use SVE for 64-bit vectors.
+define <8 x i8> @splat_v8i8(i8 %a) #0 {
+; CHECK-LABEL: splat_v8i8:
+; CHECK: dup v0.8b, w0
+; CHECK-NEXT: ret
+  %insert = insertelement <8 x i8> undef, i8 %a, i64 0
+  %splat = shufflevector <8 x i8> %insert, <8 x i8> undef, <8 x i32> zeroinitializer
+  ret <8 x i8> %splat
+}
+
+; Don't use SVE for 128-bit vectors.
+define <16 x i8> @splat_v16i8(i8 %a) #0 {
+; CHECK-LABEL: splat_v16i8:
+; CHECK: dup v0.16b, w0
+; CHECK-NEXT: ret
+  %insert = insertelement <16 x i8> undef, i8 %a, i64 0
+  %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
+  ret <16 x i8> %splat
+}
+
+define void @splat_v32i8(i8 %a, <32 x i8>* %b) #0 {
+; CHECK-LABEL: splat_v32i8:
+; CHECK-DAG: mov [[RES:z[0-9]+]].b, w0
+; CHECK-DAG: ptrue [[PG:p[0-9]+]].b, vl32
+; CHECK-NEXT: st1b { [[RES]].b }, [[PG]], [x1]
+; CHECK-NEXT: ret
+  %insert = insertelement <32 x i8> undef, i8 %a, i64 0
+  %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
+  store <32 x i8> %splat, <32 x i8>* %b
+  ret void
+}
+
+define void @splat_v64i8(i8 %a, <64 x i8>* %b) #0 {
+; CHECK-LABEL: splat_v64i8:
+; VBITS_GE_512-DAG: mov [[RES:z[0-9]+]].b, w0
+; VBITS_GE_512-DAG: ptrue [[PG:p[0-9]+]].b, vl64
+; VBITS_GE_512-NEXT: st1b { [[RES]].b }, [[PG]], [x1]
+; VBITS_GE_512-NEXT: ret
+
+; Ensure sensible type legalisation.
+; VBITS_EQ_256-DAG: mov [[RES:z[0-9]+]].b, w0
+; VBITS_EQ_256-DAG: ptrue [[PG:p[0-9]+]].b, vl32
+; VBITS_EQ_256-DAG: mov w[[OFFSET_HI:[0-9]+]], #32
+; VBITS_EQ_256-DAG: st1b { [[RES]].b }, [[PG]], [x1]
+; VBITS_EQ_256-DAG: st1b { [[RES]].b }, [[PG]], [x1, x[[OFFSET_HI]]
+; VBITS_EQ_256-NEXT: ret
+  %insert = insertelement <64 x i8> undef, i8 %a, i64 0
+  %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+  store <64 x i8> %splat, <64 x i8>* %b
+  ret void
+}
+
+define void @splat_v128i8(i8 %a, <128 x i8>* %b) #0 {
+; CHECK-LABEL: splat_v128i8:
+; VBITS_GE_1024-DAG: mov [[RES:z[0-9]+]].b, w0
+; VBITS_GE_1024-DAG: ptrue [[PG:p[0-9]+]].b, vl128
+; VBITS_GE_1024-NEXT: st1b { [[RES]].b }, [[PG]], [x1]
+; VBITS_GE_1024-NEXT: ret
+  %insert = insertelement <128 x i8> undef, i8 %a, i64 0
+  %splat = shufflevector <128 x i8> %insert, <128 x i8> undef, <128 x i32> zeroinitializer
+  store <128 x i8> %splat, <128 x i8>* %b
+  ret void
+}
+
+define void @splat_v256i8(i8 %a, <256 x i8>* %b) #0 {
+; CHECK-LABEL: splat_v256i8:
+; VBITS_GE_2048-DAG: mov [[RES:z[0-9]+]].b, w0
+; VBITS_GE_2048-DAG: ptrue [[PG:p[0-9]+]].b, vl256
+; VBITS_GE_2048-NEXT: st1b { [[RES]].b }, [[PG]], [x1]
+; VBITS_GE_2048-NEXT: ret
+  %insert = insertelement <256 x i8> undef, i8 %a, i64 0
+  %splat = shufflevector <256 x i8> %insert, <256 x i8> undef, <256 x i32> zeroinitializer
+  store <256 x i8> %splat, <256 x i8>* %b
+  ret void
+}
+
+; Don't use SVE for 64-bit vectors.
+define <4 x i16> @splat_v4i16(i16 %a) #0 {
+; CHECK-LABEL: splat_v4i16:
+; CHECK: dup v0.4h, w0
+; CHECK-NEXT: ret
+  %insert = insertelement <4 x i16> undef, i16 %a, i64 0
+  %splat = shufflevector <4 x i16> %insert, <4 x i16> undef, <4 x i32> zeroinitializer
+  ret <4 x i16> %splat
+}
+
+; Don't use SVE for 128-bit vectors.
