; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -mtriple=aarch64-linux-gnu -mattr=+sme2,+sme-i16i64 -enable-subreg-liveness -force-streaming -verify-machineinstrs < %s | FileCheck %s

; == FVDOT ==

define void @test_fvdot_lane_za32_vg1x2_nxv8f16(i32 %slice, <vscale x 8 x half> %zn1, <vscale x 8 x half> %zn2, <vscale x 8 x half> %zm) {
; CHECK-LABEL: test_fvdot_lane_za32_vg1x2_nxv8f16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    fvdot za.s[w8, 0, vgx2], { z0.h, z1.h }, z2.h[3]
; CHECK-NEXT:    fvdot za.s[w8, 7, vgx2], { z0.h, z1.h }, z2.h[3]
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.fvdot.lane.za32.vg1x2.nxv8f16(i32 %slice, <vscale x 8 x half> %zn1, <vscale x 8 x half> %zn2, <vscale x 8 x half> %zm, i32 3)
  %slice.7 = add i32 %slice, 7
  call void @llvm.aarch64.sme.fvdot.lane.za32.vg1x2.nxv8f16(i32 %slice.7, <vscale x 8 x half> %zn1, <vscale x 8 x half> %zn2, <vscale x 8 x half> %zm, i32 3)
  ret void
}


; == BFVDOT ==

define void @test_fvdot_lane_za32_vg1x2_nxv8bf16(i32 %slice, <vscale x 8 x bfloat> %zn1, <vscale x 8 x bfloat> %zn2, <vscale x 8 x bfloat> %zm) {
; CHECK-LABEL: test_fvdot_lane_za32_vg1x2_nxv8bf16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    bfvdot za.s[w8, 0, vgx2], { z0.h, z1.h }, z2.h[3]
; CHECK-NEXT:    bfvdot za.s[w8, 7, vgx2], { z0.h, z1.h }, z2.h[3]
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.fvdot.lane.za32.vg1x2.nxv8bf16(i32 %slice, <vscale x 8 x bfloat> %zn1, <vscale x 8 x bfloat> %zn2, <vscale x 8 x bfloat> %zm, i32 3)
  %slice.7 = add i32 %slice, 7
  call void @llvm.aarch64.sme.fvdot.lane.za32.vg1x2.nxv8bf16(i32 %slice.7, <vscale x 8 x bfloat> %zn1, <vscale x 8 x bfloat> %zn2, <vscale x 8 x bfloat> %zm, i32 3)
  ret void
}


; == SVDOT ==

define void @test_svdot_lane_za32_vg1x2_nxv8i16(i32 %slice, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, <vscale x 8 x i16> %zm) {
; CHECK-LABEL: test_svdot_lane_za32_vg1x2_nxv8i16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    svdot za.s[w8, 0, vgx2], { z0.h, z1.h }, z2.h[3]
; CHECK-NEXT:    svdot za.s[w8, 7, vgx2], { z0.h, z1.h }, z2.h[3]
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.svdot.lane.za32.vg1x2.nxv8i16(i32 %slice, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, <vscale x 8 x i16> %zm, i32 3)
  %slice.7 = add i32 %slice, 7
  call void @llvm.aarch64.sme.svdot.lane.za32.vg1x2.nxv8i16(i32 %slice.7, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, <vscale x 8 x i16> %zm, i32 3)
  ret void
}

define void @test_svdot_lane_za32_vg1x4_nxv16i8(i32 %slice, <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> %zm) {
; CHECK-LABEL: test_svdot_lane_za32_vg1x4_nxv16i8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    svdot za.s[w8, 0, vgx4], { z0.b - z3.b }, z4.b[3]
; CHECK-NEXT:    svdot za.s[w8, 7, vgx4], { z0.b - z3.b }, z4.b[3]
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.svdot.lane.za32.vg1x4.nxv16i8(i32 %slice, <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> %zm, i32 3)
  %slice.7 = add i32 %slice, 7
  call void @llvm.aarch64.sme.svdot.lane.za32.vg1x4.nxv16i8(i32 %slice.7, <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> %zm, i32 3)
  ret void
}

