llvm/test/CodeGen/AArch64/fmla.ll

; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 2
; RUN: llc -mtriple=aarch64 -verify-machineinstrs %s -o - | FileCheck %s --check-prefixes=CHECK,CHECK-SD,CHECK-SD-NOFP16
; RUN: llc -mtriple=aarch64 -mattr=+fullfp16 -verify-machineinstrs %s -o - | FileCheck %s --check-prefixes=CHECK,CHECK-SD,CHECK-SD-FP16
; RUN: llc -mtriple=aarch64 -global-isel -verify-machineinstrs %s -o - | FileCheck %s --check-prefixes=CHECK,CHECK-GI,CHECK-GI-NOFP16
; RUN: llc -mtriple=aarch64 -mattr=+fullfp16 -global-isel -verify-machineinstrs %s -o - | FileCheck %s --check-prefixes=CHECK,CHECK-GI,CHECK-GI-FP16

define double @fma_f64(double %a, double %b, double %c) {
; CHECK-LABEL: fma_f64:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    fmadd d0, d0, d1, d2
; CHECK-NEXT:    ret
entry:
  %d = call double @llvm.fma.f64(double %a, double %b, double %c)
  ret double %d
}

define float @fma_f32(float %a, float %b, float %c) {
; CHECK-LABEL: fma_f32:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    fmadd s0, s0, s1, s2
; CHECK-NEXT:    ret
entry:
  %d = call float @llvm.fma.f32(float %a, float %b, float %c)
  ret float %d
}

define half @fma_f16(half %a, half %b, half %c) {
; CHECK-SD-NOFP16-LABEL: fma_f16:
; CHECK-SD-NOFP16:       // %bb.0: // %entry
; CHECK-SD-NOFP16-NEXT:    fcvt s2, h2
; CHECK-SD-NOFP16-NEXT:    fcvt s1, h1
; CHECK-SD-NOFP16-NEXT:    fcvt s0, h0
; CHECK-SD-NOFP16-NEXT:    fmadd s0, s0, s1, s2
; CHECK-SD-NOFP16-NEXT:    fcvt h0, s0
; CHECK-SD-NOFP16-NEXT:    ret
;
; CHECK-SD-FP16-LABEL: fma_f16:
; CHECK-SD-FP16:       // %bb.0: // %entry
; CHECK-SD-FP16-NEXT:    fmadd h0, h0, h1, h2
; CHECK-SD-FP16-NEXT:    ret
;
; CHECK-GI-NOFP16-LABEL: fma_f16:
; CHECK-GI-NOFP16:       // %bb.0: // %entry
; CHECK-GI-NOFP16-NEXT:    fcvt s0, h0
; CHECK-GI-NOFP16-NEXT:    fcvt s1, h1
; CHECK-GI-NOFP16-NEXT:    fcvt s2, h2
; CHECK-GI-NOFP16-NEXT:    fmadd s0, s0, s1, s2
; CHECK-GI-NOFP16-NEXT:    fcvt h0, s0
; CHECK-GI-NOFP16-NEXT:    ret
;
; CHECK-GI-FP16-LABEL: fma_f16:
; CHECK-GI-FP16:       // %bb.0: // %entry
; CHECK-GI-FP16-NEXT:    fmadd h0, h0, h1, h2
; CHECK-GI-FP16-NEXT:    ret
entry:
  %d = call half @llvm.fma.f16(half %a, half %b, half %c)
  ret half %d
}

define <2 x double> @fma_v2f64(<2 x double> %a, <2 x double> %b, <2 x double> %c) {
; CHECK-LABEL: fma_v2f64:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    fmla v2.2d, v1.2d, v0.2d
; CHECK-NEXT:    mov v0.16b, v2.16b
; CHECK-NEXT:    ret
entry:
  %d = call <2 x double> @llvm.fma.v2f64(<2 x double> %a, <2 x double> %b, <2 x double> %c)
  ret <2 x double> %d
}

define <3 x double> @fma_v3f64(<3 x double> %a, <3 x double> %b, <3 x double> %c) {
; CHECK-SD-LABEL: fma_v3f64:
; CHECK-SD:       // %bb.0: // %entry
; CHECK-SD-NEXT:    // kill: def $d6 killed $d6 def $q6
; CHECK-SD-NEXT:    // kill: def $d3 killed $d3 def $q3
; CHECK-SD-NEXT:    // kill: def $d0 killed $d0 def $q0
; CHECK-SD-NEXT:    // kill: def $d7 killed $d7 def $q7
; CHECK-SD-NEXT:    // kill: def $d4 killed $d4 def $q4
; CHECK-SD-NEXT:    // kill: def $d1 killed $d1 def $q1
; CHECK-SD-NEXT:    // kill: def $d5 killed $d5 def $q5
; CHECK-SD-NEXT:    // kill: def $d2 killed $d2 def $q2
; CHECK-SD-NEXT:    mov v3.d[1], v4.d[0]
; CHECK-SD-NEXT:    mov v0.d[1], v1.d[0]
; CHECK-SD-NEXT:    mov v6.d[1], v7.d[0]
; CHECK-SD-NEXT:    fmla v6.2d, v3.2d, v0.2d
; CHECK-SD-NEXT:    ldr d3, [sp]
; CHECK-SD-NEXT:    fmla v3.2d, v5.2d, v2.2d
; CHECK-SD-NEXT:    fmov d0, d6
; CHECK-SD-NEXT:    ext v1.16b, v6.16b, v6.16b, #8
; CHECK-SD-NEXT:    // kill: def $d1 killed $d1 killed $q1
; CHECK-SD-NEXT:    fmov d2, d3
; CHECK-SD-NEXT:    ret
;
; CHECK-GI-LABEL: fma_v3f64:
; CHECK-GI:       // %bb.0: // %entry
; CHECK-GI-NEXT:    // kill: def $d0 killed $d0 def $q0
; CHECK-GI-NEXT:    // kill: def $d3 killed $d3 def $q3
; CHECK-GI-NEXT:    // kill: def $d6 killed $d6 def $q6
; CHECK-GI-NEXT:    // kill: def $d1 killed $d1 def $q1
; CHECK-GI-NEXT:    // kill: def $d4 killed $d4 def $q4
; CHECK-GI-NEXT:    // kill: def $d7 killed $d7 def $q7
; CHECK-GI-NEXT:    mov v0.d[1], v1.d[0]
; CHECK-GI-NEXT:    mov v3.d[1], v4.d[0]
; CHECK-GI-NEXT:    mov v6.d[1], v7.d[0]
; CHECK-GI-NEXT:    fmla v6.2d, v3.2d, v0.2d
; CHECK-GI-NEXT:    ldr d0, [sp]
; CHECK-GI-NEXT:    fmadd d2, d2, d5, d0
; CHECK-GI-NEXT:    mov d1, v6.d[1]
; CHECK-GI-NEXT:    fmov d0, d6
; CHECK-GI-NEXT:    ret
entry:
  %d = call <3 x double> @llvm.fma.v3f64(<3 x double> %a, <3 x double> %b, <3 x double> %c)
  ret <3 x double> %d
}

define <4 x double> @fma_v4f64(<4 x double> %a, <4 x double> %b, <4 x double> %c) {
; CHECK-LABEL: fma_v4f64:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    fmla v4.2d, v2.2d, v0.2d
; CHECK-NEXT:    fmla v5.2d, v3.2d, v1.2d
; CHECK-NEXT:    mov v0.16b, v4.16b
; CHECK-NEXT:    mov v1.16b, v5.16b
; CHECK-NEXT:    ret
entry:
  %d = call <4 x double> @llvm.fma.v4f64(<4 x double> %a, <4 x double> %b, <4 x double> %c)
  ret <4 x double> %d
}

define <2 x float> @fma_v2f32(<2 x float> %a, <2 x float> %b, <2 x float> %c) {
; CHECK-LABEL: fma_v2f32:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    fmla v2.2s, v1.2s, v0.2s
; CHECK-NEXT:    fmov d0, d2
; CHECK-NEXT:    ret
entry:
  %d = call <2 x float> @llvm.fma.v2f32(<2 x float> %a, <2 x float> %b, <2 x float> %c)
  ret <2 x float> %d
}

define <3 x float> @fma_v3f32(<3 x float> %a, <3 x float> %b, <3 x float> %c) {
; CHECK-LABEL: fma_v3f32:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    fmla v2.4s, v1.4s, v0.4s
; CHECK-NEXT:    mov v0.16b, v2.16b
; CHECK-NEXT:    ret
entry:
  %d = call <3 x float> @llvm.fma.v3f32(<3 x float> %a, <3 x float> %b, <3 x float> %c)
  ret <3 x float> %d
}

define <4 x float> @fma_v4f32(<4 x float> %a, <4 x float> %b, <4 x float> %c) {
; CHECK-LABEL: fma_v4f32:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    fmla v2.4s, v1.4s, v0.4s
; CHECK-NEXT:    mov v0.16b, v2.16b
; CHECK-NEXT:    ret
entry:
  %d = call <4 x float> @llvm.fma.v4f32(<4 x float> %a, <4 x float> %b, <4 x float> %c)
  ret <4 x float> %d
}

define <8 x float> @fma_v8f32(<8 x float> %a, <8 x float> %b, <8 x float> %c) {
; CHECK-LABEL: fma_v8f32:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    fmla v4.4s, v2.4s, v0.4s
; CHECK-NEXT:    fmla v5.4s, v3.4s, v1.4s
; CHECK-NEXT:    mov v0.16b, v4.16b
; CHECK-NEXT:    mov v1.16b, v5.16b
; CHECK-NEXT:    ret
entry:
  %d = call <8 x float> @llvm.fma.v8f32(<8 x float> %a, <8 x float> %b, <8 x float> %c)
  ret <8 x float> %d
}

