[SVE2p1][SME2] Add scalar addressing mode for LD1

[MattDevereau]

Summary

Add the scalar addressing mode for multi vector LD1 instructions.

[david-arm]

LGTM!

llvm/lib/Target/AArch64/AArch64ISelDAGToDAG.cpp

public:
   void SelectPostLoad(SDNode *N, unsigned NumVecs, unsigned Opc,
                          unsigned SubRegIdx);
   void SelectLoadLane(SDNode *N, unsigned NumVecs, unsigned Opc);
   void SelectPostLoadLane(SDNode *N, unsigned NumVecs, unsigned Opc);
   void SelectPredicatedLoad(SDNode *N, unsigned NumVecs, unsigned Scale,
                             unsigned Opc_rr, unsigned Opc_ri,
                             bool IsIntr = false);
   void SelectContiguousMultiVectorLoad(SDNode *N, unsigned NumVecs,
-                                       unsigned Scale, unsigned Opc_rr,
-                                       unsigned Opc_ri);
+                                       unsigned Scale, unsigned Opc_ri,
+                                       unsigned Opc_rr);
   void SelectDestructiveMultiIntrinsic(SDNode *N, unsigned NumVecs,
                                        bool IsZmMulti, unsigned Opcode,
                                        bool HasPred = false);
   void SelectPExtPair(SDNode *N, unsigned Opc);
   void SelectWhilePair(SDNode *N, unsigned Opc);
   void SelectCVTIntrinsic(SDNode *N, unsigned NumVecs, unsigned Opcode);
   void SelectClamp(SDNode *N, unsigned NumVecs, unsigned Opcode);
   void SelectUnaryMultiIntrinsic(SDNode *N, unsigned NumOutVecs,
void AArch64DAGToDAGISel::SelectContiguousMultiVectorLoad(SDNode *N,
                                                           unsigned Scale,
                                                           unsigned Opc_ri,
                                                           unsigned Opc_rr) {
   assert(Scale < 4 && "Invalid scaling value.");
   SDLoc DL(N);
   EVT VT = N->getValueType(0);
   SDValue Chain = N->getOperand(0);
 
-  // Use simplest addressing mode for now - base + 0 offset
   SDValue PNg = N->getOperand(2);
   SDValue Base = N->getOperand(3);
   SDValue Offset = CurDAG->getTargetConstant(0, DL, MVT::i64);
+  unsigned Opc;
+  std::tie(Opc, Base, Offset) =
+      findAddrModeSVELoadStore(N, Opc_rr, Opc_ri, Base, Offset, Scale);
 
   SDValue Ops[] = {PNg,            // Predicate-as-counter
                    Base,           // Memory operand
                    Offset, Chain};
 
   const EVT ResTys[] = {MVT::Untyped, MVT::Other};
 
-  SDNode *Load = CurDAG->getMachineNode(Opc_ri, DL, ResTys, Ops);
+  SDNode *Load = CurDAG->getMachineNode(Opc, DL, ResTys, Ops);
   SDValue SuperReg = SDValue(Load, 0);
   for (unsigned i = 0; i < NumVecs; ++i)
     ReplaceUses(SDValue(N, i), CurDAG->getTargetExtractSubreg(
                                    AArch64::zsub0 + i, DL, VT, SuperReg));
 
   // Copy chain
   unsigned ChainIdx = NumVecs;
   ReplaceUses(SDValue(N, ChainIdx), SDValue(Load, 1));

llvm/test/CodeGen/AArch64/sve2p1-intrinsics-loads.ll

 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    addvl sp, sp, #1
 ; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
 ; CHECK-NEXT:    ret
   %res = call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld1.pn.x2.nxv16i8(target("aarch64.svcount") %pn, ptr %ptr);
   ret { <vscale x 16 x i8>, <vscale x 16 x i8> } %res
 }
 
+define { <vscale x 16 x i8>, <vscale x 16 x i8> } @ld1_x2_i8_scalar(target("aarch64.svcount") %pn, ptr %ptr, i64 %index) nounwind {
+; CHECK-LABEL: ld1_x2_i8_scalar:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
+; CHECK-NEXT:    addvl sp, sp, #-1
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
+; CHECK-NEXT:    mov p8.b, p0.b
+; CHECK-NEXT:    ld1b { z0.b, z1.b }, pn8/z, [x0, x1]
+; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
+; CHECK-NEXT:    addvl sp, sp, #1
+; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
+; CHECK-NEXT:    ret
+  %base = getelementptr i8, ptr %ptr, i64 %index
+  %res = call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld1.pn.x2.nxv16i8(target("aarch64.svcount") %pn, ptr %base);
+  ret { <vscale x 16 x i8>, <vscale x 16 x i8> } %res
+}
+
 define { <vscale x 8 x i16>, <vscale x 8 x i16> } @ld1_x2_i16(target("aarch64.svcount") %pn, ptr %ptr) nounwind {
 ; CHECK-LABEL: ld1_x2_i16:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
 ; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    ld1h { z0.h, z1.h }, pn8/z, [x0]
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    addvl sp, sp, #1
 ; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
 ; CHECK-NEXT:    ret
   %res = call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld1.pn.x2.nxv8i16(target("aarch64.svcount") %pn, ptr %ptr);
   ret { <vscale x 8 x i16>, <vscale x 8 x i16> } %res
 }
 
+define { <vscale x 8 x i16>, <vscale x 8 x i16> } @ld1_x2_i16_scalar(target("aarch64.svcount") %pn, ptr %ptr, i64 %index) nounwind {
+; CHECK-LABEL: ld1_x2_i16_scalar:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
+; CHECK-NEXT:    addvl sp, sp, #-1
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
+; CHECK-NEXT:    mov p8.b, p0.b
+; CHECK-NEXT:    ld1h { z0.h, z1.h }, pn8/z, [x0, x1, lsl #1]
+; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
+; CHECK-NEXT:    addvl sp, sp, #1
+; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
+; CHECK-NEXT:    ret
+  %base = getelementptr i16, ptr %ptr, i64 %index
+  %res = call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld1.pn.x2.nxv8i16(target("aarch64.svcount") %pn, ptr %base);
+  ret { <vscale x 8 x i16>, <vscale x 8 x i16> } %res
+}
+
 define { <vscale x 4 x i32>, <vscale x 4 x i32> } @ld1_x2_i32(target("aarch64.svcount") %pn, ptr %ptr) nounwind {
 ; CHECK-LABEL: ld1_x2_i32:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
 ; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    ld1w { z0.s, z1.s }, pn8/z, [x0]
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    addvl sp, sp, #1
 ; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
 ; CHECK-NEXT:    ret
   %res = call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld1.pn.x2.nxv4i32(target("aarch64.svcount") %pn, ptr %ptr);
   ret { <vscale x 4 x i32>, <vscale x 4 x i32> } %res
 }
 
