; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -force-streaming -enable-subreg-liveness -verify-machineinstrs < %s | FileCheck %s

target triple="aarch64-linux-gnu"


; == Multi, multi (unsigned) ==

define void @udot_multi_za32_u16_vg1x2(i32 %slice, <vscale x 16 x i8> %unused, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, <vscale x 8 x i16> %zn3) #0 {
; CHECK-LABEL: udot_multi_za32_u16_vg1x2:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z5.d, z4.d
; CHECK-NEXT:    mov z7.d, z2.d
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    mov z4.d, z3.d
; CHECK-NEXT:    mov z6.d, z1.d
; CHECK-NEXT:    udot za.s[w8, 0, vgx2], { z6.h, z7.h }, { z4.h, z5.h }
; CHECK-NEXT:    udot za.s[w8, 7, vgx2], { z6.h, z7.h }, { z4.h, z5.h }
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.udot.za32.vg1x2.nxv8i16(i32 %slice, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, <vscale x 8 x i16> %zn3)
  %slice2 = add i32 %slice, 7
  call void @llvm.aarch64.sme.udot.za32.vg1x2.nxv8i16(i32 %slice2, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, <vscale x 8 x i16> %zn3)
  ret void
}

define void @udot_multi_za32_u16_vg1x2_tuple(i64 %stride, ptr %ptr) #1 {
; CHECK-LABEL: udot_multi_za32_u16_vg1x2_tuple:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    ptrue pn8.b
; CHECK-NEXT:    mov w8, wzr
; CHECK-NEXT:    ld1b { z16.b, z24.b }, pn8/z, [x1]
; CHECK-NEXT:    ld1b { z17.b, z25.b }, pn8/z, [x1, x0]
; CHECK-NEXT:    udot za.s[w8, 0, vgx2], { z16.b, z17.b }, { z24.b, z25.b }
; CHECK-NEXT:    ret
entry:
  %0 = tail call target("aarch64.svcount") @llvm.aarch64.sve.ptrue.c8()
  %1 = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld1.pn.x2.nxv16i8(target("aarch64.svcount") %0, ptr %ptr)
  %2 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 0
  %3 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 1
  %arrayidx2 = getelementptr inbounds i8, ptr %ptr, i64 %stride
  %4 = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld1.pn.x2.nxv16i8(target("aarch64.svcount") %0, ptr %arrayidx2)
  %5 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %4, 0
  %6 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %4, 1
  call void @llvm.aarch64.sme.udot.za32.vg1x2.nxv16i8(i32 0, <vscale x 16 x i8> %2, <vscale x 16 x i8> %5, <vscale x 16 x i8> %3, <vscale x 16 x i8> %6)
  ret void
}

define void @udot_multi_za32_u16_vg1x4(i32 %slice, <vscale x 16 x i8> %unused, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, <vscale x 8 x i16> %zn3,
; CHECK-LABEL: udot_multi_za32_u16_vg1x4:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z26.d, z7.d
; CHECK-NEXT:    mov z25.d, z6.d
; CHECK-NEXT:    ldr z27, [x1]
; CHECK-NEXT:    mov z7.d, z4.d
; CHECK-NEXT:    mov z24.d, z5.d
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    mov z6.d, z3.d
; CHECK-NEXT:    mov z5.d, z2.d
; CHECK-NEXT:    mov z4.d, z1.d
; CHECK-NEXT:    udot za.s[w8, 0, vgx4], { z4.h - z7.h }, { z24.h - z27.h }
; CHECK-NEXT:    udot za.s[w8, 7, vgx4], { z4.h - z7.h }, { z24.h - z27.h }
; CHECK-NEXT:    ret
                                        <vscale x 8 x i16> %zn4, <vscale x 8 x i16> %zn5, <vscale x 8 x i16> %zn6, <vscale x 8 x i16> %zn7) #0 {
  call void @llvm.aarch64.sme.udot.za32.vg1x4.nxv8i16(i32 %slice, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, <vscale x 8 x i16> %zn3,
                                                      <vscale x 8 x i16> %zn4, <vscale x 8 x i16> %zn5, <vscale x 8 x i16> %zn6, <vscale x 8 x i16> %zn7)
  %slice2 = add i32 %slice, 7
  call void @llvm.aarch64.sme.udot.za32.vg1x4.nxv8i16(i32 %slice2, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, <vscale x 8 x i16> %zn3,
                                                      <vscale x 8 x i16> %zn4, <vscale x 8 x i16> %zn5, <vscale x 8 x i16> %zn6, <vscale x 8 x i16> %zn7)
  ret void
}

define void @udot_multi_za32_u16_vg1x4_tuple(i64 %stride, ptr %ptr) #1 {
; CHECK-LABEL: udot_multi_za32_u16_vg1x4_tuple:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    lsl x9, x0, #1
; CHECK-NEXT:    ptrue pn8.b
; CHECK-NEXT:    mov w8, wzr
; CHECK-NEXT:    ld1b { z16.b, z20.b, z24.b, z28.b }, pn8/z, [x1]
; CHECK-NEXT:    ld1b { z17.b, z21.b, z25.b, z29.b }, pn8/z, [x1, x0]
; CHECK-NEXT:    add x10, x9, x0
; CHECK-NEXT:    ld1b { z18.b, z22.b, z26.b, z30.b }, pn8/z, [x1, x9]
; CHECK-NEXT:    ld1b { z19.b, z23.b, z27.b, z31.b }, pn8/z, [x1, x10]
; CHECK-NEXT:    udot za.s[w8, 0, vgx4], { z16.b - z19.b }, { z20.b - z23.b }
; CHECK-NEXT:    udot za.s[w8, 0, vgx4], { z24.b - z27.b }, { z28.b - z31.b }
; CHECK-NEXT:    ret
entry:
  %0 = tail call target("aarch64.svcount") @llvm.aarch64.sve.ptrue.c8()
  %1 = tail 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") %0, ptr %ptr)
  %2 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 0
  %3 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 1
  %4 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 2
  %5 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 3
  %arrayidx2 = getelementptr inbounds i8, ptr %ptr, i64 %stride
  %6 = tail 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") %0, ptr %arrayidx2)
  %7 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %6, 0
  %8 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %6, 1
  %9 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %6, 2
  %10 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %6, 3
  %mul3 = shl i64 %stride, 1
  %arrayidx4 = getelementptr inbounds i8, ptr %ptr, i64 %mul3
  %11 = tail 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") %0, ptr %arrayidx4)
  %12 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %11, 0
  %13 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %11, 1
  %14 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %11, 2
  %15 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %11, 3
  %mul5 = mul i64 %stride, 3
  %arrayidx6 = getelementptr inbounds i8, ptr %ptr, i64 %mul5
  %16 = tail 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") %0, ptr %arrayidx6)
  %17 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %16, 0
  %18 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %16, 1
  %19 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %16, 2
  %20 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %16, 3
  call void @llvm.aarch64.sme.udot.za32.vg1x4.nxv16i8(i32 0, <vscale x 16 x i8> %2, <vscale x 16 x i8> %7, <vscale x 16 x i8> %12, <vscale x 16 x i8> %17,
                                                      <vscale x 16 x i8> %3, <vscale x 16 x i8> %8, <vscale x 16 x i8> %13, <vscale x 16 x i8> %18)
  call void @llvm.aarch64.sme.udot.za32.vg1x4.nxv16i8(i32 0, <vscale x 16 x i8> %4, <vscale x 16 x i8> %9, <vscale x 16 x i8> %14, <vscale x 16 x i8> %19,
                                                      <vscale x 16 x i8> %5, <vscale x 16 x i8> %10, <vscale x 16 x i8> %15, <vscale x 16 x i8> %20)
  ret void
}

define void @udot_multi_za32_u8_vg1x2(i32 %slice, <vscale x 16 x i8> %unused, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2, <vscale x 16 x i8> %zn3) #0 {
; CHECK-LABEL: udot_multi_za32_u8_vg1x2:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z5.d, z4.d
; CHECK-NEXT:    mov z7.d, z2.d
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    mov z4.d, z3.d
; CHECK-NEXT:    mov z6.d, z1.d
; CHECK-NEXT:    udot za.s[w8, 0, vgx2], { z6.b, z7.b }, { z4.b, z5.b }
; CHECK-NEXT:    udot za.s[w8, 7, vgx2], { z6.b, z7.b }, { z4.b, z5.b }
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.udot.za32.vg1x2.nxv16i8(i32 %slice, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2, <vscale x 16 x i8> %zn3)
  %slice2 = add i32 %slice, 7
  call void @llvm.aarch64.sme.udot.za32.vg1x2.nxv16i8(i32 %slice2, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2, <vscale x 16 x i8> %zn3)
  ret void
}

define void @udot_multi_za32_u8_vg1x4(i32 %slice, <vscale x 16 x i8> %unused, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2, <vscale x 16 x i8> %zn3,
; CHECK-LABEL: udot_multi_za32_u8_vg1x4:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z26.d, z7.d
; CHECK-NEXT:    mov z25.d, z6.d
; CHECK-NEXT:    ldr z27, [x1]
; CHECK-NEXT:    mov z7.d, z4.d
; CHECK-NEXT:    mov z24.d, z5.d
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    mov z6.d, z3.d
; CHECK-NEXT:    mov z5.d, z2.d
; CHECK-NEXT:    mov z4.d, z1.d
; CHECK-NEXT:    udot za.s[w8, 0, vgx4], { z4.b - z7.b }, { z24.b - z27.b }
; CHECK-NEXT:    udot za.s[w8, 7, vgx4], { z4.b - z7.b }, { z24.b - z27.b }
; CHECK-NEXT:    ret
                                      <vscale x 16 x i8> %zn4, <vscale x 16 x i8> %zn5, <vscale x 16 x i8> %zn6, <vscale x 16 x i8> %zn7) #0 {
  call void @llvm.aarch64.sme.udot.za32.vg1x4.nxv16i8(i32 %slice, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2, <vscale x 16 x i8> %zn3,
                                                      <vscale x 16 x i8> %zn4, <vscale x 16 x i8> %zn5, <vscale x 16 x i8> %zn6, <vscale x 16 x i8> %zn7)
  %slice2 = add i32 %slice, 7
  call void @llvm.aarch64.sme.udot.za32.vg1x4.nxv16i8(i32 %slice2, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2, <vscale x 16 x i8> %zn3,
                                                      <vscale x 16 x i8> %zn4, <vscale x 16 x i8> %zn5, <vscale x 16 x i8> %zn6, <vscale x 16 x i8> %zn7)
  ret void
}

define void @udot_multi_za64_u16_vg1x2(i32 %slice, <vscale x 16 x i8> %unused, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, <vscale x 8 x i16> %zn3) #1 {
; CHECK-LABEL: udot_multi_za64_u16_vg1x2:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z5.d, z4.d
; CHECK-NEXT:    mov z7.d, z2.d
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    mov z4.d, z3.d
; CHECK-NEXT:    mov z6.d, z1.d
; CHECK-NEXT:    udot za.d[w8, 0, vgx2], { z6.h, z7.h }, { z4.h, z5.h }
; CHECK-NEXT:    udot za.d[w8, 7, vgx2], { z6.h, z7.h }, { z4.h, z5.h }
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.udot.za64.vg1x2.nxv8i16(i32 %slice, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, <vscale x 8 x i16> %zn3)
  %slice2 = add i32 %slice, 7
  call void @llvm.aarch64.sme.udot.za64.vg1x2.nxv8i16(i32 %slice2, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, <vscale x 8 x i16> %zn3)
  ret void
}

define void @udot_multi_za64_u16_vg1x4(i32 %slice, <vscale x 16 x i8> %unused, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, <vscale x 8 x i16> %zn3,
; CHECK-LABEL: udot_multi_za64_u16_vg1x4:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z26.d, z7.d
; CHECK-NEXT:    mov z25.d, z6.d
; CHECK-NEXT:    ldr z27, [x1]
; CHECK-NEXT:    mov z7.d, z4.d
; CHECK-NEXT:    mov z24.d, z5.d
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    mov z6.d, z3.d
; CHECK-NEXT:    mov z5.d, z2.d
; CHECK-NEXT:    mov z4.d, z1.d
; CHECK-NEXT:    udot za.d[w8, 0, vgx4], { z4.h - z7.h }, { z24.h - z27.h }
; CHECK-NEXT:    udot za.d[w8, 7, vgx4], { z4.h - z7.h }, { z24.h - z27.h }
; CHECK-NEXT:    ret
                                       <vscale x 8 x i16> %zn4, <vscale x 8 x i16> %zn5, <vscale x 8 x i16> %zn6, <vscale x 8 x i16> %zn7) #1 {
  call void @llvm.aarch64.sme.udot.za64.vg1x4.nxv8i16(i32 %slice, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, <vscale x 8 x i16> %zn3,
                                                      <vscale x 8 x i16> %zn4, <vscale x 8 x i16> %zn5, <vscale x 8 x i16> %zn6, <vscale x 8 x i16> %zn7)
  %slice2 = add i32 %slice, 7
  call void @llvm.aarch64.sme.udot.za64.vg1x4.nxv8i16(i32 %slice2, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, <vscale x 8 x i16> %zn3,
                                                      <vscale x 8 x i16> %zn4, <vscale x 8 x i16> %zn5, <vscale x 8 x i16> %zn6, <vscale x 8 x i16> %zn7)
  ret void
}

define void @usdot_multi_za32_u8_vg1x2(i32 %slice, <vscale x 16 x i8> %unused, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2, <vscale x 16 x i8> %zn3) #0 {
; CHECK-LABEL: usdot_multi_za32_u8_vg1x2:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z5.d, z4.d
; CHECK-NEXT:    mov z7.d, z2.d
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    mov z4.d, z3.d
; CHECK-NEXT:    mov z6.d, z1.d
; CHECK-NEXT:    usdot za.s[w8, 0, vgx2], { z6.b, z7.b }, { z4.b, z5.b }
; CHECK-NEXT:    usdot za.s[w8, 7, vgx2], { z6.b, z7.b }, { z4.b, z5.b }
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.usdot.za32.vg1x2.nxv16i8(i32 %slice, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2, <vscale x 16 x i8> %zn3)
  %slice2 = add i32 %slice, 7
  call void @llvm.aarch64.sme.usdot.za32.vg1x2.nxv16i8(i32 %slice2, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2, <vscale x 16 x i8> %zn3)
  ret void
}

define void @usdot_multi_za32_u16_vg1x2_tuple(i64 %stride, ptr %ptr) #1 {
; CHECK-LABEL: usdot_multi_za32_u16_vg1x2_tuple:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    ptrue pn8.b
; CHECK-NEXT:    mov w8, wzr
; CHECK-NEXT:    ld1b { z16.b, z24.b }, pn8/z, [x1]
; CHECK-NEXT:    ld1b { z17.b, z25.b }, pn8/z, [x1, x0]
; CHECK-NEXT:    usdot za.s[w8, 0, vgx2], { z16.b, z17.b }, { z24.b, z25.b }
; CHECK-NEXT:    ret
entry:
  %0 = tail call target("aarch64.svcount") @llvm.aarch64.sve.ptrue.c8()
  %1 = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld1.pn.x2.nxv16i8(target("aarch64.svcount") %0, ptr %ptr)
  %2 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 0
  %3 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 1
  %arrayidx2 = getelementptr inbounds i8, ptr %ptr, i64 %stride
  %4 = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld1.pn.x2.nxv16i8(target("aarch64.svcount") %0, ptr %arrayidx2)
  %5 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %4, 0
  %6 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %4, 1
  call void @llvm.aarch64.sme.usdot.za32.vg1x2.nxv16i8(i32 0, <vscale x 16 x i8> %2, <vscale x 16 x i8> %5, <vscale x 16 x i8> %3, <vscale x 16 x i8> %6)
  ret void
}

define void @usdot_multi_za32_u8_vg1x4(i32 %slice, <vscale x 16 x i8> %unused, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2, <vscale x 16 x i8> %zn3,
; CHECK-LABEL: usdot_multi_za32_u8_vg1x4:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z26.d, z7.d
; CHECK-NEXT:    mov z25.d, z6.d
; CHECK-NEXT:    ldr z27, [x1]
; CHECK-NEXT:    mov z7.d, z4.d
; CHECK-NEXT:    mov z24.d, z5.d
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    mov z6.d, z3.d
; CHECK-NEXT:    mov z5.d, z2.d
; CHECK-NEXT:    mov z4.d, z1.d
; CHECK-NEXT:    usdot za.s[w8, 0, vgx4], { z4.b - z7.b }, { z24.b - z27.b }
; CHECK-NEXT:    usdot za.s[w8, 7, vgx4], { z4.b - z7.b }, { z24.b - z27.b }
; CHECK-NEXT:    ret
                                      <vscale x 16 x i8> %zn4, <vscale x 16 x i8> %zn5, <vscale x 16 x i8> %zn6, <vscale x 16 x i8> %zn7) #0 {
  call void @llvm.aarch64.sme.usdot.za32.vg1x4.nxv16i8(i32 %slice, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2, <vscale x 16 x i8> %zn3,
                                                      <vscale x 16 x i8> %zn4, <vscale x 16 x i8> %zn5, <vscale x 16 x i8> %zn6, <vscale x 16 x i8> %zn7)
  %slice2 = add i32 %slice, 7
  call void @llvm.aarch64.sme.usdot.za32.vg1x4.nxv16i8(i32 %slice2, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2, <vscale x 16 x i8> %zn3,
                                                      <vscale x 16 x i8> %zn4, <vscale x 16 x i8> %zn5, <vscale x 16 x i8> %zn6, <vscale x 16 x i8> %zn7)
  ret void
}

define void @usdot_multi_za32_u16_vg1x4_tuple(i64 %stride, ptr %ptr) #1 {
; CHECK-LABEL: usdot_multi_za32_u16_vg1x4_tuple:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    lsl x9, x0, #1
; CHECK-NEXT:    ptrue pn8.b
; CHECK-NEXT:    mov w8, wzr
; CHECK-NEXT:    ld1b { z16.b, z20.b, z24.b, z28.b }, pn8/z, [x1]
; CHECK-NEXT:    ld1b { z17.b, z21.b, z25.b, z29.b }, pn8/z, [x1, x0]
; CHECK-NEXT:    add x10, x9, x0
; CHECK-NEXT:    ld1b { z18.b, z22.b, z26.b, z30.b }, pn8/z, [x1, x9]
; CHECK-NEXT:    ld1b { z19.b, z23.b, z27.b, z31.b }, pn8/z, [x1, x10]
; CHECK-NEXT:    usdot za.s[w8, 0, vgx4], { z16.b - z19.b }, { z20.b - z23.b }
; CHECK-NEXT:    usdot za.s[w8, 0, vgx4], { z24.b - z27.b }, { z28.b - z31.b }
; CHECK-NEXT:    ret
entry:
  %0 = tail call target("aarch64.svcount") @llvm.aarch64.sve.ptrue.c8()
  %1 = tail 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") %0, ptr %ptr)
  %2 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 0
  %3 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 1
  %4 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 2
  %5 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 3
  %arrayidx2 = getelementptr inbounds i8, ptr %ptr, i64 %stride
  %6 = tail 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") %0, ptr %arrayidx2)
  %7 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %6, 0
  %8 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %6, 1
  %9 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %6, 2
  %10 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %6, 3
  %mul3 = shl i64 %stride, 1
  %arrayidx4 = getelementptr inbounds i8, ptr %ptr, i64 %mul3
  %11 = tail 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") %0, ptr %arrayidx4)
  %12 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %11, 0
  %13 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %11, 1
  %14 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %11, 2
  %15 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %11, 3
  %mul5 = mul i64 %stride, 3
  %arrayidx6 = getelementptr inbounds i8, ptr %ptr, i64 %mul5
  %16 = tail 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") %0, ptr %arrayidx6)
  %17 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %16, 0
  %18 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %16, 1
  %19 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %16, 2
  %20 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %16, 3
  call void @llvm.aarch64.sme.usdot.za32.vg1x4.nxv16i8(i32 0, <vscale x 16 x i8> %2, <vscale x 16 x i8> %7, <vscale x 16 x i8> %12, <vscale x 16 x i8> %17,
                                                      <vscale x 16 x i8> %3, <vscale x 16 x i8> %8, <vscale x 16 x i8> %13, <vscale x 16 x i8> %18)
  call void @llvm.aarch64.sme.usdot.za32.vg1x4.nxv16i8(i32 0, <vscale x 16 x i8> %4, <vscale x 16 x i8> %9, <vscale x 16 x i8> %14, <vscale x 16 x i8> %19,
                                                      <vscale x 16 x i8> %5, <vscale x 16 x i8> %10, <vscale x 16 x i8> %15, <vscale x 16 x i8> %20)
  ret void
}