define <7 x half> @fma_v7f16(<7 x half> %a, <7 x half> %b, <7 x half> %c) {
; CHECK-SD-NOFP16-LABEL: fma_v7f16:
; CHECK-SD-NOFP16:       // %bb.0: // %entry
; CHECK-SD-NOFP16-NEXT:    mov h3, v2.h[1]
; CHECK-SD-NOFP16-NEXT:    mov h4, v1.h[1]
; CHECK-SD-NOFP16-NEXT:    mov h5, v0.h[1]
; CHECK-SD-NOFP16-NEXT:    fcvt s6, h2
; CHECK-SD-NOFP16-NEXT:    fcvt s7, h1
; CHECK-SD-NOFP16-NEXT:    fcvt s16, h0
; CHECK-SD-NOFP16-NEXT:    mov h17, v2.h[2]
; CHECK-SD-NOFP16-NEXT:    mov h18, v1.h[2]
; CHECK-SD-NOFP16-NEXT:    mov h19, v0.h[2]
; CHECK-SD-NOFP16-NEXT:    fcvt s3, h3
; CHECK-SD-NOFP16-NEXT:    fcvt s4, h4
; CHECK-SD-NOFP16-NEXT:    fcvt s5, h5
; CHECK-SD-NOFP16-NEXT:    fmadd s6, s16, s7, s6
; CHECK-SD-NOFP16-NEXT:    fcvt s7, h17
; CHECK-SD-NOFP16-NEXT:    fcvt s16, h18
; CHECK-SD-NOFP16-NEXT:    fcvt s17, h19
; CHECK-SD-NOFP16-NEXT:    mov h18, v1.h[3]
; CHECK-SD-NOFP16-NEXT:    mov h19, v0.h[3]
; CHECK-SD-NOFP16-NEXT:    fmadd s4, s5, s4, s3
; CHECK-SD-NOFP16-NEXT:    mov h5, v2.h[3]
; CHECK-SD-NOFP16-NEXT:    fcvt h3, s6
; CHECK-SD-NOFP16-NEXT:    fmadd s6, s17, s16, s7
; CHECK-SD-NOFP16-NEXT:    mov h17, v2.h[4]
; CHECK-SD-NOFP16-NEXT:    fcvt s7, h18
; CHECK-SD-NOFP16-NEXT:    fcvt s16, h19
; CHECK-SD-NOFP16-NEXT:    mov h18, v1.h[4]
; CHECK-SD-NOFP16-NEXT:    fcvt h4, s4
; CHECK-SD-NOFP16-NEXT:    fcvt s5, h5
; CHECK-SD-NOFP16-NEXT:    mov h19, v0.h[4]
; CHECK-SD-NOFP16-NEXT:    fcvt h6, s6
; CHECK-SD-NOFP16-NEXT:    fcvt s17, h17
; CHECK-SD-NOFP16-NEXT:    fcvt s18, h18
; CHECK-SD-NOFP16-NEXT:    mov v3.h[1], v4.h[0]
; CHECK-SD-NOFP16-NEXT:    mov h4, v2.h[5]
; CHECK-SD-NOFP16-NEXT:    fmadd s5, s16, s7, s5
; CHECK-SD-NOFP16-NEXT:    mov h7, v1.h[5]
; CHECK-SD-NOFP16-NEXT:    mov h16, v0.h[5]
; CHECK-SD-NOFP16-NEXT:    fcvt s19, h19
; CHECK-SD-NOFP16-NEXT:    mov v3.h[2], v6.h[0]
; CHECK-SD-NOFP16-NEXT:    mov h6, v2.h[6]
; CHECK-SD-NOFP16-NEXT:    fcvt s4, h4
; CHECK-SD-NOFP16-NEXT:    fcvt s7, h7
; CHECK-SD-NOFP16-NEXT:    fcvt s16, h16
; CHECK-SD-NOFP16-NEXT:    fcvt h5, s5
; CHECK-SD-NOFP16-NEXT:    fmadd s17, s19, s18, s17
; CHECK-SD-NOFP16-NEXT:    mov h18, v1.h[6]
; CHECK-SD-NOFP16-NEXT:    mov h19, v0.h[6]
; CHECK-SD-NOFP16-NEXT:    mov h2, v2.h[7]
; CHECK-SD-NOFP16-NEXT:    mov h1, v1.h[7]
; CHECK-SD-NOFP16-NEXT:    mov h0, v0.h[7]
; CHECK-SD-NOFP16-NEXT:    fmadd s4, s16, s7, s4
; CHECK-SD-NOFP16-NEXT:    mov v3.h[3], v5.h[0]
; CHECK-SD-NOFP16-NEXT:    fcvt s5, h6
; CHECK-SD-NOFP16-NEXT:    fcvt s6, h18
; CHECK-SD-NOFP16-NEXT:    fcvt s7, h19
; CHECK-SD-NOFP16-NEXT:    fcvt h16, s17
; CHECK-SD-NOFP16-NEXT:    fcvt s2, h2
; CHECK-SD-NOFP16-NEXT:    fcvt s1, h1
; CHECK-SD-NOFP16-NEXT:    fcvt s0, h0
; CHECK-SD-NOFP16-NEXT:    fcvt h4, s4
; CHECK-SD-NOFP16-NEXT:    fmadd s5, s7, s6, s5
; CHECK-SD-NOFP16-NEXT:    mov v3.h[4], v16.h[0]
; CHECK-SD-NOFP16-NEXT:    fmadd s0, s0, s1, s2
; CHECK-SD-NOFP16-NEXT:    mov v3.h[5], v4.h[0]
; CHECK-SD-NOFP16-NEXT:    fcvt h4, s5
; CHECK-SD-NOFP16-NEXT:    fcvt h0, s0
; CHECK-SD-NOFP16-NEXT:    mov v3.h[6], v4.h[0]
; CHECK-SD-NOFP16-NEXT:    mov v3.h[7], v0.h[0]
; CHECK-SD-NOFP16-NEXT:    mov v0.16b, v3.16b
; CHECK-SD-NOFP16-NEXT:    ret
;
; CHECK-SD-FP16-LABEL: fma_v7f16:
; CHECK-SD-FP16:       // %bb.0: // %entry
; CHECK-SD-FP16-NEXT:    fmla v2.8h, v1.8h, v0.8h
; CHECK-SD-FP16-NEXT:    mov v0.16b, v2.16b
; CHECK-SD-FP16-NEXT:    ret
;
; CHECK-GI-NOFP16-LABEL: fma_v7f16:
; CHECK-GI-NOFP16:       // %bb.0: // %entry
; CHECK-GI-NOFP16-NEXT:    fcvtl v3.4s, v0.4h
; CHECK-GI-NOFP16-NEXT:    fcvtl v4.4s, v1.4h
; CHECK-GI-NOFP16-NEXT:    fcvtl v5.4s, v2.4h
; CHECK-GI-NOFP16-NEXT:    mov v6.h[0], v0.h[4]
; CHECK-GI-NOFP16-NEXT:    fmla v5.4s, v4.4s, v3.4s
; CHECK-GI-NOFP16-NEXT:    mov v3.h[0], v1.h[4]
; CHECK-GI-NOFP16-NEXT:    mov v4.h[0], v2.h[4]
; CHECK-GI-NOFP16-NEXT:    mov v6.h[1], v0.h[5]
; CHECK-GI-NOFP16-NEXT:    mov v3.h[1], v1.h[5]
; CHECK-GI-NOFP16-NEXT:    mov v4.h[1], v2.h[5]
; CHECK-GI-NOFP16-NEXT:    fcvtn v5.4h, v5.4s
; CHECK-GI-NOFP16-NEXT:    mov v6.h[2], v0.h[6]
; CHECK-GI-NOFP16-NEXT:    mov v3.h[2], v1.h[6]
; CHECK-GI-NOFP16-NEXT:    mov v4.h[2], v2.h[6]
; CHECK-GI-NOFP16-NEXT:    mov v0.h[0], v5.h[0]
; CHECK-GI-NOFP16-NEXT:    fcvtl v1.4s, v6.4h
; CHECK-GI-NOFP16-NEXT:    fcvtl v2.4s, v3.4h
; CHECK-GI-NOFP16-NEXT:    fcvtl v3.4s, v4.4h
; CHECK-GI-NOFP16-NEXT:    mov v0.h[1], v5.h[1]
; CHECK-GI-NOFP16-NEXT:    fmla v3.4s, v2.4s, v1.4s
; CHECK-GI-NOFP16-NEXT:    mov v0.h[2], v5.h[2]
; CHECK-GI-NOFP16-NEXT:    fcvtn v1.4h, v3.4s
; CHECK-GI-NOFP16-NEXT:    mov v0.h[3], v5.h[3]
; CHECK-GI-NOFP16-NEXT:    mov v0.h[4], v1.h[0]
; CHECK-GI-NOFP16-NEXT:    mov v0.h[5], v1.h[1]
; CHECK-GI-NOFP16-NEXT:    mov v0.h[6], v1.h[2]
; CHECK-GI-NOFP16-NEXT:    ret
;
; CHECK-GI-FP16-LABEL: fma_v7f16:
; CHECK-GI-FP16:       // %bb.0: // %entry
; CHECK-GI-FP16-NEXT:    fmla v2.8h, v1.8h, v0.8h
; CHECK-GI-FP16-NEXT:    mov v0.16b, v2.16b
; CHECK-GI-FP16-NEXT:    ret
entry:
  %d = call <7 x half> @llvm.fma.v7f16(<7 x half> %a, <7 x half> %b, <7 x half> %c)
  ret <7 x half> %d
}

define <4 x half> @fma_v4f16(<4 x half> %a, <4 x half> %b, <4 x half> %c) {
; CHECK-SD-NOFP16-LABEL: fma_v4f16:
; CHECK-SD-NOFP16:       // %bb.0: // %entry
; CHECK-SD-NOFP16-NEXT:    // kill: def $d2 killed $d2 def $q2
; CHECK-SD-NOFP16-NEXT:    // kill: def $d1 killed $d1 def $q1
; CHECK-SD-NOFP16-NEXT:    // kill: def $d0 killed $d0 def $q0
; CHECK-SD-NOFP16-NEXT:    mov h3, v2.h[1]
; CHECK-SD-NOFP16-NEXT:    mov h4, v1.h[1]
; CHECK-SD-NOFP16-NEXT:    mov h5, v0.h[1]
; CHECK-SD-NOFP16-NEXT:    fcvt s6, h2
; CHECK-SD-NOFP16-NEXT:    fcvt s7, h1
; CHECK-SD-NOFP16-NEXT:    fcvt s16, h0
; CHECK-SD-NOFP16-NEXT:    mov h17, v2.h[2]
; CHECK-SD-NOFP16-NEXT:    mov h18, v1.h[2]
; CHECK-SD-NOFP16-NEXT:    mov h19, v0.h[2]
; CHECK-SD-NOFP16-NEXT:    mov h2, v2.h[3]
; CHECK-SD-NOFP16-NEXT:    mov h1, v1.h[3]
; CHECK-SD-NOFP16-NEXT:    fcvt s3, h3
; CHECK-SD-NOFP16-NEXT:    fcvt s4, h4
; CHECK-SD-NOFP16-NEXT:    fcvt s5, h5
; CHECK-SD-NOFP16-NEXT:    fmadd s6, s16, s7, s6
; CHECK-SD-NOFP16-NEXT:    mov h16, v0.h[3]
; CHECK-SD-NOFP16-NEXT:    fcvt s7, h19
; CHECK-SD-NOFP16-NEXT:    fcvt s2, h2
; CHECK-SD-NOFP16-NEXT:    fcvt s1, h1
; CHECK-SD-NOFP16-NEXT:    fmadd s3, s5, s4, s3
; CHECK-SD-NOFP16-NEXT:    fcvt s4, h17
; CHECK-SD-NOFP16-NEXT:    fcvt s5, h18
; CHECK-SD-NOFP16-NEXT:    fcvt h0, s6
; CHECK-SD-NOFP16-NEXT:    fmadd s4, s7, s5, s4
; CHECK-SD-NOFP16-NEXT:    fcvt h3, s3
; CHECK-SD-NOFP16-NEXT:    fcvt s5, h16
; CHECK-SD-NOFP16-NEXT:    mov v0.h[1], v3.h[0]
; CHECK-SD-NOFP16-NEXT:    fcvt h3, s4
; CHECK-SD-NOFP16-NEXT:    fmadd s1, s5, s1, s2
; CHECK-SD-NOFP16-NEXT:    mov v0.h[2], v3.h[0]
; CHECK-SD-NOFP16-NEXT:    fcvt h1, s1
; CHECK-SD-NOFP16-NEXT:    mov v0.h[3], v1.h[0]
; CHECK-SD-NOFP16-NEXT:    // kill: def $d0 killed $d0 killed $q0
; CHECK-SD-NOFP16-NEXT:    ret
;
; CHECK-SD-FP16-LABEL: fma_v4f16:
; CHECK-SD-FP16:       // %bb.0: // %entry
; CHECK-SD-FP16-NEXT:    fmla v2.4h, v1.4h, v0.4h
; CHECK-SD-FP16-NEXT:    fmov d0, d2
; CHECK-SD-FP16-NEXT:    ret
;
; CHECK-GI-NOFP16-LABEL: fma_v4f16:
; CHECK-GI-NOFP16:       // %bb.0: // %entry
; CHECK-GI-NOFP16-NEXT:    fcvtl v0.4s, v0.4h
; CHECK-GI-NOFP16-NEXT:    fcvtl v1.4s, v1.4h
; CHECK-GI-NOFP16-NEXT:    fcvtl v2.4s, v2.4h
; CHECK-GI-NOFP16-NEXT:    fmla v2.4s, v1.4s, v0.4s
; CHECK-GI-NOFP16-NEXT:    fcvtn v0.4h, v2.4s
; CHECK-GI-NOFP16-NEXT:    ret
;
; CHECK-GI-FP16-LABEL: fma_v4f16:
; CHECK-GI-FP16:       // %bb.0: // %entry
; CHECK-GI-FP16-NEXT:    fmla v2.4h, v1.4h, v0.4h
; CHECK-GI-FP16-NEXT:    fmov d0, d2
; CHECK-GI-FP16-NEXT:    ret
entry:
  %d = call <4 x half> @llvm.fma.v4f16(<4 x half> %a, <4 x half> %b, <4 x half> %c)
  ret <4 x half> %d
}

define <8 x half> @fma_v8f16(<8 x half> %a, <8 x half> %b, <8 x half> %c) {
; CHECK-SD-NOFP16-LABEL: fma_v8f16:
; CHECK-SD-NOFP16:       // %bb.0: // %entry
; CHECK-SD-NOFP16-NEXT:    mov h3, v2.h[1]
; CHECK-SD-NOFP16-NEXT:    mov h4, v1.h[1]
; CHECK-SD-NOFP16-NEXT:    mov h5, v0.h[1]
; CHECK-SD-NOFP16-NEXT:    fcvt s6, h2
; CHECK-SD-NOFP16-NEXT:    fcvt s7, h1
; CHECK-SD-NOFP16-NEXT:    fcvt s16, h0
; CHECK-SD-NOFP16-NEXT:    mov h17, v2.h[2]
; CHECK-SD-NOFP16-NEXT:    mov h18, v1.h[2]
; CHECK-SD-NOFP16-NEXT:    mov h19, v0.h[2]
; CHECK-SD-NOFP16-NEXT:    fcvt s3, h3
; CHECK-SD-NOFP16-NEXT:    fcvt s4, h4
; CHECK-SD-NOFP16-NEXT:    fcvt s5, h5
; CHECK-SD-NOFP16-NEXT:    fmadd s6, s16, s7, s6
; CHECK-SD-NOFP16-NEXT:    fcvt s7, h17
; CHECK-SD-NOFP16-NEXT:    fcvt s16, h18
; CHECK-SD-NOFP16-NEXT:    fcvt s17, h19
; CHECK-SD-NOFP16-NEXT:    mov h18, v1.h[3]
; CHECK-SD-NOFP16-NEXT:    mov h19, v0.h[3]
; CHECK-SD-NOFP16-NEXT:    fmadd s4, s5, s4, s3
; CHECK-SD-NOFP16-NEXT:    mov h5, v2.h[3]
; CHECK-SD-NOFP16-NEXT:    fcvt h3, s6
; CHECK-SD-NOFP16-NEXT:    fmadd s6, s17, s16, s7
; CHECK-SD-NOFP16-NEXT:    mov h17, v2.h[4]
; CHECK-SD-NOFP16-NEXT:    fcvt s7, h18
; CHECK-SD-NOFP16-NEXT:    fcvt s16, h19
; CHECK-SD-NOFP16-NEXT:    mov h18, v1.h[4]
; CHECK-SD-NOFP16-NEXT:    fcvt h4, s4
; CHECK-SD-NOFP16-NEXT:    fcvt s5, h5
; CHECK-SD-NOFP16-NEXT:    mov h19, v0.h[4]
; CHECK-SD-NOFP16-NEXT:    fcvt h6, s6
; CHECK-SD-NOFP16-NEXT:    fcvt s17, h17
; CHECK-SD-NOFP16-NEXT:    fcvt s18, h18
; CHECK-SD-NOFP16-NEXT:    mov v3.h[1], v4.h[0]
; CHECK-SD-NOFP16-NEXT:    mov h4, v2.h[5]
; CHECK-SD-NOFP16-NEXT:    fmadd s5, s16, s7, s5
; CHECK-SD-NOFP16-NEXT:    mov h7, v1.h[5]
; CHECK-SD-NOFP16-NEXT:    mov h16, v0.h[5]
; CHECK-SD-NOFP16-NEXT:    fcvt s19, h19
; CHECK-SD-NOFP16-NEXT:    mov v3.h[2], v6.h[0]
; CHECK-SD-NOFP16-NEXT:    mov h6, v2.h[6]
; CHECK-SD-NOFP16-NEXT:    fcvt s4, h4
; CHECK-SD-NOFP16-NEXT:    fcvt s7, h7
; CHECK-SD-NOFP16-NEXT:    fcvt s16, h16
; CHECK-SD-NOFP16-NEXT:    fcvt h5, s5
; CHECK-SD-NOFP16-NEXT:    fmadd s17, s19, s18, s17
; CHECK-SD-NOFP16-NEXT:    mov h18, v1.h[6]
; CHECK-SD-NOFP16-NEXT:    mov h19, v0.h[6]
; CHECK-SD-NOFP16-NEXT:    mov h2, v2.h[7]
; CHECK-SD-NOFP16-NEXT:    mov h1, v1.h[7]
; CHECK-SD-NOFP16-NEXT:    mov h0, v0.h[7]
; CHECK-SD-NOFP16-NEXT:    fmadd s4, s16, s7, s4
; CHECK-SD-NOFP16-NEXT:    mov v3.h[3], v5.h[0]
; CHECK-SD-NOFP16-NEXT:    fcvt s5, h6
; CHECK-SD-NOFP16-NEXT:    fcvt s6, h18
; CHECK-SD-NOFP16-NEXT:    fcvt s7, h19
; CHECK-SD-NOFP16-NEXT:    fcvt h16, s17
; CHECK-SD-NOFP16-NEXT:    fcvt s2, h2
; CHECK-SD-NOFP16-NEXT:    fcvt s1, h1
; CHECK-SD-NOFP16-NEXT:    fcvt s0, h0
; CHECK-SD-NOFP16-NEXT:    fcvt h4, s4
; CHECK-SD-NOFP16-NEXT:    fmadd s5, s7, s6, s5
; CHECK-SD-NOFP16-NEXT:    mov v3.h[4], v16.h[0]
; CHECK-SD-NOFP16-NEXT:    fmadd s0, s0, s1, s2
; CHECK-SD-NOFP16-NEXT:    mov v3.h[5], v4.h[0]
; CHECK-SD-NOFP16-NEXT:    fcvt h4, s5
; CHECK-SD-NOFP16-NEXT:    fcvt h0, s0
; CHECK-SD-NOFP16-NEXT:    mov v3.h[6], v4.h[0]
; CHECK-SD-NOFP16-NEXT:    mov v3.h[7], v0.h[0]
; CHECK-SD-NOFP16-NEXT:    mov v0.16b, v3.16b
; CHECK-SD-NOFP16-NEXT:    ret
;
; CHECK-SD-FP16-LABEL: fma_v8f16:
; CHECK-SD-FP16:       // %bb.0: // %entry
; CHECK-SD-FP16-NEXT:    fmla v2.8h, v1.8h, v0.8h
; CHECK-SD-FP16-NEXT:    mov v0.16b, v2.16b
; CHECK-SD-FP16-NEXT:    ret
;
; CHECK-GI-NOFP16-LABEL: fma_v8f16:
; CHECK-GI-NOFP16:       // %bb.0: // %entry
; CHECK-GI-NOFP16-NEXT:    fcvtl v3.4s, v0.4h
; CHECK-GI-NOFP16-NEXT:    fcvtl v4.4s, v1.4h
; CHECK-GI-NOFP16-NEXT:    fcvtl v5.4s, v2.4h
; CHECK-GI-NOFP16-NEXT:    fcvtl2 v0.4s, v0.8h
; CHECK-GI-NOFP16-NEXT:    fcvtl2 v1.4s, v1.8h
; CHECK-GI-NOFP16-NEXT:    fcvtl2 v2.4s, v2.8h
; CHECK-GI-NOFP16-NEXT:    fmla v5.4s, v4.4s, v3.4s
; CHECK-GI-NOFP16-NEXT:    fmla v2.4s, v1.4s, v0.4s
; CHECK-GI-NOFP16-NEXT:    fcvtn v0.4h, v5.4s
; CHECK-GI-NOFP16-NEXT:    fcvtn2 v0.8h, v2.4s
; CHECK-GI-NOFP16-NEXT:    ret
;
; CHECK-GI-FP16-LABEL: fma_v8f16:
; CHECK-GI-FP16:       // %bb.0: // %entry
; CHECK-GI-FP16-NEXT:    fmla v2.8h, v1.8h, v0.8h
; CHECK-GI-FP16-NEXT:    mov v0.16b, v2.16b
; CHECK-GI-FP16-NEXT:    ret
entry:
  %d = call <8 x half> @llvm.fma.v8f16(<8 x half> %a, <8 x half> %b, <8 x half> %c)
  ret <8 x half> %d
}