+define { <vscale x 4 x i32>, <vscale x 4 x i32> } @ld1_x2_i32_scalar(target("aarch64.svcount") %pn, ptr %ptr, i64 %index) nounwind {
+; CHECK-LABEL: ld1_x2_i32_scalar:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
+; CHECK-NEXT:    addvl sp, sp, #-1
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
+; CHECK-NEXT:    mov p8.b, p0.b
+; CHECK-NEXT:    ld1w { z0.s, z1.s }, pn8/z, [x0, x1, lsl #2]
+; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
+; CHECK-NEXT:    addvl sp, sp, #1
+; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
+; CHECK-NEXT:    ret
+  %base = getelementptr i32, ptr %ptr, i64 %index
+  %res = call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld1.pn.x2.nxv4i32(target("aarch64.svcount") %pn, ptr %base);
+  ret { <vscale x 4 x i32>, <vscale x 4 x i32> } %res
+}
+
 define { <vscale x 2 x i64>, <vscale x 2 x i64> } @ld1_x2_i64(target("aarch64.svcount") %pn, ptr %ptr) nounwind {
 ; CHECK-LABEL: ld1_x2_i64:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
 ; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    ld1d { z0.d, z1.d }, pn8/z, [x0]
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    addvl sp, sp, #1
 ; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
 ; CHECK-NEXT:    ret
   %res = call { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld1.pn.x2.nxv2i64(target("aarch64.svcount") %pn, ptr %ptr);
   ret { <vscale x 2 x i64>, <vscale x 2 x i64> } %res
 }
 
+define { <vscale x 2 x i64>, <vscale x 2 x i64> } @ld1_x2_i64_scalar(target("aarch64.svcount") %pn, ptr %ptr, i64 %index) nounwind {
+; CHECK-LABEL: ld1_x2_i64_scalar:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
+; CHECK-NEXT:    addvl sp, sp, #-1
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
+; CHECK-NEXT:    mov p8.b, p0.b
+; CHECK-NEXT:    ld1d { z0.d, z1.d }, pn8/z, [x0, x1, lsl #3]
+; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
+; CHECK-NEXT:    addvl sp, sp, #1
+; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
+; CHECK-NEXT:    ret
+  %base = getelementptr i64, ptr %ptr, i64 %index
+  %res = call { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld1.pn.x2.nxv2i64(target("aarch64.svcount") %pn, ptr %base);
+  ret { <vscale x 2 x i64>, <vscale x 2 x i64> } %res
+}
+
 define { <vscale x 8 x half>, <vscale x 8 x half> } @ld1_x2_f16(target("aarch64.svcount") %pn, ptr %ptr) nounwind {
 ; CHECK-LABEL: ld1_x2_f16:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
 ; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    ld1h { z0.h, z1.h }, pn8/z, [x0]
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    addvl sp, sp, #1
 ; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
 ; CHECK-NEXT:    ret
   %res = call { <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.ld1.pn.x2.nxv8f16(target("aarch64.svcount") %pn, ptr %ptr);
   ret { <vscale x 8 x half>, <vscale x 8 x half> } %res
 }
 
+define { <vscale x 8 x half>, <vscale x 8 x half> } @ld1_x2_f16_scalar(target("aarch64.svcount") %pn, ptr %ptr, i64 %index) nounwind {
+; CHECK-LABEL: ld1_x2_f16_scalar:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
+; CHECK-NEXT:    addvl sp, sp, #-1
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
+; CHECK-NEXT:    mov p8.b, p0.b
+; CHECK-NEXT:    ld1h { z0.h, z1.h }, pn8/z, [x0, x1, lsl #1]
+; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
+; CHECK-NEXT:    addvl sp, sp, #1
+; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
+; CHECK-NEXT:    ret
+  %base = getelementptr half, ptr %ptr, i64 %index
+  %res = call { <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.ld1.pn.x2.nxv8f16(target("aarch64.svcount") %pn, ptr %base);
+  ret { <vscale x 8 x half>, <vscale x 8 x half> } %res
+}
+
 define { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @ld1_x2_bf16(target("aarch64.svcount") %pn, ptr %ptr) nounwind {
 ; CHECK-LABEL: ld1_x2_bf16:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
 ; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    ld1h { z0.h, z1.h }, pn8/z, [x0]
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    addvl sp, sp, #1
 ; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
 ; CHECK-NEXT:    ret
   %res = call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.ld1.pn.x2.nxv8bf16(target("aarch64.svcount") %pn, ptr %ptr);
   ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } %res
 }
 
+define { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @ld1_x2_bf16_scalar(target("aarch64.svcount") %pn, ptr %ptr, i64 %index) nounwind {
+; CHECK-LABEL: ld1_x2_bf16_scalar:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
+; CHECK-NEXT:    addvl sp, sp, #-1
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
+; CHECK-NEXT:    mov p8.b, p0.b
+; CHECK-NEXT:    ld1h { z0.h, z1.h }, pn8/z, [x0, x1, lsl #1]
+; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
+; CHECK-NEXT:    addvl sp, sp, #1
+; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
+; CHECK-NEXT:    ret
+  %base = getelementptr bfloat, ptr %ptr, i64 %index
+  %res = call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.ld1.pn.x2.nxv8bf16(target("aarch64.svcount") %pn, ptr %base);
+  ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } %res
+}
+
 define { <vscale x 4 x float>, <vscale x 4 x float> } @ld1_x2_f32(target("aarch64.svcount") %pn, ptr %ptr) nounwind {
 ; CHECK-LABEL: ld1_x2_f32:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
 ; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    ld1w { z0.s, z1.s }, pn8/z, [x0]
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    addvl sp, sp, #1
 ; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
 ; CHECK-NEXT:    ret
   %res = call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.ld1.pn.x2.nxv4f32(target("aarch64.svcount") %pn, ptr %ptr);
   ret { <vscale x 4 x float>, <vscale x 4 x float> } %res
 }
 
+define { <vscale x 4 x float>, <vscale x 4 x float> } @ld1_x2_f32_scalar(target("aarch64.svcount") %pn, ptr %ptr, i64 %index) nounwind {
+; CHECK-LABEL: ld1_x2_f32_scalar:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
+; CHECK-NEXT:    addvl sp, sp, #-1
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
+; CHECK-NEXT:    mov p8.b, p0.b
+; CHECK-NEXT:    ld1w { z0.s, z1.s }, pn8/z, [x0, x1, lsl #2]
+; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
+; CHECK-NEXT:    addvl sp, sp, #1
+; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
+; CHECK-NEXT:    ret
+  %base = getelementptr float, ptr %ptr, i64 %index
+  %res = call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.ld1.pn.x2.nxv4f32(target("aarch64.svcount") %pn, ptr %base);
+  ret { <vscale x 4 x float>, <vscale x 4 x float> } %res
+}
+
 define { <vscale x 2 x double>, <vscale x 2 x double> } @ld1_x2_f64(target("aarch64.svcount") %pn, ptr %ptr) nounwind {
 ; CHECK-LABEL: ld1_x2_f64:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
 ; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    ld1d { z0.d, z1.d }, pn8/z, [x0]
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    addvl sp, sp, #1
 ; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
 ; CHECK-NEXT:    ret
   %res = call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.ld1.pn.x2.nxv2f64(target("aarch64.svcount") %pn, ptr %ptr);
   ret { <vscale x 2 x double>, <vscale x 2 x double> } %res
 }
 