define void @test_svdot_lane_za64_vg1x4_nxv8i16(i32 %slice, <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> %zm) {
; CHECK-LABEL: test_svdot_lane_za64_vg1x4_nxv8i16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    svdot za.d[w8, 0, vgx4], { z0.h - z3.h }, z4.h[1]
; CHECK-NEXT:    svdot za.d[w8, 7, vgx4], { z0.h - z3.h }, z4.h[1]
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.svdot.lane.za64.vg1x4.nxv8i16(i32 %slice, <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> %zm, i32 1)
  %slice.7 = add i32 %slice, 7
  call void @llvm.aarch64.sme.svdot.lane.za64.vg1x4.nxv8i16(i32 %slice.7, <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> %zm, i32 1)
  ret void
}

define void @svdot_form_2x_tuple(ptr %ptr, i64 %stride) #0 {
; CHECK-LABEL: svdot_form_2x_tuple:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    ptrue pn8.b
; CHECK-NEXT:    add x9, x0, x1
; CHECK-NEXT:    mov w8, wzr
; CHECK-NEXT:    ld1h { z16.h, z24.h }, pn8/z, [x0]
; CHECK-NEXT:    ld1h { z17.h, z25.h }, pn8/z, [x9]
; CHECK-NEXT:    svdot za.s[w8, 0, vgx2], { z16.h, z17.h }, z0.h[0]
; CHECK-NEXT:    svdot za.s[w8, 0, vgx2], { z24.h, z25.h }, z0.h[0]
; 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
  tail call void @llvm.aarch64.sme.svdot.lane.za32.vg1x2.nxv8i16(i32 0, <vscale x 8 x i16> %2, <vscale x 8 x i16> %5, <vscale x 8 x i16> poison, i32 0)
  tail call void @llvm.aarch64.sme.svdot.lane.za32.vg1x2.nxv8i16(i32 0, <vscale x 8 x i16> %3, <vscale x 8 x i16> %6, <vscale x 8 x i16> poison, i32 0)
  ret void
}

define void @svdot_form_2x_tuple_svecc(ptr %ptr, i64 %stride, <vscale x 16 x i8> %scalable_arg) #0 {
; CHECK-LABEL: svdot_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:    add x9, x0, x1
; CHECK-NEXT:    str z9, [sp, #1, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    mov w8, wzr
; CHECK-NEXT:    str z8, [sp, #2, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    ld1h { z0.h, z8.h }, pn8/z, [x0]
; CHECK-NEXT:    ld1h { z1.h, z9.h }, pn8/z, [x9]
; CHECK-NEXT:    svdot za.s[w8, 0, vgx2], { z0.h, z1.h }, z0.h[0]
; CHECK-NEXT:    svdot za.s[w8, 0, vgx2], { z8.h, z9.h }, z0.h[0]
; CHECK-NEXT:    ldr z9, [sp, #1, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z8, [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
  tail call void @llvm.aarch64.sme.svdot.lane.za32.vg1x2.nxv8i16(i32 0, <vscale x 8 x i16> %2, <vscale x 8 x i16> %5, <vscale x 8 x i16> poison, i32 0)
  tail call void @llvm.aarch64.sme.svdot.lane.za32.vg1x2.nxv8i16(i32 0, <vscale x 8 x i16> %3, <vscale x 8 x i16> %6, <vscale x 8 x i16> poison, i32 0)
  ret void
}