define <16 x half> @fma_v16f16(<16 x half> %a, <16 x half> %b, <16 x half> %c) {
; CHECK-SD-NOFP16-LABEL: fma_v16f16:
; CHECK-SD-NOFP16:       // %bb.0: // %entry
; CHECK-SD-NOFP16-NEXT:    mov h6, v4.h[1]
; CHECK-SD-NOFP16-NEXT:    mov h7, v2.h[1]
; CHECK-SD-NOFP16-NEXT:    mov h16, v0.h[1]
; CHECK-SD-NOFP16-NEXT:    fcvt s17, h4
; CHECK-SD-NOFP16-NEXT:    fcvt s18, h2
; CHECK-SD-NOFP16-NEXT:    fcvt s19, h0
; CHECK-SD-NOFP16-NEXT:    mov h20, v4.h[2]
; CHECK-SD-NOFP16-NEXT:    mov h21, v2.h[2]
; CHECK-SD-NOFP16-NEXT:    mov h22, v0.h[2]
; CHECK-SD-NOFP16-NEXT:    mov h23, v4.h[3]
; CHECK-SD-NOFP16-NEXT:    mov h24, v2.h[3]
; CHECK-SD-NOFP16-NEXT:    mov h25, v0.h[3]
; CHECK-SD-NOFP16-NEXT:    fcvt s6, h6
; CHECK-SD-NOFP16-NEXT:    fcvt s7, h7
; CHECK-SD-NOFP16-NEXT:    fcvt s16, h16
; CHECK-SD-NOFP16-NEXT:    fmadd s17, s19, s18, s17
; CHECK-SD-NOFP16-NEXT:    mov h26, v1.h[1]
; CHECK-SD-NOFP16-NEXT:    fcvt s27, h5
; CHECK-SD-NOFP16-NEXT:    fcvt s18, h20
; CHECK-SD-NOFP16-NEXT:    fcvt s19, h21
; CHECK-SD-NOFP16-NEXT:    fcvt s20, h22
; CHECK-SD-NOFP16-NEXT:    fcvt s21, h23
; CHECK-SD-NOFP16-NEXT:    fcvt s22, h24
; CHECK-SD-NOFP16-NEXT:    fcvt s23, h25
; CHECK-SD-NOFP16-NEXT:    fmadd s7, s16, s7, s6
; CHECK-SD-NOFP16-NEXT:    mov h24, v5.h[1]
; CHECK-SD-NOFP16-NEXT:    mov h25, v3.h[1]
; CHECK-SD-NOFP16-NEXT:    fcvt h6, s17
; CHECK-SD-NOFP16-NEXT:    fcvt s28, h3
; CHECK-SD-NOFP16-NEXT:    fcvt s29, h1
; CHECK-SD-NOFP16-NEXT:    fmadd s19, s20, s19, s18
; CHECK-SD-NOFP16-NEXT:    fcvt s26, h26
; CHECK-SD-NOFP16-NEXT:    mov h16, v4.h[4]
; CHECK-SD-NOFP16-NEXT:    fmadd s21, s23, s22, s21
; CHECK-SD-NOFP16-NEXT:    mov h22, v3.h[2]
; CHECK-SD-NOFP16-NEXT:    mov h23, v1.h[2]
; CHECK-SD-NOFP16-NEXT:    fcvt h20, s7
; CHECK-SD-NOFP16-NEXT:    fcvt s24, h24
; CHECK-SD-NOFP16-NEXT:    fcvt s25, h25
; CHECK-SD-NOFP16-NEXT:    mov h17, v2.h[4]
; CHECK-SD-NOFP16-NEXT:    mov h18, v0.h[4]
; CHECK-SD-NOFP16-NEXT:    mov h7, v4.h[5]
; CHECK-SD-NOFP16-NEXT:    fcvt h19, s19
; CHECK-SD-NOFP16-NEXT:    mov h30, v2.h[5]
; CHECK-SD-NOFP16-NEXT:    fcvt s16, h16
; CHECK-SD-NOFP16-NEXT:    fcvt h21, s21
; CHECK-SD-NOFP16-NEXT:    mov h31, v1.h[4]
; CHECK-SD-NOFP16-NEXT:    fmadd s24, s26, s25, s24
; CHECK-SD-NOFP16-NEXT:    fmadd s25, s29, s28, s27
; CHECK-SD-NOFP16-NEXT:    mov v6.h[1], v20.h[0]
; CHECK-SD-NOFP16-NEXT:    mov h20, v5.h[2]
; CHECK-SD-NOFP16-NEXT:    mov h26, v5.h[3]
; CHECK-SD-NOFP16-NEXT:    mov h27, v3.h[3]
; CHECK-SD-NOFP16-NEXT:    mov h28, v1.h[3]
; CHECK-SD-NOFP16-NEXT:    fcvt s17, h17
; CHECK-SD-NOFP16-NEXT:    fcvt s18, h18
; CHECK-SD-NOFP16-NEXT:    fcvt s29, h7
; CHECK-SD-NOFP16-NEXT:    fcvt s30, h30
; CHECK-SD-NOFP16-NEXT:    mov v6.h[2], v19.h[0]
; CHECK-SD-NOFP16-NEXT:    fcvt h24, s24
; CHECK-SD-NOFP16-NEXT:    fcvt h7, s25
; CHECK-SD-NOFP16-NEXT:    fcvt s19, h20
; CHECK-SD-NOFP16-NEXT:    fcvt s20, h22
; CHECK-SD-NOFP16-NEXT:    fcvt s22, h23
; CHECK-SD-NOFP16-NEXT:    fmadd s16, s18, s17, s16
; CHECK-SD-NOFP16-NEXT:    mov h23, v0.h[5]
; CHECK-SD-NOFP16-NEXT:    fcvt s25, h26
; CHECK-SD-NOFP16-NEXT:    fcvt s26, h27
; CHECK-SD-NOFP16-NEXT:    fcvt s27, h28
; CHECK-SD-NOFP16-NEXT:    mov h18, v4.h[6]
; CHECK-SD-NOFP16-NEXT:    mov v6.h[3], v21.h[0]
; CHECK-SD-NOFP16-NEXT:    mov v7.h[1], v24.h[0]
; CHECK-SD-NOFP16-NEXT:    mov h24, v5.h[5]
; CHECK-SD-NOFP16-NEXT:    fmadd s19, s22, s20, s19
; CHECK-SD-NOFP16-NEXT:    mov h20, v5.h[4]
; CHECK-SD-NOFP16-NEXT:    mov h22, v3.h[4]
; CHECK-SD-NOFP16-NEXT:    fcvt s23, h23
; CHECK-SD-NOFP16-NEXT:    mov h28, v0.h[6]
; CHECK-SD-NOFP16-NEXT:    fcvt h16, s16
; CHECK-SD-NOFP16-NEXT:    fcvt s18, h18
; CHECK-SD-NOFP16-NEXT:    mov h4, v4.h[7]
; CHECK-SD-NOFP16-NEXT:    mov h0, v0.h[7]
; CHECK-SD-NOFP16-NEXT:    fcvt s20, h20
; CHECK-SD-NOFP16-NEXT:    fcvt s21, h22
; CHECK-SD-NOFP16-NEXT:    fcvt s22, h31
; CHECK-SD-NOFP16-NEXT:    fmadd s17, s23, s30, s29
; CHECK-SD-NOFP16-NEXT:    fmadd s23, s27, s26, s25
; CHECK-SD-NOFP16-NEXT:    fcvt h19, s19
; CHECK-SD-NOFP16-NEXT:    mov h25, v3.h[5]
; CHECK-SD-NOFP16-NEXT:    mov h26, v1.h[5]
; CHECK-SD-NOFP16-NEXT:    mov h27, v2.h[6]
; CHECK-SD-NOFP16-NEXT:    mov h29, v1.h[6]
; CHECK-SD-NOFP16-NEXT:    mov h2, v2.h[7]
; CHECK-SD-NOFP16-NEXT:    mov h1, v1.h[7]
; CHECK-SD-NOFP16-NEXT:    fmadd s20, s22, s21, s20
; CHECK-SD-NOFP16-NEXT:    mov h21, v5.h[6]
; CHECK-SD-NOFP16-NEXT:    mov h22, v3.h[6]
; CHECK-SD-NOFP16-NEXT:    mov v7.h[2], v19.h[0]
; CHECK-SD-NOFP16-NEXT:    fcvt h19, s23
; CHECK-SD-NOFP16-NEXT:    fcvt s23, h24
; CHECK-SD-NOFP16-NEXT:    fcvt s24, h25
; CHECK-SD-NOFP16-NEXT:    fcvt s25, h26
; CHECK-SD-NOFP16-NEXT:    fcvt s26, h27
; CHECK-SD-NOFP16-NEXT:    fcvt s27, h28
; CHECK-SD-NOFP16-NEXT:    fcvt s28, h29
; CHECK-SD-NOFP16-NEXT:    mov h5, v5.h[7]
; CHECK-SD-NOFP16-NEXT:    fcvt s21, h21
; CHECK-SD-NOFP16-NEXT:    fcvt s22, h22
; CHECK-SD-NOFP16-NEXT:    mov h3, v3.h[7]
; CHECK-SD-NOFP16-NEXT:    mov v7.h[3], v19.h[0]
; CHECK-SD-NOFP16-NEXT:    fcvt h19, s20
; CHECK-SD-NOFP16-NEXT:    mov v6.h[4], v16.h[0]
; CHECK-SD-NOFP16-NEXT:    fmadd s20, s25, s24, s23
; CHECK-SD-NOFP16-NEXT:    fcvt h16, s17
; CHECK-SD-NOFP16-NEXT:    fcvt s4, h4
; CHECK-SD-NOFP16-NEXT:    fmadd s18, s27, s26, s18
; CHECK-SD-NOFP16-NEXT:    fcvt s2, h2
; CHECK-SD-NOFP16-NEXT:    fcvt s0, h0
; CHECK-SD-NOFP16-NEXT:    fmadd s21, s28, s22, s21
; CHECK-SD-NOFP16-NEXT:    fcvt s5, h5
; CHECK-SD-NOFP16-NEXT:    fcvt s3, h3
; CHECK-SD-NOFP16-NEXT:    mov v7.h[4], v19.h[0]
; CHECK-SD-NOFP16-NEXT:    fcvt s1, h1
; CHECK-SD-NOFP16-NEXT:    fcvt h17, s20
; CHECK-SD-NOFP16-NEXT:    mov v6.h[5], v16.h[0]
; CHECK-SD-NOFP16-NEXT:    fmadd s0, s0, s2, s4
; CHECK-SD-NOFP16-NEXT:    fcvt h2, s18
; CHECK-SD-NOFP16-NEXT:    fcvt h4, s21
; CHECK-SD-NOFP16-NEXT:    fmadd s1, s1, s3, s5
; CHECK-SD-NOFP16-NEXT:    mov v7.h[5], v17.h[0]
; CHECK-SD-NOFP16-NEXT:    mov v6.h[6], v2.h[0]
; CHECK-SD-NOFP16-NEXT:    fcvt h0, s0
; CHECK-SD-NOFP16-NEXT:    fcvt h1, s1
; CHECK-SD-NOFP16-NEXT:    mov v7.h[6], v4.h[0]
; CHECK-SD-NOFP16-NEXT:    mov v6.h[7], v0.h[0]
; CHECK-SD-NOFP16-NEXT:    mov v7.h[7], v1.h[0]
; CHECK-SD-NOFP16-NEXT:    mov v0.16b, v6.16b
; CHECK-SD-NOFP16-NEXT:    mov v1.16b, v7.16b
; CHECK-SD-NOFP16-NEXT:    ret
;
; CHECK-SD-FP16-LABEL: fma_v16f16:
; CHECK-SD-FP16:       // %bb.0: // %entry
; CHECK-SD-FP16-NEXT:    fmla v4.8h, v2.8h, v0.8h
; CHECK-SD-FP16-NEXT:    fmla v5.8h, v3.8h, v1.8h
; CHECK-SD-FP16-NEXT:    mov v0.16b, v4.16b
; CHECK-SD-FP16-NEXT:    mov v1.16b, v5.16b
; CHECK-SD-FP16-NEXT:    ret
;
; CHECK-GI-NOFP16-LABEL: fma_v16f16:
; CHECK-GI-NOFP16:       // %bb.0: // %entry
; CHECK-GI-NOFP16-NEXT:    fcvtl v6.4s, v0.4h
; CHECK-GI-NOFP16-NEXT:    fcvtl v7.4s, v1.4h
; CHECK-GI-NOFP16-NEXT:    fcvtl v16.4s, v2.4h
; CHECK-GI-NOFP16-NEXT:    fcvtl v17.4s, v3.4h
; CHECK-GI-NOFP16-NEXT:    fcvtl v18.4s, v4.4h
; CHECK-GI-NOFP16-NEXT:    fcvtl v19.4s, v5.4h
; CHECK-GI-NOFP16-NEXT:    fcvtl2 v0.4s, v0.8h
; CHECK-GI-NOFP16-NEXT:    fcvtl2 v1.4s, v1.8h
; CHECK-GI-NOFP16-NEXT:    fcvtl2 v2.4s, v2.8h
; CHECK-GI-NOFP16-NEXT:    fcvtl2 v3.4s, v3.8h
; CHECK-GI-NOFP16-NEXT:    fcvtl2 v4.4s, v4.8h
; CHECK-GI-NOFP16-NEXT:    fcvtl2 v5.4s, v5.8h
; CHECK-GI-NOFP16-NEXT:    fmla v18.4s, v16.4s, v6.4s
; CHECK-GI-NOFP16-NEXT:    fmla v19.4s, v17.4s, v7.4s
; CHECK-GI-NOFP16-NEXT:    fmla v4.4s, v2.4s, v0.4s
; CHECK-GI-NOFP16-NEXT:    fmla v5.4s, v3.4s, v1.4s
; CHECK-GI-NOFP16-NEXT:    fcvtn v0.4h, v18.4s
; CHECK-GI-NOFP16-NEXT:    fcvtn v1.4h, v19.4s
; CHECK-GI-NOFP16-NEXT:    fcvtn2 v0.8h, v4.4s
; CHECK-GI-NOFP16-NEXT:    fcvtn2 v1.8h, v5.4s
; CHECK-GI-NOFP16-NEXT:    ret
;
; CHECK-GI-FP16-LABEL: fma_v16f16:
; CHECK-GI-FP16:       // %bb.0: // %entry
; CHECK-GI-FP16-NEXT:    fmla v4.8h, v2.8h, v0.8h
; CHECK-GI-FP16-NEXT:    fmla v5.8h, v3.8h, v1.8h
; CHECK-GI-FP16-NEXT:    mov v0.16b, v4.16b
; CHECK-GI-FP16-NEXT:    mov v1.16b, v5.16b
; CHECK-GI-FP16-NEXT:    ret
entry:
  %d = call <16 x half> @llvm.fma.v16f16(<16 x half> %a, <16 x half> %b, <16 x half> %c)
  ret <16 x half> %d
}