+define { <vscale x 2 x double>, <vscale x 2 x double> } @ld1_x2_f64_scalar(target("aarch64.svcount") %pn, ptr %ptr, i64 %index) nounwind {
+; CHECK-LABEL: ld1_x2_f64_scalar:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
+; CHECK-NEXT:    addvl sp, sp, #-1
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
+; CHECK-NEXT:    mov p8.b, p0.b
+; CHECK-NEXT:    ld1d { z0.d, z1.d }, pn8/z, [x0, x1, lsl #3]
+; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
+; CHECK-NEXT:    addvl sp, sp, #1
+; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
+; CHECK-NEXT:    ret
+  %base = getelementptr double, ptr %ptr, i64 %index
+  %res = call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.ld1.pn.x2.nxv2f64(target("aarch64.svcount") %pn, ptr %base);
+  ret { <vscale x 2 x double>, <vscale x 2 x double> } %res
+}
+
 ; Test to ensure we load into the correct registers for the instruction
 define <vscale x 16 x i8> @ld1_x2_i8_z0_taken(target("aarch64.svcount") %pn, ptr %ptr, <vscale x 16 x i8> %val) {
 ; CHECK-LABEL: ld1_x2_i8_z0_taken:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
 ; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    .cfi_escape 0x0f, 0x0c, 0x8f, 0x00, 0x11, 0x10, 0x22, 0x11, 0x08, 0x92, 0x2e, 0x00, 0x1e, 0x22 // sp + 16 + 8 * VG
 ; CHECK-NEXT:    .cfi_offset w29, -16
 ; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    ld1b { z2.b, z3.b }, pn8/z, [x0]
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    add z0.b, z0.b, z2.b
 ; CHECK-NEXT:    addvl sp, sp, #1
 ; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
 ; CHECK-NEXT:    ret
   %ld1 = call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld1.pn.x2.nxv16i8(target("aarch64.svcount") %pn, ptr %ptr);
   %ld1_0 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %ld1, 0
   %res = add <vscale x 16 x i8> %val, %ld1_0
   ret <vscale x 16 x i8>  %res
 }
 
+; Test to ensure we load into the correct registers for the instruction
+define <vscale x 16 x i8> @ld1_x2_i8_z0_taken_scalar(target("aarch64.svcount") %pn, ptr %ptr, <vscale x 16 x i8> %val, i64 %index) {
+; CHECK-LABEL: ld1_x2_i8_z0_taken_scalar:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
+; CHECK-NEXT:    addvl sp, sp, #-1
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
+; CHECK-NEXT:    .cfi_escape 0x0f, 0x0c, 0x8f, 0x00, 0x11, 0x10, 0x22, 0x11, 0x08, 0x92, 0x2e, 0x00, 0x1e, 0x22 // sp + 16 + 8 * VG
+; CHECK-NEXT:    .cfi_offset w29, -16
+; CHECK-NEXT:    mov p8.b, p0.b
+; CHECK-NEXT:    ld1b { z2.b, z3.b }, pn8/z, [x0, x1]
+; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
+; CHECK-NEXT:    add z0.b, z0.b, z2.b
+; CHECK-NEXT:    addvl sp, sp, #1
+; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
+; CHECK-NEXT:    ret
+  %base = getelementptr i8, ptr %ptr, i64 %index
+  %ld1 = call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld1.pn.x2.nxv16i8(target("aarch64.svcount") %pn, ptr %base);
+  %ld1_0 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %ld1, 0
+  %res = add <vscale x 16 x i8> %val, %ld1_0
+  ret <vscale x 16 x i8>  %res
+}
+
 define { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @ld1_x4_i8(target("aarch64.svcount") %pn, ptr %ptr) nounwind {
 ; CHECK-LABEL: ld1_x4_i8:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
 ; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    ld1b { z0.b - z3.b }, pn8/z, [x0]
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    addvl sp, sp, #1
 ; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
 ; CHECK-NEXT:    ret
   %res = call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld1.pn.x4.nxv16i8(target("aarch64.svcount") %pn, ptr %ptr);
   ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %res
 }
 
+define { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @ld1_x4_i8_scalar(target("aarch64.svcount") %pn, ptr %ptr, i64 %index) nounwind {
+; CHECK-LABEL: ld1_x4_i8_scalar:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
+; CHECK-NEXT:    addvl sp, sp, #-1
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
+; CHECK-NEXT:    mov p8.b, p0.b
+; CHECK-NEXT:    ld1b { z0.b - z3.b }, pn8/z, [x0, x1]
+; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
+; CHECK-NEXT:    addvl sp, sp, #1
+; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
+; CHECK-NEXT:    ret
+  %base = getelementptr i8, ptr %ptr, i64 %index
+  %res = call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld1.pn.x4.nxv16i8(target("aarch64.svcount") %pn, ptr %base);
+  ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %res
+}
+
 define { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @ld1_x4_i16(target("aarch64.svcount") %pn, ptr %ptr) nounwind {
 ; CHECK-LABEL: ld1_x4_i16:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
 ; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    ld1h { z0.h - z3.h }, pn8/z, [x0]
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    addvl sp, sp, #1
 ; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
 ; CHECK-NEXT:    ret
   %res = call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld1.pn.x4.nxv8i16(target("aarch64.svcount") %pn, ptr %ptr);
   ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %res
 }
 
+define { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @ld1_x4_i16_scalar(target("aarch64.svcount") %pn, ptr %ptr, i64 %index) nounwind {
+; CHECK-LABEL: ld1_x4_i16_scalar:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
+; CHECK-NEXT:    addvl sp, sp, #-1
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
+; CHECK-NEXT:    mov p8.b, p0.b
+; CHECK-NEXT:    ld1h { z0.h - z3.h }, pn8/z, [x0, x1, lsl #1]
+; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
+; CHECK-NEXT:    addvl sp, sp, #1
+; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
+; CHECK-NEXT:    ret
+  %base = getelementptr i16, ptr %ptr, i64 %index
+  %res = call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld1.pn.x4.nxv8i16(target("aarch64.svcount") %pn, ptr %base);
+  ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %res
+}
+
 define { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @ld1_x4_i32(target("aarch64.svcount") %pn, ptr %ptr) nounwind {
 ; CHECK-LABEL: ld1_x4_i32:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
 ; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    ld1w { z0.s - z3.s }, pn8/z, [x0]
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    addvl sp, sp, #1
 ; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
 ; CHECK-NEXT:    ret
   %res = call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld1.pn.x4.nxv4i32(target("aarch64.svcount") %pn, ptr %ptr);
   ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } %res
 }
 
+define { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @ld1_x4_i32_scalar(target("aarch64.svcount") %pn, ptr %ptr, i64 %index) nounwind {
+; CHECK-LABEL: ld1_x4_i32_scalar:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
+; CHECK-NEXT:    addvl sp, sp, #-1
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
+; CHECK-NEXT:    mov p8.b, p0.b
+; CHECK-NEXT:    ld1w { z0.s - z3.s }, pn8/z, [x0, x1, lsl #2]
+; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
+; CHECK-NEXT:    addvl sp, sp, #1
+; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
+; CHECK-NEXT:    ret
+  %base = getelementptr i32, ptr %ptr, i64 %index
+  %res = call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld1.pn.x4.nxv4i32(target("aarch64.svcount") %pn, ptr %base);
+  ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } %res
+}
+
 define { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @ld1_x4_i64(target("aarch64.svcount") %pn, ptr %ptr) nounwind {
 ; CHECK-LABEL: ld1_x4_i64:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
 ; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    ld1d { z0.d - z3.d }, pn8/z, [x0]
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    addvl sp, sp, #1
 ; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
 ; CHECK-NEXT:    ret
   %res = call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld1.pn.x4.nxv2i64(target("aarch64.svcount") %pn, ptr %ptr);
   ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } %res
 }
 