+define <8 x i16> @splat_v8i16(i16 %a) #0 {
+; CHECK-LABEL: splat_v8i16:
+; CHECK: dup v0.8h, w0
+; CHECK-NEXT: ret
+  %insert = insertelement <8 x i16> undef, i16 %a, i64 0
+  %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
+  ret <8 x i16> %splat
+}
+
+define void @splat_v16i16(i16 %a, <16 x i16>* %b) #0 {
+; CHECK-LABEL: splat_v16i16:
+; CHECK-DAG: mov [[RES:z[0-9]+]].h, w0
+; CHECK-DAG: ptrue [[PG:p[0-9]+]].h, vl16
+; CHECK-NEXT: st1h { [[RES]].h }, [[PG]], [x1]
+; CHECK-NEXT: ret
+  %insert = insertelement <16 x i16> undef, i16 %a, i64 0
+  %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
+  store <16 x i16> %splat, <16 x i16>* %b
+  ret void
+}
+
+define void @splat_v32i16(i16 %a, <32 x i16>* %b) #0 {
+; CHECK-LABEL: splat_v32i16:
+; VBITS_GE_512-DAG: mov [[RES:z[0-9]+]].h, w0
+; VBITS_GE_512-DAG: ptrue [[PG:p[0-9]+]].h, vl32
+; VBITS_GE_512-NEXT: st1h { [[RES]].h }, [[PG]], [x1]
+; VBITS_GE_512-NEXT: ret
+
+; Ensure sensible type legalisation.
+; VBITS_EQ_256-DAG: mov [[RES:z[0-9]+]].h, w0
+; VBITS_EQ_256-DAG: ptrue [[PG:p[0-9]+]].h, vl16
+; VBITS_EQ_256-DAG: add x[[B_HI:[0-9]+]], x1, #32
+; VBITS_EQ_256-DAG: st1h { [[RES]].h }, [[PG]], [x1]
+; VBITS_EQ_256-DAG: st1h { [[RES]].h }, [[PG]], [x[[B_HI]]
+; VBITS_EQ_256-NEXT: ret
+  %insert = insertelement <32 x i16> undef, i16 %a, i64 0
+  %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+  store <32 x i16> %splat, <32 x i16>* %b
+  ret void
+}
+
+define void @splat_v64i16(i16 %a, <64 x i16>* %b) #0 {
+; CHECK-LABEL: splat_v64i16:
+; VBITS_GE_1024-DAG: mov [[RES:z[0-9]+]].h, w0
+; VBITS_GE_1024-DAG: ptrue [[PG:p[0-9]+]].h, vl64
+; VBITS_GE_1024-NEXT: st1h { [[RES]].h }, [[PG]], [x1]
+; VBITS_GE_1024-NEXT: ret
+  %insert = insertelement <64 x i16> undef, i16 %a, i64 0
+  %splat = shufflevector <64 x i16> %insert, <64 x i16> undef, <64 x i32> zeroinitializer
+  store <64 x i16> %splat, <64 x i16>* %b
+  ret void
+}
+
+define void @splat_v128i16(i16 %a, <128 x i16>* %b) #0 {
+; CHECK-LABEL: splat_v128i16:
+; VBITS_GE_2048-DAG: mov [[RES:z[0-9]+]].h, w0
+; VBITS_GE_2048-DAG: ptrue [[PG:p[0-9]+]].h, vl128
+; VBITS_GE_2048-NEXT: st1h { [[RES]].h }, [[PG]], [x1]
+; VBITS_GE_2048-NEXT: ret
+  %insert = insertelement <128 x i16> undef, i16 %a, i64 0
+  %splat = shufflevector <128 x i16> %insert, <128 x i16> undef, <128 x i32> zeroinitializer
+  store <128 x i16> %splat, <128 x i16>* %b
+  ret void
+}
+
+; Don't use SVE for 64-bit vectors.
+define <2 x i32> @splat_v2i32(i32 %a) #0 {
+; CHECK-LABEL: splat_v2i32:
+; CHECK: dup v0.2s, w0
+; CHECK-NEXT: ret
+  %insert = insertelement <2 x i32> undef, i32 %a, i64 0
+  %splat = shufflevector <2 x i32> %insert, <2 x i32> undef, <2 x i32> zeroinitializer
+  ret <2 x i32> %splat
+}
+
+; Don't use SVE for 128-bit vectors.
+define <4 x i32> @splat_v4i32(i32 %a) #0 {
+; CHECK-LABEL: splat_v4i32:
+; CHECK: dup v0.4s, w0
+; CHECK-NEXT: ret
+  %insert = insertelement <4 x i32> undef, i32 %a, i64 0
+  %splat = shufflevector <4 x i32> %insert, <4 x i32> undef, <4 x i32> zeroinitializer
+  ret <4 x i32> %splat
+}
+
+define void @splat_v8i32(i32 %a, <8 x i32>* %b) #0 {
+; CHECK-LABEL: splat_v8i32:
+; CHECK-DAG: mov [[RES:z[0-9]+]].s, w0
+; CHECK-DAG: ptrue [[PG:p[0-9]+]].s, vl8
+; CHECK-NEXT: st1w { [[RES]].s }, [[PG]], [x1]
+; CHECK-NEXT: ret
+  %insert = insertelement <8 x i32> undef, i32 %a, i64 0
+  %splat = shufflevector <8 x i32> %insert, <8 x i32> undef, <8 x i32> zeroinitializer
+  store <8 x i32> %splat, <8 x i32>* %b
+  ret void
+}
+
+define void @splat_v16i32(i32 %a, <16 x i32>* %b) #0 {
+; CHECK-LABEL: splat_v16i32:
+; VBITS_GE_512-DAG: mov [[RES:z[0-9]+]].s, w0
+; VBITS_GE_512-DAG: ptrue [[PG:p[0-9]+]].s, vl16
+; VBITS_GE_512-NEXT: st1w { [[RES]].s }, [[PG]], [x1]
+; VBITS_GE_512-NEXT: ret
+
+; Ensure sensible type legalisation.