define double @fmuladd_f64(double %a, double %b, double %c) {
; CHECK-LABEL: fmuladd_f64:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    fmadd d0, d0, d1, d2
; CHECK-NEXT:    ret
entry:
  %d = call double @llvm.fmuladd.f64(double %a, double %b, double %c)
  ret double %d
}

define float @fmuladd_f32(float %a, float %b, float %c) {
; CHECK-LABEL: fmuladd_f32:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    fmadd s0, s0, s1, s2
; CHECK-NEXT:    ret
entry:
  %d = call float @llvm.fmuladd.f32(float %a, float %b, float %c)
  ret float %d
}

define half @fmuladd_f16(half %a, half %b, half %c) {
; CHECK-SD-NOFP16-LABEL: fmuladd_f16:
; CHECK-SD-NOFP16:       // %bb.0: // %entry
; CHECK-SD-NOFP16-NEXT:    fcvt s1, h1
; CHECK-SD-NOFP16-NEXT:    fcvt s0, h0
; CHECK-SD-NOFP16-NEXT:    fmul s0, s0, s1
; CHECK-SD-NOFP16-NEXT:    fcvt s1, h2
; CHECK-SD-NOFP16-NEXT:    fcvt h0, s0
; CHECK-SD-NOFP16-NEXT:    fcvt s0, h0
; CHECK-SD-NOFP16-NEXT:    fadd s0, s0, s1
; CHECK-SD-NOFP16-NEXT:    fcvt h0, s0
; CHECK-SD-NOFP16-NEXT:    ret
;
; CHECK-SD-FP16-LABEL: fmuladd_f16:
; CHECK-SD-FP16:       // %bb.0: // %entry
; CHECK-SD-FP16-NEXT:    fmadd h0, h0, h1, h2
; CHECK-SD-FP16-NEXT:    ret
;
; CHECK-GI-NOFP16-LABEL: fmuladd_f16:
; CHECK-GI-NOFP16:       // %bb.0: // %entry
; CHECK-GI-NOFP16-NEXT:    fcvt s0, h0
; CHECK-GI-NOFP16-NEXT:    fcvt s1, h1
; CHECK-GI-NOFP16-NEXT:    fmul s0, s0, s1
; CHECK-GI-NOFP16-NEXT:    fcvt s1, h2
; CHECK-GI-NOFP16-NEXT:    fcvt h0, s0
; CHECK-GI-NOFP16-NEXT:    fcvt s0, h0
; CHECK-GI-NOFP16-NEXT:    fadd s0, s0, s1
; CHECK-GI-NOFP16-NEXT:    fcvt h0, s0
; CHECK-GI-NOFP16-NEXT:    ret
;
; CHECK-GI-FP16-LABEL: fmuladd_f16:
; CHECK-GI-FP16:       // %bb.0: // %entry
; CHECK-GI-FP16-NEXT:    fmadd h0, h0, h1, h2
; CHECK-GI-FP16-NEXT:    ret
entry:
  %d = call half @llvm.fmuladd.f16(half %a, half %b, half %c)
  ret half %d
}

define <2 x double> @fmuladd_v2f64(<2 x double> %a, <2 x double> %b, <2 x double> %c) {
; CHECK-LABEL: fmuladd_v2f64:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    fmla v2.2d, v1.2d, v0.2d
; CHECK-NEXT:    mov v0.16b, v2.16b
; CHECK-NEXT:    ret
entry:
  %d = call <2 x double> @llvm.fmuladd.v2f64(<2 x double> %a, <2 x double> %b, <2 x double> %c)
  ret <2 x double> %d
}

define <3 x double> @fmuladd_v3f64(<3 x double> %a, <3 x double> %b, <3 x double> %c) {
; CHECK-SD-LABEL: fmuladd_v3f64:
; CHECK-SD:       // %bb.0: // %entry
; CHECK-SD-NEXT:    // kill: def $d6 killed $d6 def $q6
; CHECK-SD-NEXT:    // kill: def $d3 killed $d3 def $q3
; CHECK-SD-NEXT:    // kill: def $d0 killed $d0 def $q0
; CHECK-SD-NEXT:    // kill: def $d7 killed $d7 def $q7
; CHECK-SD-NEXT:    // kill: def $d4 killed $d4 def $q4
; CHECK-SD-NEXT:    // kill: def $d1 killed $d1 def $q1
; CHECK-SD-NEXT:    // kill: def $d5 killed $d5 def $q5
; CHECK-SD-NEXT:    // kill: def $d2 killed $d2 def $q2
; CHECK-SD-NEXT:    mov v3.d[1], v4.d[0]
; CHECK-SD-NEXT:    mov v0.d[1], v1.d[0]
; CHECK-SD-NEXT:    mov v6.d[1], v7.d[0]
; CHECK-SD-NEXT:    fmla v6.2d, v3.2d, v0.2d
; CHECK-SD-NEXT:    ldr d3, [sp]
; CHECK-SD-NEXT:    fmla v3.2d, v5.2d, v2.2d
; CHECK-SD-NEXT:    fmov d0, d6
; CHECK-SD-NEXT:    ext v1.16b, v6.16b, v6.16b, #8
; CHECK-SD-NEXT:    // kill: def $d1 killed $d1 killed $q1
; CHECK-SD-NEXT:    fmov d2, d3
; CHECK-SD-NEXT:    ret
;
; CHECK-GI-LABEL: fmuladd_v3f64:
; CHECK-GI:       // %bb.0: // %entry
; CHECK-GI-NEXT:    // kill: def $d0 killed $d0 def $q0
; CHECK-GI-NEXT:    // kill: def $d3 killed $d3 def $q3
; CHECK-GI-NEXT:    // kill: def $d6 killed $d6 def $q6
; CHECK-GI-NEXT:    // kill: def $d1 killed $d1 def $q1
; CHECK-GI-NEXT:    // kill: def $d4 killed $d4 def $q4
; CHECK-GI-NEXT:    // kill: def $d7 killed $d7 def $q7
; CHECK-GI-NEXT:    mov v0.d[1], v1.d[0]
; CHECK-GI-NEXT:    mov v3.d[1], v4.d[0]
; CHECK-GI-NEXT:    mov v6.d[1], v7.d[0]
; CHECK-GI-NEXT:    fmla v6.2d, v3.2d, v0.2d
; CHECK-GI-NEXT:    ldr d0, [sp]
; CHECK-GI-NEXT:    fmadd d2, d2, d5, d0
; CHECK-GI-NEXT:    mov d1, v6.d[1]
; CHECK-GI-NEXT:    fmov d0, d6
; CHECK-GI-NEXT:    ret
entry:
  %d = call <3 x double> @llvm.fmuladd.v3f64(<3 x double> %a, <3 x double> %b, <3 x double> %c)
  ret <3 x double> %d
}

define <4 x double> @fmuladd_v4f64(<4 x double> %a, <4 x double> %b, <4 x double> %c) {
; CHECK-LABEL: fmuladd_v4f64:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    fmla v4.2d, v2.2d, v0.2d
; CHECK-NEXT:    fmla v5.2d, v3.2d, v1.2d
; CHECK-NEXT:    mov v0.16b, v4.16b
; CHECK-NEXT:    mov v1.16b, v5.16b
; CHECK-NEXT:    ret
entry:
  %d = call <4 x double> @llvm.fmuladd.v4f64(<4 x double> %a, <4 x double> %b, <4 x double> %c)
  ret <4 x double> %d
}

define <2 x float> @fmuladd_v2f32(<2 x float> %a, <2 x float> %b, <2 x float> %c) {
; CHECK-LABEL: fmuladd_v2f32:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    fmla v2.2s, v1.2s, v0.2s
; CHECK-NEXT:    fmov d0, d2
; CHECK-NEXT:    ret
entry:
  %d = call <2 x float> @llvm.fmuladd.v2f32(<2 x float> %a, <2 x float> %b, <2 x float> %c)
  ret <2 x float> %d
}

define <3 x float> @fmuladd_v3f32(<3 x float> %a, <3 x float> %b, <3 x float> %c) {
; CHECK-LABEL: fmuladd_v3f32:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    fmla v2.4s, v1.4s, v0.4s
; CHECK-NEXT:    mov v0.16b, v2.16b
; CHECK-NEXT:    ret
entry:
  %d = call <3 x float> @llvm.fmuladd.v3f32(<3 x float> %a, <3 x float> %b, <3 x float> %c)
  ret <3 x float> %d
}

define <4 x float> @fmuladd_v4f32(<4 x float> %a, <4 x float> %b, <4 x float> %c) {
; CHECK-LABEL: fmuladd_v4f32:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    fmla v2.4s, v1.4s, v0.4s
; CHECK-NEXT:    mov v0.16b, v2.16b
; CHECK-NEXT:    ret
entry:
  %d = call <4 x float> @llvm.fmuladd.v4f32(<4 x float> %a, <4 x float> %b, <4 x float> %c)
  ret <4 x float> %d
}

define <8 x float> @fmuladd_v8f32(<8 x float> %a, <8 x float> %b, <8 x float> %c) {
; CHECK-LABEL: fmuladd_v8f32:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    fmla v4.4s, v2.4s, v0.4s
; CHECK-NEXT:    fmla v5.4s, v3.4s, v1.4s
; CHECK-NEXT:    mov v0.16b, v4.16b
; CHECK-NEXT:    mov v1.16b, v5.16b
; CHECK-NEXT:    ret
entry:
  %d = call <8 x float> @llvm.fmuladd.v8f32(<8 x float> %a, <8 x float> %b, <8 x float> %c)
  ret <8 x float> %d
}