define void @svdot_form_4x_tuple(ptr %ptr, i64 %stride) #0 {
; CHECK-LABEL: svdot_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:    svdot za.s[w8, 0, vgx4], { z16.b - z19.b }, z0.b[0]
; CHECK-NEXT:    svdot za.s[w8, 0, vgx4], { z20.b - z23.b }, z0.b[0]
; CHECK-NEXT:    svdot za.s[w8, 0, vgx4], { z24.b - z27.b }, z0.b[0]
; CHECK-NEXT:    svdot 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.svdot.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.svdot.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.svdot.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.svdot.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 @svdot_form_4x_tuple_svecc(ptr %ptr, i64 %stride, <vscale x 16 x i8> %scalable_arg) #0 {
; CHECK-LABEL: svdot_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 z15, [sp, #1, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    mov w8, wzr
; CHECK-NEXT:    str z14, [sp, #2, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    add x10, x9, x1
; CHECK-NEXT:    str z13, [sp, #3, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z12, [sp, #4, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z11, [sp, #5, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z10, [sp, #6, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z9, [sp, #7, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z8, [sp, #8, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    ld1b { z0.b, z4.b, z8.b, z12.b }, pn8/z, [x0]
; CHECK-NEXT:    ld1b { z1.b, z5.b, z9.b, z13.b }, pn8/z, [x0, x1]
; CHECK-NEXT:    ld1b { z2.b, z6.b, z10.b, z14.b }, pn8/z, [x0, x9]
; CHECK-NEXT:    ld1b { z3.b, z7.b, z11.b, z15.b }, pn8/z, [x0, x10]
; CHECK-NEXT:    svdot za.s[w8, 0, vgx4], { z0.b - z3.b }, z0.b[0]
; CHECK-NEXT:    svdot za.s[w8, 0, vgx4], { z4.b - z7.b }, z0.b[0]
; CHECK-NEXT:    svdot za.s[w8, 0, vgx4], { z8.b - z11.b }, z0.b[0]
; CHECK-NEXT:    svdot za.s[w8, 0, vgx4], { z12.b - z15.b }, z0.b[0]
; CHECK-NEXT:    ldr z15, [sp, #1, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z14, [sp, #2, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z13, [sp, #3, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z12, [sp, #4, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z11, [sp, #5, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z10, [sp, #6, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z9, [sp, #7, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z8, [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
  tail call void @llvm.aarch64.sme.svdot.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.svdot.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.svdot.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.svdot.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
}

; == UVDOT ==

define void @test_uvdot_lane_za32_vg1x2_nxv8i16(i32 %slice, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, <vscale x 8 x i16> %zm) {
; CHECK-LABEL: test_uvdot_lane_za32_vg1x2_nxv8i16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    uvdot za.s[w8, 0, vgx2], { z0.h, z1.h }, z2.h[3]
; CHECK-NEXT:    uvdot za.s[w8, 7, vgx2], { z0.h, z1.h }, z2.h[3]
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.uvdot.lane.za32.vg1x2.nxv8i16(i32 %slice, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, <vscale x 8 x i16> %zm, i32 3)
  %slice.7 = add i32 %slice, 7
  call void @llvm.aarch64.sme.uvdot.lane.za32.vg1x2.nxv8i16(i32 %slice.7, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zn2, <vscale x 8 x i16> %zm, i32 3)
  ret void
}

define void @test_uvdot_lane_za32_vg1x4_nxv16i8(i32 %slice, <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> %zm) {
; CHECK-LABEL: test_uvdot_lane_za32_vg1x4_nxv16i8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    uvdot za.s[w8, 0, vgx4], { z0.b - z3.b }, z4.b[3]
; CHECK-NEXT:    uvdot za.s[w8, 7, vgx4], { z0.b - z3.b }, z4.b[3]
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.uvdot.lane.za32.vg1x4.nxv16i8(i32 %slice, <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> %zm, i32 3)
  %slice.7 = add i32 %slice, 7
  call void @llvm.aarch64.sme.uvdot.lane.za32.vg1x4.nxv16i8(i32 %slice.7, <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> %zm, i32 3)
  ret void
}

define void @test_uvdot_lane_za64_vg1x4_nxv8i16(i32 %slice, <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> %zm) {
; CHECK-LABEL: test_uvdot_lane_za64_vg1x4_nxv8i16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    uvdot za.d[w8, 0, vgx4], { z0.h - z3.h }, z4.h[1]
; CHECK-NEXT:    uvdot za.d[w8, 7, vgx4], { z0.h - z3.h }, z4.h[1]
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.uvdot.lane.za64.vg1x4.nxv8i16(i32 %slice, <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> %zm, i32 1)
  %slice.7 = add i32 %slice, 7
  call void @llvm.aarch64.sme.uvdot.lane.za64.vg1x4.nxv8i16(i32 %slice.7, <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> %zm, i32 1)
  ret void
}