+define { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @ld1_x4_i64_scalar(target("aarch64.svcount") %pn, ptr %ptr, i64 %index) nounwind {
+; CHECK-LABEL: ld1_x4_i64_scalar:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
+; CHECK-NEXT:    addvl sp, sp, #-1
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
+; CHECK-NEXT:    mov p8.b, p0.b
+; CHECK-NEXT:    ld1d { z0.d - z3.d }, pn8/z, [x0, x1, lsl #3]
+; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
+; CHECK-NEXT:    addvl sp, sp, #1
+; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
+; CHECK-NEXT:    ret
+  %base = getelementptr i64, ptr %ptr, i64 %index
+  %res = call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld1.pn.x4.nxv2i64(target("aarch64.svcount") %pn, ptr %base);
+  ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } %res
+}
+
 define { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @ld1_x4_f16(target("aarch64.svcount") %pn, ptr %ptr) nounwind {
 ; CHECK-LABEL: ld1_x4_f16:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
 ; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    ld1h { z0.h - z3.h }, pn8/z, [x0]
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    addvl sp, sp, #1
 ; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
 ; CHECK-NEXT:    ret
   %res = call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.ld1.pn.x4.nxv8f16(target("aarch64.svcount") %pn, ptr %ptr);
   ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } %res
 }
 
+define { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @ld1_x4_f16_scalar(target("aarch64.svcount") %pn, ptr %ptr, i64 %index) nounwind {
+; CHECK-LABEL: ld1_x4_f16_scalar:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
+; CHECK-NEXT:    addvl sp, sp, #-1
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
+; CHECK-NEXT:    mov p8.b, p0.b
+; CHECK-NEXT:    ld1h { z0.h - z3.h }, pn8/z, [x0, x1, lsl #1]
+; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
+; CHECK-NEXT:    addvl sp, sp, #1
+; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
+; CHECK-NEXT:    ret
+  %base = getelementptr half, ptr %ptr, i64 %index
+  %res = call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.ld1.pn.x4.nxv8f16(target("aarch64.svcount") %pn, ptr %base);
+  ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } %res
+}
+
 define { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @ld1_x4_bf16(target("aarch64.svcount") %pn, ptr %ptr) nounwind {
 ; CHECK-LABEL: ld1_x4_bf16:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
 ; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    ld1h { z0.h - z3.h }, pn8/z, [x0]
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    addvl sp, sp, #1
 ; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
 ; CHECK-NEXT:    ret
   %res = call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.ld1.pn.x4.nxv8bf16(target("aarch64.svcount") %pn, ptr %ptr);
   ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } %res
 }
 
+define { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @ld1_x4_bf16_scalar(target("aarch64.svcount") %pn, ptr %ptr, i64 %index) nounwind {
+; CHECK-LABEL: ld1_x4_bf16_scalar:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
+; CHECK-NEXT:    addvl sp, sp, #-1
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
+; CHECK-NEXT:    mov p8.b, p0.b
+; CHECK-NEXT:    ld1h { z0.h - z3.h }, pn8/z, [x0, x1, lsl #1]
+; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
+; CHECK-NEXT:    addvl sp, sp, #1
+; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
+; CHECK-NEXT:    ret
+  %base = getelementptr bfloat, ptr %ptr, i64 %index
+  %res = call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.ld1.pn.x4.nxv8bf16(target("aarch64.svcount") %pn, ptr %base);
+  ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } %res
+}
+
 define { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @ld1_x4_f32(target("aarch64.svcount") %pn, ptr %ptr) nounwind {
 ; CHECK-LABEL: ld1_x4_f32:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
 ; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    ld1w { z0.s - z3.s }, pn8/z, [x0]
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    addvl sp, sp, #1
 ; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
 ; CHECK-NEXT:    ret
   %res = call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.ld1.pn.x4.nxv4f32(target("aarch64.svcount") %pn, ptr %ptr);
   ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } %res
 }
 
+define { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @ld1_x4_f32_scalar(target("aarch64.svcount") %pn, ptr %ptr, i64 %index) nounwind {
+; CHECK-LABEL: ld1_x4_f32_scalar:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
+; CHECK-NEXT:    addvl sp, sp, #-1
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
+; CHECK-NEXT:    mov p8.b, p0.b
+; CHECK-NEXT:    ld1w { z0.s - z3.s }, pn8/z, [x0, x1, lsl #2]
+; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
+; CHECK-NEXT:    addvl sp, sp, #1
+; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
+; CHECK-NEXT:    ret
+  %base = getelementptr float, ptr %ptr, i64 %index
+  %res = call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.ld1.pn.x4.nxv4f32(target("aarch64.svcount") %pn, ptr %base);
+  ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } %res
+}
+
 define { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @ld1_x4_f64(target("aarch64.svcount") %pn, ptr %ptr) nounwind {
 ; CHECK-LABEL: ld1_x4_f64:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
 ; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    ld1d { z0.d - z3.d }, pn8/z, [x0]
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    addvl sp, sp, #1
 ; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
 ; CHECK-NEXT:    ret
   %res = call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.ld1.pn.x4.nxv2f64(target("aarch64.svcount") %pn, ptr %ptr);
   ret { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } %res
 }
 
+define { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @ld1_x4_f64_scalar(target("aarch64.svcount") %pn, ptr %ptr, i64 %index) nounwind {
+; CHECK-LABEL: ld1_x4_f64_scalar:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
+; CHECK-NEXT:    addvl sp, sp, #-1
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
+; CHECK-NEXT:    mov p8.b, p0.b
+; CHECK-NEXT:    ld1d { z0.d - z3.d }, pn8/z, [x0, x1, lsl #3]
+; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
+; CHECK-NEXT:    addvl sp, sp, #1
+; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
+; CHECK-NEXT:    ret
+  %base = getelementptr double, ptr %ptr, i64 %index
+  %res = call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.ld1.pn.x4.nxv2f64(target("aarch64.svcount") %pn, ptr %base);
+  ret { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } %res
+}
+
 ; Test to ensure we load into the correct registers for the instruction
 define <vscale x 8 x i16> @ld1_x4_i16_z0_taken(target("aarch64.svcount") %pn, ptr %ptr, <vscale x 8 x i16> %val) {
 ; CHECK-LABEL: ld1_x4_i16_z0_taken:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
 ; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    .cfi_escape 0x0f, 0x0c, 0x8f, 0x00, 0x11, 0x10, 0x22, 0x11, 0x08, 0x92, 0x2e, 0x00, 0x1e, 0x22 // sp + 16 + 8 * VG
 ; CHECK-NEXT:    .cfi_offset w29, -16
 ; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    ld1h { z4.h - z7.h }, pn8/z, [x0]
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    add z0.h, z0.h, z4.h
 ; CHECK-NEXT:    addvl sp, sp, #1
 ; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
 ; CHECK-NEXT:    ret
   %ld1 = call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld1.pn.x4.nxv8i16(target("aarch64.svcount") %pn, ptr %ptr);
   %ld1_0 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %ld1, 0
   %res = add <vscale x 8 x i16> %val, %ld1_0
   ret <vscale x 8 x i16>  %res
 }
 
+; Test to ensure we load into the correct registers for the instruction
+define <vscale x 8 x i16> @ld1_x4_i16_z0_taken_scalar(target("aarch64.svcount") %pn, ptr %ptr, <vscale x 8 x i16> %val, i64 %index) {
+; CHECK-LABEL: ld1_x4_i16_z0_taken_scalar:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
+; CHECK-NEXT:    addvl sp, sp, #-1
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
+; CHECK-NEXT:    .cfi_escape 0x0f, 0x0c, 0x8f, 0x00, 0x11, 0x10, 0x22, 0x11, 0x08, 0x92, 0x2e, 0x00, 0x1e, 0x22 // sp + 16 + 8 * VG
+; CHECK-NEXT:    .cfi_offset w29, -16
+; CHECK-NEXT:    mov p8.b, p0.b
+; CHECK-NEXT:    ld1h { z4.h - z7.h }, pn8/z, [x0, x1, lsl #1]
+; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
+; CHECK-NEXT:    add z0.h, z0.h, z4.h
+; CHECK-NEXT:    addvl sp, sp, #1
+; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
+; CHECK-NEXT:    ret
+  %base = getelementptr i16, ptr %ptr, i64 %index
+  %ld1 = call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld1.pn.x4.nxv8i16(target("aarch64.svcount") %pn, ptr %base);
+  %ld1_0 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %ld1, 0
+  %res = add <vscale x 8 x i16> %val, %ld1_0
+  ret <vscale x 8 x i16>  %res
+}
 