define <7 x half> @fmuladd_v7f16(<7 x half> %a, <7 x half> %b, <7 x half> %c) {
; CHECK-SD-NOFP16-LABEL: fmuladd_v7f16:
; CHECK-SD-NOFP16:       // %bb.0: // %entry
; CHECK-SD-NOFP16-NEXT:    fcvtl v3.4s, v1.4h
; CHECK-SD-NOFP16-NEXT:    fcvtl v4.4s, v0.4h
; CHECK-SD-NOFP16-NEXT:    fcvtl2 v1.4s, v1.8h
; CHECK-SD-NOFP16-NEXT:    fcvtl2 v0.4s, v0.8h
; CHECK-SD-NOFP16-NEXT:    fmul v3.4s, v4.4s, v3.4s
; CHECK-SD-NOFP16-NEXT:    fmul v0.4s, v0.4s, v1.4s
; CHECK-SD-NOFP16-NEXT:    fcvtn v1.4h, v3.4s
; CHECK-SD-NOFP16-NEXT:    fcvtl v3.4s, v2.4h
; CHECK-SD-NOFP16-NEXT:    fcvtl2 v2.4s, v2.8h
; CHECK-SD-NOFP16-NEXT:    fcvtn2 v1.8h, v0.4s
; CHECK-SD-NOFP16-NEXT:    fcvtl v0.4s, v1.4h
; CHECK-SD-NOFP16-NEXT:    fcvtl2 v1.4s, v1.8h
; CHECK-SD-NOFP16-NEXT:    fadd v0.4s, v0.4s, v3.4s
; CHECK-SD-NOFP16-NEXT:    fadd v1.4s, v1.4s, v2.4s
; CHECK-SD-NOFP16-NEXT:    fcvtn v0.4h, v0.4s
; CHECK-SD-NOFP16-NEXT:    fcvtn2 v0.8h, v1.4s
; CHECK-SD-NOFP16-NEXT:    ret
;
; CHECK-SD-FP16-LABEL: fmuladd_v7f16:
; CHECK-SD-FP16:       // %bb.0: // %entry
; CHECK-SD-FP16-NEXT:    fmla v2.8h, v1.8h, v0.8h
; CHECK-SD-FP16-NEXT:    mov v0.16b, v2.16b
; CHECK-SD-FP16-NEXT:    ret
;
; CHECK-GI-NOFP16-LABEL: fmuladd_v7f16:
; CHECK-GI-NOFP16:       // %bb.0: // %entry
; CHECK-GI-NOFP16-NEXT:    fcvtl v3.4s, v0.4h
; CHECK-GI-NOFP16-NEXT:    fcvtl v4.4s, v1.4h
; CHECK-GI-NOFP16-NEXT:    mov v5.h[0], v0.h[4]
; CHECK-GI-NOFP16-NEXT:    fmul v3.4s, v3.4s, v4.4s
; CHECK-GI-NOFP16-NEXT:    mov v4.h[0], v1.h[4]
; CHECK-GI-NOFP16-NEXT:    mov v5.h[1], v0.h[5]
; CHECK-GI-NOFP16-NEXT:    mov v4.h[1], v1.h[5]
; CHECK-GI-NOFP16-NEXT:    fcvtn v3.4h, v3.4s
; CHECK-GI-NOFP16-NEXT:    mov v5.h[2], v0.h[6]
; CHECK-GI-NOFP16-NEXT:    mov v4.h[2], v1.h[6]
; CHECK-GI-NOFP16-NEXT:    fcvtl v0.4s, v3.4h
; CHECK-GI-NOFP16-NEXT:    fcvtl v1.4s, v2.4h
; CHECK-GI-NOFP16-NEXT:    fcvtl v3.4s, v5.4h
; CHECK-GI-NOFP16-NEXT:    mov v5.h[0], v2.h[4]
; CHECK-GI-NOFP16-NEXT:    fcvtl v4.4s, v4.4h
; CHECK-GI-NOFP16-NEXT:    fadd v0.4s, v0.4s, v1.4s
; CHECK-GI-NOFP16-NEXT:    mov v5.h[1], v2.h[5]
; CHECK-GI-NOFP16-NEXT:    fmul v1.4s, v3.4s, v4.4s
; CHECK-GI-NOFP16-NEXT:    fcvtn v3.4h, v0.4s
; CHECK-GI-NOFP16-NEXT:    mov v5.h[2], v2.h[6]
; CHECK-GI-NOFP16-NEXT:    fcvtn v1.4h, v1.4s
; CHECK-GI-NOFP16-NEXT:    mov v0.h[0], v3.h[0]
; CHECK-GI-NOFP16-NEXT:    fcvtl v2.4s, v5.4h
; CHECK-GI-NOFP16-NEXT:    fcvtl v1.4s, v1.4h
; CHECK-GI-NOFP16-NEXT:    mov v0.h[1], v3.h[1]
; CHECK-GI-NOFP16-NEXT:    fadd v1.4s, v1.4s, v2.4s
; CHECK-GI-NOFP16-NEXT:    mov v0.h[2], v3.h[2]
; CHECK-GI-NOFP16-NEXT:    fcvtn v1.4h, v1.4s
; CHECK-GI-NOFP16-NEXT:    mov v0.h[3], v3.h[3]
; CHECK-GI-NOFP16-NEXT:    mov v0.h[4], v1.h[0]
; CHECK-GI-NOFP16-NEXT:    mov v0.h[5], v1.h[1]
; CHECK-GI-NOFP16-NEXT:    mov v0.h[6], v1.h[2]
; CHECK-GI-NOFP16-NEXT:    ret
;
; CHECK-GI-FP16-LABEL: fmuladd_v7f16:
; CHECK-GI-FP16:       // %bb.0: // %entry
; CHECK-GI-FP16-NEXT:    fmla v2.8h, v1.8h, v0.8h
; CHECK-GI-FP16-NEXT:    mov v0.16b, v2.16b
; CHECK-GI-FP16-NEXT:    ret
entry:
  %d = call <7 x half> @llvm.fmuladd.v7f16(<7 x half> %a, <7 x half> %b, <7 x half> %c)
  ret <7 x half> %d
}

define <4 x half> @fmuladd_v4f16(<4 x half> %a, <4 x half> %b, <4 x half> %c) {
; CHECK-SD-NOFP16-LABEL: fmuladd_v4f16:
; CHECK-SD-NOFP16:       // %bb.0: // %entry
; CHECK-SD-NOFP16-NEXT:    fcvtl v1.4s, v1.4h
; CHECK-SD-NOFP16-NEXT:    fcvtl v0.4s, v0.4h
; CHECK-SD-NOFP16-NEXT:    fmul v0.4s, v0.4s, v1.4s
; CHECK-SD-NOFP16-NEXT:    fcvtl v1.4s, v2.4h
; CHECK-SD-NOFP16-NEXT:    fcvtn v0.4h, v0.4s
; CHECK-SD-NOFP16-NEXT:    fcvtl v0.4s, v0.4h
; CHECK-SD-NOFP16-NEXT:    fadd v0.4s, v0.4s, v1.4s
; CHECK-SD-NOFP16-NEXT:    fcvtn v0.4h, v0.4s
; CHECK-SD-NOFP16-NEXT:    ret
;
; CHECK-SD-FP16-LABEL: fmuladd_v4f16:
; CHECK-SD-FP16:       // %bb.0: // %entry
; CHECK-SD-FP16-NEXT:    fmla v2.4h, v1.4h, v0.4h
; CHECK-SD-FP16-NEXT:    fmov d0, d2
; CHECK-SD-FP16-NEXT:    ret
;
; CHECK-GI-NOFP16-LABEL: fmuladd_v4f16:
; CHECK-GI-NOFP16:       // %bb.0: // %entry
; CHECK-GI-NOFP16-NEXT:    fcvtl v0.4s, v0.4h
; CHECK-GI-NOFP16-NEXT:    fcvtl v1.4s, v1.4h
; CHECK-GI-NOFP16-NEXT:    fmul v0.4s, v0.4s, v1.4s
; CHECK-GI-NOFP16-NEXT:    fcvtl v1.4s, v2.4h
; CHECK-GI-NOFP16-NEXT:    fcvtn v0.4h, v0.4s
; CHECK-GI-NOFP16-NEXT:    fcvtl v0.4s, v0.4h
; CHECK-GI-NOFP16-NEXT:    fadd v0.4s, v0.4s, v1.4s
; CHECK-GI-NOFP16-NEXT:    fcvtn v0.4h, v0.4s
; CHECK-GI-NOFP16-NEXT:    ret
;
; CHECK-GI-FP16-LABEL: fmuladd_v4f16:
; CHECK-GI-FP16:       // %bb.0: // %entry
; CHECK-GI-FP16-NEXT:    fmla v2.4h, v1.4h, v0.4h
; CHECK-GI-FP16-NEXT:    fmov d0, d2
; CHECK-GI-FP16-NEXT:    ret
entry:
  %d = call <4 x half> @llvm.fmuladd.v4f16(<4 x half> %a, <4 x half> %b, <4 x half> %c)
  ret <4 x half> %d
}

define <8 x half> @fmuladd_v8f16(<8 x half> %a, <8 x half> %b, <8 x half> %c) {
; CHECK-SD-NOFP16-LABEL: fmuladd_v8f16:
; CHECK-SD-NOFP16:       // %bb.0: // %entry
; CHECK-SD-NOFP16-NEXT:    fcvtl v3.4s, v1.4h
; CHECK-SD-NOFP16-NEXT:    fcvtl v4.4s, v0.4h
; CHECK-SD-NOFP16-NEXT:    fcvtl2 v1.4s, v1.8h
; CHECK-SD-NOFP16-NEXT:    fcvtl2 v0.4s, v0.8h
; CHECK-SD-NOFP16-NEXT:    fmul v3.4s, v4.4s, v3.4s
; CHECK-SD-NOFP16-NEXT:    fmul v0.4s, v0.4s, v1.4s
; CHECK-SD-NOFP16-NEXT:    fcvtn v1.4h, v3.4s
; CHECK-SD-NOFP16-NEXT:    fcvtl v3.4s, v2.4h
; CHECK-SD-NOFP16-NEXT:    fcvtl2 v2.4s, v2.8h
; CHECK-SD-NOFP16-NEXT:    fcvtn2 v1.8h, v0.4s
; CHECK-SD-NOFP16-NEXT:    fcvtl v0.4s, v1.4h
; CHECK-SD-NOFP16-NEXT:    fcvtl2 v1.4s, v1.8h
; CHECK-SD-NOFP16-NEXT:    fadd v0.4s, v0.4s, v3.4s
; CHECK-SD-NOFP16-NEXT:    fadd v1.4s, v1.4s, v2.4s
; CHECK-SD-NOFP16-NEXT:    fcvtn v0.4h, v0.4s
; CHECK-SD-NOFP16-NEXT:    fcvtn2 v0.8h, v1.4s
; CHECK-SD-NOFP16-NEXT:    ret
;
; CHECK-SD-FP16-LABEL: fmuladd_v8f16:
; CHECK-SD-FP16:       // %bb.0: // %entry
; CHECK-SD-FP16-NEXT:    fmla v2.8h, v1.8h, v0.8h
; CHECK-SD-FP16-NEXT:    mov v0.16b, v2.16b
; CHECK-SD-FP16-NEXT:    ret
;
; CHECK-GI-NOFP16-LABEL: fmuladd_v8f16:
; CHECK-GI-NOFP16:       // %bb.0: // %entry
; CHECK-GI-NOFP16-NEXT:    fcvtl v3.4s, v0.4h
; CHECK-GI-NOFP16-NEXT:    fcvtl v4.4s, v1.4h
; CHECK-GI-NOFP16-NEXT:    fcvtl2 v0.4s, v0.8h
; CHECK-GI-NOFP16-NEXT:    fcvtl2 v1.4s, v1.8h
; CHECK-GI-NOFP16-NEXT:    fmul v3.4s, v3.4s, v4.4s
; CHECK-GI-NOFP16-NEXT:    fmul v0.4s, v0.4s, v1.4s
; CHECK-GI-NOFP16-NEXT:    fcvtn v1.4h, v3.4s
; CHECK-GI-NOFP16-NEXT:    fcvtl v3.4s, v2.4h
; CHECK-GI-NOFP16-NEXT:    fcvtl2 v2.4s, v2.8h
; CHECK-GI-NOFP16-NEXT:    fcvtn v0.4h, v0.4s
; CHECK-GI-NOFP16-NEXT:    fcvtl v1.4s, v1.4h
; CHECK-GI-NOFP16-NEXT:    fcvtl v0.4s, v0.4h
; CHECK-GI-NOFP16-NEXT:    fadd v1.4s, v1.4s, v3.4s
; CHECK-GI-NOFP16-NEXT:    fadd v2.4s, v0.4s, v2.4s
; CHECK-GI-NOFP16-NEXT:    fcvtn v0.4h, v1.4s
; CHECK-GI-NOFP16-NEXT:    fcvtn2 v0.8h, v2.4s
; CHECK-GI-NOFP16-NEXT:    ret
;
; CHECK-GI-FP16-LABEL: fmuladd_v8f16:
; CHECK-GI-FP16:       // %bb.0: // %entry
; CHECK-GI-FP16-NEXT:    fmla v2.8h, v1.8h, v0.8h
; CHECK-GI-FP16-NEXT:    mov v0.16b, v2.16b
; CHECK-GI-FP16-NEXT:    ret
entry:
  %d = call <8 x half> @llvm.fmuladd.v8f16(<8 x half> %a, <8 x half> %b, <8 x half> %c)
  ret <8 x half> %d
}

