; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
; RUN: llc -mtriple=aarch64-linux-gnu -mattr=+sme -mattr=+sme2 -mattr=+faminmax -force-streaming -verify-machineinstrs < %s | FileCheck %s

; FAMAX (Multi, x2)

define { <vscale x 8 x half>, <vscale x 8 x half> } @multi_vec_max_multi_x2_f16(<vscale x 8 x half> %unused, <vscale x 8 x half> %zdn1, <vscale x 8 x half> %zdn2, <vscale x 8 x half> %zm1, <vscale x 8 x half> %zm2) {
; CHECK-LABEL: multi_vec_max_multi_x2_f16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z7.d, z4.d
; CHECK-NEXT:    mov z5.d, z2.d
; CHECK-NEXT:    mov z6.d, z3.d
; CHECK-NEXT:    mov z4.d, z1.d
; CHECK-NEXT:    famax { z4.h, z5.h }, { z4.h, z5.h }, { z6.h, z7.h }
; CHECK-NEXT:    mov z0.d, z4.d
; CHECK-NEXT:    mov z1.d, z5.d
; CHECK-NEXT:    ret
  %res = call { <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sme.famax.x2.nxv8f16(<vscale x 8 x half> %zdn1, <vscale x 8 x half> %zdn2, <vscale x 8 x half> %zm1, <vscale x 8 x half> %zm2)
  ret { <vscale x 8 x half>, <vscale x 8 x half> } %res
}

define { <vscale x 4 x float>, <vscale x 4 x float> } @multi_vec_max_multi_x2_f32(<vscale x 4 x float> %unused, <vscale x 4 x float> %zdn1, <vscale x 4 x float> %zdn2, <vscale x 4 x float> %zm1, <vscale x 4 x float> %zm2) {
; CHECK-LABEL: multi_vec_max_multi_x2_f32:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z7.d, z4.d
; CHECK-NEXT:    mov z5.d, z2.d
; CHECK-NEXT:    mov z6.d, z3.d
; CHECK-NEXT:    mov z4.d, z1.d
; CHECK-NEXT:    famax { z4.s, z5.s }, { z4.s, z5.s }, { z6.s, z7.s }
; CHECK-NEXT:    mov z0.d, z4.d
; CHECK-NEXT:    mov z1.d, z5.d
; CHECK-NEXT:    ret
  %res = call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sme.famax.x2.nxv4f32(<vscale x 4 x float> %zdn1, <vscale x 4 x float> %zdn2, <vscale x 4 x float> %zm1, <vscale x 4 x float> %zm2)
  ret { <vscale x 4 x float>, <vscale x 4 x float> } %res
}

define { <vscale x 2 x double>, <vscale x 2 x double> } @multi_vec_max_multi_x2_f64(<vscale x 2 x double> %unused, <vscale x 2 x double> %zdn1, <vscale x 2 x double> %zdn2, <vscale x 2 x double> %zm1, <vscale x 2 x double> %zm2) {
; CHECK-LABEL: multi_vec_max_multi_x2_f64:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z7.d, z4.d
; CHECK-NEXT:    mov z5.d, z2.d
; CHECK-NEXT:    mov z6.d, z3.d
; CHECK-NEXT:    mov z4.d, z1.d
; CHECK-NEXT:    famax { z4.d, z5.d }, { z4.d, z5.d }, { z6.d, z7.d }
; CHECK-NEXT:    mov z0.d, z4.d
; CHECK-NEXT:    mov z1.d, z5.d
; CHECK-NEXT:    ret
  %res = call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sme.famax.x2.nxv2f64(<vscale x 2 x double> %zdn1, <vscale x 2 x double> %zdn2, <vscale x 2 x double> %zm1, <vscale x 2 x double> %zm2)
  ret { <vscale x 2 x double>, <vscale x 2 x double> } %res
}

; FAMAX (Multi, x4)

define { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> }@multi_vec_max_multi_x4_f16(<vscale x 8 x half> %unused, <vscale x 8 x half> %zdn1, <vscale x 8 x half> %zdn2, <vscale x 8 x half> %zdn3, <vscale x 8 x half> %zdn4, <vscale x 8 x half> %zm1, <vscale x 8 x half> %zm2, <vscale x 8 x half> %zm3, <vscale x 8 x half> %zm4) {
; CHECK-LABEL: multi_vec_max_multi_x4_f16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z30.d, z7.d
; CHECK-NEXT:    mov z27.d, z4.d
; CHECK-NEXT:    mov z29.d, z6.d
; CHECK-NEXT:    mov z26.d, z3.d
; CHECK-NEXT:    mov z28.d, z5.d
; CHECK-NEXT:    mov z25.d, z2.d
; CHECK-NEXT:    ldr z31, [x0]
; CHECK-NEXT:    mov z24.d, z1.d
; CHECK-NEXT:    famax { z24.h - z27.h }, { z24.h - z27.h }, { z28.h - z31.h }
; CHECK-NEXT:    mov z0.d, z24.d
; CHECK-NEXT:    mov z1.d, z25.d
; CHECK-NEXT:    mov z2.d, z26.d
; CHECK-NEXT:    mov z3.d, z27.d
; CHECK-NEXT:    ret
  %res = call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sme.famax.x4.nxv8f16(<vscale x 8 x half> %zdn1, <vscale x 8 x half> %zdn2, <vscale x 8 x half> %zdn3, <vscale x 8 x half> %zdn4, <vscale x 8 x half> %zm1,  <vscale x 8 x half> %zm2, <vscale x 8 x half> %zm3, <vscale x 8 x half> %zm4)
  ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } %res
}

define { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @multi_vec_max_multi_x4_f32(<vscale x 4 x float> %unused, <vscale x 4 x float> %zdn1, <vscale x 4 x float> %zdn2, <vscale x 4 x float> %zdn3, <vscale x 4 x float> %zdn4, <vscale x 4 x float> %zm1, <vscale x 4 x float> %zm2, <vscale x 4 x float> %zm3, <vscale x 4 x float> %zm4) {
; CHECK-LABEL: multi_vec_max_multi_x4_f32:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z30.d, z7.d
; CHECK-NEXT:    mov z27.d, z4.d
; CHECK-NEXT:    mov z29.d, z6.d
; CHECK-NEXT:    mov z26.d, z3.d
; CHECK-NEXT:    mov z28.d, z5.d
; CHECK-NEXT:    mov z25.d, z2.d
; CHECK-NEXT:    ldr z31, [x0]
; CHECK-NEXT:    mov z24.d, z1.d
; CHECK-NEXT:    famax { z24.s - z27.s }, { z24.s - z27.s }, { z28.s - z31.s }
; CHECK-NEXT:    mov z0.d, z24.d
; CHECK-NEXT:    mov z1.d, z25.d
; CHECK-NEXT:    mov z2.d, z26.d
; CHECK-NEXT:    mov z3.d, z27.d
; CHECK-NEXT:    ret
  %res = call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sme.famax.x4.nxv4f32(<vscale x 4 x float> %zdn1, <vscale x 4 x float> %zdn2, <vscale x 4 x float> %zdn3, <vscale x 4 x float> %zdn4, <vscale x 4 x float> %zm1,  <vscale x 4 x float> %zm2,  <vscale x 4 x float> %zm3,  <vscale x 4 x float> %zm4)
  ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } %res
}

define { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @multi_vec_max_multi_x4_f64(<vscale x 2 x double> %unused, <vscale x 2 x double> %zdn1, <vscale x 2 x double> %zdn2, <vscale x 2 x double> %zdn3, <vscale x 2 x double> %zdn4, <vscale x 2 x double> %zm1, <vscale x 2 x double> %zm2, <vscale x 2 x double> %zm3, <vscale x 2 x double> %zm4) {
; CHECK-LABEL: multi_vec_max_multi_x4_f64:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z30.d, z7.d
; CHECK-NEXT:    mov z27.d, z4.d
; CHECK-NEXT:    mov z29.d, z6.d
; CHECK-NEXT:    mov z26.d, z3.d
; CHECK-NEXT:    mov z28.d, z5.d
; CHECK-NEXT:    mov z25.d, z2.d
; CHECK-NEXT:    ldr z31, [x0]
; CHECK-NEXT:    mov z24.d, z1.d
; CHECK-NEXT:    famax { z24.d - z27.d }, { z24.d - z27.d }, { z28.d - z31.d }
; CHECK-NEXT:    mov z0.d, z24.d
; CHECK-NEXT:    mov z1.d, z25.d
; CHECK-NEXT:    mov z2.d, z26.d
; CHECK-NEXT:    mov z3.d, z27.d
; CHECK-NEXT:    ret
  %res = call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> }
              @llvm.aarch64.sme.famax.x4.nxv2f64(<vscale x 2 x double> %zdn1, <vscale x 2 x double> %zdn2, <vscale x 2 x double> %zdn3, <vscale x 2 x double> %zdn4,
                                                <vscale x 2 x double> %zm1,  <vscale x 2 x double> %zm2,  <vscale x 2 x double> %zm3,  <vscale x 2 x double> %zm4)
  ret { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } %res
}