 ; == Non-temporal Multi-Vector Consecutive Loads ==
 
 define { <vscale x 16 x i8>, <vscale x 16 x i8> } @ldnt1_x2_i8(target("aarch64.svcount") %pn, ptr %ptr) nounwind {
 ; CHECK-LABEL: ldnt1_x2_i8:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
 ; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    ldnt1b { z0.b, z1.b }, pn8/z, [x0]
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    addvl sp, sp, #1
 ; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
 ; CHECK-NEXT:    ret
   %res = call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ldnt1.pn.x2.nxv16i8(target("aarch64.svcount") %pn, ptr %ptr);
   ret { <vscale x 16 x i8>, <vscale x 16 x i8> } %res
 }
 
+define { <vscale x 16 x i8>, <vscale x 16 x i8> } @ldnt1_x2_i8_scalar(target("aarch64.svcount") %pn, ptr %ptr, i64 %index) nounwind {
+; CHECK-LABEL: ldnt1_x2_i8_scalar:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
+; CHECK-NEXT:    addvl sp, sp, #-1
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
+; CHECK-NEXT:    mov p8.b, p0.b
+; CHECK-NEXT:    ldnt1b { z0.b, z1.b }, pn8/z, [x0, x1]
+; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
+; CHECK-NEXT:    addvl sp, sp, #1
+; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
+; CHECK-NEXT:    ret
+  %base = getelementptr i8, ptr %ptr, i64 %index
+  %res = call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ldnt1.pn.x2.nxv16i8(target("aarch64.svcount") %pn, ptr %base);
+  ret { <vscale x 16 x i8>, <vscale x 16 x i8> } %res
+}
+
 define { <vscale x 8 x i16>, <vscale x 8 x i16> } @ldnt1_x2_i16(target("aarch64.svcount") %pn, ptr %ptr) nounwind {
 ; CHECK-LABEL: ldnt1_x2_i16:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
 ; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    ldnt1h { z0.h, z1.h }, pn8/z, [x0]
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    addvl sp, sp, #1
 ; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
 ; CHECK-NEXT:    ret
   %res = call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ldnt1.pn.x2.nxv8i16(target("aarch64.svcount") %pn, ptr %ptr);
   ret { <vscale x 8 x i16>, <vscale x 8 x i16> } %res
 }
 
+define { <vscale x 8 x i16>, <vscale x 8 x i16> } @ldnt1_x2_i16_scalar(target("aarch64.svcount") %pn, ptr %ptr, i64 %index) nounwind {
+; CHECK-LABEL: ldnt1_x2_i16_scalar:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
+; CHECK-NEXT:    addvl sp, sp, #-1
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
+; CHECK-NEXT:    mov p8.b, p0.b
+; CHECK-NEXT:    ldnt1h { z0.h, z1.h }, pn8/z, [x0, x1, lsl #1]
+; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
+; CHECK-NEXT:    addvl sp, sp, #1
+; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
+; CHECK-NEXT:    ret
+  %base = getelementptr i16, ptr %ptr, i64 %index
+  %res = call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ldnt1.pn.x2.nxv8i16(target("aarch64.svcount") %pn, ptr %base);
+  ret { <vscale x 8 x i16>, <vscale x 8 x i16> } %res
+}
+
 define { <vscale x 4 x i32>, <vscale x 4 x i32> } @ldnt1_x2_i32(target("aarch64.svcount") %pn, ptr %ptr) nounwind {
 ; CHECK-LABEL: ldnt1_x2_i32:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
 ; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    ldnt1w { z0.s, z1.s }, pn8/z, [x0]
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    addvl sp, sp, #1
 ; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
 ; CHECK-NEXT:    ret
   %res = call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ldnt1.pn.x2.nxv4i32(target("aarch64.svcount") %pn, ptr %ptr);
   ret { <vscale x 4 x i32>, <vscale x 4 x i32> } %res
 }
 
+define { <vscale x 4 x i32>, <vscale x 4 x i32> } @ldnt1_x2_i32_scalar(target("aarch64.svcount") %pn, ptr %ptr, i64 %index) nounwind {
+; CHECK-LABEL: ldnt1_x2_i32_scalar:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
+; CHECK-NEXT:    addvl sp, sp, #-1
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
+; CHECK-NEXT:    mov p8.b, p0.b
+; CHECK-NEXT:    ldnt1w { z0.s, z1.s }, pn8/z, [x0, x1, lsl #2]
+; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
+; CHECK-NEXT:    addvl sp, sp, #1
+; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
+; CHECK-NEXT:    ret
+  %base = getelementptr i32, ptr %ptr, i64 %index
+  %res = call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ldnt1.pn.x2.nxv4i32(target("aarch64.svcount") %pn, ptr %base);
+  ret { <vscale x 4 x i32>, <vscale x 4 x i32> } %res
+}
+
 define { <vscale x 2 x i64>, <vscale x 2 x i64> } @ldnt1_x2_i64(target("aarch64.svcount") %pn, ptr %ptr) nounwind {
 ; CHECK-LABEL: ldnt1_x2_i64:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
 ; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    ldnt1d { z0.d, z1.d }, pn8/z, [x0]
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    addvl sp, sp, #1
 ; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
 ; CHECK-NEXT:    ret
   %res = call { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ldnt1.pn.x2.nxv2i64(target("aarch64.svcount") %pn, ptr %ptr);
   ret { <vscale x 2 x i64>, <vscale x 2 x i64> } %res
 }
 
+define { <vscale x 2 x i64>, <vscale x 2 x i64> } @ldnt1_x2_i64_scalar(target("aarch64.svcount") %pn, ptr %ptr, i64 %index) nounwind {
+; CHECK-LABEL: ldnt1_x2_i64_scalar:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
+; CHECK-NEXT:    addvl sp, sp, #-1
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
+; CHECK-NEXT:    mov p8.b, p0.b
+; CHECK-NEXT:    ldnt1d { z0.d, z1.d }, pn8/z, [x0, x1, lsl #3]
+; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
+; CHECK-NEXT:    addvl sp, sp, #1
+; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
+; CHECK-NEXT:    ret
+  %base = getelementptr i64, ptr %ptr, i64 %index
+  %res = call { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ldnt1.pn.x2.nxv2i64(target("aarch64.svcount") %pn, ptr %base);
+  ret { <vscale x 2 x i64>, <vscale x 2 x i64> } %res
+}
+
 define { <vscale x 8 x half>, <vscale x 8 x half> } @ldnt1_x2_f16(target("aarch64.svcount") %pn, ptr %ptr) nounwind {
 ; CHECK-LABEL: ldnt1_x2_f16:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
 ; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    ldnt1h { z0.h, z1.h }, pn8/z, [x0]
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    addvl sp, sp, #1
 ; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
 ; CHECK-NEXT:    ret
   %res = call { <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.ldnt1.pn.x2.nxv8f16(target("aarch64.svcount") %pn, ptr %ptr);
   ret { <vscale x 8 x half>, <vscale x 8 x half> } %res
 }
 