define <16 x half> @fmuladd_v16f16(<16 x half> %a, <16 x half> %b, <16 x half> %c) {
; CHECK-SD-NOFP16-LABEL: fmuladd_v16f16:
; CHECK-SD-NOFP16:       // %bb.0: // %entry
; CHECK-SD-NOFP16-NEXT:    fcvtl v6.4s, v2.4h
; CHECK-SD-NOFP16-NEXT:    fcvtl v7.4s, v0.4h
; CHECK-SD-NOFP16-NEXT:    fcvtl v16.4s, v3.4h
; CHECK-SD-NOFP16-NEXT:    fcvtl v17.4s, v1.4h
; CHECK-SD-NOFP16-NEXT:    fcvtl2 v2.4s, v2.8h
; CHECK-SD-NOFP16-NEXT:    fcvtl2 v0.4s, v0.8h
; CHECK-SD-NOFP16-NEXT:    fcvtl2 v3.4s, v3.8h
; CHECK-SD-NOFP16-NEXT:    fcvtl2 v1.4s, v1.8h
; CHECK-SD-NOFP16-NEXT:    fmul v6.4s, v7.4s, v6.4s
; CHECK-SD-NOFP16-NEXT:    fmul v7.4s, v17.4s, v16.4s
; CHECK-SD-NOFP16-NEXT:    fmul v0.4s, v0.4s, v2.4s
; CHECK-SD-NOFP16-NEXT:    fmul v1.4s, v1.4s, v3.4s
; CHECK-SD-NOFP16-NEXT:    fcvtn v2.4h, v6.4s
; CHECK-SD-NOFP16-NEXT:    fcvtl v6.4s, v5.4h
; CHECK-SD-NOFP16-NEXT:    fcvtn v3.4h, v7.4s
; CHECK-SD-NOFP16-NEXT:    fcvtn2 v2.8h, v0.4s
; CHECK-SD-NOFP16-NEXT:    fcvtl v0.4s, v4.4h
; CHECK-SD-NOFP16-NEXT:    fcvtl2 v4.4s, v4.8h
; CHECK-SD-NOFP16-NEXT:    fcvtn2 v3.8h, v1.4s
; CHECK-SD-NOFP16-NEXT:    fcvtl v1.4s, v2.4h
; CHECK-SD-NOFP16-NEXT:    fcvtl2 v2.4s, v2.8h
; CHECK-SD-NOFP16-NEXT:    fcvtl v7.4s, v3.4h
; CHECK-SD-NOFP16-NEXT:    fadd v0.4s, v1.4s, v0.4s
; CHECK-SD-NOFP16-NEXT:    fcvtl2 v1.4s, v3.8h
; CHECK-SD-NOFP16-NEXT:    fcvtl2 v3.4s, v5.8h
; CHECK-SD-NOFP16-NEXT:    fadd v5.4s, v7.4s, v6.4s
; CHECK-SD-NOFP16-NEXT:    fadd v2.4s, v2.4s, v4.4s
; CHECK-SD-NOFP16-NEXT:    fcvtn v0.4h, v0.4s
; CHECK-SD-NOFP16-NEXT:    fadd v3.4s, v1.4s, v3.4s
; CHECK-SD-NOFP16-NEXT:    fcvtn v1.4h, v5.4s
; CHECK-SD-NOFP16-NEXT:    fcvtn2 v0.8h, v2.4s
; CHECK-SD-NOFP16-NEXT:    fcvtn2 v1.8h, v3.4s
; CHECK-SD-NOFP16-NEXT:    ret
;
; CHECK-SD-FP16-LABEL: fmuladd_v16f16:
; CHECK-SD-FP16:       // %bb.0: // %entry
; CHECK-SD-FP16-NEXT:    fmla v4.8h, v2.8h, v0.8h
; CHECK-SD-FP16-NEXT:    fmla v5.8h, v3.8h, v1.8h
; CHECK-SD-FP16-NEXT:    mov v0.16b, v4.16b
; CHECK-SD-FP16-NEXT:    mov v1.16b, v5.16b
; CHECK-SD-FP16-NEXT:    ret
;
; CHECK-GI-NOFP16-LABEL: fmuladd_v16f16:
; CHECK-GI-NOFP16:       // %bb.0: // %entry
; CHECK-GI-NOFP16-NEXT:    fcvtl v6.4s, v0.4h
; CHECK-GI-NOFP16-NEXT:    fcvtl v7.4s, v2.4h
; CHECK-GI-NOFP16-NEXT:    fcvtl v16.4s, v1.4h
; CHECK-GI-NOFP16-NEXT:    fcvtl v17.4s, v3.4h
; CHECK-GI-NOFP16-NEXT:    fcvtl2 v0.4s, v0.8h
; CHECK-GI-NOFP16-NEXT:    fcvtl2 v1.4s, v1.8h
; CHECK-GI-NOFP16-NEXT:    fcvtl2 v2.4s, v2.8h
; CHECK-GI-NOFP16-NEXT:    fcvtl2 v3.4s, v3.8h
; CHECK-GI-NOFP16-NEXT:    fmul v6.4s, v6.4s, v7.4s
; CHECK-GI-NOFP16-NEXT:    fmul v7.4s, v16.4s, v17.4s
; CHECK-GI-NOFP16-NEXT:    fmul v0.4s, v0.4s, v2.4s
; CHECK-GI-NOFP16-NEXT:    fmul v1.4s, v1.4s, v3.4s
; CHECK-GI-NOFP16-NEXT:    fcvtn v2.4h, v6.4s
; CHECK-GI-NOFP16-NEXT:    fcvtl v6.4s, v4.4h
; CHECK-GI-NOFP16-NEXT:    fcvtl2 v4.4s, v4.8h
; CHECK-GI-NOFP16-NEXT:    fcvtn v3.4h, v7.4s
; CHECK-GI-NOFP16-NEXT:    fcvtl v7.4s, v5.4h
; CHECK-GI-NOFP16-NEXT:    fcvtl2 v5.4s, v5.8h
; CHECK-GI-NOFP16-NEXT:    fcvtn v0.4h, v0.4s
; CHECK-GI-NOFP16-NEXT:    fcvtn v1.4h, v1.4s
; CHECK-GI-NOFP16-NEXT:    fcvtl v2.4s, v2.4h
; CHECK-GI-NOFP16-NEXT:    fcvtl v3.4s, v3.4h
; CHECK-GI-NOFP16-NEXT:    fcvtl v0.4s, v0.4h
; CHECK-GI-NOFP16-NEXT:    fcvtl v1.4s, v1.4h
; CHECK-GI-NOFP16-NEXT:    fadd v2.4s, v2.4s, v6.4s
; CHECK-GI-NOFP16-NEXT:    fadd v3.4s, v3.4s, v7.4s
; CHECK-GI-NOFP16-NEXT:    fadd v4.4s, v0.4s, v4.4s
; CHECK-GI-NOFP16-NEXT:    fadd v5.4s, v1.4s, v5.4s
; CHECK-GI-NOFP16-NEXT:    fcvtn v0.4h, v2.4s
; CHECK-GI-NOFP16-NEXT:    fcvtn v1.4h, v3.4s
; CHECK-GI-NOFP16-NEXT:    fcvtn2 v0.8h, v4.4s
; CHECK-GI-NOFP16-NEXT:    fcvtn2 v1.8h, v5.4s
; CHECK-GI-NOFP16-NEXT:    ret
;
; CHECK-GI-FP16-LABEL: fmuladd_v16f16:
; CHECK-GI-FP16:       // %bb.0: // %entry
; CHECK-GI-FP16-NEXT:    fmla v4.8h, v2.8h, v0.8h
; CHECK-GI-FP16-NEXT:    fmla v5.8h, v3.8h, v1.8h
; CHECK-GI-FP16-NEXT:    mov v0.16b, v4.16b
; CHECK-GI-FP16-NEXT:    mov v1.16b, v5.16b
; CHECK-GI-FP16-NEXT:    ret
entry:
  %d = call <16 x half> @llvm.fmuladd.v16f16(<16 x half> %a, <16 x half> %b, <16 x half> %c)
  ret <16 x half> %d
}

define double @fmul_f64(double %a, double %b, double %c) {
; CHECK-LABEL: fmul_f64:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    fmadd d0, d0, d1, d2
; CHECK-NEXT:    ret
entry:
  %d = fmul fast double %a, %b
  %e = fadd fast double %d, %c
  ret double %e
}

define float @fmul_f32(float %a, float %b, float %c) {
; CHECK-LABEL: fmul_f32:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    fmadd s0, s0, s1, s2
; CHECK-NEXT:    ret
entry:
  %d = fmul fast float %a, %b
  %e = fadd fast float %d, %c
  ret float %e
}

define half @fmul_f16(half %a, half %b, half %c) {
; CHECK-SD-NOFP16-LABEL: fmul_f16:
; CHECK-SD-NOFP16:       // %bb.0: // %entry
; CHECK-SD-NOFP16-NEXT:    fcvt s1, h1
; CHECK-SD-NOFP16-NEXT:    fcvt s0, h0
; CHECK-SD-NOFP16-NEXT:    fmul s0, s0, s1
; CHECK-SD-NOFP16-NEXT:    fcvt s1, h2
; CHECK-SD-NOFP16-NEXT:    fcvt h0, s0
; CHECK-SD-NOFP16-NEXT:    fcvt s0, h0
; CHECK-SD-NOFP16-NEXT:    fadd s0, s0, s1
; CHECK-SD-NOFP16-NEXT:    fcvt h0, s0
; CHECK-SD-NOFP16-NEXT:    ret
;
; CHECK-SD-FP16-LABEL: fmul_f16:
; CHECK-SD-FP16:       // %bb.0: // %entry
; CHECK-SD-FP16-NEXT:    fmadd h0, h0, h1, h2
; CHECK-SD-FP16-NEXT:    ret
;
; CHECK-GI-NOFP16-LABEL: fmul_f16:
; CHECK-GI-NOFP16:       // %bb.0: // %entry
; CHECK-GI-NOFP16-NEXT:    fcvt s0, h0
; CHECK-GI-NOFP16-NEXT:    fcvt s1, h1
; CHECK-GI-NOFP16-NEXT:    fmul s0, s0, s1
; CHECK-GI-NOFP16-NEXT:    fcvt s1, h2
; CHECK-GI-NOFP16-NEXT:    fcvt h0, s0
; CHECK-GI-NOFP16-NEXT:    fcvt s0, h0
; CHECK-GI-NOFP16-NEXT:    fadd s0, s0, s1
; CHECK-GI-NOFP16-NEXT:    fcvt h0, s0
; CHECK-GI-NOFP16-NEXT:    ret
;
; CHECK-GI-FP16-LABEL: fmul_f16:
; CHECK-GI-FP16:       // %bb.0: // %entry
; CHECK-GI-FP16-NEXT:    fmadd h0, h0, h1, h2
; CHECK-GI-FP16-NEXT:    ret
entry:
  %d = fmul fast half %a, %b
  %e = fadd fast half %d, %c
  ret half %e
}

define <2 x double> @fmul_v2f64(<2 x double> %a, <2 x double> %b, <2 x double> %c) {
; CHECK-SD-LABEL: fmul_v2f64:
; CHECK-SD:       // %bb.0: // %entry
; CHECK-SD-NEXT:    fmla v2.2d, v1.2d, v0.2d
; CHECK-SD-NEXT:    mov v0.16b, v2.16b
; CHECK-SD-NEXT:    ret
;
; CHECK-GI-LABEL: fmul_v2f64:
; CHECK-GI:       // %bb.0: // %entry
; CHECK-GI-NEXT:    fmla v2.2d, v0.2d, v1.2d
; CHECK-GI-NEXT:    mov v0.16b, v2.16b
; CHECK-GI-NEXT:    ret
entry:
  %d = fmul fast <2 x double> %a, %b
  %e = fadd fast <2 x double> %d, %c
  ret <2 x double> %e
}

define <3 x double> @fmul_v3f64(<3 x double> %a, <3 x double> %b, <3 x double> %c) {
; CHECK-SD-LABEL: fmul_v3f64:
; CHECK-SD:       // %bb.0: // %entry
; CHECK-SD-NEXT:    // kill: def $d6 killed $d6 def $q6
; CHECK-SD-NEXT:    // kill: def $d3 killed $d3 def $q3
; CHECK-SD-NEXT:    // kill: def $d0 killed $d0 def $q0
; CHECK-SD-NEXT:    // kill: def $d7 killed $d7 def $q7
; CHECK-SD-NEXT:    // kill: def $d4 killed $d4 def $q4
; CHECK-SD-NEXT:    // kill: def $d1 killed $d1 def $q1
; CHECK-SD-NEXT:    // kill: def $d5 killed $d5 def $q5
; CHECK-SD-NEXT:    // kill: def $d2 killed $d2 def $q2
; CHECK-SD-NEXT:    mov v3.d[1], v4.d[0]
; CHECK-SD-NEXT:    mov v0.d[1], v1.d[0]
; CHECK-SD-NEXT:    mov v6.d[1], v7.d[0]
; CHECK-SD-NEXT:    fmla v6.2d, v3.2d, v0.2d
; CHECK-SD-NEXT:    ldr d3, [sp]
; CHECK-SD-NEXT:    fmla v3.2d, v5.2d, v2.2d
; CHECK-SD-NEXT:    fmov d0, d6
; CHECK-SD-NEXT:    ext v1.16b, v6.16b, v6.16b, #8
; CHECK-SD-NEXT:    // kill: def $d1 killed $d1 killed $q1
; CHECK-SD-NEXT:    fmov d2, d3
; CHECK-SD-NEXT:    ret
;
; CHECK-GI-LABEL: fmul_v3f64:
; CHECK-GI:       // %bb.0: // %entry
; CHECK-GI-NEXT:    // kill: def $d0 killed $d0 def $q0
; CHECK-GI-NEXT:    // kill: def $d3 killed $d3 def $q3
; CHECK-GI-NEXT:    // kill: def $d6 killed $d6 def $q6
; CHECK-GI-NEXT:    // kill: def $d1 killed $d1 def $q1
; CHECK-GI-NEXT:    // kill: def $d4 killed $d4 def $q4
; CHECK-GI-NEXT:    // kill: def $d7 killed $d7 def $q7
; CHECK-GI-NEXT:    mov v0.d[1], v1.d[0]
; CHECK-GI-NEXT:    mov v3.d[1], v4.d[0]
; CHECK-GI-NEXT:    mov v6.d[1], v7.d[0]
; CHECK-GI-NEXT:    fmla v6.2d, v0.2d, v3.2d
; CHECK-GI-NEXT:    ldr d0, [sp]
; CHECK-GI-NEXT:    fmadd d2, d2, d5, d0
; CHECK-GI-NEXT:    mov d1, v6.d[1]
; CHECK-GI-NEXT:    fmov d0, d6
; CHECK-GI-NEXT:    ret
entry:
  %d = fmul fast <3 x double> %a, %b
  %e = fadd fast <3 x double> %d, %c
  ret <3 x double> %e
}

define <4 x double> @fmul_v4f64(<4 x double> %a, <4 x double> %b, <4 x double> %c) {
; CHECK-SD-LABEL: fmul_v4f64:
; CHECK-SD:       // %bb.0: // %entry
; CHECK-SD-NEXT:    fmla v4.2d, v2.2d, v0.2d
; CHECK-SD-NEXT:    fmla v5.2d, v3.2d, v1.2d
; CHECK-SD-NEXT:    mov v0.16b, v4.16b
; CHECK-SD-NEXT:    mov v1.16b, v5.16b
; CHECK-SD-NEXT:    ret
;
; CHECK-GI-LABEL: fmul_v4f64:
; CHECK-GI:       // %bb.0: // %entry
; CHECK-GI-NEXT:    fmla v4.2d, v0.2d, v2.2d
; CHECK-GI-NEXT:    fmla v5.2d, v1.2d, v3.2d
; CHECK-GI-NEXT:    mov v0.16b, v4.16b
; CHECK-GI-NEXT:    mov v1.16b, v5.16b
; CHECK-GI-NEXT:    ret
entry:
  %d = fmul fast <4 x double> %a, %b
  %e = fadd fast <4 x double> %d, %c
  ret <4 x double> %e
}