; == Multi, multi (signed) ==

define void @sdot_multi_za32_u16_vg1x2(i32 %slice, <vscale x 16 x i8> %unused, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, <vscale x 8 x i16> %zn3) #0 {
; CHECK-LABEL: sdot_multi_za32_u16_vg1x2:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z5.d, z4.d
; CHECK-NEXT:    mov z7.d, z2.d
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    mov z4.d, z3.d
; CHECK-NEXT:    mov z6.d, z1.d
; CHECK-NEXT:    sdot za.s[w8, 0, vgx2], { z6.h, z7.h }, { z4.h, z5.h }
; CHECK-NEXT:    sdot za.s[w8, 7, vgx2], { z6.h, z7.h }, { z4.h, z5.h }
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.sdot.za32.vg1x2.nxv8i16(i32 %slice, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, <vscale x 8 x i16> %zn3)
  %slice2 = add i32 %slice, 7
  call void @llvm.aarch64.sme.sdot.za32.vg1x2.nxv8i16(i32 %slice2, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, <vscale x 8 x i16> %zn3)
  ret void
}

define void @sdot_multi_za32_u16_vg1x2_tuple(i64 %stride, ptr %ptr) #0 {
; CHECK-LABEL: sdot_multi_za32_u16_vg1x2_tuple:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    ptrue pn8.b
; CHECK-NEXT:    mov w8, wzr
; CHECK-NEXT:    ld1b { z16.b, z24.b }, pn8/z, [x1]
; CHECK-NEXT:    ld1b { z17.b, z25.b }, pn8/z, [x1, x0]
; CHECK-NEXT:    sdot za.s[w8, 0, vgx2], { z16.b, z17.b }, { z24.b, z25.b }
; CHECK-NEXT:    ret
entry:
  %0 = tail call target("aarch64.svcount") @llvm.aarch64.sve.ptrue.c8()
  %1 = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld1.pn.x2.nxv16i8(target("aarch64.svcount") %0, ptr %ptr)
  %2 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 0
  %3 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 1
  %arrayidx2 = getelementptr inbounds i8, ptr %ptr, i64 %stride
  %4 = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld1.pn.x2.nxv16i8(target("aarch64.svcount") %0, ptr %arrayidx2)
  %5 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %4, 0
  %6 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %4, 1
  call void @llvm.aarch64.sme.sdot.za32.vg1x2.nxv16i8(i32 0, <vscale x 16 x i8> %2, <vscale x 16 x i8> %5, <vscale x 16 x i8> %3, <vscale x 16 x i8> %6)
  ret void
}

define void @sdot_multi_za32_u16_vg1x4(i32 %slice, <vscale x 16 x i8> %unused, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, <vscale x 8 x i16> %zn3,
; CHECK-LABEL: sdot_multi_za32_u16_vg1x4:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z26.d, z7.d
; CHECK-NEXT:    mov z25.d, z6.d
; CHECK-NEXT:    ldr z27, [x1]
; CHECK-NEXT:    mov z7.d, z4.d
; CHECK-NEXT:    mov z24.d, z5.d
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    mov z6.d, z3.d
; CHECK-NEXT:    mov z5.d, z2.d
; CHECK-NEXT:    mov z4.d, z1.d
; CHECK-NEXT:    sdot za.s[w8, 0, vgx4], { z4.h - z7.h }, { z24.h - z27.h }
; CHECK-NEXT:    sdot za.s[w8, 7, vgx4], { z4.h - z7.h }, { z24.h - z27.h }
; CHECK-NEXT:    ret
                                        <vscale x 8 x i16> %zn4, <vscale x 8 x i16> %zn5, <vscale x 8 x i16> %zn6, <vscale x 8 x i16> %zn7) #0 {
  call void @llvm.aarch64.sme.sdot.za32.vg1x4.nxv8i16(i32 %slice, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, <vscale x 8 x i16> %zn3,
                                                      <vscale x 8 x i16> %zn4, <vscale x 8 x i16> %zn5, <vscale x 8 x i16> %zn6, <vscale x 8 x i16> %zn7)
  %slice2 = add i32 %slice, 7
  call void @llvm.aarch64.sme.sdot.za32.vg1x4.nxv8i16(i32 %slice2, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, <vscale x 8 x i16> %zn3,
                                                      <vscale x 8 x i16> %zn4, <vscale x 8 x i16> %zn5, <vscale x 8 x i16> %zn6, <vscale x 8 x i16> %zn7)
  ret void
}

define void @sdot_multi_za32_u16_vg1x4_tuple(i64 %stride, ptr %ptr) #0 {
; CHECK-LABEL: sdot_multi_za32_u16_vg1x4_tuple:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    lsl x9, x0, #1
; CHECK-NEXT:    ptrue pn8.b
; CHECK-NEXT:    mov w8, wzr
; CHECK-NEXT:    ld1b { z16.b, z20.b, z24.b, z28.b }, pn8/z, [x1]
; CHECK-NEXT:    ld1b { z17.b, z21.b, z25.b, z29.b }, pn8/z, [x1, x0]
; CHECK-NEXT:    add x10, x9, x0
; CHECK-NEXT:    ld1b { z18.b, z22.b, z26.b, z30.b }, pn8/z, [x1, x9]
; CHECK-NEXT:    ld1b { z19.b, z23.b, z27.b, z31.b }, pn8/z, [x1, x10]
; CHECK-NEXT:    sdot za.s[w8, 0, vgx4], { z16.b - z19.b }, { z20.b - z23.b }
; CHECK-NEXT:    sdot za.s[w8, 0, vgx4], { z24.b - z27.b }, { z28.b - z31.b }
; CHECK-NEXT:    ret
entry:
  %0 = tail call target("aarch64.svcount") @llvm.aarch64.sve.ptrue.c8()
  %1 = tail 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") %0, ptr %ptr)
  %2 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 0
  %3 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 1
  %4 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 2
  %5 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 3
  %arrayidx2 = getelementptr inbounds i8, ptr %ptr, i64 %stride
  %6 = tail 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") %0, ptr %arrayidx2)
  %7 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %6, 0
  %8 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %6, 1
  %9 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %6, 2
  %10 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %6, 3
  %mul3 = shl i64 %stride, 1
  %arrayidx4 = getelementptr inbounds i8, ptr %ptr, i64 %mul3
  %11 = tail 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") %0, ptr %arrayidx4)
  %12 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %11, 0
  %13 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %11, 1
  %14 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %11, 2
  %15 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %11, 3
  %mul5 = mul i64 %stride, 3
  %arrayidx6 = getelementptr inbounds i8, ptr %ptr, i64 %mul5
  %16 = tail 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") %0, ptr %arrayidx6)
  %17 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %16, 0
  %18 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %16, 1
  %19 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %16, 2
  %20 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %16, 3
  call void @llvm.aarch64.sme.sdot.za32.vg1x4.nxv16i8(i32 0, <vscale x 16 x i8> %2, <vscale x 16 x i8> %7, <vscale x 16 x i8> %12, <vscale x 16 x i8> %17,
                                                      <vscale x 16 x i8> %3, <vscale x 16 x i8> %8, <vscale x 16 x i8> %13, <vscale x 16 x i8> %18)
  call void @llvm.aarch64.sme.sdot.za32.vg1x4.nxv16i8(i32 0, <vscale x 16 x i8> %4, <vscale x 16 x i8> %9, <vscale x 16 x i8> %14, <vscale x 16 x i8> %19,
                                                      <vscale x 16 x i8> %5, <vscale x 16 x i8> %10, <vscale x 16 x i8> %15, <vscale x 16 x i8> %20)
  ret void
}

define void @sdot_multi_za32_u8_vg1x2(i32 %slice, <vscale x 16 x i8> %unused, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2, <vscale x 16 x i8> %zn3) #0 {
; CHECK-LABEL: sdot_multi_za32_u8_vg1x2:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z5.d, z4.d
; CHECK-NEXT:    mov z7.d, z2.d
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    mov z4.d, z3.d
; CHECK-NEXT:    mov z6.d, z1.d
; CHECK-NEXT:    sdot za.s[w8, 0, vgx2], { z6.b, z7.b }, { z4.b, z5.b }
; CHECK-NEXT:    sdot za.s[w8, 7, vgx2], { z6.b, z7.b }, { z4.b, z5.b }
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.sdot.za32.vg1x2.nxv16i8(i32 %slice, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2, <vscale x 16 x i8> %zn3)
  %slice2 = add i32 %slice, 7
  call void @llvm.aarch64.sme.sdot.za32.vg1x2.nxv16i8(i32 %slice2, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2, <vscale x 16 x i8> %zn3)
  ret void
}

define void @sdot_multi_za32_u8_vg1x4(i32 %slice, <vscale x 16 x i8> %unused, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2, <vscale x 16 x i8> %zn3,
; CHECK-LABEL: sdot_multi_za32_u8_vg1x4:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z26.d, z7.d
; CHECK-NEXT:    mov z25.d, z6.d
; CHECK-NEXT:    ldr z27, [x1]
; CHECK-NEXT:    mov z7.d, z4.d
; CHECK-NEXT:    mov z24.d, z5.d
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    mov z6.d, z3.d
; CHECK-NEXT:    mov z5.d, z2.d
; CHECK-NEXT:    mov z4.d, z1.d
; CHECK-NEXT:    sdot za.s[w8, 0, vgx4], { z4.b - z7.b }, { z24.b - z27.b }
; CHECK-NEXT:    sdot za.s[w8, 7, vgx4], { z4.b - z7.b }, { z24.b - z27.b }
; CHECK-NEXT:    ret
                                      <vscale x 16 x i8> %zn4, <vscale x 16 x i8> %zn5, <vscale x 16 x i8> %zn6, <vscale x 16 x i8> %zn7) #0 {
  call void @llvm.aarch64.sme.sdot.za32.vg1x4.nxv16i8(i32 %slice, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2, <vscale x 16 x i8> %zn3,
                                                      <vscale x 16 x i8> %zn4, <vscale x 16 x i8> %zn5, <vscale x 16 x i8> %zn6, <vscale x 16 x i8> %zn7)
  %slice2 = add i32 %slice, 7
  call void @llvm.aarch64.sme.sdot.za32.vg1x4.nxv16i8(i32 %slice2, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2, <vscale x 16 x i8> %zn3,
                                                      <vscale x 16 x i8> %zn4, <vscale x 16 x i8> %zn5, <vscale x 16 x i8> %zn6, <vscale x 16 x i8> %zn7)
  ret void
}

define void @sdot_multi_za64_u16_vg1x2(i32 %slice, <vscale x 16 x i8> %unused, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, <vscale x 8 x i16> %zn3) #1 {
; CHECK-LABEL: sdot_multi_za64_u16_vg1x2:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z5.d, z4.d
; CHECK-NEXT:    mov z7.d, z2.d
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    mov z4.d, z3.d
; CHECK-NEXT:    mov z6.d, z1.d
; CHECK-NEXT:    sdot za.d[w8, 0, vgx2], { z6.h, z7.h }, { z4.h, z5.h }
; CHECK-NEXT:    sdot za.d[w8, 7, vgx2], { z6.h, z7.h }, { z4.h, z5.h }
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.sdot.za64.vg1x2.nxv8i16(i32 %slice, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, <vscale x 8 x i16> %zn3)
  %slice2 = add i32 %slice, 7
  call void @llvm.aarch64.sme.sdot.za64.vg1x2.nxv8i16(i32 %slice2, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, <vscale x 8 x i16> %zn3)
  ret void
}

define void @sdot_multi_za64_u16_vg1x2_tuple(i64 %stride, ptr %ptr) #1 {
; CHECK-LABEL: sdot_multi_za64_u16_vg1x2_tuple:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    ptrue pn8.b
; CHECK-NEXT:    add x9, x1, x0
; CHECK-NEXT:    mov w8, wzr
; CHECK-NEXT:    ld1h { z16.h, z24.h }, pn8/z, [x1]
; CHECK-NEXT:    ld1h { z17.h, z25.h }, pn8/z, [x9]
; CHECK-NEXT:    sdot za.d[w8, 0, vgx2], { z16.h, z17.h }, { z24.h, z25.h }
; CHECK-NEXT:    ret
entry:
  %0 = tail call target("aarch64.svcount") @llvm.aarch64.sve.ptrue.c8()
  %1 = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld1.pn.x2.nxv8i16(target("aarch64.svcount") %0, ptr %ptr)
  %2 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } %1, 0
  %3 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } %1, 1
  %arrayidx2 = getelementptr inbounds i8, ptr %ptr, i64 %stride
  %4 = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld1.pn.x2.nxv8i16(target("aarch64.svcount") %0, ptr %arrayidx2)
  %5 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } %4, 0
  %6 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } %4, 1
  call void @llvm.aarch64.sme.sdot.za64.vg1x2.nxv8i16(i32 0, <vscale x 8 x i16> %2, <vscale x 8 x i16> %5, <vscale x 8 x i16> %3, <vscale x 8 x i16> %6)
  ret void
}

define void @sdot_multi_za64_u16_vg1x4(i32 %slice, <vscale x 16 x i8> %unused, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, <vscale x 8 x i16> %zn3,
; CHECK-LABEL: sdot_multi_za64_u16_vg1x4:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z26.d, z7.d
; CHECK-NEXT:    mov z25.d, z6.d
; CHECK-NEXT:    ldr z27, [x1]
; CHECK-NEXT:    mov z7.d, z4.d
; CHECK-NEXT:    mov z24.d, z5.d
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    mov z6.d, z3.d
; CHECK-NEXT:    mov z5.d, z2.d
; CHECK-NEXT:    mov z4.d, z1.d
; CHECK-NEXT:    sdot za.d[w8, 0, vgx4], { z4.h - z7.h }, { z24.h - z27.h }
; CHECK-NEXT:    sdot za.d[w8, 7, vgx4], { z4.h - z7.h }, { z24.h - z27.h }
; CHECK-NEXT:    ret
                                       <vscale x 8 x i16> %zn4, <vscale x 8 x i16> %zn5, <vscale x 8 x i16> %zn6, <vscale x 8 x i16> %zn7) #1 {
  call void @llvm.aarch64.sme.sdot.za64.vg1x4.nxv8i16(i32 %slice, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, <vscale x 8 x i16> %zn3,
                                                      <vscale x 8 x i16> %zn4, <vscale x 8 x i16> %zn5, <vscale x 8 x i16> %zn6, <vscale x 8 x i16> %zn7)
  %slice2 = add i32 %slice, 7
  call void @llvm.aarch64.sme.sdot.za64.vg1x4.nxv8i16(i32 %slice2, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, <vscale x 8 x i16> %zn3,
                                                      <vscale x 8 x i16> %zn4, <vscale x 8 x i16> %zn5, <vscale x 8 x i16> %zn6, <vscale x 8 x i16> %zn7)
  ret void
}

define void @sdot_multi_za64_u16_vg1x4_tuple(i64 %stride, ptr %ptr) #1 {
; CHECK-LABEL: sdot_multi_za64_u16_vg1x4_tuple:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    add x9, x0, x0, lsl #1
; CHECK-NEXT:    ptrue pn8.b
; CHECK-NEXT:    add x10, x1, x0
; CHECK-NEXT:    ld1h { z16.h, z20.h, z24.h, z28.h }, pn8/z, [x1]
; CHECK-NEXT:    ld1h { z17.h, z21.h, z25.h, z29.h }, pn8/z, [x10]
; CHECK-NEXT:    ld1h { z18.h, z22.h, z26.h, z30.h }, pn8/z, [x1, x0, lsl #1]
; CHECK-NEXT:    add x9, x1, x9
; CHECK-NEXT:    mov w8, wzr
; CHECK-NEXT:    ld1h { z19.h, z23.h, z27.h, z31.h }, pn8/z, [x9]
; CHECK-NEXT:    sdot za.d[w8, 0, vgx4], { z16.h - z19.h }, { z20.h - z23.h }
; CHECK-NEXT:    sdot za.d[w8, 0, vgx4], { z24.h - z27.h }, { z28.h - z31.h }
; CHECK-NEXT:    ret
entry:
  %0 = tail call target("aarch64.svcount") @llvm.aarch64.sve.ptrue.c8()
  %1 = tail 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") %0, ptr %ptr)
  %2 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %1, 0
  %3 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %1, 1
  %4 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %1, 2
  %5 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %1, 3
  %arrayidx2 = getelementptr inbounds i8, ptr %ptr, i64 %stride
  %6 = tail 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") %0, ptr %arrayidx2)
  %7 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %6, 0
  %8 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %6, 1
  %9 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %6, 2
  %10 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %6, 3
  %mul3 = shl i64 %stride, 1
  %arrayidx4 = getelementptr inbounds i8, ptr %ptr, i64 %mul3
  %11 = tail 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") %0, ptr %arrayidx4)
  %12 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %11, 0
  %13 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %11, 1
  %14 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %11, 2
  %15 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %11, 3
  %mul5 = mul i64 %stride, 3
  %arrayidx6 = getelementptr inbounds i8, ptr %ptr, i64 %mul5
  %16 = tail 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") %0, ptr %arrayidx6)
  %17 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %16, 0
  %18 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %16, 1
  %19 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %16, 2
  %20 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %16, 3
  call void @llvm.aarch64.sme.sdot.za64.vg1x4.nxv8i16(i32 0, <vscale x 8 x i16> %2, <vscale x 8 x i16> %7, <vscale x 8 x i16> %12, <vscale x 8 x i16> %17,
                                                      <vscale x 8 x i16> %3, <vscale x 8 x i16> %8, <vscale x 8 x i16> %13, <vscale x 8 x i16> %18)
  call void @llvm.aarch64.sme.sdot.za64.vg1x4.nxv8i16(i32 0, <vscale x 8 x i16> %4, <vscale x 8 x i16> %9, <vscale x 8 x i16> %14, <vscale x 8 x i16> %19,
                                                      <vscale x 8 x i16> %5, <vscale x 8 x i16> %10, <vscale x 8 x i16> %15, <vscale x 8 x i16> %20)
  ret void
}