+define { <vscale x 8 x half>, <vscale x 8 x half> } @ldnt1_x2_f16_scalar(target("aarch64.svcount") %pn, ptr %ptr, i64 %index) nounwind {
+; CHECK-LABEL: ldnt1_x2_f16_scalar:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
+; CHECK-NEXT:    addvl sp, sp, #-1
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
+; CHECK-NEXT:    mov p8.b, p0.b
+; CHECK-NEXT:    ldnt1h { z0.h, z1.h }, pn8/z, [x0, x1, lsl #1]
+; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
+; CHECK-NEXT:    addvl sp, sp, #1
+; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
+; CHECK-NEXT:    ret
+  %base = getelementptr i16, ptr %ptr, i64 %index
+  %res = call { <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.ldnt1.pn.x2.nxv8f16(target("aarch64.svcount") %pn, ptr %base);
+  ret { <vscale x 8 x half>, <vscale x 8 x half> } %res
+}
+
 define { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @ldnt1_x2_bf16(target("aarch64.svcount") %pn, ptr %ptr) nounwind {
 ; CHECK-LABEL: ldnt1_x2_bf16:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
 ; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    ldnt1h { z0.h, z1.h }, pn8/z, [x0]
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    addvl sp, sp, #1
 ; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
 ; CHECK-NEXT:    ret
   %res = call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.ldnt1.pn.x2.nxv8bf16(target("aarch64.svcount") %pn, ptr %ptr);
   ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } %res
 }
 
+define { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @ldnt1_x2_bf16_scalar(target("aarch64.svcount") %pn, ptr %ptr, i64 %index) nounwind {
+; CHECK-LABEL: ldnt1_x2_bf16_scalar:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
+; CHECK-NEXT:    addvl sp, sp, #-1
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
+; CHECK-NEXT:    mov p8.b, p0.b
+; CHECK-NEXT:    ldnt1h { z0.h, z1.h }, pn8/z, [x0, x1, lsl #1]
+; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
+; CHECK-NEXT:    addvl sp, sp, #1
+; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
+; CHECK-NEXT:    ret
+  %base = getelementptr bfloat, ptr %ptr, i64 %index
+  %res = call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.ldnt1.pn.x2.nxv8bf16(target("aarch64.svcount") %pn, ptr %base);
+  ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } %res
+}
+
 define { <vscale x 4 x float>, <vscale x 4 x float> } @ldnt1_x2_f32(target("aarch64.svcount") %pn, ptr %ptr) nounwind {
 ; CHECK-LABEL: ldnt1_x2_f32:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
 ; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    ldnt1w { z0.s, z1.s }, pn8/z, [x0]
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    addvl sp, sp, #1
 ; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
 ; CHECK-NEXT:    ret
   %res = call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.ldnt1.pn.x2.nxv4f32(target("aarch64.svcount") %pn, ptr %ptr);
   ret { <vscale x 4 x float>, <vscale x 4 x float> } %res
 }
 
+define { <vscale x 4 x float>, <vscale x 4 x float> } @ldnt1_x2_f32_scalar(target("aarch64.svcount") %pn, ptr %ptr, i64 %index) nounwind {
+; CHECK-LABEL: ldnt1_x2_f32_scalar:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
+; CHECK-NEXT:    addvl sp, sp, #-1
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
+; CHECK-NEXT:    mov p8.b, p0.b
+; CHECK-NEXT:    ldnt1w { z0.s, z1.s }, pn8/z, [x0, x1, lsl #2]
+; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
+; CHECK-NEXT:    addvl sp, sp, #1
+; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
+; CHECK-NEXT:    ret
+  %base = getelementptr float, ptr %ptr, i64 %index
+  %res = call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.ldnt1.pn.x2.nxv4f32(target("aarch64.svcount") %pn, ptr %base);
+  ret { <vscale x 4 x float>, <vscale x 4 x float> } %res
+}
+
 define { <vscale x 2 x double>, <vscale x 2 x double> } @ldnt1_x2_f64(target("aarch64.svcount") %pn, ptr %ptr) nounwind {
 ; CHECK-LABEL: ldnt1_x2_f64:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
 ; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    ldnt1d { z0.d, z1.d }, pn8/z, [x0]
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    addvl sp, sp, #1
 ; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
 ; CHECK-NEXT:    ret
   %res = call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.ldnt1.pn.x2.nxv2f64(target("aarch64.svcount") %pn, ptr %ptr);
   ret { <vscale x 2 x double>, <vscale x 2 x double> } %res
 }
 
+define { <vscale x 2 x double>, <vscale x 2 x double> } @ldnt1_x2_f64_scalar(target("aarch64.svcount") %pn, ptr %ptr, i64 %index) nounwind {
+; CHECK-LABEL: ldnt1_x2_f64_scalar:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
+; CHECK-NEXT:    addvl sp, sp, #-1
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
+; CHECK-NEXT:    mov p8.b, p0.b
+; CHECK-NEXT:    ldnt1d { z0.d, z1.d }, pn8/z, [x0, x1, lsl #3]
+; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
+; CHECK-NEXT:    addvl sp, sp, #1
+; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
+; CHECK-NEXT:    ret
+  %base = getelementptr double, ptr %ptr, i64 %index
+  %res = call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.ldnt1.pn.x2.nxv2f64(target("aarch64.svcount") %pn, ptr %base);
+  ret { <vscale x 2 x double>, <vscale x 2 x double> } %res
+}
+
 ; Test to ensure we load into the correct registers for the instruction
 define <vscale x 4 x i32> @ldnt1_x2_i32_z0_taken(target("aarch64.svcount") %pn, ptr %ptr, <vscale x 4 x i32> %val) {
 ; CHECK-LABEL: ldnt1_x2_i32_z0_taken:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
 ; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    .cfi_escape 0x0f, 0x0c, 0x8f, 0x00, 0x11, 0x10, 0x22, 0x11, 0x08, 0x92, 0x2e, 0x00, 0x1e, 0x22 // sp + 16 + 8 * VG
 ; CHECK-NEXT:    .cfi_offset w29, -16
 ; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    ldnt1w { z2.s, z3.s }, pn8/z, [x0]
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    add z0.s, z0.s, z2.s
 ; CHECK-NEXT:    addvl sp, sp, #1
 ; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
 ; CHECK-NEXT:    ret
   %ld1 = call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ldnt1.pn.x2.nxv4i32(target("aarch64.svcount") %pn, ptr %ptr);
   %ld1_0 = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } %ld1, 0
   %res = add <vscale x 4 x i32> %val, %ld1_0
   ret <vscale x 4 x i32>  %res
 }
 
