llvm/llvm/test/CodeGen/AArch64/sme2-intrinsics-mlall.ll

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

;
; SMLALL
;

; Single x1

define void @multi_vector_mul_add_single_long_vg4x1_s8(i32 %slice, <vscale x 16 x i8> %dummy, <vscale x 16 x i8> %zn, <vscale x 16 x i8> %zm) {
; CHECK-LABEL: multi_vector_mul_add_single_long_vg4x1_s8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    smlall za.s[w8, 0:3], z1.b, z2.b
; CHECK-NEXT:    smlall za.s[w8, 12:15], z1.b, z2.b
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.smla.za32.single.vg4x1.nxv16i8(i32 %slice, <vscale x 16 x i8> %zn, <vscale x 16 x i8> %zm)
  %slice.12 = add i32 %slice, 12
  call void @llvm.aarch64.sme.smla.za32.single.vg4x1.nxv16i8(i32 %slice.12, <vscale x 16 x i8> %zn, <vscale x 16 x i8> %zm)
  ret void
}

define void @multi_vector_mul_add_single_long_vg4x1_s16(i32 %slice, <vscale x 8 x i16> %dummy, <vscale x 8 x i16> %zn, <vscale x 8 x i16> %zm) {
; CHECK-LABEL: multi_vector_mul_add_single_long_vg4x1_s16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    smlall za.d[w8, 0:3], z1.h, z2.h
; CHECK-NEXT:    smlall za.d[w8, 12:15], z1.h, z2.h
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.smla.za64.single.vg4x1.nxv8i16(i32 %slice, <vscale x 8 x i16> %zn, <vscale x 8 x i16> %zm)
  %slice.12 = add i32 %slice, 12
  call void @llvm.aarch64.sme.smla.za64.single.vg4x1.nxv8i16(i32 %slice.12, <vscale x 8 x i16> %zn, <vscale x 8 x i16> %zm)
  ret void
}

; Single x2

define void @multi_vector_mul_add_single_long_vg4x2_s8(i32 %slice, <vscale x 16 x i8> %dummy, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zm) {
; CHECK-LABEL: multi_vector_mul_add_single_long_vg4x2_s8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    // kill: def $z2 killed $z2 killed $z1_z2 def $z1_z2
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    // kill: def $z1 killed $z1 killed $z1_z2 def $z1_z2
; CHECK-NEXT:    smlall za.s[w8, 0:3, vgx2], { z1.b, z2.b }, z3.b
; CHECK-NEXT:    smlall za.s[w8, 4:7, vgx2], { z1.b, z2.b }, z3.b
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.smla.za32.single.vg4x2.nxv16i8(i32 %slice, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zm)
  %slice.4 = add i32 %slice, 4
  call void @llvm.aarch64.sme.smla.za32.single.vg4x2.nxv16i8(i32 %slice.4, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zm)
  ret void
}

define void @multi_vector_mul_add_single_long_vg4x2_s16(i32 %slice, <vscale x 8 x i16> %dummy, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zm) {
; CHECK-LABEL: multi_vector_mul_add_single_long_vg4x2_s16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    // kill: def $z2 killed $z2 killed $z1_z2 def $z1_z2
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    // kill: def $z1 killed $z1 killed $z1_z2 def $z1_z2
; CHECK-NEXT:    smlall za.d[w8, 0:3, vgx2], { z1.h, z2.h }, z3.h
; CHECK-NEXT:    smlall za.d[w8, 4:7, vgx2], { z1.h, z2.h }, z3.h
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.smla.za64.single.vg4x2.nxv8i16(i32 %slice, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zm)
  %slice.4 = add i32 %slice, 4
  call void @llvm.aarch64.sme.smla.za64.single.vg4x2.nxv8i16(i32 %slice.4, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zm)
  ret void
}

; Single x4

define void @multi_vector_mul_add_single_long_vg4x4_s8(i32 %slice, <vscale x 16 x i8> %dummy, <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> %zm) {
; CHECK-LABEL: multi_vector_mul_add_single_long_vg4x4_s8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    // kill: def $z4 killed $z4 killed $z1_z2_z3_z4 def $z1_z2_z3_z4
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    // kill: def $z3 killed $z3 killed $z1_z2_z3_z4 def $z1_z2_z3_z4
; CHECK-NEXT:    // kill: def $z2 killed $z2 killed $z1_z2_z3_z4 def $z1_z2_z3_z4
; CHECK-NEXT:    // kill: def $z1 killed $z1 killed $z1_z2_z3_z4 def $z1_z2_z3_z4
; CHECK-NEXT:    smlall za.s[w8, 0:3, vgx4], { z1.b - z4.b }, z5.b
; CHECK-NEXT:    smlall za.s[w8, 4:7, vgx4], { z1.b - z4.b }, z5.b
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.smla.za32.single.vg4x4.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> %zm)
  %slice.4 = add i32 %slice, 4
  call void @llvm.aarch64.sme.smla.za32.single.vg4x4.nxv16i8(i32 %slice.4, <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> %zm)
  ret void
}

define void @multi_vector_mul_add_single_long_vg4x4_s16(i32 %slice, <vscale x 8 x i16> %dummy, <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> %zm) {
; CHECK-LABEL: multi_vector_mul_add_single_long_vg4x4_s16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    // kill: def $z4 killed $z4 killed $z1_z2_z3_z4 def $z1_z2_z3_z4
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    // kill: def $z3 killed $z3 killed $z1_z2_z3_z4 def $z1_z2_z3_z4
; CHECK-NEXT:    // kill: def $z2 killed $z2 killed $z1_z2_z3_z4 def $z1_z2_z3_z4
; CHECK-NEXT:    // kill: def $z1 killed $z1 killed $z1_z2_z3_z4 def $z1_z2_z3_z4
; CHECK-NEXT:    smlall za.d[w8, 0:3, vgx4], { z1.h - z4.h }, z5.h
; CHECK-NEXT:    smlall za.d[w8, 4:7, vgx4], { z1.h - z4.h }, z5.h
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.smla.za64.single.vg4x4.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> %zm)
  %slice.4 = add i32 %slice, 4
  call void @llvm.aarch64.sme.smla.za64.single.vg4x4.nxv8i16(i32 %slice.4, <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> %zm)
  ret void
}

; Multi x2

define void @multi_vector_mul_add_multi_long_vg4x2_s8(i32 %slice, <vscale x 16 x i8> %dummy, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zm0, <vscale x 16 x i8> %zm1) {
; CHECK-LABEL: multi_vector_mul_add_multi_long_vg4x2_s8:
; 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:    smlall za.s[w8, 0:3, vgx2], { z6.b, z7.b }, { z4.b, z5.b }
; CHECK-NEXT:    smlall za.s[w8, 4:7, vgx2], { z6.b, z7.b }, { z4.b, z5.b }
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.smla.za32.vg4x2.nxv16i8(i32 %slice, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zm0, <vscale x 16 x i8> %zm1)
  %slice.4 = add i32 %slice, 4
  call void @llvm.aarch64.sme.smla.za32.vg4x2.nxv16i8(i32 %slice.4, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zm0, <vscale x 16 x i8> %zm1)
  ret void
}

define void @multi_vector_mul_add_multi_long_vg4x2_s16(i32 %slice, <vscale x 8 x i16> %dummy, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zm0, <vscale x 8 x i16> %zm1) {
; CHECK-LABEL: multi_vector_mul_add_multi_long_vg4x2_s16:
; 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:    smlall za.d[w8, 0:3, vgx2], { z6.h, z7.h }, { z4.h, z5.h }
; CHECK-NEXT:    smlall za.d[w8, 4:7, vgx2], { z6.h, z7.h }, { z4.h, z5.h }
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.smla.za64.vg4x2.nxv8i16(i32 %slice, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zm0, <vscale x 8 x i16> %zm1)
  %slice.4 = add i32 %slice, 4
  call void @llvm.aarch64.sme.smla.za64.vg4x2.nxv8i16(i32 %slice.4, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zm0, <vscale x 8 x i16> %zm1)
  ret void
}

; Multi x4

define void @multi_vector_mul_add_multi_long_vg4x4_s8(i32 %slice, <vscale x 16 x i8> %dummy, <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> %zm0, <vscale x 16 x i8> %zm1, <vscale x 16 x i8> %zm2, <vscale x 16 x i8> %zm3) {
; CHECK-LABEL: multi_vector_mul_add_multi_long_vg4x4_s8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z26.d, z7.d
; CHECK-NEXT:    mov z31.d, z4.d
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    ptrue p0.b
; CHECK-NEXT:    mov z25.d, z6.d
; CHECK-NEXT:    mov z30.d, z3.d
; CHECK-NEXT:    mov z24.d, z5.d
; CHECK-NEXT:    mov z29.d, z2.d
; CHECK-NEXT:    ld1b { z27.b }, p0/z, [x1]
; CHECK-NEXT:    mov z28.d, z1.d
; CHECK-NEXT:    smlall za.s[w8, 0:3, vgx4], { z28.b - z31.b }, { z24.b - z27.b }
; CHECK-NEXT:    smlall za.s[w8, 4:7, vgx4], { z28.b - z31.b }, { z24.b - z27.b }
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.smla.za32.vg4x4.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> %zm0, <vscale x 16 x i8> %zm1, <vscale x 16 x i8> %zm2, <vscale x 16 x i8> %zm3)
  %slice.4 = add i32 %slice, 4
  call void @llvm.aarch64.sme.smla.za32.vg4x4.nxv16i8(i32 %slice.4, <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> %zm0, <vscale x 16 x i8> %zm1, <vscale x 16 x i8> %zm2, <vscale x 16 x i8> %zm3)
  ret void
}

define void @multi_vector_mul_add_multi_long_vg4x4_s16(i32 %slice, <vscale x 8 x i16> %dummy, <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> %zm0, <vscale x 8 x i16> %zm1, <vscale x 8 x i16> %zm2, <vscale x 8 x i16> %zm3) {
; CHECK-LABEL: multi_vector_mul_add_multi_long_vg4x4_s16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z26.d, z7.d
; CHECK-NEXT:    mov z31.d, z4.d
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    ptrue p0.h
; CHECK-NEXT:    mov z25.d, z6.d
; CHECK-NEXT:    mov z30.d, z3.d
; CHECK-NEXT:    mov z24.d, z5.d
; CHECK-NEXT:    mov z29.d, z2.d
; CHECK-NEXT:    ld1h { z27.h }, p0/z, [x1]
; CHECK-NEXT:    mov z28.d, z1.d
; CHECK-NEXT:    smlall za.d[w8, 0:3, vgx4], { z28.h - z31.h }, { z24.h - z27.h }
; CHECK-NEXT:    smlall za.d[w8, 4:7, vgx4], { z28.h - z31.h }, { z24.h - z27.h }
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.smla.za64.vg4x4.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> %zm0, <vscale x 8 x i16> %zm1, <vscale x 8 x i16> %zm2, <vscale x 8 x i16> %zm3)
  %slice.4 = add i32 %slice, 4
  call void @llvm.aarch64.sme.smla.za64.vg4x4.nxv8i16(i32 %slice.4, <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> %zm0, <vscale x 8 x i16> %zm1, <vscale x 8 x i16> %zm2, <vscale x 8 x i16> %zm3)
  ret void
}