; == Multi, single (unsigned) ==

define void @udot_single_za32_u16_vg1x2(i32 %slice, <vscale x 16 x i8> %unused, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2) #0 {
; CHECK-LABEL: udot_single_za32_u16_vg1x2:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    udot za.s[w8, 0, vgx2], { z1.h, z2.h }, z3.h
; CHECK-NEXT:    udot za.s[w8, 7, vgx2], { z1.h, z2.h }, z3.h
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.udot.single.za32.vg1x2.nxv8i16(i32 %slice, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2)
  %slice2 = add i32 %slice, 7
  call void @llvm.aarch64.sme.udot.single.za32.vg1x2.nxv8i16(i32 %slice2, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2)
  ret void
}

define void @udot_single_za32_u16_vg1x2_tuple(ptr %ptr, i64 %stride, <vscale x 8 x i16> %zn) #0 {
; CHECK-LABEL: udot_single_za32_u16_vg1x2_tuple:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
; CHECK-NEXT:    addvl sp, sp, #-3
; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
; CHECK-NEXT:    ptrue pn8.b
; CHECK-NEXT:    add x9, x0, x1
; CHECK-NEXT:    str z10, [sp, #1, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    mov w8, wzr
; CHECK-NEXT:    str z9, [sp, #2, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    ld1h { z1.h, z9.h }, pn8/z, [x0]
; CHECK-NEXT:    ld1h { z2.h, z10.h }, pn8/z, [x9]
; CHECK-NEXT:    udot za.s[w8, 0, vgx2], { z1.h, z2.h }, z0.h
; CHECK-NEXT:    udot za.s[w8, 0, vgx2], { z9.h, z10.h }, z0.h
; CHECK-NEXT:    ldr z10, [sp, #1, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z9, [sp, #2, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
; CHECK-NEXT:    addvl sp, sp, #3
; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
; CHECK-NEXT:    ret
entry:
  %0 = tail call target("aarch64.svcount") @llvm.aarch64.sve.ptrue.c8()
  %1 = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld1.pn.x2.nxv8i16(target("aarch64.svcount") %0, ptr %ptr)
  %2 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } %1, 0
  %3 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } %1, 1
  %arrayidx2 = getelementptr inbounds i8, ptr %ptr, i64 %stride
  %4 = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld1.pn.x2.nxv8i16(target("aarch64.svcount") %0, ptr %arrayidx2)
  %5 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } %4, 0
  %6 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } %4, 1
  call void @llvm.aarch64.sme.udot.single.za32.vg1x2.nxv8i16(i32 0, <vscale x 8 x i16> %2, <vscale x 8 x i16> %5, <vscale x 8 x i16> %zn)
  call void @llvm.aarch64.sme.udot.single.za32.vg1x2.nxv8i16(i32 0, <vscale x 8 x i16> %3, <vscale x 8 x i16> %6, <vscale x 8 x i16> %zn)
  ret void
}

define void @udot_single_za32_u16_vg1x2_x4load_x2tuple(ptr %ptr, i64 %stride, <vscale x 8 x i16> %zn) #0 {
; CHECK-LABEL: udot_single_za32_u16_vg1x2_x4load_x2tuple:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
; CHECK-NEXT:    addvl sp, sp, #-5
; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
; CHECK-NEXT:    ptrue pn8.b
; CHECK-NEXT:    add x9, x0, x1
; CHECK-NEXT:    str z14, [sp, #1, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    mov w8, wzr
; CHECK-NEXT:    str z13, [sp, #2, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z10, [sp, #3, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z9, [sp, #4, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    ld1h { z1.h, z5.h, z9.h, z13.h }, pn8/z, [x0]
; CHECK-NEXT:    ld1h { z2.h, z6.h, z10.h, z14.h }, pn8/z, [x9]
; CHECK-NEXT:    udot za.s[w8, 0, vgx2], { z1.h, z2.h }, z0.h
; CHECK-NEXT:    udot za.s[w8, 0, vgx2], { z5.h, z6.h }, z0.h
; CHECK-NEXT:    udot za.s[w8, 0, vgx2], { z9.h, z10.h }, z0.h
; CHECK-NEXT:    udot za.s[w8, 0, vgx2], { z13.h, z14.h }, z0.h
; CHECK-NEXT:    ldr z14, [sp, #1, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z13, [sp, #2, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z10, [sp, #3, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z9, [sp, #4, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
; CHECK-NEXT:    addvl sp, sp, #5
; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
; CHECK-NEXT:    ret
entry:
  %0 = tail call target("aarch64.svcount") @llvm.aarch64.sve.ptrue.c8()
  %1 = tail 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") %0, ptr %ptr)
  %2 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %1, 0
  %3 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %1, 1
  %4 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %1, 2
  %5 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %1, 3
  %arrayidx2 = getelementptr inbounds i8, ptr %ptr, i64 %stride
  %6 = tail 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") %0, ptr %arrayidx2)
  %7 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %6, 0
  %8 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %6, 1
  %9 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %6, 2
  %10 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %6, 3
  call void @llvm.aarch64.sme.udot.single.za32.vg1x2.nxv8i16(i32 0, <vscale x 8 x i16> %2, <vscale x 8 x i16> %7, <vscale x 8 x i16> %zn)
  call void @llvm.aarch64.sme.udot.single.za32.vg1x2.nxv8i16(i32 0, <vscale x 8 x i16> %3, <vscale x 8 x i16> %8, <vscale x 8 x i16> %zn)
  call void @llvm.aarch64.sme.udot.single.za32.vg1x2.nxv8i16(i32 0, <vscale x 8 x i16> %4, <vscale x 8 x i16> %9, <vscale x 8 x i16> %zn)
  call void @llvm.aarch64.sme.udot.single.za32.vg1x2.nxv8i16(i32 0, <vscale x 8 x i16> %5, <vscale x 8 x i16> %10, <vscale x 8 x i16> %zn)
  ret void
}

define void @udot_single_za32_u16_vg1x4(i32 %slice, <vscale x 16 x i8> %unused, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, <vscale x 8 x i16> %zn3, <vscale x 8 x i16> %zn4) #0 {
; CHECK-LABEL: udot_single_za32_u16_vg1x4:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    udot za.s[w8, 0, vgx4], { z1.h - z4.h }, z5.h
; CHECK-NEXT:    udot za.s[w8, 7, vgx4], { z1.h - z4.h }, z5.h
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.udot.single.za32.vg1x4.nxv8i16(i32 %slice, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, <vscale x 8 x i16> %zn3, <vscale x 8 x i16> %zn4)
  %slice2 = add i32 %slice, 7
  call void @llvm.aarch64.sme.udot.single.za32.vg1x4.nxv8i16(i32 %slice2, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, <vscale x 8 x i16> %zn3, <vscale x 8 x i16> %zn4)
  ret void
}

define void @udot_single_za32_u16_vg1x4_tuple(ptr %ptr, i64 %stride, <vscale x 8 x i16> %zn) #0 {
; CHECK-LABEL: udot_single_za32_u16_vg1x4_tuple:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
; CHECK-NEXT:    addvl sp, sp, #-9
; CHECK-NEXT:    add x9, x1, x1, lsl #1
; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
; CHECK-NEXT:    ptrue pn8.b
; CHECK-NEXT:    str z23, [sp, #1, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    add x10, x0, x1
; CHECK-NEXT:    mov w8, wzr
; CHECK-NEXT:    str z22, [sp, #2, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    add x9, x0, x9
; CHECK-NEXT:    str z21, [sp, #3, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z20, [sp, #4, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z19, [sp, #5, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z18, [sp, #6, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z17, [sp, #7, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z16, [sp, #8, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    ld1h { z16.h, z20.h, z24.h, z28.h }, pn8/z, [x0]
; CHECK-NEXT:    ld1h { z17.h, z21.h, z25.h, z29.h }, pn8/z, [x10]
; CHECK-NEXT:    ld1h { z18.h, z22.h, z26.h, z30.h }, pn8/z, [x0, x1, lsl #1]
; CHECK-NEXT:    ld1h { z19.h, z23.h, z27.h, z31.h }, pn8/z, [x9]
; CHECK-NEXT:    udot za.s[w8, 0, vgx4], { z16.h - z19.h }, z0.h
; CHECK-NEXT:    udot za.s[w8, 0, vgx4], { z20.h - z23.h }, z0.h
; CHECK-NEXT:    udot za.s[w8, 0, vgx4], { z24.h - z27.h }, z0.h
; CHECK-NEXT:    udot za.s[w8, 0, vgx4], { z28.h - z31.h }, z0.h
; CHECK-NEXT:    ldr z23, [sp, #1, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z22, [sp, #2, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z21, [sp, #3, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z20, [sp, #4, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z19, [sp, #5, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z18, [sp, #6, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z17, [sp, #7, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z16, [sp, #8, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
; CHECK-NEXT:    addvl sp, sp, #9
; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
; CHECK-NEXT:    ret
entry:
  %0 = tail call target("aarch64.svcount") @llvm.aarch64.sve.ptrue.c8()
  %1 = tail 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") %0, ptr %ptr)
  %2 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %1, 0
  %3 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %1, 1
  %4 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %1, 2
  %5 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %1, 3
  %arrayidx2 = getelementptr inbounds i8, ptr %ptr, i64 %stride
  %6 = tail 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") %0, ptr %arrayidx2)
  %7 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %6, 0
  %8 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %6, 1
  %9 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %6, 2
  %10 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %6, 3
  %mul3 = shl i64 %stride, 1
  %arrayidx4 = getelementptr inbounds i8, ptr %ptr, i64 %mul3
  %11 = tail 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") %0, ptr %arrayidx4)
  %12 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %11, 0
  %13 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %11, 1
  %14 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %11, 2
  %15 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %11, 3
  %mul5 = mul i64 %stride, 3
  %arrayidx6 = getelementptr inbounds i8, ptr %ptr, i64 %mul5
  %16 = tail 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") %0, ptr %arrayidx6)
  %17 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %16, 0
  %18 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %16, 1
  %19 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %16, 2
  %20 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %16, 3
  call void @llvm.aarch64.sme.udot.single.za32.vg1x4.nxv8i16(i32 0, <vscale x 8 x i16> %2, <vscale x 8 x i16> %7, <vscale x 8 x i16> %12, <vscale x 8 x i16> %17, <vscale x 8 x i16> %zn)
  call void @llvm.aarch64.sme.udot.single.za32.vg1x4.nxv8i16(i32 0, <vscale x 8 x i16> %3, <vscale x 8 x i16> %8, <vscale x 8 x i16> %13, <vscale x 8 x i16> %18, <vscale x 8 x i16> %zn)
  call void @llvm.aarch64.sme.udot.single.za32.vg1x4.nxv8i16(i32 0, <vscale x 8 x i16> %4, <vscale x 8 x i16> %9, <vscale x 8 x i16> %14, <vscale x 8 x i16> %19, <vscale x 8 x i16> %zn)
  call void @llvm.aarch64.sme.udot.single.za32.vg1x4.nxv8i16(i32 0, <vscale x 8 x i16> %5, <vscale x 8 x i16> %10, <vscale x 8 x i16> %15, <vscale x 8 x i16> %20, <vscale x 8 x i16> %zn)
  ret void
}

define void @udot_single_za32_u8_vg1x2(i32 %slice, <vscale x 16 x i8> %unused, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2) #0 {
; CHECK-LABEL: udot_single_za32_u8_vg1x2:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    udot za.s[w8, 0, vgx2], { z1.b, z2.b }, z3.b
; CHECK-NEXT:    udot za.s[w8, 7, vgx2], { z1.b, z2.b }, z3.b
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.udot.single.za32.vg1x2.nxv16i8(i32 %slice, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2)
  %slice2 = add i32 %slice, 7
  call void @llvm.aarch64.sme.udot.single.za32.vg1x2.nxv16i8(i32 %slice2, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2)
  ret void
}

define void @udot_single_za32_u8_vg1x4(i32 %slice, <vscale x 16 x i8> %unused, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2, <vscale x 16 x i8> %zn3, <vscale x 16 x i8> %zn4) #0 {
; CHECK-LABEL: udot_single_za32_u8_vg1x4:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    udot za.s[w8, 0, vgx4], { z1.b - z4.b }, z5.b
; CHECK-NEXT:    udot za.s[w8, 7, vgx4], { z1.b - z4.b }, z5.b
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.udot.single.za32.vg1x4.nxv16i8(i32 %slice, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2, <vscale x 16 x i8> %zn3, <vscale x 16 x i8> %zn4)
  %slice2 = add i32 %slice, 7
  call void @llvm.aarch64.sme.udot.single.za32.vg1x4.nxv16i8(i32 %slice2, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2, <vscale x 16 x i8> %zn3, <vscale x 16 x i8> %zn4)
  ret void
}

define void @udot_single_za64_u16_vg1x2(i32 %slice, <vscale x 16 x i8> %unused, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2) #1 {
; CHECK-LABEL: udot_single_za64_u16_vg1x2:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    udot za.d[w8, 0, vgx2], { z1.h, z2.h }, z3.h
; CHECK-NEXT:    udot za.d[w8, 7, vgx2], { z1.h, z2.h }, z3.h
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.udot.single.za64.vg1x2.nxv8i16(i32 %slice, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2)
  %slice2 = add i32 %slice, 7
  call void @llvm.aarch64.sme.udot.single.za64.vg1x2.nxv8i16(i32 %slice2, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2)
  ret void
}

define void @udot_single_za64_u16_vg1x4(i32 %slice, <vscale x 16 x i8> %unused, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, <vscale x 8 x i16> %zn3, <vscale x 8 x i16> %zn4) #1 {
; CHECK-LABEL: udot_single_za64_u16_vg1x4:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    udot za.d[w8, 0, vgx4], { z1.h - z4.h }, z5.h
; CHECK-NEXT:    udot za.d[w8, 7, vgx4], { z1.h - z4.h }, z5.h
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.udot.single.za64.vg1x4.nxv8i16(i32 %slice, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, <vscale x 8 x i16> %zn3, <vscale x 8 x i16> %zn4)
  %slice2 = add i32 %slice, 7
  call void @llvm.aarch64.sme.udot.single.za64.vg1x4.nxv8i16(i32 %slice2, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, <vscale x 8 x i16> %zn3, <vscale x 8 x i16> %zn4)
  ret void
}

define void @usdot_single_za32_u8_vg1x2(i32 %slice, <vscale x 16 x i8> %unused, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2) #0 {
; CHECK-LABEL: usdot_single_za32_u8_vg1x2:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    usdot za.s[w8, 0, vgx2], { z1.b, z2.b }, z3.b
; CHECK-NEXT:    usdot za.s[w8, 7, vgx2], { z1.b, z2.b }, z3.b
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.usdot.single.za32.vg1x2.nxv16i8(i32 %slice, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2)
  %slice2 = add i32 %slice, 7
  call void @llvm.aarch64.sme.usdot.single.za32.vg1x2.nxv16i8(i32 %slice2, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2)
  ret void
}

define void @usdot_single_za32_u16_vg1x2_tuple(ptr %ptr, i64 %stride, <vscale x 16 x i8> %zn) #0 {
; CHECK-LABEL: usdot_single_za32_u16_vg1x2_tuple:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
; CHECK-NEXT:    addvl sp, sp, #-3
; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
; CHECK-NEXT:    ptrue pn8.b
; CHECK-NEXT:    mov w8, wzr
; CHECK-NEXT:    str z10, [sp, #1, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z9, [sp, #2, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    ld1b { z1.b, z9.b }, pn8/z, [x0]
; CHECK-NEXT:    ld1b { z2.b, z10.b }, pn8/z, [x0, x1]
; CHECK-NEXT:    usdot za.s[w8, 0, vgx2], { z1.b, z2.b }, z0.b
; CHECK-NEXT:    usdot za.s[w8, 0, vgx2], { z9.b, z10.b }, z0.b
; CHECK-NEXT:    ldr z10, [sp, #1, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z9, [sp, #2, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
; CHECK-NEXT:    addvl sp, sp, #3
; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
; CHECK-NEXT:    ret
entry:
  %0 = tail call target("aarch64.svcount") @llvm.aarch64.sve.ptrue.c8()
  %1 = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld1.pn.x2.nxv16i8(target("aarch64.svcount") %0, ptr %ptr)
  %2 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 0
  %3 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 1
  %arrayidx2 = getelementptr inbounds i8, ptr %ptr, i64 %stride
  %4 = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld1.pn.x2.nxv16i8(target("aarch64.svcount") %0, ptr %arrayidx2)
  %5 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %4, 0
  %6 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %4, 1
  call void @llvm.aarch64.sme.usdot.single.za32.vg1x2.nxv16i8(i32 0, <vscale x 16 x i8> %2, <vscale x 16 x i8> %5, <vscale x 16 x i8> %zn)
  call void @llvm.aarch64.sme.usdot.single.za32.vg1x2.nxv16i8(i32 0, <vscale x 16 x i8> %3, <vscale x 16 x i8> %6, <vscale x 16 x i8> %zn)
  ret void
}

define void @usdot_single_za32_u8_vg1x4(i32 %slice, <vscale x 16 x i8> %unused, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2, <vscale x 16 x i8> %zn3, <vscale x 16 x i8> %zn4) #0 {
; CHECK-LABEL: usdot_single_za32_u8_vg1x4:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    usdot za.s[w8, 0, vgx4], { z1.b - z4.b }, z5.b
; CHECK-NEXT:    usdot za.s[w8, 7, vgx4], { z1.b - z4.b }, z5.b
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.usdot.single.za32.vg1x4.nxv16i8(i32 %slice, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2, <vscale x 16 x i8> %zn3, <vscale x 16 x i8> %zn4)
  %slice2 = add i32 %slice, 7
  call void @llvm.aarch64.sme.usdot.single.za32.vg1x4.nxv16i8(i32 %slice2, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2, <vscale x 16 x i8> %zn3, <vscale x 16 x i8> %zn4)
  ret void
}