+; VBITS_EQ_256-DAG: mov [[RES:z[0-9]+]].s, w0
+; VBITS_EQ_256-DAG: ptrue [[PG:p[0-9]+]].s, vl8
+; VBITS_EQ_256-DAG: add x[[B_HI:[0-9]+]], x1, #32
+; VBITS_EQ_256-DAG: st1w { [[RES]].s }, [[PG]], [x1]
+; VBITS_EQ_256-DAG: st1w { [[RES]].s }, [[PG]], [x[[B_HI]]
+; VBITS_EQ_256-NEXT: ret
+  %insert = insertelement <16 x i32> undef, i32 %a, i64 0
+  %splat = shufflevector <16 x i32> %insert, <16 x i32> undef, <16 x i32> zeroinitializer
+  store <16 x i32> %splat, <16 x i32>* %b
+  ret void
+}
+
+define void @splat_v32i32(i32 %a, <32 x i32>* %b) #0 {
+; CHECK-LABEL: splat_v32i32:
+; VBITS_GE_1024-DAG: mov [[RES:z[0-9]+]].s, w0
+; VBITS_GE_1024-DAG: ptrue [[PG:p[0-9]+]].s, vl32
+; VBITS_GE_1024-NEXT: st1w { [[RES]].s }, [[PG]], [x1]
+; VBITS_GE_1024-NEXT: ret
+  %insert = insertelement <32 x i32> undef, i32 %a, i64 0
+  %splat = shufflevector <32 x i32> %insert, <32 x i32> undef, <32 x i32> zeroinitializer
+  store <32 x i32> %splat, <32 x i32>* %b
+  ret void
+}
+
+define void @splat_v64i32(i32 %a, <64 x i32>* %b) #0 {
+; CHECK-LABEL: splat_v64i32:
+; VBITS_GE_2048-DAG: mov [[RES:z[0-9]+]].s, w0
+; VBITS_GE_2048-DAG: ptrue [[PG:p[0-9]+]].s, vl64
+; VBITS_GE_2048-NEXT: st1w { [[RES]].s }, [[PG]], [x1]
+; VBITS_GE_2048-NEXT: ret
+  %insert = insertelement <64 x i32> undef, i32 %a, i64 0
+  %splat = shufflevector <64 x i32> %insert, <64 x i32> undef, <64 x i32> zeroinitializer
+  store <64 x i32> %splat, <64 x i32>* %b
+  ret void
+}
+
+; Don't use SVE for 64-bit vectors.
+define <1 x i64> @splat_v1i64(i64 %a) #0 {
+; CHECK-LABEL: splat_v1i64:
+; CHECK: fmov d0, x0
+; CHECK-NEXT: ret
+  %insert = insertelement <1 x i64> undef, i64 %a, i64 0
+  %splat = shufflevector <1 x i64> %insert, <1 x i64> undef, <1 x i32> zeroinitializer
+  ret <1 x i64> %splat
+}
+
+; Don't use SVE for 128-bit vectors.
+define <2 x i64> @splat_v2i64(i64 %a) #0 {
+; CHECK-LABEL: splat_v2i64:
+; CHECK: dup v0.2d, x0
+; CHECK-NEXT: ret
+  %insert = insertelement <2 x i64> undef, i64 %a, i64 0
+  %splat = shufflevector <2 x i64> %insert, <2 x i64> undef, <2 x i32> zeroinitializer
+  ret <2 x i64> %splat
+}
+
+define void @splat_v4i64(i64 %a, <4 x i64>* %b) #0 {
+; CHECK-LABEL: splat_v4i64:
+; CHECK-DAG: mov [[RES:z[0-9]+]].d, x0
+; CHECK-DAG: ptrue [[PG:p[0-9]+]].d, vl4
+; CHECK-NEXT: st1d { [[RES]].d }, [[PG]], [x1]
+; CHECK-NEXT: ret
+  %insert = insertelement <4 x i64> undef, i64 %a, i64 0
+  %splat = shufflevector <4 x i64> %insert, <4 x i64> undef, <4 x i32> zeroinitializer
+  store <4 x i64> %splat, <4 x i64>* %b
+  ret void
+}
+
+define void @splat_v8i64(i64 %a, <8 x i64>* %b) #0 {
+; CHECK-LABEL: splat_v8i64:
+; VBITS_GE_512-DAG: mov [[RES:z[0-9]+]].d, x0
+; VBITS_GE_512-DAG: ptrue [[PG:p[0-9]+]].d, vl8
+; VBITS_GE_512-NEXT: st1d { [[RES]].d }, [[PG]], [x1]
+; VBITS_GE_512-NEXT: ret
+
+; Ensure sensible type legalisation.