+; Test to ensure we load into the correct registers for the instruction
+define <vscale x 4 x i32> @ldnt1_x2_i32_z0_taken_scalar(target("aarch64.svcount") %pn, ptr %ptr, <vscale x 4 x i32> %val, i64 %index) {
+; CHECK-LABEL: ldnt1_x2_i32_z0_taken_scalar:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
+; CHECK-NEXT:    addvl sp, sp, #-1
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
+; CHECK-NEXT:    .cfi_escape 0x0f, 0x0c, 0x8f, 0x00, 0x11, 0x10, 0x22, 0x11, 0x08, 0x92, 0x2e, 0x00, 0x1e, 0x22 // sp + 16 + 8 * VG
+; CHECK-NEXT:    .cfi_offset w29, -16
+; CHECK-NEXT:    mov p8.b, p0.b
+; CHECK-NEXT:    ldnt1w { z2.s, z3.s }, pn8/z, [x0, x1, lsl #2]
+; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
+; CHECK-NEXT:    add z0.s, z0.s, z2.s
+; CHECK-NEXT:    addvl sp, sp, #1
+; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
+; CHECK-NEXT:    ret
+  %base = getelementptr i32, ptr %ptr, i64 %index
+  %ld1 = call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ldnt1.pn.x2.nxv4i32(target("aarch64.svcount") %pn, ptr %base);
+  %ld1_0 = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } %ld1, 0
+  %res = add <vscale x 4 x i32> %val, %ld1_0
+  ret <vscale x 4 x i32>  %res
+}
+
 define { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @ldnt1_x4_i8(target("aarch64.svcount") %pn, ptr %ptr) nounwind {
 ; CHECK-LABEL: ldnt1_x4_i8:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
 ; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    ldnt1b { z0.b - z3.b }, pn8/z, [x0]
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    addvl sp, sp, #1
 ; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
 ; CHECK-NEXT:    ret
   %res = call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ldnt1.pn.x4.nxv16i8(target("aarch64.svcount") %pn, ptr %ptr);
   ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %res
 }
 
+define { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @ldnt1_x4_i8_scalar(target("aarch64.svcount") %pn, ptr %ptr, i64 %index) nounwind {
+; CHECK-LABEL: ldnt1_x4_i8_scalar:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
+; CHECK-NEXT:    addvl sp, sp, #-1
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
+; CHECK-NEXT:    mov p8.b, p0.b
+; CHECK-NEXT:    ldnt1b { z0.b - z3.b }, pn8/z, [x0, x1]
+; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
+; CHECK-NEXT:    addvl sp, sp, #1
+; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
+; CHECK-NEXT:    ret
+  %base = getelementptr i8, ptr %ptr, i64 %index
+  %res = call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ldnt1.pn.x4.nxv16i8(target("aarch64.svcount") %pn, ptr %base);
+  ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %res
+}
+
 define { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @ldnt1_x4_i16(target("aarch64.svcount") %pn, ptr %ptr) nounwind {
 ; CHECK-LABEL: ldnt1_x4_i16:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
 ; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    ldnt1h { z0.h - z3.h }, pn8/z, [x0]
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    addvl sp, sp, #1
 ; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
 ; CHECK-NEXT:    ret
   %res = call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ldnt1.pn.x4.nxv8i16(target("aarch64.svcount") %pn, ptr %ptr);
   ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %res
 }
 
+define { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @ldnt1_x4_i16_scalar(target("aarch64.svcount") %pn, ptr %ptr, i64 %index) nounwind {
+; CHECK-LABEL: ldnt1_x4_i16_scalar:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
+; CHECK-NEXT:    addvl sp, sp, #-1
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
+; CHECK-NEXT:    mov p8.b, p0.b
+; CHECK-NEXT:    ldnt1h { z0.h - z3.h }, pn8/z, [x0, x1, lsl #1]
+; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
+; CHECK-NEXT:    addvl sp, sp, #1
+; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
+; CHECK-NEXT:    ret
+  %base = getelementptr i16, ptr %ptr, i64 %index
+  %res = call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ldnt1.pn.x4.nxv8i16(target("aarch64.svcount") %pn, ptr %base);
+  ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %res
+}
+
 define { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @ldnt1_x4_i32(target("aarch64.svcount") %pn, ptr %ptr) nounwind {
 ; CHECK-LABEL: ldnt1_x4_i32:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
 ; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    ldnt1w { z0.s - z3.s }, pn8/z, [x0]
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    addvl sp, sp, #1
 ; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
 ; CHECK-NEXT:    ret
   %res = call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ldnt1.pn.x4.nxv4i32(target("aarch64.svcount") %pn, ptr %ptr);
   ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } %res
 }
 
+define { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @ldnt1_x4_i32_scalar(target("aarch64.svcount") %pn, ptr %ptr, i64 %index) nounwind {
+; CHECK-LABEL: ldnt1_x4_i32_scalar:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
+; CHECK-NEXT:    addvl sp, sp, #-1
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
+; CHECK-NEXT:    mov p8.b, p0.b
+; CHECK-NEXT:    ldnt1w { z0.s - z3.s }, pn8/z, [x0, x1, lsl #2]
+; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
+; CHECK-NEXT:    addvl sp, sp, #1
+; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
+; CHECK-NEXT:    ret
+  %base = getelementptr i32, ptr %ptr, i64 %index
+  %res = call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ldnt1.pn.x4.nxv4i32(target("aarch64.svcount") %pn, ptr %base);
+  ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } %res
+}
+
 define { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @ldnt1_x4_i64(target("aarch64.svcount") %pn, ptr %ptr) nounwind {
 ; CHECK-LABEL: ldnt1_x4_i64:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
 ; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    ldnt1d { z0.d - z3.d }, pn8/z, [x0]
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    addvl sp, sp, #1
 ; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
 ; CHECK-NEXT:    ret
   %res = call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ldnt1.pn.x4.nxv2i64(target("aarch64.svcount") %pn, ptr %ptr);
   ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } %res
 }
 
+define { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @ldnt1_x4_i64_scalar(target("aarch64.svcount") %pn, ptr %ptr, i64 %index) nounwind {
+; CHECK-LABEL: ldnt1_x4_i64_scalar:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
+; CHECK-NEXT:    addvl sp, sp, #-1
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
+; CHECK-NEXT:    mov p8.b, p0.b
+; CHECK-NEXT:    ldnt1d { z0.d - z3.d }, pn8/z, [x0, x1, lsl #3]
+; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
+; CHECK-NEXT:    addvl sp, sp, #1
+; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
+; CHECK-NEXT:    ret
+  %base = getelementptr i64, ptr %ptr, i64 %index
+  %res = call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ldnt1.pn.x4.nxv2i64(target("aarch64.svcount") %pn, ptr %base);
+  ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } %res
+}
+
 define { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @ldnt1_x4_f16(target("aarch64.svcount") %pn, ptr %ptr) nounwind {
 ; CHECK-LABEL: ldnt1_x4_f16:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
 ; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    ldnt1h { z0.h - z3.h }, pn8/z, [x0]
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    addvl sp, sp, #1
 ; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
 ; CHECK-NEXT:    ret
   %res = call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.ldnt1.pn.x4.nxv8f16(target("aarch64.svcount") %pn, ptr %ptr);
   ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } %res
 }
 
+define { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @ldnt1_x4_f16_scalar(target("aarch64.svcount") %pn, ptr %ptr, i64 %index) nounwind {
+; CHECK-LABEL: ldnt1_x4_f16_scalar:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
+; CHECK-NEXT:    addvl sp, sp, #-1
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
+; CHECK-NEXT:    mov p8.b, p0.b
+; CHECK-NEXT:    ldnt1h { z0.h - z3.h }, pn8/z, [x0, x1, lsl #1]
+; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
+; CHECK-NEXT:    addvl sp, sp, #1
+; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
+; CHECK-NEXT:    ret
+  %base = getelementptr half, ptr %ptr, i64 %index
+  %res = call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.ldnt1.pn.x4.nxv8f16(target("aarch64.svcount") %pn, ptr %base);
+  ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } %res
+}
+
 define { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @ldnt1_x4_bf16(target("aarch64.svcount") %pn, ptr %ptr) nounwind {
 ; CHECK-LABEL: ldnt1_x4_bf16:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
 ; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    ldnt1h { z0.h - z3.h }, pn8/z, [x0]
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    addvl sp, sp, #1
 ; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
 ; CHECK-NEXT:    ret
   %res = call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.ldnt1.pn.x4.nxv8bf16(target("aarch64.svcount") %pn, ptr %ptr);
   ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } %res
 }
 