define void @usdot_single_za32_u16_vg1x4_tuple(ptr %ptr, i64 %stride, <vscale x 16 x i8> %zn) #0 {
; CHECK-LABEL: usdot_single_za32_u16_vg1x4_tuple:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
; CHECK-NEXT:    addvl sp, sp, #-9
; CHECK-NEXT:    lsl x9, x1, #1
; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
; CHECK-NEXT:    ptrue pn8.b
; CHECK-NEXT:    str z23, [sp, #1, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    mov w8, wzr
; CHECK-NEXT:    str z22, [sp, #2, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    add x10, x9, x1
; CHECK-NEXT:    str z21, [sp, #3, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z20, [sp, #4, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z19, [sp, #5, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z18, [sp, #6, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z17, [sp, #7, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z16, [sp, #8, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    ld1b { z16.b, z20.b, z24.b, z28.b }, pn8/z, [x0]
; CHECK-NEXT:    ld1b { z17.b, z21.b, z25.b, z29.b }, pn8/z, [x0, x1]
; CHECK-NEXT:    ld1b { z18.b, z22.b, z26.b, z30.b }, pn8/z, [x0, x9]
; CHECK-NEXT:    ld1b { z19.b, z23.b, z27.b, z31.b }, pn8/z, [x0, x10]
; CHECK-NEXT:    usdot za.s[w8, 0, vgx4], { z16.b - z19.b }, z0.b
; CHECK-NEXT:    usdot za.s[w8, 0, vgx4], { z20.b - z23.b }, z0.b
; CHECK-NEXT:    usdot za.s[w8, 0, vgx4], { z24.b - z27.b }, z0.b
; CHECK-NEXT:    usdot za.s[w8, 0, vgx4], { z28.b - z31.b }, z0.b
; CHECK-NEXT:    ldr z23, [sp, #1, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z22, [sp, #2, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z21, [sp, #3, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z20, [sp, #4, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z19, [sp, #5, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z18, [sp, #6, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z17, [sp, #7, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z16, [sp, #8, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
; CHECK-NEXT:    addvl sp, sp, #9
; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
; CHECK-NEXT:    ret
entry:
  %0 = tail call target("aarch64.svcount") @llvm.aarch64.sve.ptrue.c8()
  %1 = tail 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") %0, ptr %ptr)
  %2 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 0
  %3 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 1
  %4 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 2
  %5 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 3
  %arrayidx2 = getelementptr inbounds i8, ptr %ptr, i64 %stride
  %6 = tail 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") %0, ptr %arrayidx2)
  %7 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %6, 0
  %8 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %6, 1
  %9 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %6, 2
  %10 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %6, 3
  %mul3 = shl i64 %stride, 1
  %arrayidx4 = getelementptr inbounds i8, ptr %ptr, i64 %mul3
  %11 = tail 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") %0, ptr %arrayidx4)
  %12 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %11, 0
  %13 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %11, 1
  %14 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %11, 2
  %15 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %11, 3
  %mul5 = mul i64 %stride, 3
  %arrayidx6 = getelementptr inbounds i8, ptr %ptr, i64 %mul5
  %16 = tail 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") %0, ptr %arrayidx6)
  %17 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %16, 0
  %18 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %16, 1
  %19 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %16, 2
  %20 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %16, 3
  call void @llvm.aarch64.sme.usdot.single.za32.vg1x4.nxv16i8(i32 0, <vscale x 16 x i8> %2, <vscale x 16 x i8> %7, <vscale x 16 x i8> %12, <vscale x 16 x i8> %17, <vscale x 16 x i8> %zn)
  call void @llvm.aarch64.sme.usdot.single.za32.vg1x4.nxv16i8(i32 0, <vscale x 16 x i8> %3, <vscale x 16 x i8> %8, <vscale x 16 x i8> %13, <vscale x 16 x i8> %18, <vscale x 16 x i8> %zn)
  call void @llvm.aarch64.sme.usdot.single.za32.vg1x4.nxv16i8(i32 0, <vscale x 16 x i8> %4, <vscale x 16 x i8> %9, <vscale x 16 x i8> %14, <vscale x 16 x i8> %19, <vscale x 16 x i8> %zn)
  call void @llvm.aarch64.sme.usdot.single.za32.vg1x4.nxv16i8(i32 0, <vscale x 16 x i8> %5, <vscale x 16 x i8> %10, <vscale x 16 x i8> %15, <vscale x 16 x i8> %20, <vscale x 16 x i8> %zn)
  ret void
}

; == Multi, single (signed) ==

define void @sdot_single_za32_u16_vg1x2(i32 %slice, <vscale x 16 x i8> %unused, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2) #0 {
; CHECK-LABEL: sdot_single_za32_u16_vg1x2:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    sdot za.s[w8, 0, vgx2], { z1.h, z2.h }, z3.h
; CHECK-NEXT:    sdot za.s[w8, 7, vgx2], { z1.h, z2.h }, z3.h
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.sdot.single.za32.vg1x2.nxv8i16(i32 %slice, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2)
  %slice2 = add i32 %slice, 7
  call void @llvm.aarch64.sme.sdot.single.za32.vg1x2.nxv8i16(i32 %slice2, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2)
  ret void
}

define void @sdot_single_za32_u16_vg1x2_tuple(ptr %ptr, i64 %stride, <vscale x 8 x i16> %zn) #0 {
; CHECK-LABEL: sdot_single_za32_u16_vg1x2_tuple:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
; CHECK-NEXT:    addvl sp, sp, #-3
; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
; CHECK-NEXT:    ptrue pn8.b
; CHECK-NEXT:    add x9, x0, x1
; CHECK-NEXT:    str z10, [sp, #1, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    mov w8, wzr
; CHECK-NEXT:    str z9, [sp, #2, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    ld1h { z1.h, z9.h }, pn8/z, [x0]
; CHECK-NEXT:    ld1h { z2.h, z10.h }, pn8/z, [x9]
; CHECK-NEXT:    sdot za.s[w8, 0, vgx2], { z1.h, z2.h }, z0.h
; CHECK-NEXT:    sdot za.s[w8, 0, vgx2], { z9.h, z10.h }, z0.h
; CHECK-NEXT:    ldr z10, [sp, #1, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z9, [sp, #2, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
; CHECK-NEXT:    addvl sp, sp, #3
; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
; CHECK-NEXT:    ret
entry:
  %0 = tail call target("aarch64.svcount") @llvm.aarch64.sve.ptrue.c8()
  %1 = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld1.pn.x2.nxv8i16(target("aarch64.svcount") %0, ptr %ptr)
  %2 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } %1, 0
  %3 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } %1, 1
  %arrayidx2 = getelementptr inbounds i8, ptr %ptr, i64 %stride
  %4 = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld1.pn.x2.nxv8i16(target("aarch64.svcount") %0, ptr %arrayidx2)
  %5 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } %4, 0
  %6 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } %4, 1
  call void @llvm.aarch64.sme.sdot.single.za32.vg1x2.nxv8i16(i32 0, <vscale x 8 x i16> %2, <vscale x 8 x i16> %5, <vscale x 8 x i16> %zn)
  call void @llvm.aarch64.sme.sdot.single.za32.vg1x2.nxv8i16(i32 0, <vscale x 8 x i16> %3, <vscale x 8 x i16> %6, <vscale x 8 x i16> %zn)
  ret void
}

define void @sdot_single_za32_u16_vg1x4(i32 %slice, <vscale x 16 x i8> %unused, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, <vscale x 8 x i16> %zn3, <vscale x 8 x i16> %zn4) #0 {
; CHECK-LABEL: sdot_single_za32_u16_vg1x4:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    sdot za.s[w8, 0, vgx4], { z1.h - z4.h }, z5.h
; CHECK-NEXT:    sdot za.s[w8, 7, vgx4], { z1.h - z4.h }, z5.h
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.sdot.single.za32.vg1x4.nxv8i16(i32 %slice, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, <vscale x 8 x i16> %zn3, <vscale x 8 x i16> %zn4)
  %slice2 = add i32 %slice, 7
  call void @llvm.aarch64.sme.sdot.single.za32.vg1x4.nxv8i16(i32 %slice2, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, <vscale x 8 x i16> %zn3, <vscale x 8 x i16> %zn4)
  ret void
}

define void @sdot_single_za32_u16_vg1x4_tuple(ptr %ptr, i64 %stride, <vscale x 8 x i16> %zn) #0 {
; CHECK-LABEL: sdot_single_za32_u16_vg1x4_tuple:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
; CHECK-NEXT:    addvl sp, sp, #-9
; CHECK-NEXT:    add x9, x1, x1, lsl #1
; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
; CHECK-NEXT:    ptrue pn8.b
; CHECK-NEXT:    str z23, [sp, #1, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    add x10, x0, x1
; CHECK-NEXT:    mov w8, wzr
; CHECK-NEXT:    str z22, [sp, #2, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    add x9, x0, x9
; CHECK-NEXT:    str z21, [sp, #3, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z20, [sp, #4, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z19, [sp, #5, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z18, [sp, #6, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z17, [sp, #7, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z16, [sp, #8, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    ld1h { z16.h, z20.h, z24.h, z28.h }, pn8/z, [x0]
; CHECK-NEXT:    ld1h { z17.h, z21.h, z25.h, z29.h }, pn8/z, [x10]
; CHECK-NEXT:    ld1h { z18.h, z22.h, z26.h, z30.h }, pn8/z, [x0, x1, lsl #1]
; CHECK-NEXT:    ld1h { z19.h, z23.h, z27.h, z31.h }, pn8/z, [x9]
; CHECK-NEXT:    sdot za.s[w8, 0, vgx4], { z16.h - z19.h }, z0.h
; CHECK-NEXT:    sdot za.s[w8, 0, vgx4], { z20.h - z23.h }, z0.h
; CHECK-NEXT:    sdot za.s[w8, 0, vgx4], { z24.h - z27.h }, z0.h
; CHECK-NEXT:    sdot za.s[w8, 0, vgx4], { z28.h - z31.h }, z0.h
; CHECK-NEXT:    ldr z23, [sp, #1, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z22, [sp, #2, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z21, [sp, #3, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z20, [sp, #4, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z19, [sp, #5, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z18, [sp, #6, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z17, [sp, #7, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z16, [sp, #8, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
; CHECK-NEXT:    addvl sp, sp, #9
; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
; CHECK-NEXT:    ret
entry:
  %0 = tail call target("aarch64.svcount") @llvm.aarch64.sve.ptrue.c8()
  %1 = tail 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") %0, ptr %ptr)
  %2 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %1, 0
  %3 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %1, 1
  %4 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %1, 2
  %5 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %1, 3
  %arrayidx2 = getelementptr inbounds i8, ptr %ptr, i64 %stride
  %6 = tail 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") %0, ptr %arrayidx2)
  %7 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %6, 0
  %8 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %6, 1
  %9 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %6, 2
  %10 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %6, 3
  %mul3 = shl i64 %stride, 1
  %arrayidx4 = getelementptr inbounds i8, ptr %ptr, i64 %mul3
  %11 = tail 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") %0, ptr %arrayidx4)
  %12 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %11, 0
  %13 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %11, 1
  %14 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %11, 2
  %15 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %11, 3
  %mul5 = mul i64 %stride, 3
  %arrayidx6 = getelementptr inbounds i8, ptr %ptr, i64 %mul5
  %16 = tail 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") %0, ptr %arrayidx6)
  %17 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %16, 0
  %18 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %16, 1
  %19 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %16, 2
  %20 = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } %16, 3
  call void @llvm.aarch64.sme.sdot.single.za32.vg1x4.nxv8i16(i32 0, <vscale x 8 x i16> %2, <vscale x 8 x i16> %7, <vscale x 8 x i16> %12, <vscale x 8 x i16> %17, <vscale x 8 x i16> %zn)
  call void @llvm.aarch64.sme.sdot.single.za32.vg1x4.nxv8i16(i32 0, <vscale x 8 x i16> %3, <vscale x 8 x i16> %8, <vscale x 8 x i16> %13, <vscale x 8 x i16> %18, <vscale x 8 x i16> %zn)
  call void @llvm.aarch64.sme.sdot.single.za32.vg1x4.nxv8i16(i32 0, <vscale x 8 x i16> %4, <vscale x 8 x i16> %9, <vscale x 8 x i16> %14, <vscale x 8 x i16> %19, <vscale x 8 x i16> %zn)
  call void @llvm.aarch64.sme.sdot.single.za32.vg1x4.nxv8i16(i32 0, <vscale x 8 x i16> %5, <vscale x 8 x i16> %10, <vscale x 8 x i16> %15, <vscale x 8 x i16> %20, <vscale x 8 x i16> %zn)
  ret void
}

define void @sdot_single_za32_u8_vg1x2(i32 %slice, <vscale x 16 x i8> %unused, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2) #0 {
; CHECK-LABEL: sdot_single_za32_u8_vg1x2:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    sdot za.s[w8, 0, vgx2], { z1.b, z2.b }, z3.b
; CHECK-NEXT:    sdot za.s[w8, 7, vgx2], { z1.b, z2.b }, z3.b
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.sdot.single.za32.vg1x2.nxv16i8(i32 %slice, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2)
  %slice2 = add i32 %slice, 7
  call void @llvm.aarch64.sme.sdot.single.za32.vg1x2.nxv16i8(i32 %slice2, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2)
  ret void
}

define void @sdot_single_za32_u8_vg1x4(i32 %slice, <vscale x 16 x i8> %unused, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2, <vscale x 16 x i8> %zn3, <vscale x 16 x i8> %zn4) #0 {
; CHECK-LABEL: sdot_single_za32_u8_vg1x4:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    sdot za.s[w8, 0, vgx4], { z1.b - z4.b }, z5.b
; CHECK-NEXT:    sdot za.s[w8, 7, vgx4], { z1.b - z4.b }, z5.b
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.sdot.single.za32.vg1x4.nxv16i8(i32 %slice, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2, <vscale x 16 x i8> %zn3, <vscale x 16 x i8> %zn4)
  %slice2 = add i32 %slice, 7
  call void @llvm.aarch64.sme.sdot.single.za32.vg1x4.nxv16i8(i32 %slice2, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2, <vscale x 16 x i8> %zn3, <vscale x 16 x i8> %zn4)
  ret void
}

define void @sdot_single_za64_u16_vg1x2(i32 %slice, <vscale x 16 x i8> %unused, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2) #1 {
; CHECK-LABEL: sdot_single_za64_u16_vg1x2:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    sdot za.d[w8, 0, vgx2], { z1.h, z2.h }, z3.h
; CHECK-NEXT:    sdot za.d[w8, 7, vgx2], { z1.h, z2.h }, z3.h
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.sdot.single.za64.vg1x2.nxv8i16(i32 %slice, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2)
  %slice2 = add i32 %slice, 7
  call void @llvm.aarch64.sme.sdot.single.za64.vg1x2.nxv8i16(i32 %slice2, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2)
  ret void
}

define void @sdot_single_za64_u16_vg1x4(i32 %slice, <vscale x 16 x i8> %unused, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, <vscale x 8 x i16> %zn3, <vscale x 8 x i16> %zn4) #1 {
; CHECK-LABEL: sdot_single_za64_u16_vg1x4:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    sdot za.d[w8, 0, vgx4], { z1.h - z4.h }, z5.h
; CHECK-NEXT:    sdot za.d[w8, 7, vgx4], { z1.h - z4.h }, z5.h
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.sdot.single.za64.vg1x4.nxv8i16(i32 %slice, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, <vscale x 8 x i16> %zn3, <vscale x 8 x i16> %zn4)
  %slice2 = add i32 %slice, 7
  call void @llvm.aarch64.sme.sdot.single.za64.vg1x4.nxv8i16(i32 %slice2, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, <vscale x 8 x i16> %zn3, <vscale x 8 x i16> %zn4)
  ret void
}

define void @sudot_single_za32_u8_vg1x2(i32 %slice, <vscale x 16 x i8> %unused, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2) #0 {
; CHECK-LABEL: sudot_single_za32_u8_vg1x2:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    sudot za.s[w8, 0, vgx2], { z1.b, z2.b }, z3.b
; CHECK-NEXT:    sudot za.s[w8, 7, vgx2], { z1.b, z2.b }, z3.b
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.sudot.single.za32.vg1x2.nxv16i8(i32 %slice, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2)
  %slice2 = add i32 %slice, 7
  call void @llvm.aarch64.sme.sudot.single.za32.vg1x2.nxv16i8(i32 %slice2, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2)
  ret void
}

define void @sudot_single_za32_u16_vg1x2_tuple(ptr %ptr, i64 %stride, <vscale x 16 x i8> %zn) #0 {
; CHECK-LABEL: sudot_single_za32_u16_vg1x2_tuple:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
; CHECK-NEXT:    addvl sp, sp, #-3
; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
; CHECK-NEXT:    ptrue pn8.b
; CHECK-NEXT:    mov w8, wzr
; CHECK-NEXT:    str z10, [sp, #1, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z9, [sp, #2, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    ld1b { z1.b, z9.b }, pn8/z, [x0]
; CHECK-NEXT:    ld1b { z2.b, z10.b }, pn8/z, [x0, x1]
; CHECK-NEXT:    sudot za.s[w8, 0, vgx2], { z1.b, z2.b }, z0.b
; CHECK-NEXT:    sudot za.s[w8, 0, vgx2], { z9.b, z10.b }, z0.b
; CHECK-NEXT:    ldr z10, [sp, #1, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z9, [sp, #2, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
; CHECK-NEXT:    addvl sp, sp, #3
; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
; CHECK-NEXT:    ret
entry:
  %0 = tail call target("aarch64.svcount") @llvm.aarch64.sve.ptrue.c8()
  %1 = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld1.pn.x2.nxv16i8(target("aarch64.svcount") %0, ptr %ptr)
  %2 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 0
  %3 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 1
  %arrayidx2 = getelementptr inbounds i8, ptr %ptr, i64 %stride
  %4 = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld1.pn.x2.nxv16i8(target("aarch64.svcount") %0, ptr %arrayidx2)
  %5 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %4, 0
  %6 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %4, 1
  call void @llvm.aarch64.sme.sudot.single.za32.vg1x2.nxv16i8(i32 0, <vscale x 16 x i8> %2, <vscale x 16 x i8> %5, <vscale x 16 x i8> %zn)
  call void @llvm.aarch64.sme.sudot.single.za32.vg1x2.nxv16i8(i32 0, <vscale x 16 x i8> %3, <vscale x 16 x i8> %6, <vscale x 16 x i8> %zn)
  ret void
}

define void @sudot_single_za32_u8_vg1x4(i32 %slice, <vscale x 16 x i8> %unused, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2, <vscale x 16 x i8> %zn3, <vscale x 16 x i8> %zn4) #0 {
; CHECK-LABEL: sudot_single_za32_u8_vg1x4:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    sudot za.s[w8, 0, vgx4], { z1.b - z4.b }, z5.b
; CHECK-NEXT:    sudot za.s[w8, 7, vgx4], { z1.b - z4.b }, z5.b
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.sudot.single.za32.vg1x4.nxv16i8(i32 %slice, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2, <vscale x 16 x i8> %zn3, <vscale x 16 x i8> %zn4)
  %slice2 = add i32 %slice, 7
  call void @llvm.aarch64.sme.sudot.single.za32.vg1x4.nxv16i8(i32 %slice2, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2, <vscale x 16 x i8> %zn3, <vscale x 16 x i8> %zn4)
  ret void
}