define void @uvdot_form_2x_tuple(ptr %ptr, i64 %stride) #0 {
; CHECK-LABEL: uvdot_form_2x_tuple:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    ptrue pn8.b
; CHECK-NEXT:    add x9, x0, x1
; CHECK-NEXT:    mov w8, wzr
; CHECK-NEXT:    ld1h { z16.h, z24.h }, pn8/z, [x0]
; CHECK-NEXT:    ld1h { z17.h, z25.h }, pn8/z, [x9]
; CHECK-NEXT:    uvdot za.s[w8, 0, vgx2], { z16.h, z17.h }, z0.h[0]
; CHECK-NEXT:    uvdot za.s[w8, 0, vgx2], { z24.h, z25.h }, z0.h[0]
; 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
  tail call void @llvm.aarch64.sme.uvdot.lane.za32.vg1x2.nxv8i16(i32 0, <vscale x 8 x i16> %2, <vscale x 8 x i16> %5, <vscale x 8 x i16> poison, i32 0)
  tail call void @llvm.aarch64.sme.uvdot.lane.za32.vg1x2.nxv8i16(i32 0, <vscale x 8 x i16> %3, <vscale x 8 x i16> %6, <vscale x 8 x i16> poison, i32 0)
  ret void
}

define void @uvdot_form_2x_tuple_svecc(ptr %ptr, i64 %stride, <vscale x 16 x i8> %scalable_arg) #0 {
; CHECK-LABEL: uvdot_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:    add x9, x0, x1
; CHECK-NEXT:    str z9, [sp, #1, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    mov w8, wzr
; CHECK-NEXT:    str z8, [sp, #2, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    ld1h { z0.h, z8.h }, pn8/z, [x0]
; CHECK-NEXT:    ld1h { z1.h, z9.h }, pn8/z, [x9]
; CHECK-NEXT:    uvdot za.s[w8, 0, vgx2], { z0.h, z1.h }, z0.h[0]
; CHECK-NEXT:    uvdot za.s[w8, 0, vgx2], { z8.h, z9.h }, z0.h[0]
; CHECK-NEXT:    ldr z9, [sp, #1, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z8, [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
  tail call void @llvm.aarch64.sme.uvdot.lane.za32.vg1x2.nxv8i16(i32 0, <vscale x 8 x i16> %2, <vscale x 8 x i16> %5, <vscale x 8 x i16> poison, i32 0)
  tail call void @llvm.aarch64.sme.uvdot.lane.za32.vg1x2.nxv8i16(i32 0, <vscale x 8 x i16> %3, <vscale x 8 x i16> %6, <vscale x 8 x i16> poison, i32 0)
  ret void
}

define void @uvdot_form_4x_tuple(ptr %ptr, i64 %stride) #0 {
; CHECK-LABEL: uvdot_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:    uvdot za.s[w8, 0, vgx4], { z16.b - z19.b }, z0.b[0]
; CHECK-NEXT:    uvdot za.s[w8, 0, vgx4], { z20.b - z23.b }, z0.b[0]
; CHECK-NEXT:    uvdot za.s[w8, 0, vgx4], { z24.b - z27.b }, z0.b[0]
; CHECK-NEXT:    uvdot 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.uvdot.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.uvdot.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.uvdot.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.uvdot.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 @uvdot_form_4x_tuple_svecc(ptr %ptr, i64 %stride, <vscale x 16 x i8> %scalable_arg) #0 {
; CHECK-LABEL: uvdot_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 z15, [sp, #1, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    mov w8, wzr
; CHECK-NEXT:    str z14, [sp, #2, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    add x10, x9, x1
; CHECK-NEXT:    str z13, [sp, #3, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z12, [sp, #4, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z11, [sp, #5, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z10, [sp, #6, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z9, [sp, #7, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z8, [sp, #8, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    ld1b { z0.b, z4.b, z8.b, z12.b }, pn8/z, [x0]
; CHECK-NEXT:    ld1b { z1.b, z5.b, z9.b, z13.b }, pn8/z, [x0, x1]
; CHECK-NEXT:    ld1b { z2.b, z6.b, z10.b, z14.b }, pn8/z, [x0, x9]
; CHECK-NEXT:    ld1b { z3.b, z7.b, z11.b, z15.b }, pn8/z, [x0, x10]
; CHECK-NEXT:    uvdot za.s[w8, 0, vgx4], { z0.b - z3.b }, z0.b[0]
; CHECK-NEXT:    uvdot za.s[w8, 0, vgx4], { z4.b - z7.b }, z0.b[0]
; CHECK-NEXT:    uvdot za.s[w8, 0, vgx4], { z8.b - z11.b }, z0.b[0]
; CHECK-NEXT:    uvdot za.s[w8, 0, vgx4], { z12.b - z15.b }, z0.b[0]
; CHECK-NEXT:    ldr z15, [sp, #1, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z14, [sp, #2, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z13, [sp, #3, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z12, [sp, #4, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z11, [sp, #5, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z10, [sp, #6, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z9, [sp, #7, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z8, [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
  tail call void @llvm.aarch64.sme.uvdot.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.uvdot.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.uvdot.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.uvdot.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
}

