; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4
; RUN: llc -force-streaming -verify-machineinstrs < %s | FileCheck %s

target triple = "aarch64-linux"

define void @add_f16_vg1x2(i32 %slice, <vscale x 8 x half> %zn0, <vscale x 8 x half> %zn1) #0 {
; CHECK-LABEL: add_f16_vg1x2:
; CHECK:       // %bb.0:
; CHECK-NEXT:    // kill: def $z1 killed $z1 killed $z0_z1 def $z0_z1
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    // kill: def $z0 killed $z0 killed $z0_z1 def $z0_z1
; CHECK-NEXT:    fadd za.h[w8, 0, vgx2], { z0.h, z1.h }
; CHECK-NEXT:    fadd za.h[w8, 7, vgx2], { z0.h, z1.h }
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.add.za16.vg1x2.nxv8f16(i32 %slice, <vscale x 8 x half> %zn0, <vscale x 8 x half> %zn1)
  %slice.7 = add i32 %slice, 7
  call void @llvm.aarch64.sme.add.za16.vg1x2.nxv8f16(i32 %slice.7, <vscale x 8 x half> %zn0, <vscale x 8 x half> %zn1)
  ret void
}

define void @add_f16_vg1x4(i32 %slice, <vscale x 8 x half> %zn0, <vscale x 8 x half> %zn1,
; CHECK-LABEL: add_f16_vg1x4:
; CHECK:       // %bb.0:
; CHECK-NEXT:    // kill: def $z3 killed $z3 killed $z0_z1_z2_z3 def $z0_z1_z2_z3
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    // kill: def $z2 killed $z2 killed $z0_z1_z2_z3 def $z0_z1_z2_z3
; CHECK-NEXT:    // kill: def $z1 killed $z1 killed $z0_z1_z2_z3 def $z0_z1_z2_z3
; CHECK-NEXT:    // kill: def $z0 killed $z0 killed $z0_z1_z2_z3 def $z0_z1_z2_z3
; CHECK-NEXT:    fadd za.h[w8, 0, vgx4], { z0.h - z3.h }
; CHECK-NEXT:    fadd za.h[w8, 7, vgx4], { z0.h - z3.h }
; CHECK-NEXT:    ret
                            <vscale x 8 x half> %zn2, <vscale x 8 x half> %zn3) #1 {
  call void @llvm.aarch64.sme.add.za16.vg1x4.nxv8f16(i32 %slice, <vscale x 8 x half> %zn0, <vscale x 8 x half> %zn1,
                                                     <vscale x 8 x half> %zn2, <vscale x 8 x half> %zn3);
  %slice.7 = add i32 %slice, 7
  call void @llvm.aarch64.sme.add.za16.vg1x4.nxv8f16(i32 %slice.7, <vscale x 8 x half> %zn0, <vscale x 8 x half> %zn1,
                                                     <vscale x 8 x half> %zn2, <vscale x 8 x half> %zn3);
  ret void
}

define void @sub_f16_vg1x2(i32 %slice, <vscale x 8 x half> %zn0, <vscale x 8 x half> %zn1) #1 {
; CHECK-LABEL: sub_f16_vg1x2:
; CHECK:       // %bb.0:
; CHECK-NEXT:    // kill: def $z1 killed $z1 killed $z0_z1 def $z0_z1
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    // kill: def $z0 killed $z0 killed $z0_z1 def $z0_z1
; CHECK-NEXT:    fsub za.h[w8, 0, vgx2], { z0.h, z1.h }
; CHECK-NEXT:    fsub za.h[w8, 7, vgx2], { z0.h, z1.h }
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.sub.za16.vg1x2.nxv8f16(i32 %slice, <vscale x 8 x half> %zn0, <vscale x 8 x half> %zn1)
  %slice.7 = add i32 %slice, 7
  call void @llvm.aarch64.sme.sub.za16.vg1x2.nxv8f16(i32 %slice.7, <vscale x 8 x half> %zn0, <vscale x 8 x half> %zn1)
  ret void
}