define void @sudot_single_za32_u16_vg1x4_tuple(ptr %ptr, i64 %stride, <vscale x 16 x i8> %zn) #0 {
; CHECK-LABEL: sudot_single_za32_u16_vg1x4_tuple:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
; CHECK-NEXT:    addvl sp, sp, #-9
; CHECK-NEXT:    lsl x9, x1, #1
; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
; CHECK-NEXT:    ptrue pn8.b
; CHECK-NEXT:    str z23, [sp, #1, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    mov w8, wzr
; CHECK-NEXT:    str z22, [sp, #2, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    add x10, x9, x1
; CHECK-NEXT:    str z21, [sp, #3, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z20, [sp, #4, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z19, [sp, #5, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z18, [sp, #6, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z17, [sp, #7, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z16, [sp, #8, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    ld1b { z16.b, z20.b, z24.b, z28.b }, pn8/z, [x0]
; CHECK-NEXT:    ld1b { z17.b, z21.b, z25.b, z29.b }, pn8/z, [x0, x1]
; CHECK-NEXT:    ld1b { z18.b, z22.b, z26.b, z30.b }, pn8/z, [x0, x9]
; CHECK-NEXT:    ld1b { z19.b, z23.b, z27.b, z31.b }, pn8/z, [x0, x10]
; CHECK-NEXT:    sudot za.s[w8, 0, vgx4], { z16.b - z19.b }, z0.b
; CHECK-NEXT:    sudot za.s[w8, 0, vgx4], { z20.b - z23.b }, z0.b
; CHECK-NEXT:    sudot za.s[w8, 0, vgx4], { z24.b - z27.b }, z0.b
; CHECK-NEXT:    sudot za.s[w8, 0, vgx4], { z28.b - z31.b }, z0.b
; CHECK-NEXT:    ldr z23, [sp, #1, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z22, [sp, #2, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z21, [sp, #3, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z20, [sp, #4, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z19, [sp, #5, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z18, [sp, #6, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z17, [sp, #7, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z16, [sp, #8, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
; CHECK-NEXT:    addvl sp, sp, #9
; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
; CHECK-NEXT:    ret
entry:
  %0 = tail call target("aarch64.svcount") @llvm.aarch64.sve.ptrue.c8()
  %1 = tail 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") %0, ptr %ptr)
  %2 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 0
  %3 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 1
  %4 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 2
  %5 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 3
  %arrayidx2 = getelementptr inbounds i8, ptr %ptr, i64 %stride
  %6 = tail 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") %0, ptr %arrayidx2)
  %7 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %6, 0
  %8 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %6, 1
  %9 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %6, 2
  %10 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %6, 3
  %mul3 = shl i64 %stride, 1
  %arrayidx4 = getelementptr inbounds i8, ptr %ptr, i64 %mul3
  %11 = tail 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") %0, ptr %arrayidx4)
  %12 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %11, 0
  %13 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %11, 1
  %14 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %11, 2
  %15 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %11, 3
  %mul5 = mul i64 %stride, 3
  %arrayidx6 = getelementptr inbounds i8, ptr %ptr, i64 %mul5
  %16 = tail 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") %0, ptr %arrayidx6)
  %17 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %16, 0
  %18 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %16, 1
  %19 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %16, 2
  %20 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %16, 3
  call void @llvm.aarch64.sme.sudot.single.za32.vg1x4.nxv16i8(i32 0, <vscale x 16 x i8> %2, <vscale x 16 x i8> %7, <vscale x 16 x i8> %12, <vscale x 16 x i8> %17, <vscale x 16 x i8> %zn)
  call void @llvm.aarch64.sme.sudot.single.za32.vg1x4.nxv16i8(i32 0, <vscale x 16 x i8> %3, <vscale x 16 x i8> %8, <vscale x 16 x i8> %13, <vscale x 16 x i8> %18, <vscale x 16 x i8> %zn)
  call void @llvm.aarch64.sme.sudot.single.za32.vg1x4.nxv16i8(i32 0, <vscale x 16 x i8> %4, <vscale x 16 x i8> %9, <vscale x 16 x i8> %14, <vscale x 16 x i8> %19, <vscale x 16 x i8> %zn)
  call void @llvm.aarch64.sme.sudot.single.za32.vg1x4.nxv16i8(i32 0, <vscale x 16 x i8> %5, <vscale x 16 x i8> %10, <vscale x 16 x i8> %15, <vscale x 16 x i8> %20, <vscale x 16 x i8> %zn)
  ret void
}

; == Multi, indexed (unsigned) ==

define void @udot_lane_za32_u16_vg1x2(i32 %slice, <vscale x 16 x i8> %unused, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2) #0 {
; CHECK-LABEL: udot_lane_za32_u16_vg1x2:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z5.d, z2.d
; CHECK-NEXT:    mov z4.d, z1.d
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    udot za.s[w8, 0, vgx2], { z4.h, z5.h }, z3.h[3]
; CHECK-NEXT:    udot za.s[w8, 7, vgx2], { z4.h, z5.h }, z3.h[3]
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.udot.lane.za32.vg1x2.nxv8i16(i32 %slice, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, i32 3)
  %slice2 = add i32 %slice, 7
  call void @llvm.aarch64.sme.udot.lane.za32.vg1x2.nxv8i16(i32 %slice2, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, i32 3)
  ret void
}

define void @udot_lane_za32_u16_vg1x4(i32 %slice, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, <vscale x 8 x i16> %zn3, <vscale x 8 x i16> %zn4) #0 {
; CHECK-LABEL: udot_lane_za32_u16_vg1x4:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    udot za.s[w8, 0, vgx4], { z0.h - z3.h }, z4.h[3]
; CHECK-NEXT:    udot za.s[w8, 7, vgx4], { z0.h - z3.h }, z4.h[3]
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.udot.lane.za32.vg1x4.nxv8i16(i32 %slice, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, <vscale x 8 x i16> %zn3,
                                                           <vscale x 8 x i16> %zn4, i32 3)
  %slice2 = add i32 %slice, 7
  call void @llvm.aarch64.sme.udot.lane.za32.vg1x4.nxv8i16(i32 %slice2, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, <vscale x 8 x i16> %zn3,
                                                           <vscale x 8 x i16> %zn4, i32 3)
  ret void
}

define void @udot_lane_za32_u8_vg1x2(i32 %slice, <vscale x 16 x i8> %unused, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2) #0 {
; CHECK-LABEL: udot_lane_za32_u8_vg1x2:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z5.d, z2.d
; CHECK-NEXT:    mov z4.d, z1.d
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    udot za.s[w8, 0, vgx2], { z4.b, z5.b }, z3.b[3]
; CHECK-NEXT:    udot za.s[w8, 7, vgx2], { z4.b, z5.b }, z3.b[3]
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.udot.lane.za32.vg1x2.nxv16i8(i32 %slice, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2, i32 3)
  %slice2 = add i32 %slice, 7
  call void @llvm.aarch64.sme.udot.lane.za32.vg1x2.nxv16i8(i32 %slice2, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2, i32 3)
  ret void
}

define void @udot_lane_za32_u8_vg1x4(i32 %slice, <vscale x 16 x i8> %unused, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2, <vscale x 16 x i8> %zn3, <vscale x 16 x i8> %zn4) #0 {
; CHECK-LABEL: udot_lane_za32_u8_vg1x4:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z27.d, z4.d
; CHECK-NEXT:    mov z26.d, z3.d
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    mov z25.d, z2.d
; CHECK-NEXT:    mov z24.d, z1.d
; CHECK-NEXT:    udot za.s[w8, 0, vgx4], { z24.b - z27.b }, z5.b[3]
; CHECK-NEXT:    udot za.s[w8, 7, vgx4], { z24.b - z27.b }, z5.b[3]
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.udot.lane.za32.vg1x4.nxv16i8(i32 %slice, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2, <vscale x 16 x i8> %zn3,
                                                           <vscale x 16 x i8> %zn4, i32 3)
  %slice2 = add i32 %slice, 7
  call void @llvm.aarch64.sme.udot.lane.za32.vg1x4.nxv16i8(i32 %slice2, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2, <vscale x 16 x i8> %zn3,
                                                           <vscale x 16 x i8> %zn4, i32 3)
  ret void
}

define void @udot_form_2x_tuple(ptr %ptr, i64 %stride) #0 {
; CHECK-LABEL: udot_form_2x_tuple:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    ptrue pn8.b
; CHECK-NEXT:    mov w8, wzr
; CHECK-NEXT:    ld1b { z16.b, z24.b }, pn8/z, [x0]
; CHECK-NEXT:    ld1b { z17.b, z25.b }, pn8/z, [x0, x1]
; CHECK-NEXT:    udot za.s[w8, 0, vgx2], { z16.b, z17.b }, z0.b[0]
; CHECK-NEXT:    udot za.s[w8, 0, vgx2], { z24.b, z25.b }, z0.b[0]
; CHECK-NEXT:    ret
entry:
  %0 = tail call target("aarch64.svcount") @llvm.aarch64.sve.ptrue.c8()
  %1 = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld1.pn.x2.nxv16i8(target("aarch64.svcount") %0, ptr %ptr)
  %2 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 0
  %3 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 1
  %arrayidx2 = getelementptr inbounds i8, ptr %ptr, i64 %stride
  %4 = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld1.pn.x2.nxv16i8(target("aarch64.svcount") %0, ptr %arrayidx2)
  %5 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %4, 0
  %6 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %4, 1
  tail call void @llvm.aarch64.sme.udot.lane.za32.vg1x2.nxv16i8(i32 0, <vscale x 16 x i8> %2, <vscale x 16 x i8> %5, <vscale x 16 x i8> poison, i32 0)
  tail call void @llvm.aarch64.sme.udot.lane.za32.vg1x2.nxv16i8(i32 0, <vscale x 16 x i8> %3, <vscale x 16 x i8> %6, <vscale x 16 x i8> poison, i32 0)
  ret void
}

define void @udot_form_2x_tuple_svecc(ptr %ptr, i64 %stride, <vscale x 16 x i8> %scalable_arg) #0 {
; CHECK-LABEL: udot_form_2x_tuple_svecc:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
; CHECK-NEXT:    addvl sp, sp, #-3
; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
; CHECK-NEXT:    ptrue pn8.b
; CHECK-NEXT:    mov w8, wzr
; CHECK-NEXT:    str z11, [sp, #1, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z10, [sp, #2, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    ld1b { z2.b, z10.b }, pn8/z, [x0]
; CHECK-NEXT:    ld1b { z3.b, z11.b }, pn8/z, [x0, x1]
; CHECK-NEXT:    udot za.s[w8, 0, vgx2], { z2.b, z3.b }, z0.b[0]
; CHECK-NEXT:    udot za.s[w8, 0, vgx2], { z10.b, z11.b }, z0.b[0]
; CHECK-NEXT:    ldr z11, [sp, #1, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z10, [sp, #2, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
; CHECK-NEXT:    str z0, [x0]
; CHECK-NEXT:    addvl sp, sp, #3
; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
; CHECK-NEXT:    ret
entry:
  %0 = tail call target("aarch64.svcount") @llvm.aarch64.sve.ptrue.c8()
  %1 = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld1.pn.x2.nxv16i8(target("aarch64.svcount") %0, ptr %ptr)
  %2 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 0
  %3 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 1
  %arrayidx2 = getelementptr inbounds i8, ptr %ptr, i64 %stride
  %4 = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld1.pn.x2.nxv16i8(target("aarch64.svcount") %0, ptr %arrayidx2)
  %5 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %4, 0
  %6 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %4, 1
  tail call void @llvm.aarch64.sme.udot.lane.za32.vg1x2.nxv16i8(i32 0, <vscale x 16 x i8> %2, <vscale x 16 x i8> %5, <vscale x 16 x i8> poison, i32 0)
  tail call void @llvm.aarch64.sme.udot.lane.za32.vg1x2.nxv16i8(i32 0, <vscale x 16 x i8> %3, <vscale x 16 x i8> %6, <vscale x 16 x i8> poison, i32 0)
  store <vscale x 16 x i8> %scalable_arg, ptr %ptr
  ret void
}

define void @udot_form_4x_tuple(ptr %ptr, i64 %stride) #0 {
; CHECK-LABEL: udot_form_4x_tuple:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    lsl x9, x1, #1
; CHECK-NEXT:    ptrue pn8.b
; CHECK-NEXT:    mov w8, wzr
; CHECK-NEXT:    ld1b { z16.b, z20.b, z24.b, z28.b }, pn8/z, [x0]
; CHECK-NEXT:    ld1b { z17.b, z21.b, z25.b, z29.b }, pn8/z, [x0, x1]
; CHECK-NEXT:    add x10, x9, x1
; CHECK-NEXT:    ld1b { z18.b, z22.b, z26.b, z30.b }, pn8/z, [x0, x9]
; CHECK-NEXT:    ld1b { z19.b, z23.b, z27.b, z31.b }, pn8/z, [x0, x10]
; CHECK-NEXT:    udot za.s[w8, 0, vgx4], { z16.b - z19.b }, z0.b[0]
; CHECK-NEXT:    udot za.s[w8, 0, vgx4], { z20.b - z23.b }, z0.b[0]
; CHECK-NEXT:    udot za.s[w8, 0, vgx4], { z24.b - z27.b }, z0.b[0]
; CHECK-NEXT:    udot za.s[w8, 0, vgx4], { z28.b - z31.b }, z0.b[0]
; CHECK-NEXT:    ret
entry:
  %0 = tail call target("aarch64.svcount") @llvm.aarch64.sve.ptrue.c8()
  %1 = tail 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") %0, ptr %ptr)
  %2 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 0
  %3 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 1
  %4 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 2
  %5 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 3
  %arrayidx2 = getelementptr inbounds i8, ptr %ptr, i64 %stride
  %6 = tail 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") %0, ptr %arrayidx2)
  %7 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %6, 0
  %8 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %6, 1
  %9 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %6, 2
  %10 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %6, 3
  %mul3 = shl i64 %stride, 1
  %arrayidx4 = getelementptr inbounds i8, ptr %ptr, i64 %mul3
  %11 = tail 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") %0, ptr %arrayidx4)
  %12 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %11, 0
  %13 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %11, 1
  %14 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %11, 2
  %15 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %11, 3
  %mul5 = mul i64 %stride, 3
  %arrayidx6 = getelementptr inbounds i8, ptr %ptr, i64 %mul5
  %16 = tail 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") %0, ptr %arrayidx6)
  %17 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %16, 0
  %18 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %16, 1
  %19 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %16, 2
  %20 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %16, 3
  tail call void @llvm.aarch64.sme.udot.lane.za32.vg1x4.nxv16i8(i32 0, <vscale x 16 x i8> %2, <vscale x 16 x i8> %7, <vscale x 16 x i8> %12, <vscale x 16 x i8> %17, <vscale x 16 x i8> poison, i32 0)
  tail call void @llvm.aarch64.sme.udot.lane.za32.vg1x4.nxv16i8(i32 0, <vscale x 16 x i8> %3, <vscale x 16 x i8> %8, <vscale x 16 x i8> %13, <vscale x 16 x i8> %18, <vscale x 16 x i8> poison, i32 0)
  tail call void @llvm.aarch64.sme.udot.lane.za32.vg1x4.nxv16i8(i32 0, <vscale x 16 x i8> %4, <vscale x 16 x i8> %9, <vscale x 16 x i8> %14, <vscale x 16 x i8> %19, <vscale x 16 x i8> poison, i32 0)
  tail call void @llvm.aarch64.sme.udot.lane.za32.vg1x4.nxv16i8(i32 0, <vscale x 16 x i8> %5, <vscale x 16 x i8> %10, <vscale x 16 x i8> %15, <vscale x 16 x i8> %20, <vscale x 16 x i8> poison, i32 0)
  ret void
}

define void @udot_form_4x_tuple_svecc(ptr %ptr, i64 %stride, <vscale x 16 x i8> %scalable_arg) #0 {
; CHECK-LABEL: udot_form_4x_tuple_svecc:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
; CHECK-NEXT:    addvl sp, sp, #-9
; CHECK-NEXT:    lsl x9, x1, #1
; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
; CHECK-NEXT:    ptrue pn8.b
; CHECK-NEXT:    str z23, [sp, #1, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    mov w8, wzr
; CHECK-NEXT:    str z22, [sp, #2, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    add x10, x9, x1
; CHECK-NEXT:    str z21, [sp, #3, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z20, [sp, #4, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z19, [sp, #5, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z18, [sp, #6, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z17, [sp, #7, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z16, [sp, #8, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    ld1b { z16.b, z20.b, z24.b, z28.b }, pn8/z, [x0]
; CHECK-NEXT:    ld1b { z17.b, z21.b, z25.b, z29.b }, pn8/z, [x0, x1]
; CHECK-NEXT:    ld1b { z18.b, z22.b, z26.b, z30.b }, pn8/z, [x0, x9]
; CHECK-NEXT:    ld1b { z19.b, z23.b, z27.b, z31.b }, pn8/z, [x0, x10]
; CHECK-NEXT:    udot za.s[w8, 0, vgx4], { z16.b - z19.b }, z0.b[0]
; CHECK-NEXT:    udot za.s[w8, 0, vgx4], { z20.b - z23.b }, z0.b[0]
; CHECK-NEXT:    udot za.s[w8, 0, vgx4], { z24.b - z27.b }, z0.b[0]
; CHECK-NEXT:    udot za.s[w8, 0, vgx4], { z28.b - z31.b }, z0.b[0]
; CHECK-NEXT:    ldr z23, [sp, #1, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z22, [sp, #2, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z21, [sp, #3, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z20, [sp, #4, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z19, [sp, #5, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z18, [sp, #6, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z17, [sp, #7, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z16, [sp, #8, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
; CHECK-NEXT:    str z0, [x0]
; CHECK-NEXT:    addvl sp, sp, #9
; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
; CHECK-NEXT:    ret
entry:
  %0 = tail call target("aarch64.svcount") @llvm.aarch64.sve.ptrue.c8()
  %1 = tail 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") %0, ptr %ptr)
  %2 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 0
  %3 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 1
  %4 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 2
  %5 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 3
  %arrayidx2 = getelementptr inbounds i8, ptr %ptr, i64 %stride
  %6 = tail 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") %0, ptr %arrayidx2)
  %7 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %6, 0
  %8 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %6, 1
  %9 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %6, 2
  %10 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %6, 3
  %mul3 = shl i64 %stride, 1
  %arrayidx4 = getelementptr inbounds i8, ptr %ptr, i64 %mul3
  %11 = tail 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") %0, ptr %arrayidx4)
  %12 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %11, 0
  %13 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %11, 1
  %14 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %11, 2
  %15 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %11, 3
  %mul5 = mul i64 %stride, 3
  %arrayidx6 = getelementptr inbounds i8, ptr %ptr, i64 %mul5
  %16 = tail 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") %0, ptr %arrayidx6)
  %17 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %16, 0
  %18 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %16, 1
  %19 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %16, 2
  %20 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %16, 3
  tail call void @llvm.aarch64.sme.udot.lane.za32.vg1x4.nxv16i8(i32 0, <vscale x 16 x i8> %2, <vscale x 16 x i8> %7, <vscale x 16 x i8> %12, <vscale x 16 x i8> %17, <vscale x 16 x i8> poison, i32 0)
  tail call void @llvm.aarch64.sme.udot.lane.za32.vg1x4.nxv16i8(i32 0, <vscale x 16 x i8> %3, <vscale x 16 x i8> %8, <vscale x 16 x i8> %13, <vscale x 16 x i8> %18, <vscale x 16 x i8> poison, i32 0)
  tail call void @llvm.aarch64.sme.udot.lane.za32.vg1x4.nxv16i8(i32 0, <vscale x 16 x i8> %4, <vscale x 16 x i8> %9, <vscale x 16 x i8> %14, <vscale x 16 x i8> %19, <vscale x 16 x i8> poison, i32 0)
  tail call void @llvm.aarch64.sme.udot.lane.za32.vg1x4.nxv16i8(i32 0, <vscale x 16 x i8> %5, <vscale x 16 x i8> %10, <vscale x 16 x i8> %15, <vscale x 16 x i8> %20, <vscale x 16 x i8> poison, i32 0)
  store <vscale x 16 x i8> %scalable_arg, ptr %ptr
  ret void
}