; == SUVDOT ==

define void @test_suvdot_lane_za32_vg1x4_nxv16i8(i32 %slice, <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> %zm) {
; CHECK-LABEL: test_suvdot_lane_za32_vg1x4_nxv16i8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    suvdot za.s[w8, 0, vgx4], { z0.b - z3.b }, z4.b[3]
; CHECK-NEXT:    suvdot za.s[w8, 7, vgx4], { z0.b - z3.b }, z4.b[3]
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.suvdot.lane.za32.vg1x4.nxv16i8(i32 %slice, <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> %zm, i32 3)
  %slice.7 = add i32 %slice, 7
  call void @llvm.aarch64.sme.suvdot.lane.za32.vg1x4.nxv16i8(i32 %slice.7, <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> %zm, i32 3)
  ret void
}

define void @suvdot_form_4x_tuple(ptr %ptr, i64 %stride) #0 {
; CHECK-LABEL: suvdot_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:    suvdot za.s[w8, 0, vgx4], { z16.b - z19.b }, z0.b[0]
; CHECK-NEXT:    suvdot za.s[w8, 0, vgx4], { z20.b - z23.b }, z0.b[0]
; CHECK-NEXT:    suvdot za.s[w8, 0, vgx4], { z24.b - z27.b }, z0.b[0]
; CHECK-NEXT:    suvdot 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.suvdot.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.suvdot.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.suvdot.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.suvdot.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 @suvdot_form_4x_tuple_svecc(ptr %ptr, i64 %stride, <vscale x 16 x i8> %scalable_arg) #0 {
; CHECK-LABEL: suvdot_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 z15, [sp, #1, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    mov w8, wzr
; CHECK-NEXT:    str z14, [sp, #2, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    add x10, x9, x1
; CHECK-NEXT:    str z13, [sp, #3, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z12, [sp, #4, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z11, [sp, #5, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z10, [sp, #6, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z9, [sp, #7, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z8, [sp, #8, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    ld1b { z0.b, z4.b, z8.b, z12.b }, pn8/z, [x0]
; CHECK-NEXT:    ld1b { z1.b, z5.b, z9.b, z13.b }, pn8/z, [x0, x1]
; CHECK-NEXT:    ld1b { z2.b, z6.b, z10.b, z14.b }, pn8/z, [x0, x9]
; CHECK-NEXT:    ld1b { z3.b, z7.b, z11.b, z15.b }, pn8/z, [x0, x10]
; CHECK-NEXT:    suvdot za.s[w8, 0, vgx4], { z0.b - z3.b }, z0.b[0]
; CHECK-NEXT:    suvdot za.s[w8, 0, vgx4], { z4.b - z7.b }, z0.b[0]
; CHECK-NEXT:    suvdot za.s[w8, 0, vgx4], { z8.b - z11.b }, z0.b[0]
; CHECK-NEXT:    suvdot za.s[w8, 0, vgx4], { z12.b - z15.b }, z0.b[0]
; CHECK-NEXT:    ldr z15, [sp, #1, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z14, [sp, #2, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z13, [sp, #3, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z12, [sp, #4, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z11, [sp, #5, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z10, [sp, #6, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z9, [sp, #7, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z8, [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
  tail call void @llvm.aarch64.sme.suvdot.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.suvdot.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.suvdot.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.suvdot.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
}

; == USVDOT ==

define void @test_usvdot_lane_za32_vg1x4_nxv16i8(i32 %slice, <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> %zm) {
; CHECK-LABEL: test_usvdot_lane_za32_vg1x4_nxv16i8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    usvdot za.s[w8, 0, vgx4], { z0.b - z3.b }, z4.b[3]
; CHECK-NEXT:    usvdot za.s[w8, 7, vgx4], { z0.b - z3.b }, z4.b[3]
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.usvdot.lane.za32.vg1x4.nxv16i8(i32 %slice, <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> %zm, i32 3)
  %slice.7 = add i32 %slice, 7
  call void @llvm.aarch64.sme.usvdot.lane.za32.vg1x4.nxv16i8(i32 %slice.7, <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> %zm, i32 3)
  ret void
}

define void @usvdot_form_4x_tuple(ptr %ptr, i64 %stride) #0 {
; CHECK-LABEL: usvdot_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:    usvdot za.s[w8, 0, vgx4], { z16.b - z19.b }, z0.b[0]
; CHECK-NEXT:    usvdot za.s[w8, 0, vgx4], { z20.b - z23.b }, z0.b[0]
; CHECK-NEXT:    usvdot za.s[w8, 0, vgx4], { z24.b - z27.b }, z0.b[0]
; CHECK-NEXT:    usvdot 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.usvdot.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.usvdot.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.usvdot.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.usvdot.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 @usvdot_form_4x_tuple_svecc(ptr %ptr, i64 %stride, <vscale x 16 x i8> %scalable_arg) #0 {
; CHECK-LABEL: usvdot_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 z15, [sp, #1, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    mov w8, wzr
; CHECK-NEXT:    str z14, [sp, #2, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    add x10, x9, x1
; CHECK-NEXT:    str z13, [sp, #3, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z12, [sp, #4, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z11, [sp, #5, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z10, [sp, #6, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z9, [sp, #7, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    str z8, [sp, #8, mul vl] // 16-byte Folded Spill
; CHECK-NEXT:    ld1b { z0.b, z4.b, z8.b, z12.b }, pn8/z, [x0]
; CHECK-NEXT:    ld1b { z1.b, z5.b, z9.b, z13.b }, pn8/z, [x0, x1]
; CHECK-NEXT:    ld1b { z2.b, z6.b, z10.b, z14.b }, pn8/z, [x0, x9]
; CHECK-NEXT:    ld1b { z3.b, z7.b, z11.b, z15.b }, pn8/z, [x0, x10]
; CHECK-NEXT:    usvdot za.s[w8, 0, vgx4], { z0.b - z3.b }, z0.b[0]
; CHECK-NEXT:    usvdot za.s[w8, 0, vgx4], { z4.b - z7.b }, z0.b[0]
; CHECK-NEXT:    usvdot za.s[w8, 0, vgx4], { z8.b - z11.b }, z0.b[0]
; CHECK-NEXT:    usvdot za.s[w8, 0, vgx4], { z12.b - z15.b }, z0.b[0]
; CHECK-NEXT:    ldr z15, [sp, #1, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z14, [sp, #2, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z13, [sp, #3, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z12, [sp, #4, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z11, [sp, #5, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z10, [sp, #6, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z9, [sp, #7, mul vl] // 16-byte Folded Reload
; CHECK-NEXT:    ldr z8, [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
  tail call void @llvm.aarch64.sme.usvdot.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.usvdot.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.usvdot.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.usvdot.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
}

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

; == FVDOT ==
declare void @llvm.aarch64.sme.fvdot.lane.za32.vg1x2.nxv8f16(i32, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, i32)
declare void @llvm.aarch64.sme.fvdot.lane.za32.vg1x2.nxv8bf16(i32, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, i32)

; == SVDOT ==
declare void @llvm.aarch64.sme.svdot.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.svdot.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.svdot.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)

; == UVDOT ==
declare void @llvm.aarch64.sme.uvdot.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.uvdot.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.uvdot.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)

; == SUVDOT ==
declare void @llvm.aarch64.sme.suvdot.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)

; == USVDOT ==
declare void @llvm.aarch64.sme.usvdot.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)