; Indexed x1

define void @multi_vector_mul_add_lane_long_vg4x1_s8(i32 %slice, <vscale x 16 x i8> %dummy, <vscale x 16 x i8> %zn, <vscale x 16 x i8> %zm) {
; CHECK-LABEL: multi_vector_mul_add_lane_long_vg4x1_s8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    smlall za.s[w8, 0:3], z1.b, z2.b[0]
; CHECK-NEXT:    smlall za.s[w8, 12:15], z1.b, z2.b[15]
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.smla.za32.lane.vg4x1.nxv16i8(i32 %slice, <vscale x 16 x i8> %zn, <vscale x 16 x i8> %zm, i32 0)
  %slice.12 = add i32 %slice, 12
  call void @llvm.aarch64.sme.smla.za32.lane.vg4x1.nxv16i8(i32 %slice.12, <vscale x 16 x i8> %zn, <vscale x 16 x i8> %zm, i32 15)
  ret void
}

define void @multi_vector_mul_add_lane_long_vg4x1_s16(i32 %slice, <vscale x 8 x i16> %dummy, <vscale x 8 x i16> %zn, <vscale x 8 x i16> %zm) {
; CHECK-LABEL: multi_vector_mul_add_lane_long_vg4x1_s16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    smlall za.d[w8, 0:3], z1.h, z2.h[0]
; CHECK-NEXT:    smlall za.d[w8, 12:15], z1.h, z2.h[7]
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.smla.za64.lane.vg4x1.nxv8i16(i32 %slice, <vscale x 8 x i16> %zn, <vscale x 8 x i16> %zm, i32 0)
  %slice.12 = add i32 %slice, 12
  call void @llvm.aarch64.sme.smla.za64.lane.vg4x1.nxv8i16(i32 %slice.12, <vscale x 8 x i16> %zn, <vscale x 8 x i16> %zm, i32 7)
  ret void
}

; Indexed x2

define void @multi_vector_mul_add_lane_long_vg4x2_s8(i32 %slice, <vscale x 16 x i8> %dummy, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zm) {
; CHECK-LABEL: multi_vector_mul_add_lane_long_vg4x2_s8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z5.d, z2.d
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    mov z4.d, z1.d
; CHECK-NEXT:    smlall za.s[w8, 0:3, vgx2], { z4.b, z5.b }, z3.b[0]
; CHECK-NEXT:    smlall za.s[w8, 4:7, vgx2], { z4.b, z5.b }, z3.b[15]
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.smla.za32.lane.vg4x2.nxv16i8(i32 %slice, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zm, i32 0)
  %slice.4 = add i32 %slice, 4
  call void @llvm.aarch64.sme.smla.za32.lane.vg4x2.nxv16i8(i32 %slice.4, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zm, i32 15)
  ret void
}

define void @multi_vector_mul_add_lane_long_vg4x2_s16(i32 %slice, <vscale x 8 x i16> %dummy, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zm) {
; CHECK-LABEL: multi_vector_mul_add_lane_long_vg4x2_s16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z5.d, z2.d
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    mov z4.d, z1.d
; CHECK-NEXT:    smlall za.d[w8, 0:3, vgx2], { z4.h, z5.h }, z3.h[0]
; CHECK-NEXT:    smlall za.d[w8, 4:7, vgx2], { z4.h, z5.h }, z3.h[7]
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.smla.za64.lane.vg4x2.nxv8i16(i32 %slice, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zm, i32 0)
  %slice.4 = add i32 %slice, 4
  call void @llvm.aarch64.sme.smla.za64.lane.vg4x2.nxv8i16(i32 %slice.4, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zm, i32 7)
  ret void
}

; Indexed x4

define void @multi_vector_mul_add_lane_long_vg4x4_s8(i32 %slice, <vscale x 16 x i8> %dummy, <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> %zm) {
; CHECK-LABEL: multi_vector_mul_add_lane_long_vg4x4_s8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z27.d, z4.d
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    mov z26.d, z3.d
; CHECK-NEXT:    mov z25.d, z2.d
; CHECK-NEXT:    mov z24.d, z1.d
; CHECK-NEXT:    smlall za.s[w8, 0:3, vgx4], { z24.b - z27.b }, z5.b[0]
; CHECK-NEXT:    smlall za.s[w8, 4:7, vgx4], { z24.b - z27.b }, z5.b[15]
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.smla.za32.lane.vg4x4.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> %zm, i32 0)
  %slice.4 = add i32 %slice, 4
  call void @llvm.aarch64.sme.smla.za32.lane.vg4x4.nxv16i8(i32 %slice.4, <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> %zm, i32 15)
  ret void
}

define void @multi_vector_mul_add_lane_long_vg4x4_s16(i32 %slice, <vscale x 8 x i16> %dummy, <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> %zm) {
; CHECK-LABEL: multi_vector_mul_add_lane_long_vg4x4_s16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z27.d, z4.d
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    mov z26.d, z3.d
; CHECK-NEXT:    mov z25.d, z2.d
; CHECK-NEXT:    mov z24.d, z1.d
; CHECK-NEXT:    smlall za.d[w8, 0:3, vgx4], { z24.h - z27.h }, z5.h[0]
; CHECK-NEXT:    smlall za.d[w8, 4:7, vgx4], { z24.h - z27.h }, z5.h[7]
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.smla.za64.lane.vg4x4.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> %zm, i32 0)
  %slice.4 = add i32 %slice, 4
  call void @llvm.aarch64.sme.smla.za64.lane.vg4x4.nxv8i16(i32 %slice.4, <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> %zm, i32 7)
  ret void
}

; UMLALL

; Single x1

define void @multi_vector_mul_add_single_long_vg4x1_u8(i32 %slice, <vscale x 16 x i8> %dummy, <vscale x 16 x i8> %zn, <vscale x 16 x i8> %zm) {
; CHECK-LABEL: multi_vector_mul_add_single_long_vg4x1_u8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    umlall za.s[w8, 0:3], z1.b, z2.b
; CHECK-NEXT:    umlall za.s[w8, 12:15], z1.b, z2.b
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.umla.za32.single.vg4x1.nxv16i8(i32 %slice, <vscale x 16 x i8> %zn, <vscale x 16 x i8> %zm)
  %slice.12 = add i32 %slice, 12
  call void @llvm.aarch64.sme.umla.za32.single.vg4x1.nxv16i8(i32 %slice.12, <vscale x 16 x i8> %zn, <vscale x 16 x i8> %zm)
  ret void
}

define void @multi_vector_mul_add_single_long_vg4x1_u16(i32 %slice, <vscale x 8 x i16> %dummy, <vscale x 8 x i16> %zn, <vscale x 8 x i16> %zm) {
; CHECK-LABEL: multi_vector_mul_add_single_long_vg4x1_u16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    umlall za.d[w8, 0:3], z1.h, z2.h
; CHECK-NEXT:    umlall za.d[w8, 12:15], z1.h, z2.h
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.umla.za64.single.vg4x1.nxv8i16(i32 %slice, <vscale x 8 x i16> %zn, <vscale x 8 x i16> %zm)
  %slice.12 = add i32 %slice, 12
  call void @llvm.aarch64.sme.umla.za64.single.vg4x1.nxv8i16(i32 %slice.12, <vscale x 8 x i16> %zn, <vscale x 8 x i16> %zm)
  ret void
}

; Single x2

define void @multi_vector_mul_add_single_long_vg4x2_u8(i32 %slice, <vscale x 16 x i8> %dummy, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zm) {
; CHECK-LABEL: multi_vector_mul_add_single_long_vg4x2_u8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    // kill: def $z2 killed $z2 killed $z1_z2 def $z1_z2
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    // kill: def $z1 killed $z1 killed $z1_z2 def $z1_z2
; CHECK-NEXT:    umlall za.s[w8, 0:3, vgx2], { z1.b, z2.b }, z3.b
; CHECK-NEXT:    umlall za.s[w8, 4:7, vgx2], { z1.b, z2.b }, z3.b
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.umla.za32.single.vg4x2.nxv16i8(i32 %slice, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zm)
  %slice.4 = add i32 %slice, 4
  call void @llvm.aarch64.sme.umla.za32.single.vg4x2.nxv16i8(i32 %slice.4, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zm)
  ret void
}

define void @multi_vector_mul_add_single_long_vg4x2_u16(i32 %slice, <vscale x 8 x i16> %dummy, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zm) {
; CHECK-LABEL: multi_vector_mul_add_single_long_vg4x2_u16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    // kill: def $z2 killed $z2 killed $z1_z2 def $z1_z2
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    // kill: def $z1 killed $z1 killed $z1_z2 def $z1_z2
; CHECK-NEXT:    umlall za.d[w8, 0:3, vgx2], { z1.h, z2.h }, z3.h
; CHECK-NEXT:    umlall za.d[w8, 4:7, vgx2], { z1.h, z2.h }, z3.h
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.umla.za64.single.vg4x2.nxv8i16(i32 %slice, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zm)
  %slice.4 = add i32 %slice, 4
  call void @llvm.aarch64.sme.umla.za64.single.vg4x2.nxv8i16(i32 %slice.4, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zm)
  ret void
}

; Single x4

define void @multi_vector_mul_add_single_long_vg4x4_u8(i32 %slice, <vscale x 16 x i8> %dummy, <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> %zm) {
; CHECK-LABEL: multi_vector_mul_add_single_long_vg4x4_u8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    // kill: def $z4 killed $z4 killed $z1_z2_z3_z4 def $z1_z2_z3_z4
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    // kill: def $z3 killed $z3 killed $z1_z2_z3_z4 def $z1_z2_z3_z4
; CHECK-NEXT:    // kill: def $z2 killed $z2 killed $z1_z2_z3_z4 def $z1_z2_z3_z4
; CHECK-NEXT:    // kill: def $z1 killed $z1 killed $z1_z2_z3_z4 def $z1_z2_z3_z4
; CHECK-NEXT:    umlall za.s[w8, 0:3, vgx4], { z1.b - z4.b }, z5.b
; CHECK-NEXT:    umlall za.s[w8, 4:7, vgx4], { z1.b - z4.b }, z5.b
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.umla.za32.single.vg4x4.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> %zm)
  %slice.4 = add i32 %slice, 4
  call void @llvm.aarch64.sme.umla.za32.single.vg4x4.nxv16i8(i32 %slice.4, <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> %zm)
  ret void
}

define void @multi_vector_mul_add_single_long_vg4x4_u16(i32 %slice, <vscale x 8 x i16> %dummy, <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> %zm) {
; CHECK-LABEL: multi_vector_mul_add_single_long_vg4x4_u16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    // kill: def $z4 killed $z4 killed $z1_z2_z3_z4 def $z1_z2_z3_z4
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    // kill: def $z3 killed $z3 killed $z1_z2_z3_z4 def $z1_z2_z3_z4
; CHECK-NEXT:    // kill: def $z2 killed $z2 killed $z1_z2_z3_z4 def $z1_z2_z3_z4
; CHECK-NEXT:    // kill: def $z1 killed $z1 killed $z1_z2_z3_z4 def $z1_z2_z3_z4
; CHECK-NEXT:    umlall za.d[w8, 0:3, vgx4], { z1.h - z4.h }, z5.h
; CHECK-NEXT:    umlall za.d[w8, 4:7, vgx4], { z1.h - z4.h }, z5.h
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.umla.za64.single.vg4x4.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> %zm)
  %slice.4 = add i32 %slice, 4
  call void @llvm.aarch64.sme.umla.za64.single.vg4x4.nxv8i16(i32 %slice.4, <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> %zm)
  ret void
}