define void @udot_single_za32_u16_vg1x4_x2load_x4tuple(ptr %ptr, i64 %stride, <vscale x 16 x i8> %zn) #0 {
; CHECK-LABEL: udot_single_za32_u16_vg1x4_x2load_x4tuple:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
; CHECK-NEXT:    addvl sp, sp, #-5
; CHECK-NEXT:    lsl x9, x1, #1
; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
; CHECK-NEXT:    ptrue pn8.b
; CHECK-NEXT:    str z12, [sp, #1, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    st1b { z10.b, z11.b }, pn8, [sp, #2, mul vl] // 32-byte Folded Spill
; CHECK-NEXT:    ptrue pn8.b
; CHECK-NEXT:    str z9, [sp, #4, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    add x10, x9, x1
; CHECK-NEXT:    mov w8, wzr
; CHECK-NEXT:    ld1b { z1.b, z9.b }, pn8/z, [x0]
; CHECK-NEXT:    ld1b { z2.b, z10.b }, pn8/z, [x0, x1]
; CHECK-NEXT:    ld1b { z3.b, z11.b }, pn8/z, [x0, x9]
; CHECK-NEXT:    ld1b { z4.b, z12.b }, pn8/z, [x0, x10]
; CHECK-NEXT:    ptrue pn8.b
; CHECK-NEXT:    udot za.s[w8, 0, vgx4], { z1.b - z4.b }, z0.b
; CHECK-NEXT:    udot za.s[w8, 0, vgx4], { z9.b - z12.b }, z0.b
; CHECK-NEXT:    ldr z12, [sp, #1, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ld1b { z10.b, z11.b }, pn8/z, [sp, #2, mul vl] // 32-byte Folded Reload
; CHECK-NEXT:    ldr z9, [sp, #4, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
; CHECK-NEXT:    addvl sp, sp, #5
; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
; CHECK-NEXT:    ret
entry:
  %0 = tail call target("aarch64.svcount") @llvm.aarch64.sve.ptrue.c8()
  %1 = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld1.pn.x2.nxv16i8(target("aarch64.svcount") %0, ptr %ptr)
  %2 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 0
  %3 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 1
  %arrayidx2 = getelementptr inbounds i8, ptr %ptr, i64 %stride
  %4 = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld1.pn.x2.nxv16i8(target("aarch64.svcount") %0, ptr %arrayidx2)
  %5 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %4, 0
  %6 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %4, 1
  %mul3 = shl i64 %stride, 1
  %arrayidx4 = getelementptr inbounds i8, ptr %ptr, i64 %mul3
  %7 = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld1.pn.x2.nxv16i8(target("aarch64.svcount") %0, ptr %arrayidx4)
  %8 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %7, 0
  %9 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %7, 1
  %mul5 = mul i64 %stride, 3
  %arrayidx6 = getelementptr inbounds i8, ptr %ptr, i64 %mul5
  %10 = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld1.pn.x2.nxv16i8(target("aarch64.svcount") %0, ptr %arrayidx6)
  %11 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %10, 0
  %12 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %10, 1
  call void @llvm.aarch64.sme.udot.single.za32.vg1x4.nxv16i8(i32 0, <vscale x 16 x i8> %2, <vscale x 16 x i8> %5, <vscale x 16 x i8> %8, <vscale x 16 x i8> %11, <vscale x 16 x i8> %zn)
  call void @llvm.aarch64.sme.udot.single.za32.vg1x4.nxv16i8(i32 0, <vscale x 16 x i8> %3, <vscale x 16 x i8> %6, <vscale x 16 x i8> %9, <vscale x 16 x i8> %12, <vscale x 16 x i8> %zn)
  ret void
}

define void @udot_lane_za64_u16_vg1x2(i32 %slice, <vscale x 16 x i8> %unused, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2) #1 {
; CHECK-LABEL: udot_lane_za64_u16_vg1x2:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z5.d, z2.d
; CHECK-NEXT:    mov z4.d, z1.d
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    udot za.d[w8, 0, vgx2], { z4.h, z5.h }, z3.h[1]
; CHECK-NEXT:    udot za.d[w8, 7, vgx2], { z4.h, z5.h }, z3.h[1]
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.udot.lane.za64.vg1x2.nxv8i16(i32 %slice, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, i32 1)
  %slice2 = add i32 %slice, 7
  call void @llvm.aarch64.sme.udot.lane.za64.vg1x2.nxv8i16(i32 %slice2, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, i32 1)
  ret void
}

define void @udot_lane_za64_u16_vg1x4(i32 %slice, <vscale x 16 x i8> %unused, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, <vscale x 8 x i16> %zn3, <vscale x 8 x i16> %zn4) #1 {
; CHECK-LABEL: udot_lane_za64_u16_vg1x4:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z27.d, z4.d
; CHECK-NEXT:    mov z26.d, z3.d
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    mov z25.d, z2.d
; CHECK-NEXT:    mov z24.d, z1.d
; CHECK-NEXT:    udot za.d[w8, 0, vgx4], { z24.h - z27.h }, z5.h[1]
; CHECK-NEXT:    udot za.d[w8, 7, vgx4], { z24.h - z27.h }, z5.h[1]
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.udot.lane.za64.vg1x4.nxv8i16(i32 %slice, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, <vscale x 8 x i16> %zn3,
                                                           <vscale x 8 x i16> %zn4, i32 1)
  %slice2 = add i32 %slice, 7
  call void @llvm.aarch64.sme.udot.lane.za64.vg1x4.nxv8i16(i32 %slice2, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, <vscale x 8 x i16> %zn3,
                                                           <vscale x 8 x i16> %zn4, i32 1)
  ret void
}

define void @usdot_lane_za32_u8_vg1x2(i32 %slice, <vscale x 16 x i8> %unused, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2) #0 {
; CHECK-LABEL: usdot_lane_za32_u8_vg1x2:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z5.d, z2.d
; CHECK-NEXT:    mov z4.d, z1.d
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    usdot za.s[w8, 0, vgx2], { z4.b, z5.b }, z3.b[3]
; CHECK-NEXT:    usdot za.s[w8, 7, vgx2], { z4.b, z5.b }, z3.b[3]
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.usdot.lane.za32.vg1x2.nxv16i8(i32 %slice, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2, i32 3)
  %slice2 = add i32 %slice, 7
  call void @llvm.aarch64.sme.usdot.lane.za32.vg1x2.nxv16i8(i32 %slice2, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2, i32 3)
  ret void
}

define void @usdot_lane_za32_u8_vg1x4(i32 %slice, <vscale x 16 x i8> %unused, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2, <vscale x 16 x i8> %zn3, <vscale x 16 x i8> %zn4) #0 {
; CHECK-LABEL: usdot_lane_za32_u8_vg1x4:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z27.d, z4.d
; CHECK-NEXT:    mov z26.d, z3.d
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    mov z25.d, z2.d
; CHECK-NEXT:    mov z24.d, z1.d
; CHECK-NEXT:    usdot za.s[w8, 0, vgx4], { z24.b - z27.b }, z5.b[3]
; CHECK-NEXT:    usdot za.s[w8, 7, vgx4], { z24.b - z27.b }, z5.b[3]
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.usdot.lane.za32.vg1x4.nxv16i8(i32 %slice, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2, <vscale x 16 x i8> %zn3,
                                                            <vscale x 16 x i8> %zn4, i32 3)
  %slice2 = add i32 %slice, 7
  call void @llvm.aarch64.sme.usdot.lane.za32.vg1x4.nxv16i8(i32 %slice2, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2, <vscale x 16 x i8> %zn3,
                                                            <vscale x 16 x i8> %zn4, i32 3)
  ret void
}

define void @usdot_form_2x_tuple(ptr %ptr, i64 %stride) #0 {
; CHECK-LABEL: usdot_form_2x_tuple:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    ptrue pn8.b
; CHECK-NEXT:    mov w8, wzr
; CHECK-NEXT:    ld1b { z16.b, z24.b }, pn8/z, [x0]
; CHECK-NEXT:    ld1b { z17.b, z25.b }, pn8/z, [x0, x1]
; CHECK-NEXT:    usdot za.s[w8, 0, vgx2], { z16.b, z17.b }, z0.b[0]
; CHECK-NEXT:    usdot za.s[w8, 0, vgx2], { z24.b, z25.b }, z0.b[0]
; CHECK-NEXT:    ret
entry:
  %0 = tail call target("aarch64.svcount") @llvm.aarch64.sve.ptrue.c8()
  %1 = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld1.pn.x2.nxv16i8(target("aarch64.svcount") %0, ptr %ptr)
  %2 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 0
  %3 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 1
  %arrayidx2 = getelementptr inbounds i8, ptr %ptr, i64 %stride
  %4 = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld1.pn.x2.nxv16i8(target("aarch64.svcount") %0, ptr %arrayidx2)
  %5 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %4, 0
  %6 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %4, 1
  tail call void @llvm.aarch64.sme.usdot.lane.za32.vg1x2.nxv16i8(i32 0, <vscale x 16 x i8> %2, <vscale x 16 x i8> %5, <vscale x 16 x i8> poison, i32 0)
  tail call void @llvm.aarch64.sme.usdot.lane.za32.vg1x2.nxv16i8(i32 0, <vscale x 16 x i8> %3, <vscale x 16 x i8> %6, <vscale x 16 x i8> poison, i32 0)
  ret void
}

define void @usdot_form_2x_tuple_svecc(ptr %ptr, i64 %stride, <vscale x 16 x i8> %scalable_arg) #0 {
; CHECK-LABEL: usdot_form_2x_tuple_svecc:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
; CHECK-NEXT:    addvl sp, sp, #-3
; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
; CHECK-NEXT:    ptrue pn8.b
; CHECK-NEXT:    mov w8, wzr
; CHECK-NEXT:    str z11, [sp, #1, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z10, [sp, #2, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    ld1b { z2.b, z10.b }, pn8/z, [x0]
; CHECK-NEXT:    ld1b { z3.b, z11.b }, pn8/z, [x0, x1]
; CHECK-NEXT:    usdot za.s[w8, 0, vgx2], { z2.b, z3.b }, z0.b[0]
; CHECK-NEXT:    usdot za.s[w8, 0, vgx2], { z10.b, z11.b }, z0.b[0]
; CHECK-NEXT:    ldr z11, [sp, #1, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z10, [sp, #2, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
; CHECK-NEXT:    str z0, [x0]
; CHECK-NEXT:    addvl sp, sp, #3
; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
; CHECK-NEXT:    ret
entry:
  %0 = tail call target("aarch64.svcount") @llvm.aarch64.sve.ptrue.c8()
  %1 = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld1.pn.x2.nxv16i8(target("aarch64.svcount") %0, ptr %ptr)
  %2 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 0
  %3 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 1
  %arrayidx2 = getelementptr inbounds i8, ptr %ptr, i64 %stride
  %4 = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld1.pn.x2.nxv16i8(target("aarch64.svcount") %0, ptr %arrayidx2)
  %5 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %4, 0
  %6 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %4, 1
  tail call void @llvm.aarch64.sme.usdot.lane.za32.vg1x2.nxv16i8(i32 0, <vscale x 16 x i8> %2, <vscale x 16 x i8> %5, <vscale x 16 x i8> poison, i32 0)
  tail call void @llvm.aarch64.sme.usdot.lane.za32.vg1x2.nxv16i8(i32 0, <vscale x 16 x i8> %3, <vscale x 16 x i8> %6, <vscale x 16 x i8> poison, i32 0)
  store <vscale x 16 x i8> %scalable_arg, ptr %ptr
  ret void
}

define void @usdot_form_4x_tuple(ptr %ptr, i64 %stride) #0 {
; CHECK-LABEL: usdot_form_4x_tuple:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    lsl x9, x1, #1
; CHECK-NEXT:    ptrue pn8.b
; CHECK-NEXT:    mov w8, wzr
; CHECK-NEXT:    ld1b { z16.b, z20.b, z24.b, z28.b }, pn8/z, [x0]
; CHECK-NEXT:    ld1b { z17.b, z21.b, z25.b, z29.b }, pn8/z, [x0, x1]
; CHECK-NEXT:    add x10, x9, x1
; CHECK-NEXT:    ld1b { z18.b, z22.b, z26.b, z30.b }, pn8/z, [x0, x9]
; CHECK-NEXT:    ld1b { z19.b, z23.b, z27.b, z31.b }, pn8/z, [x0, x10]
; CHECK-NEXT:    usdot za.s[w8, 0, vgx4], { z16.b - z19.b }, z0.b[0]
; CHECK-NEXT:    usdot za.s[w8, 0, vgx4], { z20.b - z23.b }, z0.b[0]
; CHECK-NEXT:    usdot za.s[w8, 0, vgx4], { z24.b - z27.b }, z0.b[0]
; CHECK-NEXT:    usdot za.s[w8, 0, vgx4], { z28.b - z31.b }, z0.b[0]
; CHECK-NEXT:    ret
entry:
  %0 = tail call target("aarch64.svcount") @llvm.aarch64.sve.ptrue.c8()
  %1 = tail 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") %0, ptr %ptr)
  %2 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 0
  %3 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 1
  %4 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 2
  %5 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 3
  %arrayidx2 = getelementptr inbounds i8, ptr %ptr, i64 %stride
  %6 = tail 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") %0, ptr %arrayidx2)
  %7 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %6, 0
  %8 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %6, 1
  %9 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %6, 2
  %10 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %6, 3
  %mul3 = shl i64 %stride, 1
  %arrayidx4 = getelementptr inbounds i8, ptr %ptr, i64 %mul3
  %11 = tail 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") %0, ptr %arrayidx4)
  %12 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %11, 0
  %13 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %11, 1
  %14 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %11, 2
  %15 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %11, 3
  %mul5 = mul i64 %stride, 3
  %arrayidx6 = getelementptr inbounds i8, ptr %ptr, i64 %mul5
  %16 = tail 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") %0, ptr %arrayidx6)
  %17 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %16, 0
  %18 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %16, 1
  %19 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %16, 2
  %20 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %16, 3
  tail call void @llvm.aarch64.sme.usdot.lane.za32.vg1x4.nxv16i8(i32 0, <vscale x 16 x i8> %2, <vscale x 16 x i8> %7, <vscale x 16 x i8> %12, <vscale x 16 x i8> %17, <vscale x 16 x i8> poison, i32 0)
  tail call void @llvm.aarch64.sme.usdot.lane.za32.vg1x4.nxv16i8(i32 0, <vscale x 16 x i8> %3, <vscale x 16 x i8> %8, <vscale x 16 x i8> %13, <vscale x 16 x i8> %18, <vscale x 16 x i8> poison, i32 0)
  tail call void @llvm.aarch64.sme.usdot.lane.za32.vg1x4.nxv16i8(i32 0, <vscale x 16 x i8> %4, <vscale x 16 x i8> %9, <vscale x 16 x i8> %14, <vscale x 16 x i8> %19, <vscale x 16 x i8> poison, i32 0)
  tail call void @llvm.aarch64.sme.usdot.lane.za32.vg1x4.nxv16i8(i32 0, <vscale x 16 x i8> %5, <vscale x 16 x i8> %10, <vscale x 16 x i8> %15, <vscale x 16 x i8> %20, <vscale x 16 x i8> poison, i32 0)
  ret void
}

define void @usdot_form_4x_tuple_svecc(ptr %ptr, i64 %stride, <vscale x 16 x i8> %scalable_arg) #0 {
; CHECK-LABEL: usdot_form_4x_tuple_svecc:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
; CHECK-NEXT:    addvl sp, sp, #-9
; CHECK-NEXT:    lsl x9, x1, #1
; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
; CHECK-NEXT:    ptrue pn8.b
; CHECK-NEXT:    str z23, [sp, #1, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    mov w8, wzr
; CHECK-NEXT:    str z22, [sp, #2, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    add x10, x9, x1
; CHECK-NEXT:    str z21, [sp, #3, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z20, [sp, #4, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z19, [sp, #5, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z18, [sp, #6, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z17, [sp, #7, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z16, [sp, #8, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    ld1b { z16.b, z20.b, z24.b, z28.b }, pn8/z, [x0]
; CHECK-NEXT:    ld1b { z17.b, z21.b, z25.b, z29.b }, pn8/z, [x0, x1]
; CHECK-NEXT:    ld1b { z18.b, z22.b, z26.b, z30.b }, pn8/z, [x0, x9]
; CHECK-NEXT:    ld1b { z19.b, z23.b, z27.b, z31.b }, pn8/z, [x0, x10]
; CHECK-NEXT:    usdot za.s[w8, 0, vgx4], { z16.b - z19.b }, z0.b[0]
; CHECK-NEXT:    usdot za.s[w8, 0, vgx4], { z20.b - z23.b }, z0.b[0]
; CHECK-NEXT:    usdot za.s[w8, 0, vgx4], { z24.b - z27.b }, z0.b[0]
; CHECK-NEXT:    usdot za.s[w8, 0, vgx4], { z28.b - z31.b }, z0.b[0]
; CHECK-NEXT:    ldr z23, [sp, #1, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z22, [sp, #2, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z21, [sp, #3, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z20, [sp, #4, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z19, [sp, #5, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z18, [sp, #6, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z17, [sp, #7, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z16, [sp, #8, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
; CHECK-NEXT:    str z0, [x0]
; CHECK-NEXT:    addvl sp, sp, #9
; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
; CHECK-NEXT:    ret
entry:
  %0 = tail call target("aarch64.svcount") @llvm.aarch64.sve.ptrue.c8()
  %1 = tail 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") %0, ptr %ptr)
  %2 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 0
  %3 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 1
  %4 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 2
  %5 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 3
  %arrayidx2 = getelementptr inbounds i8, ptr %ptr, i64 %stride
  %6 = tail 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") %0, ptr %arrayidx2)
  %7 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %6, 0
  %8 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %6, 1
  %9 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %6, 2
  %10 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %6, 3
  %mul3 = shl i64 %stride, 1
  %arrayidx4 = getelementptr inbounds i8, ptr %ptr, i64 %mul3
  %11 = tail 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") %0, ptr %arrayidx4)
  %12 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %11, 0
  %13 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %11, 1
  %14 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %11, 2
  %15 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %11, 3
  %mul5 = mul i64 %stride, 3
  %arrayidx6 = getelementptr inbounds i8, ptr %ptr, i64 %mul5
  %16 = tail 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") %0, ptr %arrayidx6)
  %17 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %16, 0
  %18 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %16, 1
  %19 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %16, 2
  %20 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %16, 3
  tail call void @llvm.aarch64.sme.usdot.lane.za32.vg1x4.nxv16i8(i32 0, <vscale x 16 x i8> %2, <vscale x 16 x i8> %7, <vscale x 16 x i8> %12, <vscale x 16 x i8> %17, <vscale x 16 x i8> poison, i32 0)
  tail call void @llvm.aarch64.sme.usdot.lane.za32.vg1x4.nxv16i8(i32 0, <vscale x 16 x i8> %3, <vscale x 16 x i8> %8, <vscale x 16 x i8> %13, <vscale x 16 x i8> %18, <vscale x 16 x i8> poison, i32 0)
  tail call void @llvm.aarch64.sme.usdot.lane.za32.vg1x4.nxv16i8(i32 0, <vscale x 16 x i8> %4, <vscale x 16 x i8> %9, <vscale x 16 x i8> %14, <vscale x 16 x i8> %19, <vscale x 16 x i8> poison, i32 0)
  tail call void @llvm.aarch64.sme.usdot.lane.za32.vg1x4.nxv16i8(i32 0, <vscale x 16 x i8> %5, <vscale x 16 x i8> %10, <vscale x 16 x i8> %15, <vscale x 16 x i8> %20, <vscale x 16 x i8> poison, i32 0)
  store <vscale x 16 x i8> %scalable_arg, ptr %ptr
  ret void
}