define void @sub_f16_vg1x4(i32 %slice, <vscale x 8 x half> %zn0, <vscale x 8 x half> %zn1,
; CHECK-LABEL: sub_f16_vg1x4:
; CHECK:       // %bb.0:
; CHECK-NEXT:    // kill: def $z3 killed $z3 killed $z0_z1_z2_z3 def $z0_z1_z2_z3
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    // kill: def $z2 killed $z2 killed $z0_z1_z2_z3 def $z0_z1_z2_z3
; CHECK-NEXT:    // kill: def $z1 killed $z1 killed $z0_z1_z2_z3 def $z0_z1_z2_z3
; CHECK-NEXT:    // kill: def $z0 killed $z0 killed $z0_z1_z2_z3 def $z0_z1_z2_z3
; CHECK-NEXT:    fsub za.h[w8, 0, vgx4], { z0.h - z3.h }
; CHECK-NEXT:    fsub za.h[w8, 7, vgx4], { z0.h - z3.h }
; CHECK-NEXT:    ret
                            <vscale x 8 x half> %zn2, <vscale x 8 x half> %zn3) #0 {
  call void @llvm.aarch64.sme.sub.za16.vg1x4.nxv8f16(i32 %slice, <vscale x 8 x half> %zn0, <vscale x 8 x half> %zn1,
                                                     <vscale x 8 x half> %zn2, <vscale x 8 x half> %zn3);
  %slice.7 = add i32 %slice, 7
  call void @llvm.aarch64.sme.sub.za16.vg1x4.nxv8f16(i32 %slice.7, <vscale x 8 x half> %zn0, <vscale x 8 x half> %zn1,
                                                     <vscale x 8 x half> %zn2, <vscale x 8 x half> %zn3);
  ret void
}

define void @add_bf16_vg1x2(i32 %slice, <vscale x 8 x bfloat> %zn0, <vscale x 8 x bfloat> %zn1) #2 {
; CHECK-LABEL: add_bf16_vg1x2:
; CHECK:       // %bb.0:
; CHECK-NEXT:    // kill: def $z1 killed $z1 killed $z0_z1 def $z0_z1
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    // kill: def $z0 killed $z0 killed $z0_z1 def $z0_z1
; CHECK-NEXT:    bfadd za.h[w8, 0, vgx2], { z0.h, z1.h }
; CHECK-NEXT:    bfadd za.h[w8, 7, vgx2], { z0.h, z1.h }
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.add.za16.vg1x2.nxv8bf16(i32 %slice, <vscale x 8 x bfloat> %zn0, <vscale x 8 x bfloat> %zn1)
  %slice.7 = add i32 %slice, 7
  call void @llvm.aarch64.sme.add.za16.vg1x2.nxv8bf16(i32 %slice.7, <vscale x 8 x bfloat> %zn0, <vscale x 8 x bfloat> %zn1)
  ret void
}

define void @add_bf16_vg1x4(i32 %slice, <vscale x 8 x bfloat> %zn0, <vscale x 8 x bfloat> %zn1,
; CHECK-LABEL: add_bf16_vg1x4:
; CHECK:       // %bb.0:
; CHECK-NEXT:    // kill: def $z3 killed $z3 killed $z0_z1_z2_z3 def $z0_z1_z2_z3
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    // kill: def $z2 killed $z2 killed $z0_z1_z2_z3 def $z0_z1_z2_z3
; CHECK-NEXT:    // kill: def $z1 killed $z1 killed $z0_z1_z2_z3 def $z0_z1_z2_z3
; CHECK-NEXT:    // kill: def $z0 killed $z0 killed $z0_z1_z2_z3 def $z0_z1_z2_z3
; CHECK-NEXT:    bfadd za.h[w8, 0, vgx4], { z0.h - z3.h }
; CHECK-NEXT:    bfadd za.h[w8, 7, vgx4], { z0.h - z3.h }
; CHECK-NEXT:    ret
                            <vscale x 8 x bfloat> %zn2, <vscale x 8 x bfloat> %zn3) #2 {
  call void @llvm.aarch64.sme.add.za16.vg1x4.nxv8bf16(i32 %slice, <vscale x 8 x bfloat> %zn0, <vscale x 8 x bfloat> %zn1,
                                                     <vscale x 8 x bfloat> %zn2, <vscale x 8 x bfloat> %zn3);
  %slice.7 = add i32 %slice, 7
  call void @llvm.aarch64.sme.add.za16.vg1x4.nxv8bf16(i32 %slice.7, <vscale x 8 x bfloat> %zn0, <vscale x 8 x bfloat> %zn1,
                                                     <vscale x 8 x bfloat> %zn2, <vscale x 8 x bfloat> %zn3);
  ret void
}