+; VBITS_EQ_256-DAG: mov [[RES:z[0-9]+]].d, x0
+; VBITS_EQ_256-DAG: ptrue [[PG:p[0-9]+]].d, vl4
+; VBITS_EQ_256-DAG: add x[[B_HI:[0-9]+]], x1, #32
+; VBITS_EQ_256-DAG: st1d { [[RES]].d }, [[PG]], [x1]
+; VBITS_EQ_256-DAG: st1d { [[RES]].d }, [[PG]], [x[[B_HI]]
+; VBITS_EQ_256-NEXT: ret
+  %insert = insertelement <8 x i64> undef, i64 %a, i64 0
+  %splat = shufflevector <8 x i64> %insert, <8 x i64> undef, <8 x i32> zeroinitializer
+  store <8 x i64> %splat, <8 x i64>* %b
+  ret void
+}
+
+define void @splat_v16i64(i64 %a, <16 x i64>* %b) #0 {
+; CHECK-LABEL: splat_v16i64:
+; VBITS_GE_1024-DAG: mov [[RES:z[0-9]+]].d, x0
+; VBITS_GE_1024-DAG: ptrue [[PG:p[0-9]+]].d, vl16
+; VBITS_GE_1024-NEXT: st1d { [[RES]].d }, [[PG]], [x1]
+; VBITS_GE_1024-NEXT: ret
+  %insert = insertelement <16 x i64> undef, i64 %a, i64 0
+  %splat = shufflevector <16 x i64> %insert, <16 x i64> undef, <16 x i32> zeroinitializer
+  store <16 x i64> %splat, <16 x i64>* %b
+  ret void
+}
+
+define void @splat_v32i64(i64 %a, <32 x i64>* %b) #0 {
+; CHECK-LABEL: splat_v32i64:
+; VBITS_GE_2048-DAG: mov [[RES:z[0-9]+]].d, x0
+; VBITS_GE_2048-DAG: ptrue [[PG:p[0-9]+]].d, vl32
+; VBITS_GE_2048-NEXT: st1d { [[RES]].d }, [[PG]], [x1]
+; VBITS_GE_2048-NEXT: ret
+  %insert = insertelement <32 x i64> undef, i64 %a, i64 0
+  %splat = shufflevector <32 x i64> %insert, <32 x i64> undef, <32 x i32> zeroinitializer
+  store <32 x i64> %splat, <32 x i64>* %b
+  ret void
+}
+
+;
+; DUP (floating-point)
+;
+
+; Don't use SVE for 64-bit vectors.
+define <4 x half> @splat_v4f16(half %a) #0 {
+; CHECK-LABEL: splat_v4f16:
+; CHECK: dup v0.4h, v0.h[0]
+; CHECK-NEXT: ret
+  %insert = insertelement <4 x half> undef, half %a, i64 0
+  %splat = shufflevector <4 x half> %insert, <4 x half> undef, <4 x i32> zeroinitializer
+  ret <4 x half> %splat
+}
+
+; Don't use SVE for 128-bit vectors.
+define <8 x half> @splat_v8f16(half %a) #0 {
+; CHECK-LABEL: splat_v8f16:
+; CHECK: dup v0.8h, v0.h[0]
+; CHECK-NEXT: ret
+  %insert = insertelement <8 x half> undef, half %a, i64 0
+  %splat = shufflevector <8 x half> %insert, <8 x half> undef, <8 x i32> zeroinitializer
+  ret <8 x half> %splat
+}
+
+define void @splat_v16f16(half %a, <16 x half>* %b) #0 {
+; CHECK-LABEL: splat_v16f16:
+; CHECK-DAG: mov [[RES:z[0-9]+]].h, h0
+; CHECK-DAG: ptrue [[PG:p[0-9]+]].h, vl16
+; CHECK-NEXT: st1h { [[RES]].h }, [[PG]], [x0]
+; CHECK-NEXT: ret
+  %insert = insertelement <16 x half> undef, half %a, i64 0
+  %splat = shufflevector <16 x half> %insert, <16 x half> undef, <16 x i32> zeroinitializer
+  store <16 x half> %splat, <16 x half>* %b
+  ret void
+}
+
+define void @splat_v32f16(half %a, <32 x half>* %b) #0 {
+; CHECK-LABEL: splat_v32f16:
+; VBITS_GE_512-DAG: mov [[RES:z[0-9]+]].h, h0
+; VBITS_GE_512-DAG: ptrue [[PG:p[0-9]+]].h, vl32
+; VBITS_GE_512-NEXT: st1h { [[RES]].h }, [[PG]], [x0]
+; VBITS_GE_512-NEXT: ret
+
+; Ensure sensible type legalisation.