; == Multi, indexed (signed) ==

define void @sdot_lane_za32_u16_vg1x2(i32 %slice, <vscale x 16 x i8> %unused, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2) #0 {
; CHECK-LABEL: sdot_lane_za32_u16_vg1x2:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z5.d, z2.d
; CHECK-NEXT:    mov z4.d, z1.d
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    sdot za.s[w8, 0, vgx2], { z4.h, z5.h }, z3.h[3]
; CHECK-NEXT:    sdot za.s[w8, 7, vgx2], { z4.h, z5.h }, z3.h[3]
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.sdot.lane.za32.vg1x2.nxv8i16(i32 %slice, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, i32 3)
  %slice2 = add i32 %slice, 7
  call void @llvm.aarch64.sme.sdot.lane.za32.vg1x2.nxv8i16(i32 %slice2, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, i32 3)
  ret void
}

define void @sdot_lane_za32_u16_vg1x4(i32 %slice, <vscale x 16 x i8> %unused, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, <vscale x 8 x i16> %zn3, <vscale x 8 x i16> %zn4) #0 {
; CHECK-LABEL: sdot_lane_za32_u16_vg1x4:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z27.d, z4.d
; CHECK-NEXT:    mov z26.d, z3.d
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    mov z25.d, z2.d
; CHECK-NEXT:    mov z24.d, z1.d
; CHECK-NEXT:    sdot za.s[w8, 0, vgx4], { z24.h - z27.h }, z5.h[3]
; CHECK-NEXT:    sdot za.s[w8, 7, vgx4], { z24.h - z27.h }, z5.h[3]
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.sdot.lane.za32.vg1x4.nxv8i16(i32 %slice, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, <vscale x 8 x i16> %zn3,
                                                           <vscale x 8 x i16> %zn4, i32 3)
  %slice2 = add i32 %slice, 7
  call void @llvm.aarch64.sme.sdot.lane.za32.vg1x4.nxv8i16(i32 %slice2, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, <vscale x 8 x i16> %zn3,
                                                           <vscale x 8 x i16> %zn4, i32 3)
  ret void
}

define void @sdot_lane_za32_u8_vg1x2(i32 %slice, <vscale x 16 x i8> %unused, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2) #0 {
; CHECK-LABEL: sdot_lane_za32_u8_vg1x2:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z5.d, z2.d
; CHECK-NEXT:    mov z4.d, z1.d
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    sdot za.s[w8, 0, vgx2], { z4.b, z5.b }, z3.b[3]
; CHECK-NEXT:    sdot za.s[w8, 7, vgx2], { z4.b, z5.b }, z3.b[3]
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.sdot.lane.za32.vg1x2.nxv16i8(i32 %slice, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2, i32 3)
  %slice2 = add i32 %slice, 7
  call void @llvm.aarch64.sme.sdot.lane.za32.vg1x2.nxv16i8(i32 %slice2, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2, i32 3)
  ret void
}

define void @sdot_lane_za32_u8_vg1x4(i32 %slice, <vscale x 16 x i8> %unused, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2, <vscale x 16 x i8> %zn3, <vscale x 16 x i8> %zn4) #0 {
; CHECK-LABEL: sdot_lane_za32_u8_vg1x4:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z27.d, z4.d
; CHECK-NEXT:    mov z26.d, z3.d
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    mov z25.d, z2.d
; CHECK-NEXT:    mov z24.d, z1.d
; CHECK-NEXT:    sdot za.s[w8, 0, vgx4], { z24.b - z27.b }, z5.b[3]
; CHECK-NEXT:    sdot za.s[w8, 7, vgx4], { z24.b - z27.b }, z5.b[3]
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.sdot.lane.za32.vg1x4.nxv16i8(i32 %slice, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2, <vscale x 16 x i8> %zn3,
                                                           <vscale x 16 x i8> %zn4, i32 3)
  %slice2 = add i32 %slice, 7
  call void @llvm.aarch64.sme.sdot.lane.za32.vg1x4.nxv16i8(i32 %slice2, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2, <vscale x 16 x i8> %zn3,
                                                           <vscale x 16 x i8> %zn4, i32 3)
  ret void
}

define void @sdot_form_2x_tuple(ptr %ptr, i64 %stride) #0 {
; CHECK-LABEL: sdot_form_2x_tuple:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    ptrue pn8.b
; CHECK-NEXT:    mov w8, wzr
; CHECK-NEXT:    ld1b { z16.b, z24.b }, pn8/z, [x0]
; CHECK-NEXT:    ld1b { z17.b, z25.b }, pn8/z, [x0, x1]
; CHECK-NEXT:    sdot za.s[w8, 0, vgx2], { z16.b, z17.b }, z0.b[0]
; CHECK-NEXT:    sdot za.s[w8, 0, vgx2], { z24.b, z25.b }, z0.b[0]
; CHECK-NEXT:    ret
entry:
  %0 = tail call target("aarch64.svcount") @llvm.aarch64.sve.ptrue.c8()
  %1 = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld1.pn.x2.nxv16i8(target("aarch64.svcount") %0, ptr %ptr)
  %2 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 0
  %3 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 1
  %arrayidx2 = getelementptr inbounds i8, ptr %ptr, i64 %stride
  %4 = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld1.pn.x2.nxv16i8(target("aarch64.svcount") %0, ptr %arrayidx2)
  %5 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %4, 0
  %6 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %4, 1
  tail call void @llvm.aarch64.sme.sdot.lane.za32.vg1x2.nxv16i8(i32 0, <vscale x 16 x i8> %2, <vscale x 16 x i8> %5, <vscale x 16 x i8> poison, i32 0)
  tail call void @llvm.aarch64.sme.sdot.lane.za32.vg1x2.nxv16i8(i32 0, <vscale x 16 x i8> %3, <vscale x 16 x i8> %6, <vscale x 16 x i8> poison, i32 0)
  ret void
}

define void @sdot_form_2x_tuple_svecc(ptr %ptr, i64 %stride, <vscale x 16 x i8> %scalable_arg) #0 {
; CHECK-LABEL: sdot_form_2x_tuple_svecc:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
; CHECK-NEXT:    addvl sp, sp, #-3
; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
; CHECK-NEXT:    ptrue pn8.b
; CHECK-NEXT:    mov w8, wzr
; CHECK-NEXT:    str z11, [sp, #1, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z10, [sp, #2, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    ld1b { z2.b, z10.b }, pn8/z, [x0]
; CHECK-NEXT:    ld1b { z3.b, z11.b }, pn8/z, [x0, x1]
; CHECK-NEXT:    sdot za.s[w8, 0, vgx2], { z2.b, z3.b }, z0.b[0]
; CHECK-NEXT:    sdot za.s[w8, 0, vgx2], { z10.b, z11.b }, z0.b[0]
; CHECK-NEXT:    ldr z11, [sp, #1, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z10, [sp, #2, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
; CHECK-NEXT:    str z0, [x0]
; CHECK-NEXT:    addvl sp, sp, #3
; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
; CHECK-NEXT:    ret
entry:
  %0 = tail call target("aarch64.svcount") @llvm.aarch64.sve.ptrue.c8()
  %1 = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld1.pn.x2.nxv16i8(target("aarch64.svcount") %0, ptr %ptr)
  %2 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 0
  %3 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 1
  %arrayidx2 = getelementptr inbounds i8, ptr %ptr, i64 %stride
  %4 = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld1.pn.x2.nxv16i8(target("aarch64.svcount") %0, ptr %arrayidx2)
  %5 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %4, 0
  %6 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %4, 1
  tail call void @llvm.aarch64.sme.sdot.lane.za32.vg1x2.nxv16i8(i32 0, <vscale x 16 x i8> %2, <vscale x 16 x i8> %5, <vscale x 16 x i8> poison, i32 0)
  tail call void @llvm.aarch64.sme.sdot.lane.za32.vg1x2.nxv16i8(i32 0, <vscale x 16 x i8> %3, <vscale x 16 x i8> %6, <vscale x 16 x i8> poison, i32 0)
  store <vscale x 16 x i8> %scalable_arg, ptr %ptr
  ret void
}

define void @sdot_form_4x_tuple(ptr %ptr, i64 %stride) #0 {
; CHECK-LABEL: sdot_form_4x_tuple:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    lsl x9, x1, #1
; CHECK-NEXT:    ptrue pn8.b
; CHECK-NEXT:    mov w8, wzr
; CHECK-NEXT:    ld1b { z16.b, z20.b, z24.b, z28.b }, pn8/z, [x0]
; CHECK-NEXT:    ld1b { z17.b, z21.b, z25.b, z29.b }, pn8/z, [x0, x1]
; CHECK-NEXT:    add x10, x9, x1
; CHECK-NEXT:    ld1b { z18.b, z22.b, z26.b, z30.b }, pn8/z, [x0, x9]
; CHECK-NEXT:    ld1b { z19.b, z23.b, z27.b, z31.b }, pn8/z, [x0, x10]
; CHECK-NEXT:    sdot za.s[w8, 0, vgx4], { z16.b - z19.b }, z0.b[0]
; CHECK-NEXT:    sdot za.s[w8, 0, vgx4], { z20.b - z23.b }, z0.b[0]
; CHECK-NEXT:    sdot za.s[w8, 0, vgx4], { z24.b - z27.b }, z0.b[0]
; CHECK-NEXT:    sdot za.s[w8, 0, vgx4], { z28.b - z31.b }, z0.b[0]
; CHECK-NEXT:    ret
entry:
  %0 = tail call target("aarch64.svcount") @llvm.aarch64.sve.ptrue.c8()
  %1 = tail 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") %0, ptr %ptr)
  %2 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 0
  %3 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 1
  %4 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 2
  %5 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 3
  %arrayidx2 = getelementptr inbounds i8, ptr %ptr, i64 %stride
  %6 = tail 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") %0, ptr %arrayidx2)
  %7 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %6, 0
  %8 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %6, 1
  %9 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %6, 2
  %10 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %6, 3
  %mul3 = shl i64 %stride, 1
  %arrayidx4 = getelementptr inbounds i8, ptr %ptr, i64 %mul3
  %11 = tail 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") %0, ptr %arrayidx4)
  %12 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %11, 0
  %13 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %11, 1
  %14 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %11, 2
  %15 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %11, 3
  %mul5 = mul i64 %stride, 3
  %arrayidx6 = getelementptr inbounds i8, ptr %ptr, i64 %mul5
  %16 = tail 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") %0, ptr %arrayidx6)
  %17 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %16, 0
  %18 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %16, 1
  %19 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %16, 2
  %20 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %16, 3
  tail call void @llvm.aarch64.sme.sdot.lane.za32.vg1x4.nxv16i8(i32 0, <vscale x 16 x i8> %2, <vscale x 16 x i8> %7, <vscale x 16 x i8> %12, <vscale x 16 x i8> %17, <vscale x 16 x i8> poison, i32 0)
  tail call void @llvm.aarch64.sme.sdot.lane.za32.vg1x4.nxv16i8(i32 0, <vscale x 16 x i8> %3, <vscale x 16 x i8> %8, <vscale x 16 x i8> %13, <vscale x 16 x i8> %18, <vscale x 16 x i8> poison, i32 0)
  tail call void @llvm.aarch64.sme.sdot.lane.za32.vg1x4.nxv16i8(i32 0, <vscale x 16 x i8> %4, <vscale x 16 x i8> %9, <vscale x 16 x i8> %14, <vscale x 16 x i8> %19, <vscale x 16 x i8> poison, i32 0)
  tail call void @llvm.aarch64.sme.sdot.lane.za32.vg1x4.nxv16i8(i32 0, <vscale x 16 x i8> %5, <vscale x 16 x i8> %10, <vscale x 16 x i8> %15, <vscale x 16 x i8> %20, <vscale x 16 x i8> poison, i32 0)
  ret void
}

define void @sdot_form_4x_tuple_svecc(ptr %ptr, i64 %stride, <vscale x 16 x i8> %scalable_arg) #0 {
; CHECK-LABEL: sdot_form_4x_tuple_svecc:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
; CHECK-NEXT:    addvl sp, sp, #-9
; CHECK-NEXT:    lsl x9, x1, #1
; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
; CHECK-NEXT:    ptrue pn8.b
; CHECK-NEXT:    str z23, [sp, #1, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    mov w8, wzr
; CHECK-NEXT:    str z22, [sp, #2, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    add x10, x9, x1
; CHECK-NEXT:    str z21, [sp, #3, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z20, [sp, #4, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z19, [sp, #5, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z18, [sp, #6, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z17, [sp, #7, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z16, [sp, #8, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    ld1b { z16.b, z20.b, z24.b, z28.b }, pn8/z, [x0]
; CHECK-NEXT:    ld1b { z17.b, z21.b, z25.b, z29.b }, pn8/z, [x0, x1]
; CHECK-NEXT:    ld1b { z18.b, z22.b, z26.b, z30.b }, pn8/z, [x0, x9]
; CHECK-NEXT:    ld1b { z19.b, z23.b, z27.b, z31.b }, pn8/z, [x0, x10]
; CHECK-NEXT:    sdot za.s[w8, 0, vgx4], { z16.b - z19.b }, z0.b[0]
; CHECK-NEXT:    sdot za.s[w8, 0, vgx4], { z20.b - z23.b }, z0.b[0]
; CHECK-NEXT:    sdot za.s[w8, 0, vgx4], { z24.b - z27.b }, z0.b[0]
; CHECK-NEXT:    sdot za.s[w8, 0, vgx4], { z28.b - z31.b }, z0.b[0]
; CHECK-NEXT:    ldr z23, [sp, #1, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z22, [sp, #2, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z21, [sp, #3, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z20, [sp, #4, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z19, [sp, #5, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z18, [sp, #6, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z17, [sp, #7, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z16, [sp, #8, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
; CHECK-NEXT:    str z0, [x0]
; CHECK-NEXT:    addvl sp, sp, #9
; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
; CHECK-NEXT:    ret
entry:
  %0 = tail call target("aarch64.svcount") @llvm.aarch64.sve.ptrue.c8()
  %1 = tail 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") %0, ptr %ptr)
  %2 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 0
  %3 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 1
  %4 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 2
  %5 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 3
  %arrayidx2 = getelementptr inbounds i8, ptr %ptr, i64 %stride
  %6 = tail 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") %0, ptr %arrayidx2)
  %7 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %6, 0
  %8 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %6, 1
  %9 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %6, 2
  %10 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %6, 3
  %mul3 = shl i64 %stride, 1
  %arrayidx4 = getelementptr inbounds i8, ptr %ptr, i64 %mul3
  %11 = tail 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") %0, ptr %arrayidx4)
  %12 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %11, 0
  %13 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %11, 1
  %14 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %11, 2
  %15 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %11, 3
  %mul5 = mul i64 %stride, 3
  %arrayidx6 = getelementptr inbounds i8, ptr %ptr, i64 %mul5
  %16 = tail 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") %0, ptr %arrayidx6)
  %17 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %16, 0
  %18 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %16, 1
  %19 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %16, 2
  %20 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %16, 3
  tail call void @llvm.aarch64.sme.sdot.lane.za32.vg1x4.nxv16i8(i32 0, <vscale x 16 x i8> %2, <vscale x 16 x i8> %7, <vscale x 16 x i8> %12, <vscale x 16 x i8> %17, <vscale x 16 x i8> poison, i32 0)
  tail call void @llvm.aarch64.sme.sdot.lane.za32.vg1x4.nxv16i8(i32 0, <vscale x 16 x i8> %3, <vscale x 16 x i8> %8, <vscale x 16 x i8> %13, <vscale x 16 x i8> %18, <vscale x 16 x i8> poison, i32 0)
  tail call void @llvm.aarch64.sme.sdot.lane.za32.vg1x4.nxv16i8(i32 0, <vscale x 16 x i8> %4, <vscale x 16 x i8> %9, <vscale x 16 x i8> %14, <vscale x 16 x i8> %19, <vscale x 16 x i8> poison, i32 0)
  tail call void @llvm.aarch64.sme.sdot.lane.za32.vg1x4.nxv16i8(i32 0, <vscale x 16 x i8> %5, <vscale x 16 x i8> %10, <vscale x 16 x i8> %15, <vscale x 16 x i8> %20, <vscale x 16 x i8> poison, i32 0)
  store <vscale x 16 x i8> %scalable_arg, ptr %ptr
  ret void
}

define void @sdot_lane_za64_u16_vg1x2(i32 %slice, <vscale x 16 x i8> %unused, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2) #1 {
; CHECK-LABEL: sdot_lane_za64_u16_vg1x2:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z5.d, z2.d
; CHECK-NEXT:    mov z4.d, z1.d
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    sdot za.d[w8, 0, vgx2], { z4.h, z5.h }, z3.h[1]
; CHECK-NEXT:    sdot za.d[w8, 7, vgx2], { z4.h, z5.h }, z3.h[1]
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.sdot.lane.za64.vg1x2.nxv8i16(i32 %slice, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, i32 1)
  %slice2 = add i32 %slice, 7
  call void @llvm.aarch64.sme.sdot.lane.za64.vg1x2.nxv8i16(i32 %slice2, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, i32 1)
  ret void
}

define void @sdot_lane_za64_u16_vg1x4(i32 %slice, <vscale x 16 x i8> %unused, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, <vscale x 8 x i16> %zn3, <vscale x 8 x i16> %zn4) #1 {
; CHECK-LABEL: sdot_lane_za64_u16_vg1x4:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z27.d, z4.d
; CHECK-NEXT:    mov z26.d, z3.d
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    mov z25.d, z2.d
; CHECK-NEXT:    mov z24.d, z1.d
; CHECK-NEXT:    sdot za.d[w8, 0, vgx4], { z24.h - z27.h }, z5.h[1]
; CHECK-NEXT:    sdot za.d[w8, 7, vgx4], { z24.h - z27.h }, z5.h[1]
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.sdot.lane.za64.vg1x4.nxv8i16(i32 %slice, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, <vscale x 8 x i16> %zn3,
                                                           <vscale x 8 x i16> %zn4, i32 1)
  %slice2 = add i32 %slice, 7
  call void @llvm.aarch64.sme.sdot.lane.za64.vg1x4.nxv8i16(i32 %slice2, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, <vscale x 8 x i16> %zn3,
                                                           <vscale x 8 x i16> %zn4, i32 1)
  ret void
}



define void @sudot_lane_za32_u8_vg1x2(i32 %slice, <vscale x 16 x i8> %unused, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2) #0 {
; CHECK-LABEL: sudot_lane_za32_u8_vg1x2:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z5.d, z2.d
; CHECK-NEXT:    mov z4.d, z1.d
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    sudot za.s[w8, 0, vgx2], { z4.b, z5.b }, z3.b[3]
; CHECK-NEXT:    sudot za.s[w8, 7, vgx2], { z4.b, z5.b }, z3.b[3]
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.sudot.lane.za32.vg1x2.nxv16i8(i32 %slice, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2, i32 3)
  %slice2 = add i32 %slice, 7
  call void @llvm.aarch64.sme.sudot.lane.za32.vg1x2.nxv16i8(i32 %slice2, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2, i32 3)
  ret void
}