+define { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @ldnt1_x4_bf16_scalar(target("aarch64.svcount") %pn, ptr %ptr, i64 %index) nounwind {
+; CHECK-LABEL: ldnt1_x4_bf16_scalar:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
+; CHECK-NEXT:    addvl sp, sp, #-1
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
+; CHECK-NEXT:    mov p8.b, p0.b
+; CHECK-NEXT:    ldnt1h { z0.h - z3.h }, pn8/z, [x0, x1, lsl #1]
+; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
+; CHECK-NEXT:    addvl sp, sp, #1
+; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
+; CHECK-NEXT:    ret
+  %base = getelementptr bfloat, ptr %ptr, i64 %index
+  %res = call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.ldnt1.pn.x4.nxv8bf16(target("aarch64.svcount") %pn, ptr %base);
+  ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } %res
+}
+
 define { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @ldnt1_x4_f32(target("aarch64.svcount") %pn, ptr %ptr) nounwind {
 ; CHECK-LABEL: ldnt1_x4_f32:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
 ; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    ldnt1w { z0.s - z3.s }, pn8/z, [x0]
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    addvl sp, sp, #1
 ; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
 ; CHECK-NEXT:    ret
   %res = call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.ldnt1.pn.x4.nxv4f32(target("aarch64.svcount") %pn, ptr %ptr);
   ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } %res
 }
 
+define { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @ldnt1_x4_f32_scalar(target("aarch64.svcount") %pn, ptr %ptr, i64 %index) nounwind {
+; CHECK-LABEL: ldnt1_x4_f32_scalar:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
+; CHECK-NEXT:    addvl sp, sp, #-1
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
+; CHECK-NEXT:    mov p8.b, p0.b
+; CHECK-NEXT:    ldnt1w { z0.s - z3.s }, pn8/z, [x0, x1, lsl #2]
+; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
+; CHECK-NEXT:    addvl sp, sp, #1
+; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
+; CHECK-NEXT:    ret
+  %base = getelementptr float, ptr %ptr, i64 %index
+  %res = call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.ldnt1.pn.x4.nxv4f32(target("aarch64.svcount") %pn, ptr %base);
+  ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } %res
+}
+
 define { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @ldnt1_x4_f64(target("aarch64.svcount") %pn, ptr %ptr) nounwind {
 ; CHECK-LABEL: ldnt1_x4_f64:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
 ; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    ldnt1d { z0.d - z3.d }, pn8/z, [x0]
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    addvl sp, sp, #1
 ; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
 ; CHECK-NEXT:    ret
   %res = call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.ldnt1.pn.x4.nxv2f64(target("aarch64.svcount") %pn, ptr %ptr);
   ret { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } %res
 }
 
+define { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @ldnt1_x4_f64_scalar(target("aarch64.svcount") %pn, ptr %ptr, i64 %index) nounwind {
+; CHECK-LABEL: ldnt1_x4_f64_scalar:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
+; CHECK-NEXT:    addvl sp, sp, #-1
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
+; CHECK-NEXT:    mov p8.b, p0.b
+; CHECK-NEXT:    ldnt1d { z0.d - z3.d }, pn8/z, [x0, x1, lsl #3]
+; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
+; CHECK-NEXT:    addvl sp, sp, #1
+; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
+; CHECK-NEXT:    ret
+  %base = getelementptr double, ptr %ptr, i64 %index
+  %res = call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.ldnt1.pn.x4.nxv2f64(target("aarch64.svcount") %pn, ptr %base);
+  ret { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } %res
+}
+
 ; Test to ensure we load into the correct registers for the instruction
 define <vscale x 2 x i64> @ldnt1_x4_i64_z0_taken(target("aarch64.svcount") %pn, ptr %ptr, <vscale x 2 x i64> %val) {
 ; CHECK-LABEL: ldnt1_x4_i64_z0_taken:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-NEXT:    addvl sp, sp, #-1
 ; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
 ; CHECK-NEXT:    .cfi_escape 0x0f, 0x0c, 0x8f, 0x00, 0x11, 0x10, 0x22, 0x11, 0x08, 0x92, 0x2e, 0x00, 0x1e, 0x22 // sp + 16 + 8 * VG
 ; CHECK-NEXT:    .cfi_offset w29, -16
 ; CHECK-NEXT:    mov p8.b, p0.b
 ; CHECK-NEXT:    ldnt1d { z4.d - z7.d }, pn8/z, [x0]
 ; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
 ; CHECK-NEXT:    add z0.d, z0.d, z4.d
 ; CHECK-NEXT:    addvl sp, sp, #1
 ; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
 ; CHECK-NEXT:    ret
   %ld1 = call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ldnt1.pn.x4.nxv2i64(target("aarch64.svcount") %pn, ptr %ptr);
   %ld1_0 = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } %ld1, 0
   %res = add <vscale x 2 x i64> %val, %ld1_0
   ret <vscale x 2 x i64>  %res
 }
 
+; Test to ensure we load into the correct registers for the instruction
+define <vscale x 2 x i64> @ldnt1_x4_i64_z0_taken_scalar(target("aarch64.svcount") %pn, ptr %ptr, <vscale x 2 x i64> %val, i64 %index) {
+; CHECK-LABEL: ldnt1_x4_i64_z0_taken_scalar:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
+; CHECK-NEXT:    addvl sp, sp, #-1
+; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
+; CHECK-NEXT:    .cfi_escape 0x0f, 0x0c, 0x8f, 0x00, 0x11, 0x10, 0x22, 0x11, 0x08, 0x92, 0x2e, 0x00, 0x1e, 0x22 // sp + 16 + 8 * VG
+; CHECK-NEXT:    .cfi_offset w29, -16
+; CHECK-NEXT:    mov p8.b, p0.b
+; CHECK-NEXT:    ldnt1d { z4.d - z7.d }, pn8/z, [x0, x1, lsl #3]
+; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
+; CHECK-NEXT:    add z0.d, z0.d, z4.d
+; CHECK-NEXT:    addvl sp, sp, #1
+; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
+; CHECK-NEXT:    ret
+  %base = getelementptr i64, ptr %ptr, i64 %index
+  %ld1 = call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ldnt1.pn.x4.nxv2i64(target("aarch64.svcount") %pn, ptr %base);
+  %ld1_0 = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } %ld1, 0
+  %res = add <vscale x 2 x i64> %val, %ld1_0
+  ret <vscale x 2 x i64>  %res
+}
+
 declare { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld1.pn.x2.nxv2i64(target("aarch64.svcount"), ptr)
 declare { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld1.pn.x2.nxv4i32(target("aarch64.svcount"), ptr)
 declare { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld1.pn.x2.nxv8i16(target("aarch64.svcount"), ptr)
 declare { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld1.pn.x2.nxv16i8(target("aarch64.svcount"), ptr)
 declare { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.ld1.pn.x2.nxv2f64(target("aarch64.svcount"), ptr)
 declare { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.ld1.pn.x2.nxv4f32(target("aarch64.svcount"), ptr)
 declare { <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.ld1.pn.x2.nxv8f16(target("aarch64.svcount"), ptr)
 declare { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.ld1.pn.x2.nxv8bf16(target("aarch64.svcount"), ptr)