+; VBITS_EQ_256-DAG: mov [[RES:z[0-9]+]].h, h0
+; VBITS_EQ_256-DAG: ptrue [[PG:p[0-9]+]].h, vl16
+; VBITS_EQ_256-DAG: add x[[B_HI:[0-9]+]], x0, #32
+; VBITS_EQ_256-DAG: st1h { [[RES]].h }, [[PG]], [x0]
+; VBITS_EQ_256-DAG: st1h { [[RES]].h }, [[PG]], [x[[B_HI]]
+; VBITS_EQ_256-NEXT: ret
+  %insert = insertelement <32 x half> undef, half %a, i64 0
+  %splat = shufflevector <32 x half> %insert, <32 x half> undef, <32 x i32> zeroinitializer
+  store <32 x half> %splat, <32 x half>* %b
+  ret void
+}
+
+define void @splat_v64f16(half %a, <64 x half>* %b) #0 {
+; CHECK-LABEL: splat_v64f16:
+; VBITS_GE_1024-DAG: mov [[RES:z[0-9]+]].h, h0
+; VBITS_GE_1024-DAG: ptrue [[PG:p[0-9]+]].h, vl64
+; VBITS_GE_1024-NEXT: st1h { [[RES]].h }, [[PG]], [x0]
+; VBITS_GE_1024-NEXT: ret
+  %insert = insertelement <64 x half> undef, half %a, i64 0
+  %splat = shufflevector <64 x half> %insert, <64 x half> undef, <64 x i32> zeroinitializer
+  store <64 x half> %splat, <64 x half>* %b
+  ret void
+}
+
+define void @splat_v128f16(half %a, <128 x half>* %b) #0 {
+; CHECK-LABEL: splat_v128f16:
+; VBITS_GE_2048-DAG: mov [[RES:z[0-9]+]].h, h0
+; VBITS_GE_2048-DAG: ptrue [[PG:p[0-9]+]].h, vl128
+; VBITS_GE_2048-NEXT: st1h { [[RES]].h }, [[PG]], [x0]
+; VBITS_GE_2048-NEXT: ret
+  %insert = insertelement <128 x half> undef, half %a, i64 0
+  %splat = shufflevector <128 x half> %insert, <128 x half> undef, <128 x i32> zeroinitializer
+  store <128 x half> %splat, <128 x half>* %b
+  ret void
+}
+
+; Don't use SVE for 64-bit vectors.
+define <2 x float> @splat_v2f32(float %a, <2 x float> %op2) #0 {
+; CHECK-LABEL: splat_v2f32:
+; CHECK: dup v0.2s, v0.s[0]
+; CHECK-NEXT: ret
+  %insert = insertelement <2 x float> undef, float %a, i64 0
+  %splat = shufflevector <2 x float> %insert, <2 x float> undef, <2 x i32> zeroinitializer
+  ret <2 x float> %splat
+}
+
+; Don't use SVE for 128-bit vectors.
+define <4 x float> @splat_v4f32(float %a, <4 x float> %op2) #0 {
+; CHECK-LABEL: splat_v4f32:
+; CHECK: dup v0.4s, v0.s[0]
+; CHECK-NEXT: ret
+  %insert = insertelement <4 x float> undef, float %a, i64 0
+  %splat = shufflevector <4 x float> %insert, <4 x float> undef, <4 x i32> zeroinitializer
+  ret <4 x float> %splat
+}
+
+define void @splat_v8f32(float %a, <8 x float>* %b) #0 {
+; CHECK-LABEL: splat_v8f32:
+; CHECK-DAG: mov [[RES:z[0-9]+]].s, s0
+; CHECK-DAG: ptrue [[PG:p[0-9]+]].s, vl8
+; CHECK-NEXT: st1w { [[RES]].s }, [[PG]], [x0]
+; CHECK-NEXT: ret
+  %insert = insertelement <8 x float> undef, float %a, i64 0
+  %splat = shufflevector <8 x float> %insert, <8 x float> undef, <8 x i32> zeroinitializer
+  store <8 x float> %splat, <8 x float>* %b
+  ret void
+}
+
+define void @splat_v16f32(float %a, <16 x float>* %b) #0 {
+; CHECK-LABEL: splat_v16f32:
+; VBITS_GE_512-DAG: mov [[RES:z[0-9]+]].s, s0
+; VBITS_GE_512-DAG: ptrue [[PG:p[0-9]+]].s, vl16
+; VBITS_GE_512-NEXT: st1w { [[RES]].s }, [[PG]], [x0]
+; VBITS_GE_512-NEXT: ret
+
+; Ensure sensible type legalisation.
+; VBITS_EQ_256-DAG: mov [[RES:z[0-9]+]].s, s0
+; VBITS_EQ_256-DAG: ptrue [[PG:p[0-9]+]].s, vl8
+; VBITS_EQ_256-DAG: add x[[B_HI:[0-9]+]], x0, #32
+; VBITS_EQ_256-DAG: st1w { [[RES]].s }, [[PG]], [x0]
+; VBITS_EQ_256-DAG: st1w { [[RES]].s }, [[PG]], [x[[B_HI]]
+; VBITS_EQ_256-NEXT: ret
+  %insert = insertelement <16 x float> undef, float %a, i64 0
+  %splat = shufflevector <16 x float> %insert, <16 x float> undef, <16 x i32> zeroinitializer
+  store <16 x float> %splat, <16 x float>* %b
+  ret void
+}
+
+define void @splat_v32f32(float %a, <32 x float>* %b) #0 {
+; CHECK-LABEL: splat_v32f32:
+; VBITS_GE_1024-DAG: mov [[RES:z[0-9]+]].s, s0
+; VBITS_GE_1024-DAG: ptrue [[PG:p[0-9]+]].s, vl32
+; VBITS_GE_1024-NEXT: st1w { [[RES]].s }, [[PG]], [x0]
+; VBITS_GE_1024-NEXT: ret
+  %insert = insertelement <32 x float> undef, float %a, i64 0
+  %splat = shufflevector <32 x float> %insert, <32 x float> undef, <32 x i32> zeroinitializer
+  store <32 x float> %splat, <32 x float>* %b
+  ret void
+}
+
+define void @splat_v64f32(float %a, <64 x float>* %b) #0 {
+; CHECK-LABEL: splat_v64f32:
+; VBITS_GE_2048-DAG: mov [[RES:z[0-9]+]].s, s0
+; VBITS_GE_2048-DAG: ptrue [[PG:p[0-9]+]].s, vl64
+; VBITS_GE_2048-NEXT: st1w { [[RES]].s }, [[PG]], [x0]
+; VBITS_GE_2048-NEXT: ret
+  %insert = insertelement <64 x float> undef, float %a, i64 0
+  %splat = shufflevector <64 x float> %insert, <64 x float> undef, <64 x i32> zeroinitializer
+  store <64 x float> %splat, <64 x float>* %b
+  ret void
+}
+
+; Don't use SVE for 64-bit vectors.