define void @sudot_lane_za32_u8_vg1x4(i32 %slice, <vscale x 16 x i8> %unused, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2, <vscale x 16 x i8> %zn3, <vscale x 16 x i8> %zn4) #0 {
; CHECK-LABEL: sudot_lane_za32_u8_vg1x4:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z27.d, z4.d
; CHECK-NEXT:    mov z26.d, z3.d
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    mov z25.d, z2.d
; CHECK-NEXT:    mov z24.d, z1.d
; CHECK-NEXT:    sudot za.s[w8, 0, vgx4], { z24.b - z27.b }, z5.b[3]
; CHECK-NEXT:    sudot za.s[w8, 7, vgx4], { z24.b - z27.b }, z5.b[3]
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.sudot.lane.za32.vg1x4.nxv16i8(i32 %slice, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2, <vscale x 16 x i8> %zn3,
                                                            <vscale x 16 x i8> %zn4, i32 3)
  %slice2 = add i32 %slice, 7
  call void @llvm.aarch64.sme.sudot.lane.za32.vg1x4.nxv16i8(i32 %slice2, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zn2, <vscale x 16 x i8> %zn3,
                                                            <vscale x 16 x i8> %zn4, i32 3)
  ret void
}

define void @sudot_form_2x_tuple(ptr %ptr, i64 %stride) #0 {
; CHECK-LABEL: sudot_form_2x_tuple:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    ptrue pn8.b
; CHECK-NEXT:    mov w8, wzr
; CHECK-NEXT:    ld1b { z16.b, z24.b }, pn8/z, [x0]
; CHECK-NEXT:    ld1b { z17.b, z25.b }, pn8/z, [x0, x1]
; CHECK-NEXT:    sudot za.s[w8, 0, vgx2], { z16.b, z17.b }, z0.b[0]
; CHECK-NEXT:    sudot za.s[w8, 0, vgx2], { z24.b, z25.b }, z0.b[0]
; CHECK-NEXT:    ret
entry:
  %0 = tail call target("aarch64.svcount") @llvm.aarch64.sve.ptrue.c8()
  %1 = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld1.pn.x2.nxv16i8(target("aarch64.svcount") %0, ptr %ptr)
  %2 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 0
  %3 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 1
  %arrayidx2 = getelementptr inbounds i8, ptr %ptr, i64 %stride
  %4 = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld1.pn.x2.nxv16i8(target("aarch64.svcount") %0, ptr %arrayidx2)
  %5 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %4, 0
  %6 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %4, 1
  tail call void @llvm.aarch64.sme.sudot.lane.za32.vg1x2.nxv16i8(i32 0, <vscale x 16 x i8> %2, <vscale x 16 x i8> %5, <vscale x 16 x i8> poison, i32 0)
  tail call void @llvm.aarch64.sme.sudot.lane.za32.vg1x2.nxv16i8(i32 0, <vscale x 16 x i8> %3, <vscale x 16 x i8> %6, <vscale x 16 x i8> poison, i32 0)
  ret void
}

define void @sudot_form_2x_tuple_svecc(ptr %ptr, i64 %stride, <vscale x 16 x i8> %scalable_arg) #0 {
; CHECK-LABEL: sudot_form_2x_tuple_svecc:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
; CHECK-NEXT:    addvl sp, sp, #-3
; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
; CHECK-NEXT:    ptrue pn8.b
; CHECK-NEXT:    mov w8, wzr
; CHECK-NEXT:    str z11, [sp, #1, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z10, [sp, #2, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    ld1b { z2.b, z10.b }, pn8/z, [x0]
; CHECK-NEXT:    ld1b { z3.b, z11.b }, pn8/z, [x0, x1]
; CHECK-NEXT:    sudot za.s[w8, 0, vgx2], { z2.b, z3.b }, z0.b[0]
; CHECK-NEXT:    sudot za.s[w8, 0, vgx2], { z10.b, z11.b }, z0.b[0]
; CHECK-NEXT:    ldr z11, [sp, #1, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z10, [sp, #2, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
; CHECK-NEXT:    str z0, [x0]
; CHECK-NEXT:    addvl sp, sp, #3
; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
; CHECK-NEXT:    ret
entry:
  %0 = tail call target("aarch64.svcount") @llvm.aarch64.sve.ptrue.c8()
  %1 = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld1.pn.x2.nxv16i8(target("aarch64.svcount") %0, ptr %ptr)
  %2 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 0
  %3 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 1
  %arrayidx2 = getelementptr inbounds i8, ptr %ptr, i64 %stride
  %4 = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld1.pn.x2.nxv16i8(target("aarch64.svcount") %0, ptr %arrayidx2)
  %5 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %4, 0
  %6 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } %4, 1
  tail call void @llvm.aarch64.sme.sudot.lane.za32.vg1x2.nxv16i8(i32 0, <vscale x 16 x i8> %2, <vscale x 16 x i8> %5, <vscale x 16 x i8> poison, i32 0)
  tail call void @llvm.aarch64.sme.sudot.lane.za32.vg1x2.nxv16i8(i32 0, <vscale x 16 x i8> %3, <vscale x 16 x i8> %6, <vscale x 16 x i8> poison, i32 0)
  store <vscale x 16 x i8> %scalable_arg, ptr %ptr
  ret void
}

define void @sudot_form_4x_tuple(ptr %ptr, i64 %stride) #0 {
; CHECK-LABEL: sudot_form_4x_tuple:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    lsl x9, x1, #1
; CHECK-NEXT:    ptrue pn8.b
; CHECK-NEXT:    mov w8, wzr
; CHECK-NEXT:    ld1b { z16.b, z20.b, z24.b, z28.b }, pn8/z, [x0]
; CHECK-NEXT:    ld1b { z17.b, z21.b, z25.b, z29.b }, pn8/z, [x0, x1]
; CHECK-NEXT:    add x10, x9, x1
; CHECK-NEXT:    ld1b { z18.b, z22.b, z26.b, z30.b }, pn8/z, [x0, x9]
; CHECK-NEXT:    ld1b { z19.b, z23.b, z27.b, z31.b }, pn8/z, [x0, x10]
; CHECK-NEXT:    sudot za.s[w8, 0, vgx4], { z16.b - z19.b }, z0.b[0]
; CHECK-NEXT:    sudot za.s[w8, 0, vgx4], { z20.b - z23.b }, z0.b[0]
; CHECK-NEXT:    sudot za.s[w8, 0, vgx4], { z24.b - z27.b }, z0.b[0]
; CHECK-NEXT:    sudot za.s[w8, 0, vgx4], { z28.b - z31.b }, z0.b[0]
; CHECK-NEXT:    ret
entry:
  %0 = tail call target("aarch64.svcount") @llvm.aarch64.sve.ptrue.c8()
  %1 = tail 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") %0, ptr %ptr)
  %2 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 0
  %3 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 1
  %4 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 2
  %5 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 3
  %arrayidx2 = getelementptr inbounds i8, ptr %ptr, i64 %stride
  %6 = tail 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") %0, ptr %arrayidx2)
  %7 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %6, 0
  %8 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %6, 1
  %9 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %6, 2
  %10 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %6, 3
  %mul3 = shl i64 %stride, 1
  %arrayidx4 = getelementptr inbounds i8, ptr %ptr, i64 %mul3
  %11 = tail 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") %0, ptr %arrayidx4)
  %12 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %11, 0
  %13 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %11, 1
  %14 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %11, 2
  %15 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %11, 3
  %mul5 = mul i64 %stride, 3
  %arrayidx6 = getelementptr inbounds i8, ptr %ptr, i64 %mul5
  %16 = tail 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") %0, ptr %arrayidx6)
  %17 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %16, 0
  %18 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %16, 1
  %19 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %16, 2
  %20 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %16, 3
  tail call void @llvm.aarch64.sme.sudot.lane.za32.vg1x4.nxv16i8(i32 0, <vscale x 16 x i8> %2, <vscale x 16 x i8> %7, <vscale x 16 x i8> %12, <vscale x 16 x i8> %17, <vscale x 16 x i8> poison, i32 0)
  tail call void @llvm.aarch64.sme.sudot.lane.za32.vg1x4.nxv16i8(i32 0, <vscale x 16 x i8> %3, <vscale x 16 x i8> %8, <vscale x 16 x i8> %13, <vscale x 16 x i8> %18, <vscale x 16 x i8> poison, i32 0)
  tail call void @llvm.aarch64.sme.sudot.lane.za32.vg1x4.nxv16i8(i32 0, <vscale x 16 x i8> %4, <vscale x 16 x i8> %9, <vscale x 16 x i8> %14, <vscale x 16 x i8> %19, <vscale x 16 x i8> poison, i32 0)
  tail call void @llvm.aarch64.sme.sudot.lane.za32.vg1x4.nxv16i8(i32 0, <vscale x 16 x i8> %5, <vscale x 16 x i8> %10, <vscale x 16 x i8> %15, <vscale x 16 x i8> %20, <vscale x 16 x i8> poison, i32 0)
  ret void
}

define void @sudot_form_4x_tuple_svecc(ptr %ptr, i64 %stride, <vscale x 16 x i8> %scalable_arg) #0 {
; CHECK-LABEL: sudot_form_4x_tuple_svecc:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
; CHECK-NEXT:    addvl sp, sp, #-9
; CHECK-NEXT:    lsl x9, x1, #1
; CHECK-NEXT:    str p8, [sp, #7, mul vl] // 2-byte Folded Spill
; CHECK-NEXT:    ptrue pn8.b
; CHECK-NEXT:    str z23, [sp, #1, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    mov w8, wzr
; CHECK-NEXT:    str z22, [sp, #2, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    add x10, x9, x1
; CHECK-NEXT:    str z21, [sp, #3, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z20, [sp, #4, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z19, [sp, #5, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z18, [sp, #6, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z17, [sp, #7, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z16, [sp, #8, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    ld1b { z16.b, z20.b, z24.b, z28.b }, pn8/z, [x0]
; CHECK-NEXT:    ld1b { z17.b, z21.b, z25.b, z29.b }, pn8/z, [x0, x1]
; CHECK-NEXT:    ld1b { z18.b, z22.b, z26.b, z30.b }, pn8/z, [x0, x9]
; CHECK-NEXT:    ld1b { z19.b, z23.b, z27.b, z31.b }, pn8/z, [x0, x10]
; CHECK-NEXT:    sudot za.s[w8, 0, vgx4], { z16.b - z19.b }, z0.b[0]
; CHECK-NEXT:    sudot za.s[w8, 0, vgx4], { z20.b - z23.b }, z0.b[0]
; CHECK-NEXT:    sudot za.s[w8, 0, vgx4], { z24.b - z27.b }, z0.b[0]
; CHECK-NEXT:    sudot za.s[w8, 0, vgx4], { z28.b - z31.b }, z0.b[0]
; CHECK-NEXT:    ldr z23, [sp, #1, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z22, [sp, #2, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z21, [sp, #3, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z20, [sp, #4, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z19, [sp, #5, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z18, [sp, #6, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z17, [sp, #7, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z16, [sp, #8, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr p8, [sp, #7, mul vl] // 2-byte Folded Reload
; CHECK-NEXT:    str z0, [x0]
; CHECK-NEXT:    addvl sp, sp, #9
; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
; CHECK-NEXT:    ret
entry:
  %0 = tail call target("aarch64.svcount") @llvm.aarch64.sve.ptrue.c8()
  %1 = tail 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") %0, ptr %ptr)
  %2 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 0
  %3 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 1
  %4 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 2
  %5 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %1, 3
  %arrayidx2 = getelementptr inbounds i8, ptr %ptr, i64 %stride
  %6 = tail 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") %0, ptr %arrayidx2)
  %7 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %6, 0
  %8 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %6, 1
  %9 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %6, 2
  %10 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %6, 3
  %mul3 = shl i64 %stride, 1
  %arrayidx4 = getelementptr inbounds i8, ptr %ptr, i64 %mul3
  %11 = tail 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") %0, ptr %arrayidx4)
  %12 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %11, 0
  %13 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %11, 1
  %14 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %11, 2
  %15 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %11, 3
  %mul5 = mul i64 %stride, 3
  %arrayidx6 = getelementptr inbounds i8, ptr %ptr, i64 %mul5
  %16 = tail 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") %0, ptr %arrayidx6)
  %17 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %16, 0
  %18 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %16, 1
  %19 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %16, 2
  %20 = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } %16, 3
  tail call void @llvm.aarch64.sme.sudot.lane.za32.vg1x4.nxv16i8(i32 0, <vscale x 16 x i8> %2, <vscale x 16 x i8> %7, <vscale x 16 x i8> %12, <vscale x 16 x i8> %17, <vscale x 16 x i8> poison, i32 0)
  tail call void @llvm.aarch64.sme.sudot.lane.za32.vg1x4.nxv16i8(i32 0, <vscale x 16 x i8> %3, <vscale x 16 x i8> %8, <vscale x 16 x i8> %13, <vscale x 16 x i8> %18, <vscale x 16 x i8> poison, i32 0)
  tail call void @llvm.aarch64.sme.sudot.lane.za32.vg1x4.nxv16i8(i32 0, <vscale x 16 x i8> %4, <vscale x 16 x i8> %9, <vscale x 16 x i8> %14, <vscale x 16 x i8> %19, <vscale x 16 x i8> poison, i32 0)
  tail call void @llvm.aarch64.sme.sudot.lane.za32.vg1x4.nxv16i8(i32 0, <vscale x 16 x i8> %5, <vscale x 16 x i8> %10, <vscale x 16 x i8> %15, <vscale x 16 x i8> %20, <vscale x 16 x i8> poison, i32 0)
  store <vscale x 16 x i8> %scalable_arg, ptr %ptr
  ret void
}


attributes #0 = { nounwind "target-features"="+sme2" }
attributes #1 = { nounwind "target-features"="+sme2,+sme-i16i64" }

; == Multi, multi (unsigned)

declare void @llvm.aarch64.sme.udot.za32.vg1x2.nxv8i16(i32, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>)
declare void @llvm.aarch64.sme.udot.za32.vg1x4.nxv8i16(i32, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>,
                                                       <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>)
declare void @llvm.aarch64.sme.udot.za32.vg1x2.nxv16i8(i32, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>)
declare void @llvm.aarch64.sme.udot.za32.vg1x4.nxv16i8(i32, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>,
                                                       <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>)
declare void @llvm.aarch64.sme.udot.za64.vg1x2.nxv8i16(i32, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>)
declare void @llvm.aarch64.sme.udot.za64.vg1x4.nxv8i16(i32, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>,
                                                       <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>)
declare void @llvm.aarch64.sme.usdot.za32.vg1x2.nxv16i8(i32, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>)
declare void @llvm.aarch64.sme.usdot.za32.vg1x4.nxv16i8(i32, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>,
                                                        <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>)

; == Multi, multi (signed)

declare void @llvm.aarch64.sme.sdot.za32.vg1x2.nxv8i16(i32, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>)
declare void @llvm.aarch64.sme.sdot.za32.vg1x4.nxv8i16(i32, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>,
                                                       <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>)
declare void @llvm.aarch64.sme.sdot.za32.vg1x2.nxv16i8(i32, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>)
declare void @llvm.aarch64.sme.sdot.za32.vg1x4.nxv16i8(i32, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>,
                                                       <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>)
declare void @llvm.aarch64.sme.sdot.za64.vg1x2.nxv8i16(i32, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>)
declare void @llvm.aarch64.sme.sdot.za64.vg1x4.nxv8i16(i32, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>,
                                                       <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>)

; == Multi, single (unsigned)

declare void @llvm.aarch64.sme.udot.single.za32.vg1x2.nxv8i16(i32, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>)
declare void @llvm.aarch64.sme.udot.single.za32.vg1x4.nxv8i16(i32, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>)
declare void @llvm.aarch64.sme.udot.single.za32.vg1x2.nxv16i8(i32, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>)
declare void @llvm.aarch64.sme.udot.single.za32.vg1x4.nxv16i8(i32, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>)
declare void @llvm.aarch64.sme.udot.single.za64.vg1x2.nxv8i16(i32, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>)
declare void @llvm.aarch64.sme.udot.single.za64.vg1x4.nxv8i16(i32, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>)
declare void @llvm.aarch64.sme.usdot.single.za32.vg1x2.nxv16i8(i32, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>)
declare void @llvm.aarch64.sme.usdot.single.za32.vg1x4.nxv16i8(i32, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>)

; == Multi, single (signed)

declare void @llvm.aarch64.sme.sdot.single.za32.vg1x2.nxv8i16(i32, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>)
declare void @llvm.aarch64.sme.sdot.single.za32.vg1x4.nxv8i16(i32, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>)
declare void @llvm.aarch64.sme.sdot.single.za32.vg1x2.nxv16i8(i32, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>)
declare void @llvm.aarch64.sme.sdot.single.za32.vg1x4.nxv16i8(i32, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>)
declare void @llvm.aarch64.sme.sdot.single.za64.vg1x2.nxv8i16(i32, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>)
declare void @llvm.aarch64.sme.sdot.single.za64.vg1x4.nxv8i16(i32, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>)
declare void @llvm.aarch64.sme.sudot.single.za32.vg1x2.nxv16i8(i32, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>)
declare void @llvm.aarch64.sme.sudot.single.za32.vg1x4.nxv16i8(i32, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>)

; == Multi, indexed (unsigned)

declare void @llvm.aarch64.sme.udot.lane.za32.vg1x2.nxv8i16(i32, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, i32)
declare void @llvm.aarch64.sme.udot.lane.za32.vg1x4.nxv8i16(i32, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, i32)
declare void @llvm.aarch64.sme.udot.lane.za32.vg1x2.nxv16i8(i32, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, i32)
declare void @llvm.aarch64.sme.udot.lane.za32.vg1x4.nxv16i8(i32, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, i32)
declare void @llvm.aarch64.sme.udot.lane.za64.vg1x2.nxv8i16(i32, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, i32)
declare void @llvm.aarch64.sme.udot.lane.za64.vg1x4.nxv8i16(i32, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, i32)
declare void @llvm.aarch64.sme.usdot.lane.za32.vg1x2.nxv16i8(i32, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, i32)
declare void @llvm.aarch64.sme.usdot.lane.za32.vg1x4.nxv16i8(i32, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, i32)

; == Multi, indexed (signed)

declare void @llvm.aarch64.sme.sdot.lane.za32.vg1x2.nxv8i16(i32, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, i32)
declare void @llvm.aarch64.sme.sdot.lane.za32.vg1x4.nxv8i16(i32, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, i32)
declare void @llvm.aarch64.sme.sdot.lane.za32.vg1x2.nxv16i8(i32, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, i32)
declare void @llvm.aarch64.sme.sdot.lane.za32.vg1x4.nxv16i8(i32, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, i32)
declare void @llvm.aarch64.sme.sdot.lane.za64.vg1x2.nxv8i16(i32, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, i32)
declare void @llvm.aarch64.sme.sdot.lane.za64.vg1x4.nxv8i16(i32, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, i32)
declare void @llvm.aarch64.sme.sudot.lane.za32.vg1x2.nxv16i8(i32, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, i32)
declare void @llvm.aarch64.sme.sudot.lane.za32.vg1x4.nxv16i8(i32, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, i32)