; Multi x2

define void @multi_vector_mul_add_multi_long_vg4x2_u8(i32 %slice, <vscale x 16 x i8> %dummy, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zm0, <vscale x 16 x i8> %zm1) {
; CHECK-LABEL: multi_vector_mul_add_multi_long_vg4x2_u8:
; 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:    umlall za.s[w8, 0:3, vgx2], { z6.b, z7.b }, { z4.b, z5.b }
; CHECK-NEXT:    umlall za.s[w8, 4:7, vgx2], { z6.b, z7.b }, { z4.b, z5.b }
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.umla.za32.vg4x2.nxv16i8(i32 %slice, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zm0, <vscale x 16 x i8> %zm1)
  %slice.4 = add i32 %slice, 4
  call void @llvm.aarch64.sme.umla.za32.vg4x2.nxv16i8(i32 %slice.4, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zm0, <vscale x 16 x i8> %zm1)
  ret void
}

define void @multi_vector_mul_add_multi_long_vg4x2_u16(i32 %slice, <vscale x 8 x i16> %dummy, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zm0, <vscale x 8 x i16> %zm1) {
; CHECK-LABEL: multi_vector_mul_add_multi_long_vg4x2_u16:
; 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:    umlall za.d[w8, 0:3, vgx2], { z6.h, z7.h }, { z4.h, z5.h }
; CHECK-NEXT:    umlall za.d[w8, 4:7, vgx2], { z6.h, z7.h }, { z4.h, z5.h }
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.umla.za64.vg4x2.nxv8i16(i32 %slice, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zm0, <vscale x 8 x i16> %zm1)
  %slice.4 = add i32 %slice, 4
  call void @llvm.aarch64.sme.umla.za64.vg4x2.nxv8i16(i32 %slice.4, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zm0, <vscale x 8 x i16> %zm1)
  ret void
}

; Multi x4

define void @multi_vector_mul_add_multi_long_vg4x4_u8(i32 %slice, <vscale x 16 x i8> %dummy, <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> %zm0, <vscale x 16 x i8> %zm1, <vscale x 16 x i8> %zm2, <vscale x 16 x i8> %zm3) {
; CHECK-LABEL: multi_vector_mul_add_multi_long_vg4x4_u8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z26.d, z7.d
; CHECK-NEXT:    mov z31.d, z4.d
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    ptrue p0.b
; CHECK-NEXT:    mov z25.d, z6.d
; CHECK-NEXT:    mov z30.d, z3.d
; CHECK-NEXT:    mov z24.d, z5.d
; CHECK-NEXT:    mov z29.d, z2.d
; CHECK-NEXT:    ld1b { z27.b }, p0/z, [x1]
; CHECK-NEXT:    mov z28.d, z1.d
; CHECK-NEXT:    umlall za.s[w8, 0:3, vgx4], { z28.b - z31.b }, { z24.b - z27.b }
; CHECK-NEXT:    umlall za.s[w8, 4:7, vgx4], { z28.b - z31.b }, { z24.b - z27.b }
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.umla.za32.vg4x4.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> %zm0, <vscale x 16 x i8> %zm1, <vscale x 16 x i8> %zm2, <vscale x 16 x i8> %zm3)
  %slice.4 = add i32 %slice, 4
  call void @llvm.aarch64.sme.umla.za32.vg4x4.nxv16i8(i32 %slice.4, <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> %zm0, <vscale x 16 x i8> %zm1, <vscale x 16 x i8> %zm2, <vscale x 16 x i8> %zm3)
  ret void
}

define void @multi_vector_mul_add_multi_long_vg4x4_u16(i32 %slice, <vscale x 8 x i16> %dummy, <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> %zm0, <vscale x 8 x i16> %zm1, <vscale x 8 x i16> %zm2, <vscale x 8 x i16> %zm3) {
; CHECK-LABEL: multi_vector_mul_add_multi_long_vg4x4_u16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z26.d, z7.d
; CHECK-NEXT:    mov z31.d, z4.d
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    ptrue p0.h
; CHECK-NEXT:    mov z25.d, z6.d
; CHECK-NEXT:    mov z30.d, z3.d
; CHECK-NEXT:    mov z24.d, z5.d
; CHECK-NEXT:    mov z29.d, z2.d
; CHECK-NEXT:    ld1h { z27.h }, p0/z, [x1]
; CHECK-NEXT:    mov z28.d, z1.d
; CHECK-NEXT:    umlall za.d[w8, 0:3, vgx4], { z28.h - z31.h }, { z24.h - z27.h }
; CHECK-NEXT:    umlall za.d[w8, 4:7, vgx4], { z28.h - z31.h }, { z24.h - z27.h }
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.umla.za64.vg4x4.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> %zm0, <vscale x 8 x i16> %zm1, <vscale x 8 x i16> %zm2, <vscale x 8 x i16> %zm3)
  %slice.4 = add i32 %slice, 4
  call void @llvm.aarch64.sme.umla.za64.vg4x4.nxv8i16(i32 %slice.4, <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> %zm0, <vscale x 8 x i16> %zm1, <vscale x 8 x i16> %zm2, <vscale x 8 x i16> %zm3)
  ret void
}

; Indexed x1

define void @multi_vector_mul_add_lane_long_vg4x1_u8(i32 %slice, <vscale x 16 x i8> %dummy, <vscale x 16 x i8> %zn, <vscale x 16 x i8> %zm) {
; CHECK-LABEL: multi_vector_mul_add_lane_long_vg4x1_u8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    umlall za.s[w8, 0:3], z1.b, z2.b[0]
; CHECK-NEXT:    umlall za.s[w8, 12:15], z1.b, z2.b[15]
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.umla.za32.lane.vg4x1.nxv16i8(i32 %slice, <vscale x 16 x i8> %zn, <vscale x 16 x i8> %zm, i32 0)
  %slice.12 = add i32 %slice, 12
  call void @llvm.aarch64.sme.umla.za32.lane.vg4x1.nxv16i8(i32 %slice.12, <vscale x 16 x i8> %zn, <vscale x 16 x i8> %zm, i32 15)
  ret void
}

define void @multi_vector_mul_add_lane_long_vg4x1_u16(i32 %slice, <vscale x 8 x i16> %dummy, <vscale x 8 x i16> %zn, <vscale x 8 x i16> %zm) {
; CHECK-LABEL: multi_vector_mul_add_lane_long_vg4x1_u16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    umlall za.d[w8, 0:3], z1.h, z2.h[0]
; CHECK-NEXT:    umlall za.d[w8, 12:15], z1.h, z2.h[7]
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.umla.za64.lane.vg4x1.nxv8i16(i32 %slice, <vscale x 8 x i16> %zn, <vscale x 8 x i16> %zm, i32 0)
  %slice.12 = add i32 %slice, 12
  call void @llvm.aarch64.sme.umla.za64.lane.vg4x1.nxv8i16(i32 %slice.12, <vscale x 8 x i16> %zn, <vscale x 8 x i16> %zm, i32 7)
  ret void
}

; Indexed x2

define void @multi_vector_mul_add_lane_long_vg4x2_u8(i32 %slice, <vscale x 16 x i8> %dummy, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zm) {
; CHECK-LABEL: multi_vector_mul_add_lane_long_vg4x2_u8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z5.d, z2.d
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    mov z4.d, z1.d
; CHECK-NEXT:    umlall za.s[w8, 0:3, vgx2], { z4.b, z5.b }, z3.b[0]
; CHECK-NEXT:    umlall za.s[w8, 4:7, vgx2], { z4.b, z5.b }, z3.b[15]
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.umla.za32.lane.vg4x2.nxv16i8(i32 %slice, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zm, i32 0)
  %slice.4 = add i32 %slice, 4
  call void @llvm.aarch64.sme.umla.za32.lane.vg4x2.nxv16i8(i32 %slice.4, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zm, i32 15)
  ret void
}

define void @multi_vector_mul_add_lane_long_vg4x2_u16(i32 %slice, <vscale x 8 x i16> %dummy, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zm) {
; CHECK-LABEL: multi_vector_mul_add_lane_long_vg4x2_u16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z5.d, z2.d
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    mov z4.d, z1.d
; CHECK-NEXT:    umlall za.d[w8, 0:3, vgx2], { z4.h, z5.h }, z3.h[0]
; CHECK-NEXT:    umlall za.d[w8, 4:7, vgx2], { z4.h, z5.h }, z3.h[7]
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.umla.za64.lane.vg4x2.nxv8i16(i32 %slice, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zm, i32 0)
  %slice.4 = add i32 %slice, 4
  call void @llvm.aarch64.sme.umla.za64.lane.vg4x2.nxv8i16(i32 %slice.4, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zm, i32 7)
  ret void
}

; Indexed x4

define void @multi_vector_mul_add_lane_long_vg4x4_u8(i32 %slice, <vscale x 16 x i8> %dummy, <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> %zm) {
; CHECK-LABEL: multi_vector_mul_add_lane_long_vg4x4_u8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z27.d, z4.d
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    mov z26.d, z3.d
; CHECK-NEXT:    mov z25.d, z2.d
; CHECK-NEXT:    mov z24.d, z1.d
; CHECK-NEXT:    umlall za.s[w8, 0:3, vgx4], { z24.b - z27.b }, z5.b[0]
; CHECK-NEXT:    umlall za.s[w8, 4:7, vgx4], { z24.b - z27.b }, z5.b[15]
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.umla.za32.lane.vg4x4.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> %zm, i32 0)
  %slice.4 = add i32 %slice, 4
  call void @llvm.aarch64.sme.umla.za32.lane.vg4x4.nxv16i8(i32 %slice.4, <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> %zm, i32 15)
  ret void
}

define void @multi_vector_mul_add_lane_long_vg4x4_u16(i32 %slice, <vscale x 8 x i16> %dummy, <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> %zm) {
; CHECK-LABEL: multi_vector_mul_add_lane_long_vg4x4_u16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z27.d, z4.d
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    mov z26.d, z3.d
; CHECK-NEXT:    mov z25.d, z2.d
; CHECK-NEXT:    mov z24.d, z1.d
; CHECK-NEXT:    umlall za.d[w8, 0:3, vgx4], { z24.h - z27.h }, z5.h[0]
; CHECK-NEXT:    umlall za.d[w8, 4:7, vgx4], { z24.h - z27.h }, z5.h[7]
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.umla.za64.lane.vg4x4.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> %zm, i32 0)
  %slice.4 = add i32 %slice, 4
  call void @llvm.aarch64.sme.umla.za64.lane.vg4x4.nxv8i16(i32 %slice.4, <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> %zm, i32 7)
  ret void
}