define <2 x float> @fmul_v2f32(<2 x float> %a, <2 x float> %b, <2 x float> %c) {
; CHECK-SD-LABEL: fmul_v2f32:
; CHECK-SD:       // %bb.0: // %entry
; CHECK-SD-NEXT:    fmla v2.2s, v1.2s, v0.2s
; CHECK-SD-NEXT:    fmov d0, d2
; CHECK-SD-NEXT:    ret
;
; CHECK-GI-LABEL: fmul_v2f32:
; CHECK-GI:       // %bb.0: // %entry
; CHECK-GI-NEXT:    fmla v2.2s, v0.2s, v1.2s
; CHECK-GI-NEXT:    fmov d0, d2
; CHECK-GI-NEXT:    ret
entry:
  %d = fmul fast <2 x float> %a, %b
  %e = fadd fast <2 x float> %d, %c
  ret <2 x float> %e
}

define <3 x float> @fmul_v3f32(<3 x float> %a, <3 x float> %b, <3 x float> %c) {
; CHECK-SD-LABEL: fmul_v3f32:
; CHECK-SD:       // %bb.0: // %entry
; CHECK-SD-NEXT:    fmla v2.4s, v1.4s, v0.4s
; CHECK-SD-NEXT:    mov v0.16b, v2.16b
; CHECK-SD-NEXT:    ret
;
; CHECK-GI-LABEL: fmul_v3f32:
; CHECK-GI:       // %bb.0: // %entry
; CHECK-GI-NEXT:    fmla v2.4s, v0.4s, v1.4s
; CHECK-GI-NEXT:    mov v0.16b, v2.16b
; CHECK-GI-NEXT:    ret
entry:
  %d = fmul fast <3 x float> %a, %b
  %e = fadd fast <3 x float> %d, %c
  ret <3 x float> %e
}

define <4 x float> @fmul_v4f32(<4 x float> %a, <4 x float> %b, <4 x float> %c) {
; CHECK-SD-LABEL: fmul_v4f32:
; CHECK-SD:       // %bb.0: // %entry
; CHECK-SD-NEXT:    fmla v2.4s, v1.4s, v0.4s
; CHECK-SD-NEXT:    mov v0.16b, v2.16b
; CHECK-SD-NEXT:    ret
;
; CHECK-GI-LABEL: fmul_v4f32:
; CHECK-GI:       // %bb.0: // %entry
; CHECK-GI-NEXT:    fmla v2.4s, v0.4s, v1.4s
; CHECK-GI-NEXT:    mov v0.16b, v2.16b
; CHECK-GI-NEXT:    ret
entry:
  %d = fmul fast <4 x float> %a, %b
  %e = fadd fast <4 x float> %d, %c
  ret <4 x float> %e
}

define <8 x float> @fmul_v8f32(<8 x float> %a, <8 x float> %b, <8 x float> %c) {
; CHECK-SD-LABEL: fmul_v8f32:
; CHECK-SD:       // %bb.0: // %entry
; CHECK-SD-NEXT:    fmla v4.4s, v2.4s, v0.4s
; CHECK-SD-NEXT:    fmla v5.4s, v3.4s, v1.4s
; CHECK-SD-NEXT:    mov v0.16b, v4.16b
; CHECK-SD-NEXT:    mov v1.16b, v5.16b
; CHECK-SD-NEXT:    ret
;
; CHECK-GI-LABEL: fmul_v8f32:
; CHECK-GI:       // %bb.0: // %entry
; CHECK-GI-NEXT:    fmla v4.4s, v0.4s, v2.4s
; CHECK-GI-NEXT:    fmla v5.4s, v1.4s, v3.4s
; CHECK-GI-NEXT:    mov v0.16b, v4.16b
; CHECK-GI-NEXT:    mov v1.16b, v5.16b
; CHECK-GI-NEXT:    ret
entry:
  %d = fmul fast <8 x float> %a, %b
  %e = fadd fast <8 x float> %d, %c
  ret <8 x float> %e
}

define <7 x half> @fmul_v7f16(<7 x half> %a, <7 x half> %b, <7 x half> %c) {
; CHECK-SD-NOFP16-LABEL: fmul_v7f16:
; CHECK-SD-NOFP16:       // %bb.0: // %entry
; CHECK-SD-NOFP16-NEXT:    fcvtl v3.4s, v1.4h
; CHECK-SD-NOFP16-NEXT:    fcvtl v4.4s, v0.4h
; CHECK-SD-NOFP16-NEXT:    fcvtl2 v1.4s, v1.8h
; CHECK-SD-NOFP16-NEXT:    fcvtl2 v0.4s, v0.8h
; CHECK-SD-NOFP16-NEXT:    fmul v3.4s, v4.4s, v3.4s
; CHECK-SD-NOFP16-NEXT:    fmul v0.4s, v0.4s, v1.4s
; CHECK-SD-NOFP16-NEXT:    fcvtn v1.4h, v3.4s
; CHECK-SD-NOFP16-NEXT:    fcvtl v3.4s, v2.4h
; CHECK-SD-NOFP16-NEXT:    fcvtl2 v2.4s, v2.8h
; CHECK-SD-NOFP16-NEXT:    fcvtn2 v1.8h, v0.4s
; CHECK-SD-NOFP16-NEXT:    fcvtl v0.4s, v1.4h
; CHECK-SD-NOFP16-NEXT:    fcvtl2 v1.4s, v1.8h
; CHECK-SD-NOFP16-NEXT:    fadd v0.4s, v0.4s, v3.4s
; CHECK-SD-NOFP16-NEXT:    fadd v1.4s, v1.4s, v2.4s
; CHECK-SD-NOFP16-NEXT:    fcvtn v0.4h, v0.4s
; CHECK-SD-NOFP16-NEXT:    fcvtn2 v0.8h, v1.4s
; CHECK-SD-NOFP16-NEXT:    ret
;
; CHECK-SD-FP16-LABEL: fmul_v7f16:
; CHECK-SD-FP16:       // %bb.0: // %entry
; CHECK-SD-FP16-NEXT:    fmla v2.8h, v1.8h, v0.8h
; CHECK-SD-FP16-NEXT:    mov v0.16b, v2.16b
; CHECK-SD-FP16-NEXT:    ret
;
; CHECK-GI-NOFP16-LABEL: fmul_v7f16:
; CHECK-GI-NOFP16:       // %bb.0: // %entry
; CHECK-GI-NOFP16-NEXT:    fcvtl v3.4s, v0.4h
; CHECK-GI-NOFP16-NEXT:    fcvtl v4.4s, v1.4h
; CHECK-GI-NOFP16-NEXT:    mov v5.h[0], v0.h[4]
; CHECK-GI-NOFP16-NEXT:    fmul v3.4s, v3.4s, v4.4s
; CHECK-GI-NOFP16-NEXT:    mov v4.h[0], v1.h[4]
; CHECK-GI-NOFP16-NEXT:    mov v5.h[1], v0.h[5]
; CHECK-GI-NOFP16-NEXT:    mov v4.h[1], v1.h[5]
; CHECK-GI-NOFP16-NEXT:    fcvtn v3.4h, v3.4s
; CHECK-GI-NOFP16-NEXT:    mov v5.h[2], v0.h[6]
; CHECK-GI-NOFP16-NEXT:    mov v4.h[2], v1.h[6]
; CHECK-GI-NOFP16-NEXT:    fcvtl v0.4s, v3.4h
; CHECK-GI-NOFP16-NEXT:    fcvtl v1.4s, v2.4h
; CHECK-GI-NOFP16-NEXT:    fcvtl v3.4s, v5.4h
; CHECK-GI-NOFP16-NEXT:    mov v5.h[0], v2.h[4]
; CHECK-GI-NOFP16-NEXT:    fcvtl v4.4s, v4.4h
; CHECK-GI-NOFP16-NEXT:    fadd v0.4s, v0.4s, v1.4s
; CHECK-GI-NOFP16-NEXT:    mov v5.h[1], v2.h[5]
; CHECK-GI-NOFP16-NEXT:    fmul v1.4s, v3.4s, v4.4s
; CHECK-GI-NOFP16-NEXT:    fcvtn v3.4h, v0.4s
; CHECK-GI-NOFP16-NEXT:    mov v5.h[2], v2.h[6]
; CHECK-GI-NOFP16-NEXT:    fcvtn v1.4h, v1.4s
; CHECK-GI-NOFP16-NEXT:    mov v0.h[0], v3.h[0]
; CHECK-GI-NOFP16-NEXT:    fcvtl v2.4s, v5.4h
; CHECK-GI-NOFP16-NEXT:    fcvtl v1.4s, v1.4h
; CHECK-GI-NOFP16-NEXT:    mov v0.h[1], v3.h[1]
; CHECK-GI-NOFP16-NEXT:    fadd v1.4s, v1.4s, v2.4s
; CHECK-GI-NOFP16-NEXT:    mov v0.h[2], v3.h[2]
; CHECK-GI-NOFP16-NEXT:    fcvtn v1.4h, v1.4s
; CHECK-GI-NOFP16-NEXT:    mov v0.h[3], v3.h[3]
; CHECK-GI-NOFP16-NEXT:    mov v0.h[4], v1.h[0]
; CHECK-GI-NOFP16-NEXT:    mov v0.h[5], v1.h[1]
; CHECK-GI-NOFP16-NEXT:    mov v0.h[6], v1.h[2]
; CHECK-GI-NOFP16-NEXT:    ret
;
; CHECK-GI-FP16-LABEL: fmul_v7f16:
; CHECK-GI-FP16:       // %bb.0: // %entry
; CHECK-GI-FP16-NEXT:    fmla v2.8h, v0.8h, v1.8h
; CHECK-GI-FP16-NEXT:    mov v0.16b, v2.16b
; CHECK-GI-FP16-NEXT:    ret
entry:
  %d = fmul fast <7 x half> %a, %b
  %e = fadd fast <7 x half> %d, %c
  ret <7 x half> %e
}

define <4 x half> @fmul_v4f16(<4 x half> %a, <4 x half> %b, <4 x half> %c) {
; CHECK-SD-NOFP16-LABEL: fmul_v4f16:
; CHECK-SD-NOFP16:       // %bb.0: // %entry
; CHECK-SD-NOFP16-NEXT:    fcvtl v1.4s, v1.4h
; CHECK-SD-NOFP16-NEXT:    fcvtl v0.4s, v0.4h
; CHECK-SD-NOFP16-NEXT:    fmul v0.4s, v0.4s, v1.4s
; CHECK-SD-NOFP16-NEXT:    fcvtl v1.4s, v2.4h
; CHECK-SD-NOFP16-NEXT:    fcvtn v0.4h, v0.4s
; CHECK-SD-NOFP16-NEXT:    fcvtl v0.4s, v0.4h
; CHECK-SD-NOFP16-NEXT:    fadd v0.4s, v0.4s, v1.4s
; CHECK-SD-NOFP16-NEXT:    fcvtn v0.4h, v0.4s
; CHECK-SD-NOFP16-NEXT:    ret
;
; CHECK-SD-FP16-LABEL: fmul_v4f16:
; CHECK-SD-FP16:       // %bb.0: // %entry
; CHECK-SD-FP16-NEXT:    fmla v2.4h, v1.4h, v0.4h
; CHECK-SD-FP16-NEXT:    fmov d0, d2
; CHECK-SD-FP16-NEXT:    ret
;
; CHECK-GI-NOFP16-LABEL: fmul_v4f16:
; CHECK-GI-NOFP16:       // %bb.0: // %entry
; CHECK-GI-NOFP16-NEXT:    fcvtl v0.4s, v0.4h
; CHECK-GI-NOFP16-NEXT:    fcvtl v1.4s, v1.4h
; CHECK-GI-NOFP16-NEXT:    fmul v0.4s, v0.4s, v1.4s
; CHECK-GI-NOFP16-NEXT:    fcvtl v1.4s, v2.4h
; CHECK-GI-NOFP16-NEXT:    fcvtn v0.4h, v0.4s
; CHECK-GI-NOFP16-NEXT:    fcvtl v0.4s, v0.4h
; CHECK-GI-NOFP16-NEXT:    fadd v0.4s, v0.4s, v1.4s
; CHECK-GI-NOFP16-NEXT:    fcvtn v0.4h, v0.4s
; CHECK-GI-NOFP16-NEXT:    ret
;
; CHECK-GI-FP16-LABEL: fmul_v4f16:
; CHECK-GI-FP16:       // %bb.0: // %entry
; CHECK-GI-FP16-NEXT:    fmla v2.4h, v0.4h, v1.4h
; CHECK-GI-FP16-NEXT:    fmov d0, d2
; CHECK-GI-FP16-NEXT:    ret
entry:
  %d = fmul fast <4 x half> %a, %b
  %e = fadd fast <4 x half> %d, %c
  ret <4 x half> %e
}