+define <1 x double> @splat_v1f64(double %a, <1 x double> %op2) #0 {
+; CHECK-LABEL: splat_v1f64:
+; CHECK: // %bb.0:
+; CHECK-NEXT: ret
+  %insert = insertelement <1 x double> undef, double %a, i64 0
+  %splat = shufflevector <1 x double> %insert, <1 x double> undef, <1 x i32> zeroinitializer
+  ret <1 x double> %splat
+}
+
+; Don't use SVE for 128-bit vectors.
+define <2 x double> @splat_v2f64(double %a, <2 x double> %op2) #0 {
+; CHECK-LABEL: splat_v2f64:
+; CHECK: dup v0.2d, v0.d[0]
+; CHECK-NEXT: ret
+  %insert = insertelement <2 x double> undef, double %a, i64 0
+  %splat = shufflevector <2 x double> %insert, <2 x double> undef, <2 x i32> zeroinitializer
+  ret <2 x double> %splat
+}
+
+define void @splat_v4f64(double %a, <4 x double>* %b) #0 {
+; CHECK-LABEL: splat_v4f64:
+; CHECK-DAG: mov [[RES:z[0-9]+]].d, d0
+; CHECK-DAG: ptrue [[PG:p[0-9]+]].d, vl4
+; CHECK-NEXT: st1d { [[RES]].d }, [[PG]], [x0]
+; CHECK-NEXT: ret
+  %insert = insertelement <4 x double> undef, double %a, i64 0
+  %splat = shufflevector <4 x double> %insert, <4 x double> undef, <4 x i32> zeroinitializer
+  store <4 x double> %splat, <4 x double>* %b
+  ret void
+}
+
+define void @splat_v8f64(double %a, <8 x double>* %b) #0 {
+; CHECK-LABEL: splat_v8f64:
+; VBITS_GE_512-DAG: mov [[RES:z[0-9]+]].d, d0
+; VBITS_GE_512-DAG: ptrue [[PG:p[0-9]+]].d, vl8
+; VBITS_GE_512-NEXT: st1d { [[RES]].d }, [[PG]], [x0]
+; VBITS_GE_512-NEXT: ret
+
+; Ensure sensible type legalisation.
+; VBITS_EQ_256-DAG: mov [[RES:z[0-9]+]].d, d0
+; VBITS_EQ_256-DAG: ptrue [[PG:p[0-9]+]].d, vl4
+; VBITS_EQ_256-DAG: add x[[B_HI:[0-9]+]], x0, #32
+; VBITS_EQ_256-DAG: st1d { [[RES]].d }, [[PG]], [x0]
+; VBITS_EQ_256-DAG: st1d { [[RES]].d }, [[PG]], [x[[B_HI]]
+; VBITS_EQ_256-NEXT: ret
+  %insert = insertelement <8 x double> undef, double %a, i64 0
+  %splat = shufflevector <8 x double> %insert, <8 x double> undef, <8 x i32> zeroinitializer
+  store <8 x double> %splat, <8 x double>* %b
+  ret void
+}
+
+define void @splat_v16f64(double %a, <16 x double>* %b) #0 {
+; CHECK-LABEL: splat_v16f64:
+; VBITS_GE_1024-DAG: mov [[RES:z[0-9]+]].d, d0
+; VBITS_GE_1024-DAG: ptrue [[PG:p[0-9]+]].d, vl16
+; VBITS_GE_1024-NEXT: st1d { [[RES]].d }, [[PG]], [x0]
+; VBITS_GE_1024-NEXT: ret
+  %insert = insertelement <16 x double> undef, double %a, i64 0
+  %splat = shufflevector <16 x double> %insert, <16 x double> undef, <16 x i32> zeroinitializer
+  store <16 x double> %splat, <16 x double>* %b
+  ret void
+}
+
+define void @splat_v32f64(double %a, <32 x double>* %b) #0 {
+; CHECK-LABEL: splat_v32f64:
+; VBITS_GE_2048-DAG: mov [[RES:z[0-9]+]].d, d0
+; VBITS_GE_2048-DAG: ptrue [[PG:p[0-9]+]].d, vl32
+; VBITS_GE_2048-NEXT: st1d { [[RES]].d }, [[PG]], [x0]
+; VBITS_GE_2048-NEXT: ret
+  %insert = insertelement <32 x double> undef, double %a, i64 0
+  %splat = shufflevector <32 x double> %insert, <32 x double> undef, <32 x i32> zeroinitializer
+  store <32 x double> %splat, <32 x double>* %b
+  ret void
+}
+
+;
+; DUP (integer immediate)
+;
+
+define void @splat_imm_v64i8(<64 x i8>* %a) #0 {
+; CHECK-LABEL: splat_imm_v64i8:
+; VBITS_GE_512-DAG: mov [[RES:z[0-9]+]].b, #1
+; VBITS_GE_512-DAG: ptrue [[PG:p[0-9]+]].b, vl64