; FAMIN (Multi, x2)

define { <vscale x 8 x half>, <vscale x 8 x half> } @multi_vec_min_multi_x2_f16(<vscale x 8 x half> %unused, <vscale x 8 x half> %zdn1, <vscale x 8 x half> %zdn2, <vscale x 8 x half> %zm1, <vscale x 8 x half> %zm2) {
; CHECK-LABEL: multi_vec_min_multi_x2_f16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z7.d, z4.d
; CHECK-NEXT:    mov z5.d, z2.d
; CHECK-NEXT:    mov z6.d, z3.d
; CHECK-NEXT:    mov z4.d, z1.d
; CHECK-NEXT:    famin { z4.h, z5.h }, { z4.h, z5.h }, { z6.h, z7.h }
; CHECK-NEXT:    mov z0.d, z4.d
; CHECK-NEXT:    mov z1.d, z5.d
; CHECK-NEXT:    ret
  %res = call { <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sme.famin.x2.nxv8f16(<vscale x 8 x half> %zdn1, <vscale x 8 x half> %zdn2, <vscale x 8 x half> %zm1, <vscale x 8 x half> %zm2)
  ret { <vscale x 8 x half>, <vscale x 8 x half> } %res
}

define { <vscale x 4 x float>, <vscale x 4 x float> } @multi_vec_min_multi_x2_f32(<vscale x 4 x float> %unused, <vscale x 4 x float> %zdn1, <vscale x 4 x float> %zdn2, <vscale x 4 x float> %zm1, <vscale x 4 x float> %zm2) {
; CHECK-LABEL: multi_vec_min_multi_x2_f32:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z7.d, z4.d
; CHECK-NEXT:    mov z5.d, z2.d
; CHECK-NEXT:    mov z6.d, z3.d
; CHECK-NEXT:    mov z4.d, z1.d
; CHECK-NEXT:    famin { z4.s, z5.s }, { z4.s, z5.s }, { z6.s, z7.s }
; CHECK-NEXT:    mov z0.d, z4.d
; CHECK-NEXT:    mov z1.d, z5.d
; CHECK-NEXT:    ret
  %res = call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sme.famin.x2.nxv4f32(<vscale x 4 x float> %zdn1, <vscale x 4 x float> %zdn2, <vscale x 4 x float> %zm1, <vscale x 4 x float> %zm2)
  ret { <vscale x 4 x float>, <vscale x 4 x float> } %res
}

define { <vscale x 2 x double>, <vscale x 2 x double> } @multi_vec_main_multi_x2_f64(<vscale x 2 x double> %unused, <vscale x 2 x double> %zdn1, <vscale x 2 x double> %zdn2, <vscale x 2 x double> %zm1, <vscale x 2 x double> %zm2) {
; CHECK-LABEL: multi_vec_main_multi_x2_f64:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z7.d, z4.d
; CHECK-NEXT:    mov z5.d, z2.d
; CHECK-NEXT:    mov z6.d, z3.d
; CHECK-NEXT:    mov z4.d, z1.d
; CHECK-NEXT:    famin { z4.d, z5.d }, { z4.d, z5.d }, { z6.d, z7.d }
; CHECK-NEXT:    mov z0.d, z4.d
; CHECK-NEXT:    mov z1.d, z5.d
; CHECK-NEXT:    ret
  %res = call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sme.famin.x2.nxv2f64(<vscale x 2 x double> %zdn1, <vscale x 2 x double> %zdn2, <vscale x 2 x double> %zm1, <vscale x 2 x double> %zm2)
  ret { <vscale x 2 x double>, <vscale x 2 x double> } %res
}

; FAMIN (Multi, x4)

define { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @multi_vec_min_multi_x4_f16(<vscale x 8 x half> %unused, <vscale x 8 x half> %zdn1, <vscale x 8 x half> %zdn2, <vscale x 8 x half> %zdn3, <vscale x 8 x half> %zdn4, <vscale x 8 x half> %zm1, <vscale x 8 x half> %zm2, <vscale x 8 x half> %zm3, <vscale x 8 x half> %zm4) {
; CHECK-LABEL: multi_vec_min_multi_x4_f16:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z30.d, z7.d
; CHECK-NEXT:    mov z27.d, z4.d
; CHECK-NEXT:    mov z29.d, z6.d
; CHECK-NEXT:    mov z26.d, z3.d
; CHECK-NEXT:    mov z28.d, z5.d
; CHECK-NEXT:    mov z25.d, z2.d
; CHECK-NEXT:    ldr z31, [x0]
; CHECK-NEXT:    mov z24.d, z1.d
; CHECK-NEXT:    famin { z24.h - z27.h }, { z24.h - z27.h }, { z28.h - z31.h }
; CHECK-NEXT:    mov z0.d, z24.d
; CHECK-NEXT:    mov z1.d, z25.d
; CHECK-NEXT:    mov z2.d, z26.d
; CHECK-NEXT:    mov z3.d, z27.d
; CHECK-NEXT:    ret
  %res = call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> }
              @llvm.aarch64.sme.famin.x4.nxv8f16(<vscale x 8 x half> %zdn1, <vscale x 8 x half> %zdn2, <vscale x 8 x half> %zdn3, <vscale x 8 x half> %zdn4,
                                                <vscale x 8 x half> %zm1,  <vscale x 8 x half> %zm2, <vscale x 8 x half> %zm3, <vscale x 8 x half> %zm4)
  ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } %res
}

define { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @multi_vec_min_multi_x4_f32(<vscale x 4 x float> %unused, <vscale x 4 x float> %zdn1, <vscale x 4 x float> %zdn2, <vscale x 4 x float> %zdn3, <vscale x 4 x float> %zdn4, <vscale x 4 x float> %zm1, <vscale x 4 x float> %zm2, <vscale x 4 x float> %zm3, <vscale x 4 x float> %zm4) {
; CHECK-LABEL: multi_vec_min_multi_x4_f32:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z30.d, z7.d
; CHECK-NEXT:    mov z27.d, z4.d
; CHECK-NEXT:    mov z29.d, z6.d
; CHECK-NEXT:    mov z26.d, z3.d
; CHECK-NEXT:    mov z28.d, z5.d
; CHECK-NEXT:    mov z25.d, z2.d
; CHECK-NEXT:    ldr z31, [x0]
; CHECK-NEXT:    mov z24.d, z1.d
; CHECK-NEXT:    famin { z24.s - z27.s }, { z24.s - z27.s }, { z28.s - z31.s }
; CHECK-NEXT:    mov z0.d, z24.d
; CHECK-NEXT:    mov z1.d, z25.d
; CHECK-NEXT:    mov z2.d, z26.d
; CHECK-NEXT:    mov z3.d, z27.d
; CHECK-NEXT:    ret
  %res = call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }
              @llvm.aarch64.sme.famin.x4.nxv4f32(<vscale x 4 x float> %zdn1, <vscale x 4 x float> %zdn2, <vscale x 4 x float> %zdn3, <vscale x 4 x float> %zdn4,
                                                <vscale x 4 x float> %zm1,  <vscale x 4 x float> %zm2,  <vscale x 4 x float> %zm3,  <vscale x 4 x float> %zm4)
  ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } %res
}

define { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @multi_vec_min_multi_x4_f64(<vscale x 2 x double> %unused, <vscale x 2 x double> %zdn1, <vscale x 2 x double> %zdn2, <vscale x 2 x double> %zdn3, <vscale x 2 x double> %zdn4, <vscale x 2 x double> %zm1, <vscale x 2 x double> %zm2, <vscale x 2 x double> %zm3, <vscale x 2 x double> %zm4) {
; CHECK-LABEL: multi_vec_min_multi_x4_f64:
; CHECK:       // %bb.0:
; CHECK-NEXT:    mov z30.d, z7.d
; CHECK-NEXT:    mov z27.d, z4.d
; CHECK-NEXT:    mov z29.d, z6.d
; CHECK-NEXT:    mov z26.d, z3.d
; CHECK-NEXT:    mov z28.d, z5.d
; CHECK-NEXT:    mov z25.d, z2.d
; CHECK-NEXT:    ldr z31, [x0]
; CHECK-NEXT:    mov z24.d, z1.d
; CHECK-NEXT:    famin { z24.d - z27.d }, { z24.d - z27.d }, { z28.d - z31.d }
; CHECK-NEXT:    mov z0.d, z24.d
; CHECK-NEXT:    mov z1.d, z25.d
; CHECK-NEXT:    mov z2.d, z26.d
; CHECK-NEXT:    mov z3.d, z27.d
; CHECK-NEXT:    ret
  %res = call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> }
              @llvm.aarch64.sme.famin.x4.nxv2f64(<vscale x 2 x double> %zdn1, <vscale x 2 x double> %zdn2, <vscale x 2 x double> %zdn3, <vscale x 2 x double> %zdn4,
                                                <vscale x 2 x double> %zm1,  <vscale x 2 x double> %zm2,  <vscale x 2 x double> %zm3,  <vscale x 2 x double> %zm4)
  ret { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } %res
}