define <8 x half> @fmul_v8f16(<8 x half> %a, <8 x half> %b, <8 x half> %c) {
; CHECK-SD-NOFP16-LABEL: fmul_v8f16:
; CHECK-SD-NOFP16:       // %bb.0: // %entry
; CHECK-SD-NOFP16-NEXT:    fcvtl v3.4s, v1.4h
; CHECK-SD-NOFP16-NEXT:    fcvtl v4.4s, v0.4h
; CHECK-SD-NOFP16-NEXT:    fcvtl2 v1.4s, v1.8h
; CHECK-SD-NOFP16-NEXT:    fcvtl2 v0.4s, v0.8h
; CHECK-SD-NOFP16-NEXT:    fmul v3.4s, v4.4s, v3.4s
; CHECK-SD-NOFP16-NEXT:    fmul v0.4s, v0.4s, v1.4s
; CHECK-SD-NOFP16-NEXT:    fcvtn v1.4h, v3.4s
; CHECK-SD-NOFP16-NEXT:    fcvtl v3.4s, v2.4h
; CHECK-SD-NOFP16-NEXT:    fcvtl2 v2.4s, v2.8h
; CHECK-SD-NOFP16-NEXT:    fcvtn2 v1.8h, v0.4s
; CHECK-SD-NOFP16-NEXT:    fcvtl v0.4s, v1.4h
; CHECK-SD-NOFP16-NEXT:    fcvtl2 v1.4s, v1.8h
; CHECK-SD-NOFP16-NEXT:    fadd v0.4s, v0.4s, v3.4s
; CHECK-SD-NOFP16-NEXT:    fadd v1.4s, v1.4s, v2.4s
; CHECK-SD-NOFP16-NEXT:    fcvtn v0.4h, v0.4s
; CHECK-SD-NOFP16-NEXT:    fcvtn2 v0.8h, v1.4s
; CHECK-SD-NOFP16-NEXT:    ret
;
; CHECK-SD-FP16-LABEL: fmul_v8f16:
; CHECK-SD-FP16:       // %bb.0: // %entry
; CHECK-SD-FP16-NEXT:    fmla v2.8h, v1.8h, v0.8h
; CHECK-SD-FP16-NEXT:    mov v0.16b, v2.16b
; CHECK-SD-FP16-NEXT:    ret
;
; CHECK-GI-NOFP16-LABEL: fmul_v8f16:
; CHECK-GI-NOFP16:       // %bb.0: // %entry
; CHECK-GI-NOFP16-NEXT:    fcvtl v3.4s, v0.4h
; CHECK-GI-NOFP16-NEXT:    fcvtl v4.4s, v1.4h
; CHECK-GI-NOFP16-NEXT:    fcvtl2 v0.4s, v0.8h
; CHECK-GI-NOFP16-NEXT:    fcvtl2 v1.4s, v1.8h
; CHECK-GI-NOFP16-NEXT:    fmul v3.4s, v3.4s, v4.4s
; CHECK-GI-NOFP16-NEXT:    fmul v0.4s, v0.4s, v1.4s
; CHECK-GI-NOFP16-NEXT:    fcvtn v1.4h, v3.4s
; CHECK-GI-NOFP16-NEXT:    fcvtl v3.4s, v2.4h
; CHECK-GI-NOFP16-NEXT:    fcvtl2 v2.4s, v2.8h
; CHECK-GI-NOFP16-NEXT:    fcvtn v0.4h, v0.4s
; CHECK-GI-NOFP16-NEXT:    fcvtl v1.4s, v1.4h
; CHECK-GI-NOFP16-NEXT:    fcvtl v0.4s, v0.4h
; CHECK-GI-NOFP16-NEXT:    fadd v1.4s, v1.4s, v3.4s
; CHECK-GI-NOFP16-NEXT:    fadd v2.4s, v0.4s, v2.4s
; CHECK-GI-NOFP16-NEXT:    fcvtn v0.4h, v1.4s
; CHECK-GI-NOFP16-NEXT:    fcvtn2 v0.8h, v2.4s
; CHECK-GI-NOFP16-NEXT:    ret
;
; CHECK-GI-FP16-LABEL: fmul_v8f16:
; CHECK-GI-FP16:       // %bb.0: // %entry
; CHECK-GI-FP16-NEXT:    fmla v2.8h, v0.8h, v1.8h
; CHECK-GI-FP16-NEXT:    mov v0.16b, v2.16b
; CHECK-GI-FP16-NEXT:    ret
entry:
  %d = fmul fast <8 x half> %a, %b
  %e = fadd fast <8 x half> %d, %c
  ret <8 x half> %e
}

define <16 x half> @fmul_v16f16(<16 x half> %a, <16 x half> %b, <16 x half> %c) {
; CHECK-SD-NOFP16-LABEL: fmul_v16f16:
; CHECK-SD-NOFP16:       // %bb.0: // %entry
; CHECK-SD-NOFP16-NEXT:    fcvtl v6.4s, v2.4h
; CHECK-SD-NOFP16-NEXT:    fcvtl v7.4s, v0.4h
; CHECK-SD-NOFP16-NEXT:    fcvtl v16.4s, v3.4h
; CHECK-SD-NOFP16-NEXT:    fcvtl v17.4s, v1.4h
; CHECK-SD-NOFP16-NEXT:    fcvtl2 v2.4s, v2.8h
; CHECK-SD-NOFP16-NEXT:    fcvtl2 v0.4s, v0.8h
; CHECK-SD-NOFP16-NEXT:    fcvtl2 v3.4s, v3.8h
; CHECK-SD-NOFP16-NEXT:    fcvtl2 v1.4s, v1.8h
; CHECK-SD-NOFP16-NEXT:    fmul v6.4s, v7.4s, v6.4s
; CHECK-SD-NOFP16-NEXT:    fmul v7.4s, v17.4s, v16.4s
; CHECK-SD-NOFP16-NEXT:    fmul v0.4s, v0.4s, v2.4s
; CHECK-SD-NOFP16-NEXT:    fmul v1.4s, v1.4s, v3.4s
; CHECK-SD-NOFP16-NEXT:    fcvtn v2.4h, v6.4s
; CHECK-SD-NOFP16-NEXT:    fcvtl v6.4s, v5.4h
; CHECK-SD-NOFP16-NEXT:    fcvtn v3.4h, v7.4s
; CHECK-SD-NOFP16-NEXT:    fcvtn2 v2.8h, v0.4s
; CHECK-SD-NOFP16-NEXT:    fcvtl v0.4s, v4.4h
; CHECK-SD-NOFP16-NEXT:    fcvtl2 v4.4s, v4.8h
; CHECK-SD-NOFP16-NEXT:    fcvtn2 v3.8h, v1.4s
; CHECK-SD-NOFP16-NEXT:    fcvtl v1.4s, v2.4h
; CHECK-SD-NOFP16-NEXT:    fcvtl2 v2.4s, v2.8h
; CHECK-SD-NOFP16-NEXT:    fcvtl v7.4s, v3.4h
; CHECK-SD-NOFP16-NEXT:    fadd v0.4s, v1.4s, v0.4s
; CHECK-SD-NOFP16-NEXT:    fcvtl2 v1.4s, v3.8h
; CHECK-SD-NOFP16-NEXT:    fcvtl2 v3.4s, v5.8h
; CHECK-SD-NOFP16-NEXT:    fadd v5.4s, v7.4s, v6.4s
; CHECK-SD-NOFP16-NEXT:    fadd v2.4s, v2.4s, v4.4s
; CHECK-SD-NOFP16-NEXT:    fcvtn v0.4h, v0.4s
; CHECK-SD-NOFP16-NEXT:    fadd v3.4s, v1.4s, v3.4s
; CHECK-SD-NOFP16-NEXT:    fcvtn v1.4h, v5.4s
; CHECK-SD-NOFP16-NEXT:    fcvtn2 v0.8h, v2.4s
; CHECK-SD-NOFP16-NEXT:    fcvtn2 v1.8h, v3.4s
; CHECK-SD-NOFP16-NEXT:    ret
;
; CHECK-SD-FP16-LABEL: fmul_v16f16:
; CHECK-SD-FP16:       // %bb.0: // %entry
; CHECK-SD-FP16-NEXT:    fmla v4.8h, v2.8h, v0.8h
; CHECK-SD-FP16-NEXT:    fmla v5.8h, v3.8h, v1.8h
; CHECK-SD-FP16-NEXT:    mov v0.16b, v4.16b
; CHECK-SD-FP16-NEXT:    mov v1.16b, v5.16b
; CHECK-SD-FP16-NEXT:    ret
;
; CHECK-GI-NOFP16-LABEL: fmul_v16f16:
; CHECK-GI-NOFP16:       // %bb.0: // %entry
; CHECK-GI-NOFP16-NEXT:    fcvtl v6.4s, v0.4h
; CHECK-GI-NOFP16-NEXT:    fcvtl v7.4s, v2.4h
; CHECK-GI-NOFP16-NEXT:    fcvtl v16.4s, v1.4h
; CHECK-GI-NOFP16-NEXT:    fcvtl v17.4s, v3.4h
; CHECK-GI-NOFP16-NEXT:    fcvtl2 v0.4s, v0.8h
; CHECK-GI-NOFP16-NEXT:    fcvtl2 v1.4s, v1.8h
; CHECK-GI-NOFP16-NEXT:    fcvtl2 v2.4s, v2.8h
; CHECK-GI-NOFP16-NEXT:    fcvtl2 v3.4s, v3.8h
; CHECK-GI-NOFP16-NEXT:    fmul v6.4s, v6.4s, v7.4s
; CHECK-GI-NOFP16-NEXT:    fmul v7.4s, v16.4s, v17.4s
; CHECK-GI-NOFP16-NEXT:    fmul v0.4s, v0.4s, v2.4s
; CHECK-GI-NOFP16-NEXT:    fmul v1.4s, v1.4s, v3.4s
; CHECK-GI-NOFP16-NEXT:    fcvtn v2.4h, v6.4s
; CHECK-GI-NOFP16-NEXT:    fcvtl v6.4s, v4.4h
; CHECK-GI-NOFP16-NEXT:    fcvtl2 v4.4s, v4.8h
; CHECK-GI-NOFP16-NEXT:    fcvtn v3.4h, v7.4s
; CHECK-GI-NOFP16-NEXT:    fcvtl v7.4s, v5.4h
; CHECK-GI-NOFP16-NEXT:    fcvtl2 v5.4s, v5.8h
; CHECK-GI-NOFP16-NEXT:    fcvtn v0.4h, v0.4s
; CHECK-GI-NOFP16-NEXT:    fcvtn v1.4h, v1.4s
; CHECK-GI-NOFP16-NEXT:    fcvtl v2.4s, v2.4h
; CHECK-GI-NOFP16-NEXT:    fcvtl v3.4s, v3.4h
; CHECK-GI-NOFP16-NEXT:    fcvtl v0.4s, v0.4h
; CHECK-GI-NOFP16-NEXT:    fcvtl v1.4s, v1.4h
; CHECK-GI-NOFP16-NEXT:    fadd v2.4s, v2.4s, v6.4s
; CHECK-GI-NOFP16-NEXT:    fadd v3.4s, v3.4s, v7.4s
; CHECK-GI-NOFP16-NEXT:    fadd v4.4s, v0.4s, v4.4s
; CHECK-GI-NOFP16-NEXT:    fadd v5.4s, v1.4s, v5.4s
; CHECK-GI-NOFP16-NEXT:    fcvtn v0.4h, v2.4s
; CHECK-GI-NOFP16-NEXT:    fcvtn v1.4h, v3.4s
; CHECK-GI-NOFP16-NEXT:    fcvtn2 v0.8h, v4.4s
; CHECK-GI-NOFP16-NEXT:    fcvtn2 v1.8h, v5.4s
; CHECK-GI-NOFP16-NEXT:    ret
;
; CHECK-GI-FP16-LABEL: fmul_v16f16:
; CHECK-GI-FP16:       // %bb.0: // %entry
; CHECK-GI-FP16-NEXT:    fmla v4.8h, v0.8h, v2.8h
; CHECK-GI-FP16-NEXT:    fmla v5.8h, v1.8h, v3.8h
; CHECK-GI-FP16-NEXT:    mov v0.16b, v4.16b
; CHECK-GI-FP16-NEXT:    mov v1.16b, v5.16b
; CHECK-GI-FP16-NEXT:    ret
entry:
  %d = fmul fast <16 x half> %a, %b
  %e = fadd fast <16 x half> %d, %c
  ret <16 x half> %e
}

declare <16 x half> @llvm.fma.v16f16(<16 x half>, <16 x half>, <16 x half>)
declare <16 x half> @llvm.fmuladd.v16f16(<16 x half>, <16 x half>, <16 x half>)
declare <2 x double> @llvm.fma.v2f64(<2 x double>, <2 x double>, <2 x double>)
declare <2 x double> @llvm.fmuladd.v2f64(<2 x double>, <2 x double>, <2 x double>)
declare <2 x float> @llvm.fma.v2f32(<2 x float>, <2 x float>, <2 x float>)
declare <2 x float> @llvm.fmuladd.v2f32(<2 x float>, <2 x float>, <2 x float>)
declare <3 x double> @llvm.fma.v3f64(<3 x double>, <3 x double>, <3 x double>)
declare <3 x double> @llvm.fmuladd.v3f64(<3 x double>, <3 x double>, <3 x double>)
declare <3 x float> @llvm.fma.v3f32(<3 x float>, <3 x float>, <3 x float>)
declare <3 x float> @llvm.fmuladd.v3f32(<3 x float>, <3 x float>, <3 x float>)
declare <4 x double> @llvm.fma.v4f64(<4 x double>, <4 x double>, <4 x double>)
declare <4 x double> @llvm.fmuladd.v4f64(<4 x double>, <4 x double>, <4 x double>)
declare <4 x float> @llvm.fma.v4f32(<4 x float>, <4 x float>, <4 x float>)
declare <4 x float> @llvm.fmuladd.v4f32(<4 x float>, <4 x float>, <4 x float>)
declare <4 x half> @llvm.fma.v4f16(<4 x half>, <4 x half>, <4 x half>)
declare <4 x half> @llvm.fmuladd.v4f16(<4 x half>, <4 x half>, <4 x half>)
declare <7 x half> @llvm.fma.v7f16(<7 x half>, <7 x half>, <7 x half>)
declare <7 x half> @llvm.fmuladd.v7f16(<7 x half>, <7 x half>, <7 x half>)
declare <8 x float> @llvm.fma.v8f32(<8 x float>, <8 x float>, <8 x float>)
declare <8 x float> @llvm.fmuladd.v8f32(<8 x float>, <8 x float>, <8 x float>)
declare <8 x half> @llvm.fma.v8f16(<8 x half>, <8 x half>, <8 x half>)
declare <8 x half> @llvm.fmuladd.v8f16(<8 x half>, <8 x half>, <8 x half>)
declare double @llvm.fma.f64(double, double, double)
declare double @llvm.fmuladd.f64(double, double, double)
declare float @llvm.fma.f32(float, float, float)
declare float @llvm.fmuladd.f32(float, float, float)
declare half @llvm.fma.f16(half, half, half)
declare half @llvm.fmuladd.f16(half, half, half)