+; VBITS_GE_512-NEXT: st1b { [[RES]].b }, [[PG]], [x0]
+; VBITS_GE_512-NEXT: ret
+  %insert = insertelement <64 x i8> undef, i8 1, i64 0
+  %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+  store <64 x i8> %splat, <64 x i8>* %a
+  ret void
+}
+
+define void @splat_imm_v32i16(<32 x i16>* %a) #0 {
+; CHECK-LABEL: splat_imm_v32i16:
+; VBITS_GE_512-DAG: mov [[RES:z[0-9]+]].h, #2
+; VBITS_GE_512-DAG: ptrue [[PG:p[0-9]+]].h, vl32
+; VBITS_GE_512-NEXT: st1h { [[RES]].h }, [[PG]], [x0]
+; VBITS_GE_512-NEXT: ret
+  %insert = insertelement <32 x i16> undef, i16 2, i64 0
+  %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+  store <32 x i16> %splat, <32 x i16>* %a
+  ret void
+}
+
+define void @splat_imm_v16i32(<16 x i32>* %a) #0 {
+; CHECK-LABEL: splat_imm_v16i32:
+; VBITS_GE_512-DAG: mov [[RES:z[0-9]+]].s, #3
+; VBITS_GE_512-DAG: ptrue [[PG:p[0-9]+]].s, vl16
+; VBITS_GE_512-NEXT: st1w { [[RES]].s }, [[PG]], [x0]
+; VBITS_GE_512-NEXT: ret
+  %insert = insertelement <16 x i32> undef, i32 3, i64 0
+  %splat = shufflevector <16 x i32> %insert, <16 x i32> undef, <16 x i32> zeroinitializer
+  store <16 x i32> %splat, <16 x i32>* %a
+  ret void
+}
+
+define void @splat_imm_v8i64(<8 x i64>* %a) #0 {
+; CHECK-LABEL: splat_imm_v8i64:
+; VBITS_GE_512-DAG: mov [[RES:z[0-9]+]].d, #4
+; VBITS_GE_512-DAG: ptrue [[PG:p[0-9]+]].d, vl8
+; VBITS_GE_512-NEXT: st1d { [[RES]].d }, [[PG]], [x0]
+; VBITS_GE_512-NEXT: ret
+  %insert = insertelement <8 x i64> undef, i64 4, i64 0
+  %splat = shufflevector <8 x i64> %insert, <8 x i64> undef, <8 x i32> zeroinitializer
+  store <8 x i64> %splat, <8 x i64>* %a
+  ret void
+}
+
+;
+; DUP (floating-point immediate)
+;
+
+define void @splat_imm_v32f16(<32 x half>* %a) #0 {
+; CHECK-LABEL: splat_imm_v32f16:
+; VBITS_GE_512-DAG: fmov [[RES:z[0-9]+]].h, #5.00000000
+; VBITS_GE_512-DAG: ptrue [[PG:p[0-9]+]].h, vl32
+; VBITS_GE_512-NEXT: st1h { [[RES]].h }, [[PG]], [x0]
+; VBITS_GE_512-NEXT: ret
+  %insert = insertelement <32 x half> undef, half 5.0, i64 0
+  %splat = shufflevector <32 x half> %insert, <32 x half> undef, <32 x i32> zeroinitializer
+  store <32 x half> %splat, <32 x half>* %a
+  ret void
+}
+
+define void @splat_imm_v16f32(<16 x float>* %a) #0 {
+; CHECK-LABEL: splat_imm_v16f32:
+; VBITS_GE_512-DAG: fmov [[RES:z[0-9]+]].s, #6.00000000
+; VBITS_GE_512-DAG: ptrue [[PG:p[0-9]+]].s, vl16
+; VBITS_GE_512-NEXT: st1w { [[RES]].s }, [[PG]], [x0]
+; VBITS_GE_512-NEXT: ret
+  %insert = insertelement <16 x float> undef, float 6.0, i64 0
+  %splat = shufflevector <16 x float> %insert, <16 x float> undef, <16 x i32> zeroinitializer
+  store <16 x float> %splat, <16 x float>* %a
+  ret void
+}
+
+define void @splat_imm_v8f64(<8 x double>* %a) #0 {
+; CHECK-LABEL: splat_imm_v8f64:
+; VBITS_GE_512-DAG: fmov [[RES:z[0-9]+]].d, #7.00000000
+; VBITS_GE_512-DAG: ptrue [[PG:p[0-9]+]].d, vl8
+; VBITS_GE_512-NEXT: st1d { [[RES]].d }, [[PG]], [x0]
+; VBITS_GE_512-NEXT: ret
+  %insert = insertelement <8 x double> undef, double 7.0, i64 0
+  %splat = shufflevector <8 x double> %insert, <8 x double> undef, <8 x i32> zeroinitializer
+  store <8 x double> %splat, <8 x double>* %a
+  ret void
+}