define void @sub_bf16_vg1x2(i32 %slice, <vscale x 8 x bfloat> %zn0, <vscale x 8 x bfloat> %zn1) #2 {
; CHECK-LABEL: sub_bf16_vg1x2:
; CHECK:       // %bb.0:
; CHECK-NEXT:    // kill: def $z1 killed $z1 killed $z0_z1 def $z0_z1
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    // kill: def $z0 killed $z0 killed $z0_z1 def $z0_z1
; CHECK-NEXT:    bfsub za.h[w8, 0, vgx2], { z0.h, z1.h }
; CHECK-NEXT:    bfsub za.h[w8, 7, vgx2], { z0.h, z1.h }
; CHECK-NEXT:    ret
  call void @llvm.aarch64.sme.sub.za16.vg1x2.nxv8bf16(i32 %slice, <vscale x 8 x bfloat> %zn0, <vscale x 8 x bfloat> %zn1)
  %slice.7 = add i32 %slice, 7
  call void @llvm.aarch64.sme.sub.za16.vg1x2.nxv8bf16(i32 %slice.7, <vscale x 8 x bfloat> %zn0, <vscale x 8 x bfloat> %zn1)
  ret void
}

define void @sub_bf16_vg1x4(i32 %slice, <vscale x 8 x bfloat> %zn0, <vscale x 8 x bfloat> %zn1,
; CHECK-LABEL: sub_bf16_vg1x4:
; CHECK:       // %bb.0:
; CHECK-NEXT:    // kill: def $z3 killed $z3 killed $z0_z1_z2_z3 def $z0_z1_z2_z3
; CHECK-NEXT:    mov w8, w0
; CHECK-NEXT:    // kill: def $z2 killed $z2 killed $z0_z1_z2_z3 def $z0_z1_z2_z3
; CHECK-NEXT:    // kill: def $z1 killed $z1 killed $z0_z1_z2_z3 def $z0_z1_z2_z3
; CHECK-NEXT:    // kill: def $z0 killed $z0 killed $z0_z1_z2_z3 def $z0_z1_z2_z3
; CHECK-NEXT:    bfsub za.h[w8, 0, vgx4], { z0.h - z3.h }
; CHECK-NEXT:    bfsub za.h[w8, 7, vgx4], { z0.h - z3.h }
; CHECK-NEXT:    ret
                            <vscale x 8 x bfloat> %zn2, <vscale x 8 x bfloat> %zn3) #2 {
  call void @llvm.aarch64.sme.sub.za16.vg1x4.nxv8bf16(i32 %slice, <vscale x 8 x bfloat> %zn0, <vscale x 8 x bfloat> %zn1,
                                                     <vscale x 8 x bfloat> %zn2, <vscale x 8 x bfloat> %zn3);
  %slice.7 = add i32 %slice, 7
  call void @llvm.aarch64.sme.sub.za16.vg1x4.nxv8bf16(i32 %slice.7, <vscale x 8 x bfloat> %zn0, <vscale x 8 x bfloat> %zn1,
                                                     <vscale x 8 x bfloat> %zn2, <vscale x 8 x bfloat> %zn3);
  ret void
}

attributes #0 = { nounwind "target-features"="+sme-f16f16" }
attributes #1 = { nounwind "target-features"="+sme-f8f16" }
attributes #2 = { nounwind "target-features"="+sme-b16b16,+bf16" }