; SMLSLL

; Single x1

define void @multi_vector_mul_sub_single_long_vg4x1_s8(i32 %slice, <vscale x 16 x i8> %dummy, <vscale x 16 x i8> %zn, <vscale x 16 x i8> %zm) {
; CHECK-LABEL: multi_vector_mul_sub_single_long_vg4x1_s8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    smlsll za.s[w8, 0:3], z1.b, z2.b
; CHECK-NEXT:    smlsll za.s[w8, 12:15], z1.b, z2.b
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.smls.za32.single.vg4x1.nxv16i8(i32 %slice, <vscale x 16 x i8> %zn, <vscale x 16 x i8> %zm)
  %slice.12 = add i32 %slice, 12
  call void @llvm.aarch64.sme.smls.za32.single.vg4x1.nxv16i8(i32 %slice.12, <vscale x 16 x i8> %zn, <vscale x 16 x i8> %zm)
  ret void
}

define void @multi_vector_mul_sub_single_long_vg4x1_s16(i32 %slice, <vscale x 8 x i16> %dummy, <vscale x 8 x i16> %zn, <vscale x 8 x i16> %zm) {
; CHECK-LABEL: multi_vector_mul_sub_single_long_vg4x1_s16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    smlsll za.d[w8, 0:3], z1.h, z2.h
; CHECK-NEXT:    smlsll za.d[w8, 12:15], z1.h, z2.h
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.smls.za64.single.vg4x1.nxv8i16(i32 %slice, <vscale x 8 x i16> %zn, <vscale x 8 x i16> %zm)
  %slice.12 = add i32 %slice, 12
  call void @llvm.aarch64.sme.smls.za64.single.vg4x1.nxv8i16(i32 %slice.12, <vscale x 8 x i16> %zn, <vscale x 8 x i16> %zm)
  ret void
}

; Single x2

define void @multi_vector_mul_sub_single_long_vg4x2_s8(i32 %slice, <vscale x 16 x i8> %dummy, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zm) {
; CHECK-LABEL: multi_vector_mul_sub_single_long_vg4x2_s8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    // kill: def $z2 killed $z2 killed $z1_z2 def $z1_z2
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    // kill: def $z1 killed $z1 killed $z1_z2 def $z1_z2
; CHECK-NEXT:    smlsll za.s[w8, 0:3, vgx2], { z1.b, z2.b }, z3.b
; CHECK-NEXT:    smlsll za.s[w8, 4:7, vgx2], { z1.b, z2.b }, z3.b
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.smls.za32.single.vg4x2.nxv16i8(i32 %slice, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zm)
  %slice.4 = add i32 %slice, 4
  call void @llvm.aarch64.sme.smls.za32.single.vg4x2.nxv16i8(i32 %slice.4, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zm)
  ret void
}

define void @multi_vector_mul_sub_single_long_vg4x2_s16(i32 %slice, <vscale x 8 x i16> %dummy, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zm) {
; CHECK-LABEL: multi_vector_mul_sub_single_long_vg4x2_s16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    // kill: def $z2 killed $z2 killed $z1_z2 def $z1_z2
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    // kill: def $z1 killed $z1 killed $z1_z2 def $z1_z2
; CHECK-NEXT:    smlsll za.d[w8, 0:3, vgx2], { z1.h, z2.h }, z3.h
; CHECK-NEXT:    smlsll za.d[w8, 4:7, vgx2], { z1.h, z2.h }, z3.h
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.smls.za64.single.vg4x2.nxv8i16(i32 %slice, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zm)
  %slice.4 = add i32 %slice, 4
  call void @llvm.aarch64.sme.smls.za64.single.vg4x2.nxv8i16(i32 %slice.4, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zm)
  ret void
}

; Single x4

define void @multi_vector_mul_sub_single_long_vg4x4_s8(i32 %slice, <vscale x 16 x i8> %dummy, <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> %zm) {
; CHECK-LABEL: multi_vector_mul_sub_single_long_vg4x4_s8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    // kill: def $z4 killed $z4 killed $z1_z2_z3_z4 def $z1_z2_z3_z4
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    // kill: def $z3 killed $z3 killed $z1_z2_z3_z4 def $z1_z2_z3_z4
; CHECK-NEXT:    // kill: def $z2 killed $z2 killed $z1_z2_z3_z4 def $z1_z2_z3_z4
; CHECK-NEXT:    // kill: def $z1 killed $z1 killed $z1_z2_z3_z4 def $z1_z2_z3_z4
; CHECK-NEXT:    smlsll za.s[w8, 0:3, vgx4], { z1.b - z4.b }, z5.b
; CHECK-NEXT:    smlsll za.s[w8, 4:7, vgx4], { z1.b - z4.b }, z5.b
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.smls.za32.single.vg4x4.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> %zm)
  %slice.4 = add i32 %slice, 4
  call void @llvm.aarch64.sme.smls.za32.single.vg4x4.nxv16i8(i32 %slice.4, <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> %zm)
  ret void
}

define void @multi_vector_mul_sub_single_long_vg4x4_s16(i32 %slice, <vscale x 8 x i16> %dummy, <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> %zm) {
; CHECK-LABEL: multi_vector_mul_sub_single_long_vg4x4_s16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    // kill: def $z4 killed $z4 killed $z1_z2_z3_z4 def $z1_z2_z3_z4
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    // kill: def $z3 killed $z3 killed $z1_z2_z3_z4 def $z1_z2_z3_z4
; CHECK-NEXT:    // kill: def $z2 killed $z2 killed $z1_z2_z3_z4 def $z1_z2_z3_z4
; CHECK-NEXT:    // kill: def $z1 killed $z1 killed $z1_z2_z3_z4 def $z1_z2_z3_z4
; CHECK-NEXT:    smlsll za.d[w8, 0:3, vgx4], { z1.h - z4.h }, z5.h
; CHECK-NEXT:    smlsll za.d[w8, 4:7, vgx4], { z1.h - z4.h }, z5.h
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.smls.za64.single.vg4x4.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> %zm)
  %slice.4 = add i32 %slice, 4
  call void @llvm.aarch64.sme.smls.za64.single.vg4x4.nxv8i16(i32 %slice.4, <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> %zm)
  ret void
}

; Multi x2

define void @multi_vector_mul_sub_multi_long_vg4x2_s8(i32 %slice, <vscale x 16 x i8> %dummy, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zm0, <vscale x 16 x i8> %zm1) {
; CHECK-LABEL: multi_vector_mul_sub_multi_long_vg4x2_s8:
; 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:    smlsll za.s[w8, 0:3, vgx2], { z6.b, z7.b }, { z4.b, z5.b }
; CHECK-NEXT:    smlsll za.s[w8, 4:7, vgx2], { z6.b, z7.b }, { z4.b, z5.b }
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.smls.za32.vg4x2.nxv16i8(i32 %slice, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zm0, <vscale x 16 x i8> %zm1)
  %slice.4 = add i32 %slice, 4
  call void @llvm.aarch64.sme.smls.za32.vg4x2.nxv16i8(i32 %slice.4, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zm0, <vscale x 16 x i8> %zm1)
  ret void
}

define void @multi_vector_mul_sub_multi_long_vg4x2_s16(i32 %slice, <vscale x 8 x i16> %dummy, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zm0, <vscale x 8 x i16> %zm1) {
; CHECK-LABEL: multi_vector_mul_sub_multi_long_vg4x2_s16:
; 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:    smlsll za.d[w8, 0:3, vgx2], { z6.h, z7.h }, { z4.h, z5.h }
; CHECK-NEXT:    smlsll za.d[w8, 4:7, vgx2], { z6.h, z7.h }, { z4.h, z5.h }
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.smls.za64.vg4x2.nxv8i16(i32 %slice, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zm0, <vscale x 8 x i16> %zm1)
  %slice.4 = add i32 %slice, 4
  call void @llvm.aarch64.sme.smls.za64.vg4x2.nxv8i16(i32 %slice.4, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zm0, <vscale x 8 x i16> %zm1)
  ret void
}

; Multi x4

define void @multi_vector_mul_sub_multi_long_vg4x4_s8(i32 %slice, <vscale x 16 x i8> %dummy, <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> %zm0, <vscale x 16 x i8> %zm1, <vscale x 16 x i8> %zm2, <vscale x 16 x i8> %zm3) {
; CHECK-LABEL: multi_vector_mul_sub_multi_long_vg4x4_s8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z26.d, z7.d
; CHECK-NEXT:    mov z31.d, z4.d
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    ptrue p0.b
; CHECK-NEXT:    mov z25.d, z6.d
; CHECK-NEXT:    mov z30.d, z3.d
; CHECK-NEXT:    mov z24.d, z5.d
; CHECK-NEXT:    mov z29.d, z2.d
; CHECK-NEXT:    ld1b { z27.b }, p0/z, [x1]
; CHECK-NEXT:    mov z28.d, z1.d
; CHECK-NEXT:    smlsll za.s[w8, 0:3, vgx4], { z28.b - z31.b }, { z24.b - z27.b }
; CHECK-NEXT:    smlsll za.s[w8, 4:7, vgx4], { z28.b - z31.b }, { z24.b - z27.b }
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.smls.za32.vg4x4.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> %zm0, <vscale x 16 x i8> %zm1, <vscale x 16 x i8> %zm2, <vscale x 16 x i8> %zm3)
  %slice.4 = add i32 %slice, 4
  call void @llvm.aarch64.sme.smls.za32.vg4x4.nxv16i8(i32 %slice.4, <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> %zm0, <vscale x 16 x i8> %zm1, <vscale x 16 x i8> %zm2, <vscale x 16 x i8> %zm3)
  ret void
}

define void @multi_vector_mul_sub_multi_long_vg4x4_s16(i32 %slice, <vscale x 8 x i16> %dummy, <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> %zm0, <vscale x 8 x i16> %zm1, <vscale x 8 x i16> %zm2, <vscale x 8 x i16> %zm3) {
; CHECK-LABEL: multi_vector_mul_sub_multi_long_vg4x4_s16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z26.d, z7.d
; CHECK-NEXT:    mov z31.d, z4.d
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    ptrue p0.h
; CHECK-NEXT:    mov z25.d, z6.d
; CHECK-NEXT:    mov z30.d, z3.d
; CHECK-NEXT:    mov z24.d, z5.d
; CHECK-NEXT:    mov z29.d, z2.d
; CHECK-NEXT:    ld1h { z27.h }, p0/z, [x1]
; CHECK-NEXT:    mov z28.d, z1.d
; CHECK-NEXT:    smlsll za.d[w8, 0:3, vgx4], { z28.h - z31.h }, { z24.h - z27.h }
; CHECK-NEXT:    smlsll za.d[w8, 4:7, vgx4], { z28.h - z31.h }, { z24.h - z27.h }
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.smls.za64.vg4x4.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> %zm0, <vscale x 8 x i16> %zm1, <vscale x 8 x i16> %zm2, <vscale x 8 x i16> %zm3)
  %slice.4 = add i32 %slice, 4
  call void @llvm.aarch64.sme.smls.za64.vg4x4.nxv8i16(i32 %slice.4, <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> %zm0, <vscale x 8 x i16> %zm1, <vscale x 8 x i16> %zm2, <vscale x 8 x i16> %zm3)
  ret void
}

; Indexed x1

define void @multi_vector_mul_sub_lane_long_vg4x1_s8(i32 %slice, <vscale x 16 x i8> %dummy, <vscale x 16 x i8> %zn, <vscale x 16 x i8> %zm) {
; CHECK-LABEL: multi_vector_mul_sub_lane_long_vg4x1_s8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    smlsll za.s[w8, 0:3], z1.b, z2.b[0]
; CHECK-NEXT:    smlsll za.s[w8, 12:15], z1.b, z2.b[15]
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.smls.za32.lane.vg4x1.nxv16i8(i32 %slice, <vscale x 16 x i8> %zn, <vscale x 16 x i8> %zm, i32 0)
  %slice.12 = add i32 %slice, 12
  call void @llvm.aarch64.sme.smls.za32.lane.vg4x1.nxv16i8(i32 %slice.12, <vscale x 16 x i8> %zn, <vscale x 16 x i8> %zm, i32 15)
  ret void
}

define void @multi_vector_mul_sub_lane_long_vg4x1_s16(i32 %slice, <vscale x 8 x i16> %dummy, <vscale x 8 x i16> %zn, <vscale x 8 x i16> %zm) {
; CHECK-LABEL: multi_vector_mul_sub_lane_long_vg4x1_s16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    smlsll za.d[w8, 0:3], z1.h, z2.h[0]
; CHECK-NEXT:    smlsll za.d[w8, 12:15], z1.h, z2.h[7]
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.smls.za64.lane.vg4x1.nxv8i16(i32 %slice, <vscale x 8 x i16> %zn, <vscale x 8 x i16> %zm, i32 0)
  %slice.12 = add i32 %slice, 12
  call void @llvm.aarch64.sme.smls.za64.lane.vg4x1.nxv8i16(i32 %slice.12, <vscale x 8 x i16> %zn, <vscale x 8 x i16> %zm, i32 7)
  ret void
}

; Indexed x2

define void @multi_vector_mul_sub_lane_long_vg4x2_s8(i32 %slice, <vscale x 16 x i8> %dummy, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zm) {
; CHECK-LABEL: multi_vector_mul_sub_lane_long_vg4x2_s8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z5.d, z2.d
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    mov z4.d, z1.d
; CHECK-NEXT:    smlsll za.s[w8, 0:3, vgx2], { z4.b, z5.b }, z3.b[0]
; CHECK-NEXT:    smlsll za.s[w8, 4:7, vgx2], { z4.b, z5.b }, z3.b[15]
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.smls.za32.lane.vg4x2.nxv16i8(i32 %slice, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zm, i32 0)
  %slice.4 = add i32 %slice, 4
  call void @llvm.aarch64.sme.smls.za32.lane.vg4x2.nxv16i8(i32 %slice.4, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zm, i32 15)
  ret void
}

define void @multi_vector_mul_sub_lane_long_vg4x2_s16(i32 %slice, <vscale x 8 x i16> %dummy, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zm) {
; CHECK-LABEL: multi_vector_mul_sub_lane_long_vg4x2_s16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z5.d, z2.d
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    mov z4.d, z1.d
; CHECK-NEXT:    smlsll za.d[w8, 0:3, vgx2], { z4.h, z5.h }, z3.h[0]
; CHECK-NEXT:    smlsll za.d[w8, 4:7, vgx2], { z4.h, z5.h }, z3.h[7]
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.smls.za64.lane.vg4x2.nxv8i16(i32 %slice, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zm, i32 0)
  %slice.4 = add i32 %slice, 4
  call void @llvm.aarch64.sme.smls.za64.lane.vg4x2.nxv8i16(i32 %slice.4, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zm, i32 7)
  ret void
}

; Indexed x4

define void @multi_vector_mul_sub_lane_long_vg4x4_s8(i32 %slice, <vscale x 16 x i8> %dummy, <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> %zm) {
; CHECK-LABEL: multi_vector_mul_sub_lane_long_vg4x4_s8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z27.d, z4.d
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    mov z26.d, z3.d
; CHECK-NEXT:    mov z25.d, z2.d
; CHECK-NEXT:    mov z24.d, z1.d
; CHECK-NEXT:    smlsll za.s[w8, 0:3, vgx4], { z24.b - z27.b }, z5.b[0]
; CHECK-NEXT:    smlsll za.s[w8, 4:7, vgx4], { z24.b - z27.b }, z5.b[15]
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.smls.za32.lane.vg4x4.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> %zm, i32 0)
  %slice.4 = add i32 %slice, 4
  call void @llvm.aarch64.sme.smls.za32.lane.vg4x4.nxv16i8(i32 %slice.4, <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> %zm, i32 15)
  ret void
}

define void @multi_vector_mul_sub_lane_long_vg4x4_s16(i32 %slice, <vscale x 8 x i16> %dummy, <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> %zm) {
; CHECK-LABEL: multi_vector_mul_sub_lane_long_vg4x4_s16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z27.d, z4.d
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    mov z26.d, z3.d
; CHECK-NEXT:    mov z25.d, z2.d
; CHECK-NEXT:    mov z24.d, z1.d
; CHECK-NEXT:    smlsll za.d[w8, 0:3, vgx4], { z24.h - z27.h }, z5.h[0]
; CHECK-NEXT:    smlsll za.d[w8, 4:7, vgx4], { z24.h - z27.h }, z5.h[7]
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.smls.za64.lane.vg4x4.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> %zm, i32 0)
  %slice.4 = add i32 %slice, 4
  call void @llvm.aarch64.sme.smls.za64.lane.vg4x4.nxv8i16(i32 %slice.4, <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> %zm, i32 7)
  ret void
}

; UMLSLL

; Single x1

define void @multi_vector_mul_sub_single_long_vg4x1_u8(i32 %slice, <vscale x 16 x i8> %dummy, <vscale x 16 x i8> %zn, <vscale x 16 x i8> %zm) {
; CHECK-LABEL: multi_vector_mul_sub_single_long_vg4x1_u8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    umlsll za.s[w8, 0:3], z1.b, z2.b
; CHECK-NEXT:    umlsll za.s[w8, 12:15], z1.b, z2.b
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.umls.za32.single.vg4x1.nxv16i8(i32 %slice, <vscale x 16 x i8> %zn, <vscale x 16 x i8> %zm)
  %slice.12 = add i32 %slice, 12
  call void @llvm.aarch64.sme.umls.za32.single.vg4x1.nxv16i8(i32 %slice.12, <vscale x 16 x i8> %zn, <vscale x 16 x i8> %zm)
  ret void
}

define void @multi_vector_mul_sub_single_long_vg4x1_u16(i32 %slice, <vscale x 8 x i16> %dummy, <vscale x 8 x i16> %zn, <vscale x 8 x i16> %zm) {
; CHECK-LABEL: multi_vector_mul_sub_single_long_vg4x1_u16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    umlsll za.d[w8, 0:3], z1.h, z2.h
; CHECK-NEXT:    umlsll za.d[w8, 12:15], z1.h, z2.h
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.umls.za64.single.vg4x1.nxv8i16(i32 %slice, <vscale x 8 x i16> %zn, <vscale x 8 x i16> %zm)
  %slice.12 = add i32 %slice, 12
  call void @llvm.aarch64.sme.umls.za64.single.vg4x1.nxv8i16(i32 %slice.12, <vscale x 8 x i16> %zn, <vscale x 8 x i16> %zm)
  ret void
}

; Single x2

define void @multi_vector_mul_sub_single_long_vg4x2_u8(i32 %slice, <vscale x 16 x i8> %dummy, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zm) {
; CHECK-LABEL: multi_vector_mul_sub_single_long_vg4x2_u8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    // kill: def $z2 killed $z2 killed $z1_z2 def $z1_z2
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    // kill: def $z1 killed $z1 killed $z1_z2 def $z1_z2
; CHECK-NEXT:    umlsll za.s[w8, 0:3, vgx2], { z1.b, z2.b }, z3.b
; CHECK-NEXT:    umlsll za.s[w8, 4:7, vgx2], { z1.b, z2.b }, z3.b
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.umls.za32.single.vg4x2.nxv16i8(i32 %slice, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zm)
  %slice.4 = add i32 %slice, 4
  call void @llvm.aarch64.sme.umls.za32.single.vg4x2.nxv16i8(i32 %slice.4, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zm)
  ret void
}

define void @multi_vector_mul_sub_single_long_vg4x2_u16(i32 %slice, <vscale x 8 x i16> %dummy, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zm) {
; CHECK-LABEL: multi_vector_mul_sub_single_long_vg4x2_u16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    // kill: def $z2 killed $z2 killed $z1_z2 def $z1_z2
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    // kill: def $z1 killed $z1 killed $z1_z2 def $z1_z2
; CHECK-NEXT:    umlsll za.d[w8, 0:3, vgx2], { z1.h, z2.h }, z3.h
; CHECK-NEXT:    umlsll za.d[w8, 4:7, vgx2], { z1.h, z2.h }, z3.h
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.umls.za64.single.vg4x2.nxv8i16(i32 %slice, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zm)
  %slice.4 = add i32 %slice, 4
  call void @llvm.aarch64.sme.umls.za64.single.vg4x2.nxv8i16(i32 %slice.4, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zm)
  ret void
}

; Single x4

define void @multi_vector_mul_sub_single_long_vg4x4_u8(i32 %slice, <vscale x 16 x i8> %dummy, <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> %zm) {
; CHECK-LABEL: multi_vector_mul_sub_single_long_vg4x4_u8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    // kill: def $z4 killed $z4 killed $z1_z2_z3_z4 def $z1_z2_z3_z4
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    // kill: def $z3 killed $z3 killed $z1_z2_z3_z4 def $z1_z2_z3_z4
; CHECK-NEXT:    // kill: def $z2 killed $z2 killed $z1_z2_z3_z4 def $z1_z2_z3_z4
; CHECK-NEXT:    // kill: def $z1 killed $z1 killed $z1_z2_z3_z4 def $z1_z2_z3_z4
; CHECK-NEXT:    umlsll za.s[w8, 0:3, vgx4], { z1.b - z4.b }, z5.b
; CHECK-NEXT:    umlsll za.s[w8, 4:7, vgx4], { z1.b - z4.b }, z5.b
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.umls.za32.single.vg4x4.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> %zm)
  %slice.4 = add i32 %slice, 4
  call void @llvm.aarch64.sme.umls.za32.single.vg4x4.nxv16i8(i32 %slice.4, <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> %zm)
  ret void
}

define void @multi_vector_mul_sub_single_long_vg4x4_u16(i32 %slice, <vscale x 8 x i16> %dummy, <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> %zm) {
; CHECK-LABEL: multi_vector_mul_sub_single_long_vg4x4_u16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    // kill: def $z4 killed $z4 killed $z1_z2_z3_z4 def $z1_z2_z3_z4
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    // kill: def $z3 killed $z3 killed $z1_z2_z3_z4 def $z1_z2_z3_z4
; CHECK-NEXT:    // kill: def $z2 killed $z2 killed $z1_z2_z3_z4 def $z1_z2_z3_z4
; CHECK-NEXT:    // kill: def $z1 killed $z1 killed $z1_z2_z3_z4 def $z1_z2_z3_z4
; CHECK-NEXT:    umlsll za.d[w8, 0:3, vgx4], { z1.h - z4.h }, z5.h
; CHECK-NEXT:    umlsll za.d[w8, 4:7, vgx4], { z1.h - z4.h }, z5.h
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.umls.za64.single.vg4x4.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> %zm)
  %slice.4 = add i32 %slice, 4
  call void @llvm.aarch64.sme.umls.za64.single.vg4x4.nxv8i16(i32 %slice.4, <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> %zm)
  ret void
}

; Multi x2

define void @multi_vector_mul_sub_multi_long_vg4x2_u8(i32 %slice, <vscale x 16 x i8> %dummy, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zm0, <vscale x 16 x i8> %zm1) {
; CHECK-LABEL: multi_vector_mul_sub_multi_long_vg4x2_u8:
; 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:    umlsll za.s[w8, 0:3, vgx2], { z6.b, z7.b }, { z4.b, z5.b }
; CHECK-NEXT:    umlsll za.s[w8, 4:7, vgx2], { z6.b, z7.b }, { z4.b, z5.b }
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.umls.za32.vg4x2.nxv16i8(i32 %slice, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zm0, <vscale x 16 x i8> %zm1)
  %slice.4 = add i32 %slice, 4
  call void @llvm.aarch64.sme.umls.za32.vg4x2.nxv16i8(i32 %slice.4, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zm0, <vscale x 16 x i8> %zm1)
  ret void
}

define void @multi_vector_mul_sub_multi_long_vg4x2_u16(i32 %slice, <vscale x 8 x i16> %dummy, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zm0, <vscale x 8 x i16> %zm1) {
; CHECK-LABEL: multi_vector_mul_sub_multi_long_vg4x2_u16:
; 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:    umlsll za.d[w8, 0:3, vgx2], { z6.h, z7.h }, { z4.h, z5.h }
; CHECK-NEXT:    umlsll za.d[w8, 4:7, vgx2], { z6.h, z7.h }, { z4.h, z5.h }
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.umls.za64.vg4x2.nxv8i16(i32 %slice, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zm0, <vscale x 8 x i16> %zm1)
  %slice.4 = add i32 %slice, 4
  call void @llvm.aarch64.sme.umls.za64.vg4x2.nxv8i16(i32 %slice.4, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zm0, <vscale x 8 x i16> %zm1)
  ret void
}

; Multi x4

define void @multi_vector_mul_sub_multi_long_vg4x4_u8(i32 %slice, <vscale x 16 x i8> %dummy, <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> %zm0, <vscale x 16 x i8> %zm1, <vscale x 16 x i8> %zm2, <vscale x 16 x i8> %zm3) {
; CHECK-LABEL: multi_vector_mul_sub_multi_long_vg4x4_u8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z26.d, z7.d
; CHECK-NEXT:    mov z31.d, z4.d
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    ptrue p0.b
; CHECK-NEXT:    mov z25.d, z6.d
; CHECK-NEXT:    mov z30.d, z3.d
; CHECK-NEXT:    mov z24.d, z5.d
; CHECK-NEXT:    mov z29.d, z2.d
; CHECK-NEXT:    ld1b { z27.b }, p0/z, [x1]
; CHECK-NEXT:    mov z28.d, z1.d
; CHECK-NEXT:    umlsll za.s[w8, 0:3, vgx4], { z28.b - z31.b }, { z24.b - z27.b }
; CHECK-NEXT:    umlsll za.s[w8, 4:7, vgx4], { z28.b - z31.b }, { z24.b - z27.b }
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.umls.za32.vg4x4.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> %zm0, <vscale x 16 x i8> %zm1, <vscale x 16 x i8> %zm2, <vscale x 16 x i8> %zm3)
  %slice.4 = add i32 %slice, 4
  call void @llvm.aarch64.sme.umls.za32.vg4x4.nxv16i8(i32 %slice.4, <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> %zm0, <vscale x 16 x i8> %zm1, <vscale x 16 x i8> %zm2, <vscale x 16 x i8> %zm3)
  ret void
}

define void @multi_vector_mul_sub_multi_long_vg4x4_u16(i32 %slice, <vscale x 8 x i16> %dummy, <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> %zm0, <vscale x 8 x i16> %zm1, <vscale x 8 x i16> %zm2, <vscale x 8 x i16> %zm3) {
; CHECK-LABEL: multi_vector_mul_sub_multi_long_vg4x4_u16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z26.d, z7.d
; CHECK-NEXT:    mov z31.d, z4.d
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    ptrue p0.h
; CHECK-NEXT:    mov z25.d, z6.d
; CHECK-NEXT:    mov z30.d, z3.d
; CHECK-NEXT:    mov z24.d, z5.d
; CHECK-NEXT:    mov z29.d, z2.d
; CHECK-NEXT:    ld1h { z27.h }, p0/z, [x1]
; CHECK-NEXT:    mov z28.d, z1.d
; CHECK-NEXT:    umlsll za.d[w8, 0:3, vgx4], { z28.h - z31.h }, { z24.h - z27.h }
; CHECK-NEXT:    umlsll za.d[w8, 4:7, vgx4], { z28.h - z31.h }, { z24.h - z27.h }
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.umls.za64.vg4x4.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> %zm0, <vscale x 8 x i16> %zm1, <vscale x 8 x i16> %zm2, <vscale x 8 x i16> %zm3)
  %slice.4 = add i32 %slice, 4
  call void @llvm.aarch64.sme.umls.za64.vg4x4.nxv8i16(i32 %slice.4, <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> %zm0, <vscale x 8 x i16> %zm1, <vscale x 8 x i16> %zm2, <vscale x 8 x i16> %zm3)
  ret void
}

; Indexed x1

define void @multi_vector_mul_sub_lane_long_vg4x1_u8(i32 %slice, <vscale x 16 x i8> %dummy, <vscale x 16 x i8> %zn, <vscale x 16 x i8> %zm) {
; CHECK-LABEL: multi_vector_mul_sub_lane_long_vg4x1_u8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    umlsll za.s[w8, 0:3], z1.b, z2.b[0]
; CHECK-NEXT:    umlsll za.s[w8, 12:15], z1.b, z2.b[15]
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.umls.za32.lane.vg4x1.nxv16i8(i32 %slice, <vscale x 16 x i8> %zn, <vscale x 16 x i8> %zm, i32 0)
  %slice.12 = add i32 %slice, 12
  call void @llvm.aarch64.sme.umls.za32.lane.vg4x1.nxv16i8(i32 %slice.12, <vscale x 16 x i8> %zn, <vscale x 16 x i8> %zm, i32 15)
  ret void
}

define void @multi_vector_mul_sub_lane_long_vg4x1_u16(i32 %slice, <vscale x 8 x i16> %dummy, <vscale x 8 x i16> %zn, <vscale x 8 x i16> %zm) {
; CHECK-LABEL: multi_vector_mul_sub_lane_long_vg4x1_u16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    umlsll za.d[w8, 0:3], z1.h, z2.h[0]
; CHECK-NEXT:    umlsll za.d[w8, 12:15], z1.h, z2.h[7]
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.umls.za64.lane.vg4x1.nxv8i16(i32 %slice, <vscale x 8 x i16> %zn, <vscale x 8 x i16> %zm, i32 0)
  %slice.12 = add i32 %slice, 12
  call void @llvm.aarch64.sme.umls.za64.lane.vg4x1.nxv8i16(i32 %slice.12, <vscale x 8 x i16> %zn, <vscale x 8 x i16> %zm, i32 7)
  ret void
}

; Indexed x2

define void @multi_vector_mul_sub_lane_long_vg4x2_u8(i32 %slice, <vscale x 16 x i8> %dummy, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zm) {
; CHECK-LABEL: multi_vector_mul_sub_lane_long_vg4x2_u8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z5.d, z2.d
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    mov z4.d, z1.d
; CHECK-NEXT:    umlsll za.s[w8, 0:3, vgx2], { z4.b, z5.b }, z3.b[0]
; CHECK-NEXT:    umlsll za.s[w8, 4:7, vgx2], { z4.b, z5.b }, z3.b[15]
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.umls.za32.lane.vg4x2.nxv16i8(i32 %slice, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zm, i32 0)
  %slice.4 = add i32 %slice, 4
  call void @llvm.aarch64.sme.umls.za32.lane.vg4x2.nxv16i8(i32 %slice.4, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zm, i32 15)
  ret void
}

define void @multi_vector_mul_sub_lane_long_vg4x2_u16(i32 %slice, <vscale x 8 x i16> %dummy, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zm) {
; CHECK-LABEL: multi_vector_mul_sub_lane_long_vg4x2_u16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z5.d, z2.d
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    mov z4.d, z1.d
; CHECK-NEXT:    umlsll za.d[w8, 0:3, vgx2], { z4.h, z5.h }, z3.h[0]
; CHECK-NEXT:    umlsll za.d[w8, 4:7, vgx2], { z4.h, z5.h }, z3.h[7]
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.umls.za64.lane.vg4x2.nxv8i16(i32 %slice, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zm, i32 0)
  %slice.4 = add i32 %slice, 4
  call void @llvm.aarch64.sme.umls.za64.lane.vg4x2.nxv8i16(i32 %slice.4, <vscale x 8 x i16> %zn0, <vscale x 8 x i16> %zn1, <vscale x 8 x i16> %zm, i32 7)
  ret void
}

; Indexed x4

define void @multi_vector_mul_sub_lane_long_vg4x4_u8(i32 %slice, <vscale x 16 x i8> %dummy, <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> %zm) {
; CHECK-LABEL: multi_vector_mul_sub_lane_long_vg4x4_u8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z27.d, z4.d
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    mov z26.d, z3.d
; CHECK-NEXT:    mov z25.d, z2.d
; CHECK-NEXT:    mov z24.d, z1.d
; CHECK-NEXT:    umlsll za.s[w8, 0:3, vgx4], { z24.b - z27.b }, z5.b[0]
; CHECK-NEXT:    umlsll za.s[w8, 4:7, vgx4], { z24.b - z27.b }, z5.b[15]
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.umls.za32.lane.vg4x4.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> %zm, i32 0)
  %slice.4 = add i32 %slice, 4
  call void @llvm.aarch64.sme.umls.za32.lane.vg4x4.nxv16i8(i32 %slice.4, <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> %zm, i32 15)
  ret void
}

define void @multi_vector_mul_sub_lane_long_vg4x4_u16(i32 %slice, <vscale x 8 x i16> %dummy, <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> %zm) {
; CHECK-LABEL: multi_vector_mul_sub_lane_long_vg4x4_u16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z27.d, z4.d
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    mov z26.d, z3.d
; CHECK-NEXT:    mov z25.d, z2.d
; CHECK-NEXT:    mov z24.d, z1.d
; CHECK-NEXT:    umlsll za.d[w8, 0:3, vgx4], { z24.h - z27.h }, z5.h[0]
; CHECK-NEXT:    umlsll za.d[w8, 4:7, vgx4], { z24.h - z27.h }, z5.h[7]
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.umls.za64.lane.vg4x4.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> %zm, i32 0)
  %slice.4 = add i32 %slice, 4
  call void @llvm.aarch64.sme.umls.za64.lane.vg4x4.nxv8i16(i32 %slice.4, <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> %zm, i32 7)
  ret void
}

;
; SUMLALL
;

; Single x 2

define void @multi_vector_mul_add_single_signed_long_vg4x2_s8(i32 %slice, <vscale x 16 x i8> %dummy, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zm) {
; CHECK-LABEL: multi_vector_mul_add_single_signed_long_vg4x2_s8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    // kill: def $z2 killed $z2 killed $z1_z2 def $z1_z2
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    // kill: def $z1 killed $z1 killed $z1_z2 def $z1_z2
; CHECK-NEXT:    sumlall za.s[w8, 0:3, vgx2], { z1.b, z2.b }, z3.b
; CHECK-NEXT:    sumlall za.s[w8, 4:7, vgx2], { z1.b, z2.b }, z3.b
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.sumla.za32.single.vg4x2.nxv16i8(i32 %slice, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zm)
  %slice.4 = add i32 %slice, 4
  call void @llvm.aarch64.sme.sumla.za32.single.vg4x2.nxv16i8(i32 %slice.4, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zm)
  ret void
}

; Single x 4

define void @multi_vector_mul_add_single_signed_long_vg4x4_s8(i32 %slice, <vscale x 16 x i8> %dummy, <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> %zm) {
; CHECK-LABEL: multi_vector_mul_add_single_signed_long_vg4x4_s8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    // kill: def $z4 killed $z4 killed $z1_z2_z3_z4 def $z1_z2_z3_z4
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    // kill: def $z3 killed $z3 killed $z1_z2_z3_z4 def $z1_z2_z3_z4
; CHECK-NEXT:    // kill: def $z2 killed $z2 killed $z1_z2_z3_z4 def $z1_z2_z3_z4
; CHECK-NEXT:    // kill: def $z1 killed $z1 killed $z1_z2_z3_z4 def $z1_z2_z3_z4
; CHECK-NEXT:    sumlall za.s[w8, 0:3, vgx4], { z1.b - z4.b }, z5.b
; CHECK-NEXT:    sumlall za.s[w8, 4:7, vgx4], { z1.b - z4.b }, z5.b
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.sumla.za32.single.vg4x4.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> %zm)
  %slice.4 = add i32 %slice, 4
  call void @llvm.aarch64.sme.sumla.za32.single.vg4x4.nxv16i8(i32 %slice.4, <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> %zm)
  ret void
}

; Indexed x1

define void @multi_vector_mul_add_lane_signed_long_vg4x1_s8(i32 %slice, <vscale x 16 x i8> %dummy, <vscale x 16 x i8> %zn, <vscale x 16 x i8> %zm) {
; CHECK-LABEL: multi_vector_mul_add_lane_signed_long_vg4x1_s8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    sumlall za.s[w8, 0:3], z1.b, z2.b[0]
; CHECK-NEXT:    sumlall za.s[w8, 12:15], z1.b, z2.b[15]
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.sumla.za32.lane.vg4x1.nxv16i8(i32 %slice, <vscale x 16 x i8> %zn, <vscale x 16 x i8> %zm, i32 0)
  %slice.12 = add i32 %slice, 12
  call void @llvm.aarch64.sme.sumla.za32.lane.vg4x1.nxv16i8(i32 %slice.12, <vscale x 16 x i8> %zn, <vscale x 16 x i8> %zm, i32 15)
  ret void
}

; Indexed x2

define void @multi_vector_mul_add_lane_signed_long_vg4x2_s8(i32 %slice, <vscale x 16 x i8> %dummy, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zm) {
; CHECK-LABEL: multi_vector_mul_add_lane_signed_long_vg4x2_s8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z5.d, z2.d
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    mov z4.d, z1.d
; CHECK-NEXT:    sumlall za.s[w8, 0:3, vgx2], { z4.b, z5.b }, z3.b[0]
; CHECK-NEXT:    sumlall za.s[w8, 4:7, vgx2], { z4.b, z5.b }, z3.b[15]
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.sumla.za32.lane.vg4x2.nxv16i8(i32 %slice, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zm, i32 0)
  %slice.4 = add i32 %slice, 4
  call void @llvm.aarch64.sme.sumla.za32.lane.vg4x2.nxv16i8(i32 %slice.4, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zm, i32 15)
  ret void
}

; Indexed x4

define void @multi_vector_mul_add_lane_signed_long_vg4x4_s8(i32 %slice, <vscale x 16 x i8> %dummy, <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> %zm) {
; CHECK-LABEL: multi_vector_mul_add_lane_signed_long_vg4x4_s8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z27.d, z4.d
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    mov z26.d, z3.d
; CHECK-NEXT:    mov z25.d, z2.d
; CHECK-NEXT:    mov z24.d, z1.d
; CHECK-NEXT:    sumlall za.s[w8, 0:3, vgx4], { z24.b - z27.b }, z5.b[0]
; CHECK-NEXT:    sumlall za.s[w8, 4:7, vgx4], { z24.b - z27.b }, z5.b[15]
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.sumla.za32.lane.vg4x4.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> %zm, i32 0)
  %slice.4 = add i32 %slice, 4
  call void @llvm.aarch64.sme.sumla.za32.lane.vg4x4.nxv16i8(i32 %slice.4, <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> %zm, i32 15)
  ret void
}

; USMLALL

; Single x1

define void @multi_vector_mul_add_single_unsigned_long_vg4x1_s8(i32 %slice, <vscale x 16 x i8> %dummy, <vscale x 16 x i8> %zn, <vscale x 16 x i8> %zm) {
; CHECK-LABEL: multi_vector_mul_add_single_unsigned_long_vg4x1_s8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    usmlall za.s[w8, 0:3], z1.b, z2.b
; CHECK-NEXT:    usmlall za.s[w8, 12:15], z1.b, z2.b
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.usmla.za32.single.vg4x1.nxv16i8(i32 %slice, <vscale x 16 x i8> %zn, <vscale x 16 x i8> %zm)
  %slice.12 = add i32 %slice, 12
  call void @llvm.aarch64.sme.usmla.za32.single.vg4x1.nxv16i8(i32 %slice.12, <vscale x 16 x i8> %zn, <vscale x 16 x i8> %zm)
  ret void
}

; Single x 2

define void @multi_vector_mul_add_single_unsigned_long_vg4x2_s8(i32 %slice, <vscale x 16 x i8> %dummy, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zm) {
; CHECK-LABEL: multi_vector_mul_add_single_unsigned_long_vg4x2_s8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    // kill: def $z2 killed $z2 killed $z1_z2 def $z1_z2
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    // kill: def $z1 killed $z1 killed $z1_z2 def $z1_z2
; CHECK-NEXT:    usmlall za.s[w8, 0:3, vgx2], { z1.b, z2.b }, z3.b
; CHECK-NEXT:    usmlall za.s[w8, 4:7, vgx2], { z1.b, z2.b }, z3.b
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.usmla.za32.single.vg4x2.nxv16i8(i32 %slice, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zm)
  %slice.4 = add i32 %slice, 4
  call void @llvm.aarch64.sme.usmla.za32.single.vg4x2.nxv16i8(i32 %slice.4, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zm)
  ret void
}

; Single x4

define void @multi_vector_mul_add_single_unsigned_long_vg4x4_s8(i32 %slice, <vscale x 16 x i8> %dummy, <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> %zm) {
; CHECK-LABEL: multi_vector_mul_add_single_unsigned_long_vg4x4_s8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    // kill: def $z4 killed $z4 killed $z1_z2_z3_z4 def $z1_z2_z3_z4
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    // kill: def $z3 killed $z3 killed $z1_z2_z3_z4 def $z1_z2_z3_z4
; CHECK-NEXT:    // kill: def $z2 killed $z2 killed $z1_z2_z3_z4 def $z1_z2_z3_z4
; CHECK-NEXT:    // kill: def $z1 killed $z1 killed $z1_z2_z3_z4 def $z1_z2_z3_z4
; CHECK-NEXT:    usmlall za.s[w8, 0:3, vgx4], { z1.b - z4.b }, z5.b
; CHECK-NEXT:    usmlall za.s[w8, 4:7, vgx4], { z1.b - z4.b }, z5.b
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.usmla.za32.single.vg4x4.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> %zm)
  %slice.4 = add i32 %slice, 4
  call void @llvm.aarch64.sme.usmla.za32.single.vg4x4.nxv16i8(i32 %slice.4, <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> %zm)
  ret void
}

; Multi x2

define void @multi_vector_mul_add_multi_unsigned_long_vg4x2_u8(i32 %slice, <vscale x 16 x i8> %dummy, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zm0, <vscale x 16 x i8> %zm1) {
; CHECK-LABEL: multi_vector_mul_add_multi_unsigned_long_vg4x2_u8:
; 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:    usmlall za.s[w8, 0:3, vgx2], { z6.b, z7.b }, { z4.b, z5.b }
; CHECK-NEXT:    usmlall za.s[w8, 4:7, vgx2], { z6.b, z7.b }, { z4.b, z5.b }
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.usmla.za32.vg4x2.nxv16i8(i32 %slice, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zm0, <vscale x 16 x i8> %zm1)
  %slice.4 = add i32 %slice, 4
  call void @llvm.aarch64.sme.usmla.za32.vg4x2.nxv16i8(i32 %slice.4, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zm0, <vscale x 16 x i8> %zm1)
  ret void
}

; Multi x4

define void @multi_vector_mul_add_multi_unsigned_long_vg4x4_u8(i32 %slice, <vscale x 16 x i8> %dummy, <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> %zm0, <vscale x 16 x i8> %zm1, <vscale x 16 x i8> %zm2, <vscale x 16 x i8> %zm3) {
; CHECK-LABEL: multi_vector_mul_add_multi_unsigned_long_vg4x4_u8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z26.d, z7.d
; CHECK-NEXT:    mov z31.d, z4.d
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    ptrue p0.b
; CHECK-NEXT:    mov z25.d, z6.d
; CHECK-NEXT:    mov z30.d, z3.d
; CHECK-NEXT:    mov z24.d, z5.d
; CHECK-NEXT:    mov z29.d, z2.d
; CHECK-NEXT:    ld1b { z27.b }, p0/z, [x1]
; CHECK-NEXT:    mov z28.d, z1.d
; CHECK-NEXT:    usmlall za.s[w8, 0:3, vgx4], { z28.b - z31.b }, { z24.b - z27.b }
; CHECK-NEXT:    usmlall za.s[w8, 4:7, vgx4], { z28.b - z31.b }, { z24.b - z27.b }
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.usmla.za32.vg4x4.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> %zm0, <vscale x 16 x i8> %zm1, <vscale x 16 x i8> %zm2, <vscale x 16 x i8> %zm3)
  %slice.4 = add i32 %slice, 4
  call void @llvm.aarch64.sme.usmla.za32.vg4x4.nxv16i8(i32 %slice.4, <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> %zm0, <vscale x 16 x i8> %zm1, <vscale x 16 x i8> %zm2, <vscale x 16 x i8> %zm3)
  ret void
}

; Indexed x1

define void @multi_vector_mul_add_lane_unsigned_long_vg4x1_s8(i32 %slice, <vscale x 16 x i8> %dummy, <vscale x 16 x i8> %zn, <vscale x 16 x i8> %zm) {
; CHECK-LABEL: multi_vector_mul_add_lane_unsigned_long_vg4x1_s8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    usmlall za.s[w8, 0:3], z1.b, z2.b[0]
; CHECK-NEXT:    usmlall za.s[w8, 12:15], z1.b, z2.b[15]
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.usmla.za32.lane.vg4x1.nxv16i8(i32 %slice, <vscale x 16 x i8> %zn, <vscale x 16 x i8> %zm, i32 0)
  %slice.12 = add i32 %slice, 12
  call void @llvm.aarch64.sme.usmla.za32.lane.vg4x1.nxv16i8(i32 %slice.12, <vscale x 16 x i8> %zn, <vscale x 16 x i8> %zm, i32 15)
  ret void
}

; Indexed x2

define void @multi_vector_mul_add_lane_unsigned_long_vg4x2_s8(i32 %slice, <vscale x 16 x i8> %dummy, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zm) {
; CHECK-LABEL: multi_vector_mul_add_lane_unsigned_long_vg4x2_s8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z5.d, z2.d
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    mov z4.d, z1.d
; CHECK-NEXT:    usmlall za.s[w8, 0:3, vgx2], { z4.b, z5.b }, z3.b[0]
; CHECK-NEXT:    usmlall za.s[w8, 4:7, vgx2], { z4.b, z5.b }, z3.b[15]
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.usmla.za32.lane.vg4x2.nxv16i8(i32 %slice, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zm, i32 0)
  %slice.4 = add i32 %slice, 4
  call void @llvm.aarch64.sme.usmla.za32.lane.vg4x2.nxv16i8(i32 %slice.4, <vscale x 16 x i8> %zn0, <vscale x 16 x i8> %zn1, <vscale x 16 x i8> %zm, i32 15)
  ret void
}

; Indexed x4

define void @multi_vector_mul_add_lane_unsigned_long_vg4x4_s8(i32 %slice, <vscale x 16 x i8> %dummy, <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> %zm) {
; CHECK-LABEL: multi_vector_mul_add_lane_unsigned_long_vg4x4_s8:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z27.d, z4.d
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    mov z26.d, z3.d
; CHECK-NEXT:    mov z25.d, z2.d
; CHECK-NEXT:    mov z24.d, z1.d
; CHECK-NEXT:    usmlall za.s[w8, 0:3, vgx4], { z24.b - z27.b }, z5.b[0]
; CHECK-NEXT:    usmlall za.s[w8, 4:7, vgx4], { z24.b - z27.b }, z5.b[15]
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.usmla.za32.lane.vg4x4.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> %zm, i32 0)
  %slice.4 = add i32 %slice, 4
  call void @llvm.aarch64.sme.usmla.za32.lane.vg4x4.nxv16i8(i32 %slice.4, <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> %zm, i32 15)
  ret void
}

declare void @llvm.aarch64.sme.smla.za32.single.vg4x1.nxv16i8(i32, <vscale x 16 x i8>, <vscale x 16 x i8>)
declare void @llvm.aarch64.sme.smla.za32.single.vg4x2.nxv16i8(i32, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>)
declare void @llvm.aarch64.sme.smla.za32.single.vg4x4.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.smla.za64.single.vg4x1.nxv8i16(i32, <vscale x 8 x i16>, <vscale x 8 x i16>)
declare void @llvm.aarch64.sme.smla.za64.single.vg4x2.nxv8i16(i32, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>)
declare void @llvm.aarch64.sme.smla.za64.single.vg4x4.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.smla.za32.vg4x2.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.smla.za32.vg4x4.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.smla.za64.vg4x2.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.smla.za64.vg4x4.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.smla.za32.lane.vg4x1.nxv16i8(i32, <vscale x 16 x i8>, <vscale x 16 x i8>, i32)
declare void @llvm.aarch64.sme.smla.za32.lane.vg4x2.nxv16i8(i32, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, i32)
declare void @llvm.aarch64.sme.smla.za32.lane.vg4x4.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.smla.za64.lane.vg4x1.nxv8i16(i32, <vscale x 8 x i16>, <vscale x 8 x i16>, i32)
declare void @llvm.aarch64.sme.smla.za64.lane.vg4x2.nxv8i16(i32, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, i32)
declare void @llvm.aarch64.sme.smla.za64.lane.vg4x4.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.umla.za32.single.vg4x1.nxv16i8(i32, <vscale x 16 x i8>, <vscale x 16 x i8>)
declare void @llvm.aarch64.sme.umla.za32.single.vg4x2.nxv16i8(i32, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>)
declare void @llvm.aarch64.sme.umla.za32.single.vg4x4.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.umla.za64.single.vg4x1.nxv8i16(i32, <vscale x 8 x i16>, <vscale x 8 x i16>)
declare void @llvm.aarch64.sme.umla.za64.single.vg4x2.nxv8i16(i32, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>)
declare void @llvm.aarch64.sme.umla.za64.single.vg4x4.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.umla.za32.vg4x2.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.umla.za32.vg4x4.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.umla.za64.vg4x2.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.umla.za64.vg4x4.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.umla.za32.lane.vg4x1.nxv16i8(i32, <vscale x 16 x i8>, <vscale x 16 x i8>, i32)
declare void @llvm.aarch64.sme.umla.za32.lane.vg4x2.nxv16i8(i32, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, i32)
declare void @llvm.aarch64.sme.umla.za32.lane.vg4x4.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.umla.za64.lane.vg4x1.nxv8i16(i32, <vscale x 8 x i16>, <vscale x 8 x i16>, i32)
declare void @llvm.aarch64.sme.umla.za64.lane.vg4x2.nxv8i16(i32, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, i32)
declare void @llvm.aarch64.sme.umla.za64.lane.vg4x4.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.smls.za32.single.vg4x1.nxv16i8(i32, <vscale x 16 x i8>, <vscale x 16 x i8>)
declare void @llvm.aarch64.sme.smls.za32.single.vg4x2.nxv16i8(i32, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>)
declare void @llvm.aarch64.sme.smls.za32.single.vg4x4.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.smls.za64.single.vg4x1.nxv8i16(i32, <vscale x 8 x i16>, <vscale x 8 x i16>)
declare void @llvm.aarch64.sme.smls.za64.single.vg4x2.nxv8i16(i32, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>)
declare void @llvm.aarch64.sme.smls.za64.single.vg4x4.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.smls.za32.vg4x2.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.smls.za32.vg4x4.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.smls.za64.vg4x2.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.smls.za64.vg4x4.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.smls.za32.lane.vg4x1.nxv16i8(i32, <vscale x 16 x i8>, <vscale x 16 x i8>, i32)
declare void @llvm.aarch64.sme.smls.za32.lane.vg4x2.nxv16i8(i32, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, i32)
declare void @llvm.aarch64.sme.smls.za32.lane.vg4x4.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.smls.za64.lane.vg4x1.nxv8i16(i32, <vscale x 8 x i16>, <vscale x 8 x i16>, i32)
declare void @llvm.aarch64.sme.smls.za64.lane.vg4x2.nxv8i16(i32, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, i32)
declare void @llvm.aarch64.sme.smls.za64.lane.vg4x4.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.umls.za32.single.vg4x1.nxv16i8(i32, <vscale x 16 x i8>, <vscale x 16 x i8>)
declare void @llvm.aarch64.sme.umls.za32.single.vg4x2.nxv16i8(i32, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>)
declare void @llvm.aarch64.sme.umls.za32.single.vg4x4.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.umls.za64.single.vg4x1.nxv8i16(i32, <vscale x 8 x i16>, <vscale x 8 x i16>)
declare void @llvm.aarch64.sme.umls.za64.single.vg4x2.nxv8i16(i32, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>)
declare void @llvm.aarch64.sme.umls.za64.single.vg4x4.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.umls.za32.vg4x2.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.umls.za32.vg4x4.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.umls.za64.vg4x2.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.umls.za64.vg4x4.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.umls.za32.lane.vg4x1.nxv16i8(i32, <vscale x 16 x i8>, <vscale x 16 x i8>, i32)
declare void @llvm.aarch64.sme.umls.za32.lane.vg4x2.nxv16i8(i32, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, i32)
declare void @llvm.aarch64.sme.umls.za32.lane.vg4x4.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.umls.za64.lane.vg4x1.nxv8i16(i32, <vscale x 8 x i16>, <vscale x 8 x i16>, i32)
declare void @llvm.aarch64.sme.umls.za64.lane.vg4x2.nxv8i16(i32, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, i32)
declare void @llvm.aarch64.sme.umls.za64.lane.vg4x4.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.sumla.za32.single.vg4x2.nxv16i8(i32, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>)
declare void @llvm.aarch64.sme.sumla.za32.single.vg4x4.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.sumla.za32.lane.vg4x1.nxv16i8(i32, <vscale x 16 x i8>, <vscale x 16 x i8>, i32)
declare void @llvm.aarch64.sme.sumla.za32.lane.vg4x2.nxv16i8(i32, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, i32)
declare void @llvm.aarch64.sme.sumla.za32.lane.vg4x4.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.sumla.za64.lane.vg4x1.nxv8i16(i32, <vscale x 8 x i16>, <vscale x 8 x i16>, i32)
declare void @llvm.aarch64.sme.sumls.za64.lane.vg4x2.nxv8i16(i32, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, i32)
declare void @llvm.aarch64.sme.sumls.za64.lane.vg4x4.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.usmla.za32.single.vg4x1.nxv16i8(i32, <vscale x 16 x i8>, <vscale x 16 x i8>)
declare void @llvm.aarch64.sme.usmla.za32.single.vg4x2.nxv16i8(i32, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>)
declare void @llvm.aarch64.sme.usmla.za32.single.vg4x4.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.usmla.za32.vg4x2.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.usmla.za32.vg4x4.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.usmla.za32.lane.vg4x1.nxv16i8(i32, <vscale x 16 x i8>, <vscale x 16 x i8>, i32)
declare void @llvm.aarch64.sme.usmla.za32.lane.vg4x2.nxv16i8(i32, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, i32)
declare void @llvm.aarch64.sme.usmla.za32.lane.vg4x4.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.usmla.za64.lane.vg4x1.nxv8i16(i32, <vscale x 8 x i16>, <vscale x 8 x i16>, i32)
declare void @llvm.aarch64.sme.usmls.za64.lane.vg4x2.nxv8i16(i32, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, i32)
declare void @llvm.aarch64.sme.usmls.za64.lane.vg4x4.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)