+attributes #0 = { "target-features"="+sve" }

llvm/test/CodeGen/AArch64/sve-fixed-length-subvector.ll

bb1:
   ret void
 }
 
 ;
 define <8 x i1> @no_warn_dropped_scalable(<8 x i32>* %in) #0 {
 ; CHECK-LABEL: no_warn_dropped_scalable:
 ; CHECK: ptrue [[PG:p[0-9]+]].s, vl8
 ; CHECK: ld1w { [[A:z[0-9]+]].s }, [[PG]]/z, [x0]
-; CHECK: cmpgt p{{[0-9]}}.s, [[PG]]/z, [[A]].s, z{{[0-9]+}}.s
+; CHECK: cmpgt p{{[0-9]}}.s, [[PG]]/z, [[A]].s, #0
 ; CHECK: ret
   %a = load <8 x i32>, <8 x i32>* %in
   br label %bb1
 
 bb1:
   %cond = icmp sgt <8 x i32> %a, zeroinitializer
   ret <8 x i1> %cond
 }
 
 attributes #0 = { "target-features"="+sve" }

llvm/test/CodeGen/AArch64/sve-vector-splat.ll

 ; CHECK-LABEL: splat_nxv2f64_zero:
 ; CHECK: mov z0.d, #0
 ; CHECK-NEXT: ret
   ret <vscale x 2 x double> zeroinitializer
 }
 
 define <vscale x 8 x half> @splat_nxv8f16_imm() {
 ; CHECK-LABEL: splat_nxv8f16_imm:
-; CHECK: mov z0.h, #1.0
+; CHECK: fmov z0.h, #1.00000000
 ; CHECK-NEXT: ret
   %1 = insertelement <vscale x 8 x half> undef, half 1.0, i32 0
   %2 = shufflevector <vscale x 8 x half> %1, <vscale x 8 x half> undef, <vscale x 8 x i32> zeroinitializer
   ret <vscale x 8 x half> %2
 }
 
 define <vscale x 4 x half> @splat_nxv4f16_imm() {
 ; CHECK-LABEL: splat_nxv4f16_imm:
-; CHECK: mov z0.h, #1.0
+; CHECK: fmov z0.h, #1.00000000
 ; CHECK-NEXT: ret
   %1 = insertelement <vscale x 4 x half> undef, half 1.0, i32 0
   %2 = shufflevector <vscale x 4 x half> %1, <vscale x 4 x half> undef, <vscale x 4 x i32> zeroinitializer
   ret <vscale x 4 x half> %2
 }
 
 define <vscale x 2 x half> @splat_nxv2f16_imm() {
 ; CHECK-LABEL: splat_nxv2f16_imm:
-; CHECK: mov z0.h, #1.0
+; CHECK: fmov z0.h, #1.00000000
 ; CHECK-NEXT: ret
   %1 = insertelement <vscale x 2 x half> undef, half 1.0, i32 0
   %2 = shufflevector <vscale x 2 x half> %1, <vscale x 2 x half> undef, <vscale x 2 x i32> zeroinitializer
   ret <vscale x 2 x half> %2
 }
 
 define <vscale x 4 x float> @splat_nxv4f32_imm() {
 ; CHECK-LABEL: splat_nxv4f32_imm:
-; CHECK: mov z0.s, #1.0
+; CHECK: fmov z0.s, #1.00000000
 ; CHECK-NEXT: ret
   %1 = insertelement <vscale x 4 x float> undef, float 1.0, i32 0
   %2 = shufflevector <vscale x 4 x float> %1, <vscale x 4 x float> undef, <vscale x 4 x i32> zeroinitializer
   ret <vscale x 4 x float> %2
 }
 
 define <vscale x 2 x float> @splat_nxv2f32_imm() {
 ; CHECK-LABEL: splat_nxv2f32_imm:
-; CHECK: mov z0.s, #1.0
+; CHECK: fmov z0.s, #1.00000000
 ; CHECK-NEXT: ret
   %1 = insertelement <vscale x 2 x float> undef, float 1.0, i32 0
   %2 = shufflevector <vscale x 2 x float> %1, <vscale x 2 x float> undef, <vscale x 2 x i32> zeroinitializer
   ret <vscale x 2 x float> %2
 }
 
 define <vscale x 2 x double> @splat_nxv2f64_imm() {
 ; CHECK-LABEL: splat_nxv2f64_imm:
-; CHECK: mov z0.d, #1.0
+; CHECK: fmov z0.d, #1.00000000
 ; CHECK-NEXT: ret
   %1 = insertelement <vscale x 2 x double> undef, double 1.0, i32 0
   %2 = shufflevector <vscale x 2 x double> %1, <vscale x 2 x double> undef, <vscale x 2 x i32> zeroinitializer
   ret <vscale x 2 x double> %2
 }
 
 define <vscale x 4 x i32> @splat_nxv4i32_fold(<vscale x 4 x i32> %x) {
 ; CHECK-LABEL: splat_nxv4i32_fold: