llvm-project/llvm/test/CodeGen/X86/recip-fastmath2.ll

; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse2 -print-schedule      | FileCheck %s --check-prefix=CHECK --check-prefix=SSE --check-prefix=SSE-RECIP
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx -print-schedule       | FileCheck %s --check-prefix=CHECK --check-prefix=AVX --check-prefix=AVX-RECIP
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx,+fma -print-schedule  | FileCheck %s --check-prefix=CHECK --check-prefix=AVX --check-prefix=FMA-RECIP
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=bdver2 -print-schedule      | FileCheck %s --check-prefix=CHECK --check-prefix=AVX --check-prefix=BDVER2
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=btver2 -print-schedule      | FileCheck %s --check-prefix=CHECK --check-prefix=AVX --check-prefix=BTVER2
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=sandybridge -print-schedule | FileCheck %s --check-prefix=CHECK --check-prefix=AVX --check-prefix=SANDY
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=haswell -print-schedule     | FileCheck %s --check-prefix=CHECK --check-prefix=AVX --check-prefix=HASWELL
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=haswell -print-schedule -mattr=-fma | FileCheck %s --check-prefix=CHECK --check-prefix=AVX --check-prefix=HASWELL-NO-FMA
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=knl -print-schedule         | FileCheck %s --check-prefix=CHECK --check-prefix=AVX --check-prefix=AVX512 --check-prefix=KNL
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=skx -print-schedule         | FileCheck %s --check-prefix=CHECK --check-prefix=AVX --check-prefix=AVX512 --check-prefix=SKX

; It's the extra tests coverage for recip as discussed on D26855.

define float @f32_no_step_2(float %x) #3 {
; SSE-LABEL: f32_no_step_2:
; SSE:       # %bb.0:
; SSE-NEXT:    rcpss %xmm0, %xmm0
; SSE-NEXT:    mulss {{.*}}(%rip), %xmm0
; SSE-NEXT:    retq
;
; AVX-RECIP-LABEL: f32_no_step_2:
; AVX-RECIP:       # %bb.0:
; AVX-RECIP-NEXT:    vrcpss %xmm0, %xmm0, %xmm0
; AVX-RECIP-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm0
; AVX-RECIP-NEXT:    retq
;
; FMA-RECIP-LABEL: f32_no_step_2:
; FMA-RECIP:       # %bb.0:
; FMA-RECIP-NEXT:    vrcpss %xmm0, %xmm0, %xmm0
; FMA-RECIP-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm0
; FMA-RECIP-NEXT:    retq
;
; BDVER2-LABEL: f32_no_step_2:
; BDVER2:       # %bb.0:
; BDVER2-NEXT:    vrcpss %xmm0, %xmm0, %xmm0 # sched: [5:1.00]
; BDVER2-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [10:1.00]
; BDVER2-NEXT:    retq # sched: [5:1.00]
;
; BTVER2-LABEL: f32_no_step_2:
; BTVER2:       # %bb.0:
; BTVER2-NEXT:    vrcpss %xmm0, %xmm0, %xmm0 # sched: [2:1.00]
; BTVER2-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [7:1.00]
; BTVER2-NEXT:    retq # sched: [4:1.00]
;
; SANDY-LABEL: f32_no_step_2:
; SANDY:       # %bb.0:
; SANDY-NEXT:    vrcpss %xmm0, %xmm0, %xmm0 # sched: [5:1.00]
; SANDY-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [11:1.00]
; SANDY-NEXT:    retq # sched: [1:1.00]
;
; HASWELL-LABEL: f32_no_step_2:
; HASWELL:       # %bb.0:
; HASWELL-NEXT:    vrcpss %xmm0, %xmm0, %xmm0 # sched: [5:1.00]
; HASWELL-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [10:0.50]
; HASWELL-NEXT:    retq # sched: [7:1.00]
;
; HASWELL-NO-FMA-LABEL: f32_no_step_2:
; HASWELL-NO-FMA:       # %bb.0:
; HASWELL-NO-FMA-NEXT:    vrcpss %xmm0, %xmm0, %xmm0 # sched: [5:1.00]
; HASWELL-NO-FMA-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [10:0.50]
; HASWELL-NO-FMA-NEXT:    retq # sched: [7:1.00]
;
; KNL-LABEL: f32_no_step_2:
; KNL:       # %bb.0:
; KNL-NEXT:    vrcpss %xmm0, %xmm0, %xmm0 # sched: [5:1.00]
; KNL-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [10:0.50]
; KNL-NEXT:    retq # sched: [7:1.00]
;
; SKX-LABEL: f32_no_step_2:
; SKX:       # %bb.0:
; SKX-NEXT:    vrcpss %xmm0, %xmm0, %xmm0 # sched: [4:1.00]
; SKX-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [9:0.50]
; SKX-NEXT:    retq # sched: [7:1.00]
  %div = fdiv fast float 1234.0, %x
  ret float %div
}

define float @f32_one_step_2(float %x) #1 {
; SSE-LABEL: f32_one_step_2:
; SSE:       # %bb.0:
; SSE-NEXT:    rcpss %xmm0, %xmm2
; SSE-NEXT:    mulss %xmm2, %xmm0
; SSE-NEXT:    movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; SSE-NEXT:    subss %xmm0, %xmm1
; SSE-NEXT:    mulss %xmm2, %xmm1
; SSE-NEXT:    addss %xmm2, %xmm1
; SSE-NEXT:    mulss {{.*}}(%rip), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm0
; SSE-NEXT:    retq
;
; AVX-RECIP-LABEL: f32_one_step_2:
; AVX-RECIP:       # %bb.0:
; AVX-RECIP-NEXT:    vrcpss %xmm0, %xmm0, %xmm1
; AVX-RECIP-NEXT:    vmulss %xmm1, %xmm0, %xmm0
; AVX-RECIP-NEXT:    vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero
; AVX-RECIP-NEXT:    vsubss %xmm0, %xmm2, %xmm0
; AVX-RECIP-NEXT:    vmulss %xmm0, %xmm1, %xmm0
; AVX-RECIP-NEXT:    vaddss %xmm0, %xmm1, %xmm0
; AVX-RECIP-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm0
; AVX-RECIP-NEXT:    retq
;
; FMA-RECIP-LABEL: f32_one_step_2:
; FMA-RECIP:       # %bb.0:
; FMA-RECIP-NEXT:    vrcpss %xmm0, %xmm0, %xmm1
; FMA-RECIP-NEXT:    vfnmadd213ss {{.*#+}} xmm0 = -(xmm1 * xmm0) + mem
; FMA-RECIP-NEXT:    vfmadd132ss {{.*#+}} xmm0 = (xmm0 * xmm1) + xmm1
; FMA-RECIP-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm0
; FMA-RECIP-NEXT:    retq
;
; BDVER2-LABEL: f32_one_step_2:
; BDVER2:       # %bb.0:
; BDVER2-NEXT:    vrcpss %xmm0, %xmm0, %xmm1 # sched: [5:1.00]
; BDVER2-NEXT:    vfnmaddss {{.*}}(%rip), %xmm1, %xmm0, %xmm0 # sched: [10:0.50]
; BDVER2-NEXT:    vfmaddss %xmm1, %xmm0, %xmm1, %xmm0 # sched: [5:0.50]
; BDVER2-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [10:1.00]
; BDVER2-NEXT:    retq # sched: [5:1.00]
;
; BTVER2-LABEL: f32_one_step_2:
; BTVER2:       # %bb.0:
; BTVER2-NEXT:    vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero sched: [5:1.00]
; BTVER2-NEXT:    vrcpss %xmm0, %xmm0, %xmm1 # sched: [2:1.00]
; BTVER2-NEXT:    vmulss %xmm1, %xmm0, %xmm0 # sched: [2:1.00]
; BTVER2-NEXT:    vsubss %xmm0, %xmm2, %xmm0 # sched: [3:1.00]
; BTVER2-NEXT:    vmulss %xmm0, %xmm1, %xmm0 # sched: [2:1.00]
; BTVER2-NEXT:    vaddss %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
; BTVER2-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [7:1.00]
; BTVER2-NEXT:    retq # sched: [4:1.00]
;
; SANDY-LABEL: f32_one_step_2:
; SANDY:       # %bb.0:
; SANDY-NEXT:    vrcpss %xmm0, %xmm0, %xmm1 # sched: [5:1.00]
; SANDY-NEXT:    vmulss %xmm1, %xmm0, %xmm0 # sched: [5:1.00]
; SANDY-NEXT:    vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero sched: [6:0.50]
; SANDY-NEXT:    vsubss %xmm0, %xmm2, %xmm0 # sched: [3:1.00]
; SANDY-NEXT:    vmulss %xmm0, %xmm1, %xmm0 # sched: [5:1.00]
; SANDY-NEXT:    vaddss %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
; SANDY-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [11:1.00]
; SANDY-NEXT:    retq # sched: [1:1.00]
;
; HASWELL-LABEL: f32_one_step_2:
; HASWELL:       # %bb.0:
; HASWELL-NEXT:    vrcpss %xmm0, %xmm0, %xmm1 # sched: [5:1.00]
; HASWELL-NEXT:    vfnmadd213ss {{.*#+}} xmm0 = -(xmm1 * xmm0) + mem sched: [10:0.50]
; HASWELL-NEXT:    vfmadd132ss {{.*#+}} xmm0 = (xmm0 * xmm1) + xmm1 sched: [5:0.50]
; HASWELL-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [10:0.50]
; HASWELL-NEXT:    retq # sched: [7:1.00]
;
; HASWELL-NO-FMA-LABEL: f32_one_step_2:
; HASWELL-NO-FMA:       # %bb.0:
; HASWELL-NO-FMA-NEXT:    vrcpss %xmm0, %xmm0, %xmm1 # sched: [5:1.00]
; HASWELL-NO-FMA-NEXT:    vmulss %xmm1, %xmm0, %xmm0 # sched: [5:0.50]
; HASWELL-NO-FMA-NEXT:    vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero sched: [5:0.50]
; HASWELL-NO-FMA-NEXT:    vsubss %xmm0, %xmm2, %xmm0 # sched: [3:1.00]
; HASWELL-NO-FMA-NEXT:    vmulss %xmm0, %xmm1, %xmm0 # sched: [5:0.50]
; HASWELL-NO-FMA-NEXT:    vaddss %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
; HASWELL-NO-FMA-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [10:0.50]
; HASWELL-NO-FMA-NEXT:    retq # sched: [7:1.00]
;
; KNL-LABEL: f32_one_step_2:
; KNL:       # %bb.0:
; KNL-NEXT:    vrcpss %xmm0, %xmm0, %xmm1 # sched: [5:1.00]
; KNL-NEXT:    vfnmadd213ss {{.*#+}} xmm0 = -(xmm1 * xmm0) + mem sched: [10:0.50]
; KNL-NEXT:    vfmadd132ss {{.*#+}} xmm0 = (xmm0 * xmm1) + xmm1 sched: [5:0.50]
; KNL-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [10:0.50]
; KNL-NEXT:    retq # sched: [7:1.00]
;
; SKX-LABEL: f32_one_step_2:
; SKX:       # %bb.0:
; SKX-NEXT:    vrcpss %xmm0, %xmm0, %xmm1 # sched: [4:1.00]
; SKX-NEXT:    vfnmadd213ss {{.*#+}} xmm0 = -(xmm1 * xmm0) + mem sched: [9:0.50]
; SKX-NEXT:    vfmadd132ss {{.*#+}} xmm0 = (xmm0 * xmm1) + xmm1 sched: [4:0.50]
; SKX-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [9:0.50]
; SKX-NEXT:    retq # sched: [7:1.00]
  %div = fdiv fast float 3456.0, %x
  ret float %div
}

define float @f32_one_step_2_divs(float %x) #1 {
; SSE-LABEL: f32_one_step_2_divs:
; SSE:       # %bb.0:
; SSE-NEXT:    rcpss %xmm0, %xmm1
; SSE-NEXT:    mulss %xmm1, %xmm0
; SSE-NEXT:    movss {{.*#+}} xmm2 = mem[0],zero,zero,zero
; SSE-NEXT:    subss %xmm0, %xmm2
; SSE-NEXT:    mulss %xmm1, %xmm2
; SSE-NEXT:    addss %xmm1, %xmm2
; SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; SSE-NEXT:    mulss %xmm2, %xmm0
; SSE-NEXT:    mulss %xmm2, %xmm0
; SSE-NEXT:    retq
;
; AVX-RECIP-LABEL: f32_one_step_2_divs:
; AVX-RECIP:       # %bb.0:
; AVX-RECIP-NEXT:    vrcpss %xmm0, %xmm0, %xmm1
; AVX-RECIP-NEXT:    vmulss %xmm1, %xmm0, %xmm0
; AVX-RECIP-NEXT:    vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero
; AVX-RECIP-NEXT:    vsubss %xmm0, %xmm2, %xmm0
; AVX-RECIP-NEXT:    vmulss %xmm0, %xmm1, %xmm0
; AVX-RECIP-NEXT:    vaddss %xmm0, %xmm1, %xmm0
; AVX-RECIP-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm1
; AVX-RECIP-NEXT:    vmulss %xmm0, %xmm1, %xmm0
; AVX-RECIP-NEXT:    retq
;
; FMA-RECIP-LABEL: f32_one_step_2_divs:
; FMA-RECIP:       # %bb.0:
; FMA-RECIP-NEXT:    vrcpss %xmm0, %xmm0, %xmm1
; FMA-RECIP-NEXT:    vfnmadd213ss {{.*#+}} xmm0 = -(xmm1 * xmm0) + mem
; FMA-RECIP-NEXT:    vfmadd132ss {{.*#+}} xmm0 = (xmm0 * xmm1) + xmm1
; FMA-RECIP-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm1
; FMA-RECIP-NEXT:    vmulss %xmm0, %xmm1, %xmm0
; FMA-RECIP-NEXT:    retq
;
; BDVER2-LABEL: f32_one_step_2_divs:
; BDVER2:       # %bb.0:
; BDVER2-NEXT:    vrcpss %xmm0, %xmm0, %xmm1 # sched: [5:1.00]
; BDVER2-NEXT:    vfnmaddss {{.*}}(%rip), %xmm1, %xmm0, %xmm0 # sched: [10:0.50]
; BDVER2-NEXT:    vfmaddss %xmm1, %xmm0, %xmm1, %xmm0 # sched: [5:0.50]
; BDVER2-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm1 # sched: [10:1.00]
; BDVER2-NEXT:    vmulss %xmm0, %xmm1, %xmm0 # sched: [5:1.00]
; BDVER2-NEXT:    retq # sched: [5:1.00]
;
; BTVER2-LABEL: f32_one_step_2_divs:
; BTVER2:       # %bb.0:
; BTVER2-NEXT:    vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero sched: [5:1.00]
; BTVER2-NEXT:    vrcpss %xmm0, %xmm0, %xmm1 # sched: [2:1.00]
; BTVER2-NEXT:    vmulss %xmm1, %xmm0, %xmm0 # sched: [2:1.00]
; BTVER2-NEXT:    vsubss %xmm0, %xmm2, %xmm0 # sched: [3:1.00]
; BTVER2-NEXT:    vmulss %xmm0, %xmm1, %xmm0 # sched: [2:1.00]
; BTVER2-NEXT:    vaddss %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
; BTVER2-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm1 # sched: [7:1.00]
; BTVER2-NEXT:    vmulss %xmm0, %xmm1, %xmm0 # sched: [2:1.00]
; BTVER2-NEXT:    retq # sched: [4:1.00]
;
; SANDY-LABEL: f32_one_step_2_divs:
; SANDY:       # %bb.0:
; SANDY-NEXT:    vrcpss %xmm0, %xmm0, %xmm1 # sched: [5:1.00]
; SANDY-NEXT:    vmulss %xmm1, %xmm0, %xmm0 # sched: [5:1.00]
; SANDY-NEXT:    vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero sched: [6:0.50]
; SANDY-NEXT:    vsubss %xmm0, %xmm2, %xmm0 # sched: [3:1.00]
; SANDY-NEXT:    vmulss %xmm0, %xmm1, %xmm0 # sched: [5:1.00]
; SANDY-NEXT:    vaddss %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
; SANDY-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm1 # sched: [11:1.00]
; SANDY-NEXT:    vmulss %xmm0, %xmm1, %xmm0 # sched: [5:1.00]
; SANDY-NEXT:    retq # sched: [1:1.00]
;
; HASWELL-LABEL: f32_one_step_2_divs:
; HASWELL:       # %bb.0:
; HASWELL-NEXT:    vrcpss %xmm0, %xmm0, %xmm1 # sched: [5:1.00]
; HASWELL-NEXT:    vfnmadd213ss {{.*#+}} xmm0 = -(xmm1 * xmm0) + mem sched: [10:0.50]
; HASWELL-NEXT:    vfmadd132ss {{.*#+}} xmm0 = (xmm0 * xmm1) + xmm1 sched: [5:0.50]
; HASWELL-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm1 # sched: [10:0.50]
; HASWELL-NEXT:    vmulss %xmm0, %xmm1, %xmm0 # sched: [5:0.50]
; HASWELL-NEXT:    retq # sched: [7:1.00]
;
; HASWELL-NO-FMA-LABEL: f32_one_step_2_divs:
; HASWELL-NO-FMA:       # %bb.0:
; HASWELL-NO-FMA-NEXT:    vrcpss %xmm0, %xmm0, %xmm1 # sched: [5:1.00]
; HASWELL-NO-FMA-NEXT:    vmulss %xmm1, %xmm0, %xmm0 # sched: [5:0.50]
; HASWELL-NO-FMA-NEXT:    vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero sched: [5:0.50]
; HASWELL-NO-FMA-NEXT:    vsubss %xmm0, %xmm2, %xmm0 # sched: [3:1.00]
; HASWELL-NO-FMA-NEXT:    vmulss %xmm0, %xmm1, %xmm0 # sched: [5:0.50]
; HASWELL-NO-FMA-NEXT:    vaddss %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
; HASWELL-NO-FMA-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm1 # sched: [10:0.50]
; HASWELL-NO-FMA-NEXT:    vmulss %xmm0, %xmm1, %xmm0 # sched: [5:0.50]
; HASWELL-NO-FMA-NEXT:    retq # sched: [7:1.00]
;
; KNL-LABEL: f32_one_step_2_divs:
; KNL:       # %bb.0:
; KNL-NEXT:    vrcpss %xmm0, %xmm0, %xmm1 # sched: [5:1.00]
; KNL-NEXT:    vfnmadd213ss {{.*#+}} xmm0 = -(xmm1 * xmm0) + mem sched: [10:0.50]
; KNL-NEXT:    vfmadd132ss {{.*#+}} xmm0 = (xmm0 * xmm1) + xmm1 sched: [5:0.50]
; KNL-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm1 # sched: [10:0.50]
; KNL-NEXT:    vmulss %xmm0, %xmm1, %xmm0 # sched: [5:0.50]
; KNL-NEXT:    retq # sched: [7:1.00]
;
; SKX-LABEL: f32_one_step_2_divs:
; SKX:       # %bb.0:
; SKX-NEXT:    vrcpss %xmm0, %xmm0, %xmm1 # sched: [4:1.00]
; SKX-NEXT:    vfnmadd213ss {{.*#+}} xmm0 = -(xmm1 * xmm0) + mem sched: [9:0.50]
; SKX-NEXT:    vfmadd132ss {{.*#+}} xmm0 = (xmm0 * xmm1) + xmm1 sched: [4:0.50]
; SKX-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm1 # sched: [9:0.50]
; SKX-NEXT:    vmulss %xmm0, %xmm1, %xmm0 # sched: [4:0.50]
; SKX-NEXT:    retq # sched: [7:1.00]
  %div = fdiv fast float 3456.0, %x
  %div2 = fdiv fast float %div, %x
  ret float %div2
}

define float @f32_two_step_2(float %x) #2 {
; SSE-LABEL: f32_two_step_2:
; SSE:       # %bb.0:
; SSE-NEXT:    rcpss %xmm0, %xmm2
; SSE-NEXT:    movaps %xmm0, %xmm3
; SSE-NEXT:    mulss %xmm2, %xmm3
; SSE-NEXT:    movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; SSE-NEXT:    movaps %xmm1, %xmm4
; SSE-NEXT:    subss %xmm3, %xmm4
; SSE-NEXT:    mulss %xmm2, %xmm4
; SSE-NEXT:    addss %xmm2, %xmm4
; SSE-NEXT:    mulss %xmm4, %xmm0
; SSE-NEXT:    subss %xmm0, %xmm1
; SSE-NEXT:    mulss %xmm4, %xmm1
; SSE-NEXT:    addss %xmm4, %xmm1
; SSE-NEXT:    mulss {{.*}}(%rip), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm0
; SSE-NEXT:    retq
;
; AVX-RECIP-LABEL: f32_two_step_2:
; AVX-RECIP:       # %bb.0:
; AVX-RECIP-NEXT:    vrcpss %xmm0, %xmm0, %xmm1
; AVX-RECIP-NEXT:    vmulss %xmm1, %xmm0, %xmm2
; AVX-RECIP-NEXT:    vmovss {{.*#+}} xmm3 = mem[0],zero,zero,zero
; AVX-RECIP-NEXT:    vsubss %xmm2, %xmm3, %xmm2
; AVX-RECIP-NEXT:    vmulss %xmm2, %xmm1, %xmm2
; AVX-RECIP-NEXT:    vaddss %xmm2, %xmm1, %xmm1
; AVX-RECIP-NEXT:    vmulss %xmm1, %xmm0, %xmm0
; AVX-RECIP-NEXT:    vsubss %xmm0, %xmm3, %xmm0
; AVX-RECIP-NEXT:    vmulss %xmm0, %xmm1, %xmm0
; AVX-RECIP-NEXT:    vaddss %xmm0, %xmm1, %xmm0
; AVX-RECIP-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm0
; AVX-RECIP-NEXT:    retq
;
; FMA-RECIP-LABEL: f32_two_step_2:
; FMA-RECIP:       # %bb.0:
; FMA-RECIP-NEXT:    vrcpss %xmm0, %xmm0, %xmm1
; FMA-RECIP-NEXT:    vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero
; FMA-RECIP-NEXT:    vmovaps %xmm1, %xmm3
; FMA-RECIP-NEXT:    vfnmadd213ss {{.*#+}} xmm3 = -(xmm0 * xmm3) + xmm2
; FMA-RECIP-NEXT:    vfmadd132ss {{.*#+}} xmm3 = (xmm3 * xmm1) + xmm1
; FMA-RECIP-NEXT:    vfnmadd213ss {{.*#+}} xmm0 = -(xmm3 * xmm0) + xmm2
; FMA-RECIP-NEXT:    vfmadd132ss {{.*#+}} xmm0 = (xmm0 * xmm3) + xmm3
; FMA-RECIP-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm0
; FMA-RECIP-NEXT:    retq
;
; BDVER2-LABEL: f32_two_step_2:
; BDVER2:       # %bb.0:
; BDVER2-NEXT:    vrcpss %xmm0, %xmm0, %xmm1 # sched: [5:1.00]
; BDVER2-NEXT:    vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero sched: [5:0.50]
; BDVER2-NEXT:    vfnmaddss %xmm2, %xmm1, %xmm0, %xmm3 # sched: [5:0.50]
; BDVER2-NEXT:    vfmaddss %xmm1, %xmm3, %xmm1, %xmm1 # sched: [5:0.50]
; BDVER2-NEXT:    vfnmaddss %xmm2, %xmm1, %xmm0, %xmm0 # sched: [5:0.50]
; BDVER2-NEXT:    vfmaddss %xmm1, %xmm0, %xmm1, %xmm0 # sched: [5:0.50]
; BDVER2-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [10:1.00]
; BDVER2-NEXT:    retq # sched: [5:1.00]
;
; BTVER2-LABEL: f32_two_step_2:
; BTVER2:       # %bb.0:
; BTVER2-NEXT:    vmovss {{.*#+}} xmm3 = mem[0],zero,zero,zero sched: [5:1.00]
; BTVER2-NEXT:    vrcpss %xmm0, %xmm0, %xmm1 # sched: [2:1.00]
; BTVER2-NEXT:    vmulss %xmm1, %xmm0, %xmm2 # sched: [2:1.00]
; BTVER2-NEXT:    vsubss %xmm2, %xmm3, %xmm2 # sched: [3:1.00]
; BTVER2-NEXT:    vmulss %xmm2, %xmm1, %xmm2 # sched: [2:1.00]
; BTVER2-NEXT:    vaddss %xmm2, %xmm1, %xmm1 # sched: [3:1.00]
; BTVER2-NEXT:    vmulss %xmm1, %xmm0, %xmm0 # sched: [2:1.00]
; BTVER2-NEXT:    vsubss %xmm0, %xmm3, %xmm0 # sched: [3:1.00]
; BTVER2-NEXT:    vmulss %xmm0, %xmm1, %xmm0 # sched: [2:1.00]
; BTVER2-NEXT:    vaddss %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
; BTVER2-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [7:1.00]
; BTVER2-NEXT:    retq # sched: [4:1.00]
;
; SANDY-LABEL: f32_two_step_2:
; SANDY:       # %bb.0:
; SANDY-NEXT:    vrcpss %xmm0, %xmm0, %xmm1 # sched: [5:1.00]
; SANDY-NEXT:    vmulss %xmm1, %xmm0, %xmm2 # sched: [5:1.00]
; SANDY-NEXT:    vmovss {{.*#+}} xmm3 = mem[0],zero,zero,zero sched: [6:0.50]
; SANDY-NEXT:    vsubss %xmm2, %xmm3, %xmm2 # sched: [3:1.00]
; SANDY-NEXT:    vmulss %xmm2, %xmm1, %xmm2 # sched: [5:1.00]
; SANDY-NEXT:    vaddss %xmm2, %xmm1, %xmm1 # sched: [3:1.00]
; SANDY-NEXT:    vmulss %xmm1, %xmm0, %xmm0 # sched: [5:1.00]
; SANDY-NEXT:    vsubss %xmm0, %xmm3, %xmm0 # sched: [3:1.00]
; SANDY-NEXT:    vmulss %xmm0, %xmm1, %xmm0 # sched: [5:1.00]
; SANDY-NEXT:    vaddss %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
; SANDY-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [11:1.00]
; SANDY-NEXT:    retq # sched: [1:1.00]
;
; HASWELL-LABEL: f32_two_step_2:
; HASWELL:       # %bb.0:
; HASWELL-NEXT:    vrcpss %xmm0, %xmm0, %xmm1 # sched: [5:1.00]
; HASWELL-NEXT:    vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero sched: [5:0.50]
; HASWELL-NEXT:    vmovaps %xmm1, %xmm3 # sched: [1:1.00]
; HASWELL-NEXT:    vfnmadd213ss {{.*#+}} xmm3 = -(xmm0 * xmm3) + xmm2 sched: [5:0.50]
; HASWELL-NEXT:    vfmadd132ss {{.*#+}} xmm3 = (xmm3 * xmm1) + xmm1 sched: [5:0.50]
; HASWELL-NEXT:    vfnmadd213ss {{.*#+}} xmm0 = -(xmm3 * xmm0) + xmm2 sched: [5:0.50]
; HASWELL-NEXT:    vfmadd132ss {{.*#+}} xmm0 = (xmm0 * xmm3) + xmm3 sched: [5:0.50]
; HASWELL-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [10:0.50]
; HASWELL-NEXT:    retq # sched: [7:1.00]
;
; HASWELL-NO-FMA-LABEL: f32_two_step_2:
; HASWELL-NO-FMA:       # %bb.0:
; HASWELL-NO-FMA-NEXT:    vrcpss %xmm0, %xmm0, %xmm1 # sched: [5:1.00]
; HASWELL-NO-FMA-NEXT:    vmulss %xmm1, %xmm0, %xmm2 # sched: [5:0.50]
; HASWELL-NO-FMA-NEXT:    vmovss {{.*#+}} xmm3 = mem[0],zero,zero,zero sched: [5:0.50]
; HASWELL-NO-FMA-NEXT:    vsubss %xmm2, %xmm3, %xmm2 # sched: [3:1.00]
; HASWELL-NO-FMA-NEXT:    vmulss %xmm2, %xmm1, %xmm2 # sched: [5:0.50]
; HASWELL-NO-FMA-NEXT:    vaddss %xmm2, %xmm1, %xmm1 # sched: [3:1.00]
; HASWELL-NO-FMA-NEXT:    vmulss %xmm1, %xmm0, %xmm0 # sched: [5:0.50]
; HASWELL-NO-FMA-NEXT:    vsubss %xmm0, %xmm3, %xmm0 # sched: [3:1.00]
; HASWELL-NO-FMA-NEXT:    vmulss %xmm0, %xmm1, %xmm0 # sched: [5:0.50]
; HASWELL-NO-FMA-NEXT:    vaddss %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
; HASWELL-NO-FMA-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [10:0.50]
; HASWELL-NO-FMA-NEXT:    retq # sched: [7:1.00]
;
; KNL-LABEL: f32_two_step_2:
; KNL:       # %bb.0:
; KNL-NEXT:    vrcpss %xmm0, %xmm0, %xmm1 # sched: [5:1.00]
; KNL-NEXT:    vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero sched: [5:0.50]
; KNL-NEXT:    vmovaps %xmm1, %xmm3 # sched: [1:1.00]
; KNL-NEXT:    vfnmadd213ss {{.*#+}} xmm3 = -(xmm0 * xmm3) + xmm2 sched: [5:0.50]
; KNL-NEXT:    vfmadd132ss {{.*#+}} xmm3 = (xmm3 * xmm1) + xmm1 sched: [5:0.50]
; KNL-NEXT:    vfnmadd213ss {{.*#+}} xmm0 = -(xmm3 * xmm0) + xmm2 sched: [5:0.50]
; KNL-NEXT:    vfmadd132ss {{.*#+}} xmm0 = (xmm0 * xmm3) + xmm3 sched: [5:0.50]
; KNL-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [10:0.50]
; KNL-NEXT:    retq # sched: [7:1.00]
;
; SKX-LABEL: f32_two_step_2:
; SKX:       # %bb.0:
; SKX-NEXT:    vrcpss %xmm0, %xmm0, %xmm1 # sched: [4:1.00]
; SKX-NEXT:    vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero sched: [5:0.50]
; SKX-NEXT:    vmovaps %xmm1, %xmm3 # sched: [1:0.33]
; SKX-NEXT:    vfnmadd213ss {{.*#+}} xmm3 = -(xmm0 * xmm3) + xmm2 sched: [4:0.50]
; SKX-NEXT:    vfmadd132ss {{.*#+}} xmm3 = (xmm3 * xmm1) + xmm1 sched: [4:0.50]
; SKX-NEXT:    vfnmadd213ss {{.*#+}} xmm0 = -(xmm3 * xmm0) + xmm2 sched: [4:0.50]
; SKX-NEXT:    vfmadd132ss {{.*#+}} xmm0 = (xmm0 * xmm3) + xmm3 sched: [4:0.50]
; SKX-NEXT:    vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [9:0.50]
; SKX-NEXT:    retq # sched: [7:1.00]
  %div = fdiv fast float 6789.0, %x
  ret float %div
}

define <4 x float> @v4f32_one_step2(<4 x float> %x) #1 {
; SSE-LABEL: v4f32_one_step2:
; SSE:       # %bb.0:
; SSE-NEXT:    rcpps %xmm0, %xmm2
; SSE-NEXT:    mulps %xmm2, %xmm0
; SSE-NEXT:    movaps {{.*#+}} xmm1 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0]
; SSE-NEXT:    subps %xmm0, %xmm1
; SSE-NEXT:    mulps %xmm2, %xmm1
; SSE-NEXT:    addps %xmm2, %xmm1
; SSE-NEXT:    mulps {{.*}}(%rip), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm0
; SSE-NEXT:    retq
;
; AVX-RECIP-LABEL: v4f32_one_step2:
; AVX-RECIP:       # %bb.0:
; AVX-RECIP-NEXT:    vrcpps %xmm0, %xmm1
; AVX-RECIP-NEXT:    vmulps %xmm1, %xmm0, %xmm0
; AVX-RECIP-NEXT:    vmovaps {{.*#+}} xmm2 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0]
; AVX-RECIP-NEXT:    vsubps %xmm0, %xmm2, %xmm0
; AVX-RECIP-NEXT:    vmulps %xmm0, %xmm1, %xmm0
; AVX-RECIP-NEXT:    vaddps %xmm0, %xmm1, %xmm0
; AVX-RECIP-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm0
; AVX-RECIP-NEXT:    retq
;
; FMA-RECIP-LABEL: v4f32_one_step2:
; FMA-RECIP:       # %bb.0:
; FMA-RECIP-NEXT:    vrcpps %xmm0, %xmm1
; FMA-RECIP-NEXT:    vfnmadd213ps {{.*#+}} xmm0 = -(xmm1 * xmm0) + mem
; FMA-RECIP-NEXT:    vfmadd132ps {{.*#+}} xmm0 = (xmm0 * xmm1) + xmm1
; FMA-RECIP-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm0
; FMA-RECIP-NEXT:    retq
;
; BDVER2-LABEL: v4f32_one_step2:
; BDVER2:       # %bb.0:
; BDVER2-NEXT:    vrcpps %xmm0, %xmm1 # sched: [5:1.00]
; BDVER2-NEXT:    vfnmaddps {{.*}}(%rip), %xmm1, %xmm0, %xmm0 # sched: [10:0.50]
; BDVER2-NEXT:    vfmaddps %xmm1, %xmm0, %xmm1, %xmm0 # sched: [5:0.50]
; BDVER2-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm0 # sched: [10:1.00]
; BDVER2-NEXT:    retq # sched: [5:1.00]
;
; BTVER2-LABEL: v4f32_one_step2:
; BTVER2:       # %bb.0:
; BTVER2-NEXT:    vmovaps {{.*#+}} xmm2 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0] sched: [5:1.00]
; BTVER2-NEXT:    vrcpps %xmm0, %xmm1 # sched: [2:1.00]
; BTVER2-NEXT:    vmulps %xmm1, %xmm0, %xmm0 # sched: [2:1.00]
; BTVER2-NEXT:    vsubps %xmm0, %xmm2, %xmm0 # sched: [3:1.00]
; BTVER2-NEXT:    vmulps %xmm0, %xmm1, %xmm0 # sched: [2:1.00]
; BTVER2-NEXT:    vaddps %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
; BTVER2-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm0 # sched: [7:1.00]
; BTVER2-NEXT:    retq # sched: [4:1.00]
;
; SANDY-LABEL: v4f32_one_step2:
; SANDY:       # %bb.0:
; SANDY-NEXT:    vrcpps %xmm0, %xmm1 # sched: [5:1.00]
; SANDY-NEXT:    vmulps %xmm1, %xmm0, %xmm0 # sched: [5:1.00]
; SANDY-NEXT:    vmovaps {{.*#+}} xmm2 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0] sched: [6:0.50]
; SANDY-NEXT:    vsubps %xmm0, %xmm2, %xmm0 # sched: [3:1.00]
; SANDY-NEXT:    vmulps %xmm0, %xmm1, %xmm0 # sched: [5:1.00]
; SANDY-NEXT:    vaddps %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
; SANDY-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm0 # sched: [11:1.00]
; SANDY-NEXT:    retq # sched: [1:1.00]
;
; HASWELL-LABEL: v4f32_one_step2:
; HASWELL:       # %bb.0:
; HASWELL-NEXT:    vrcpps %xmm0, %xmm1 # sched: [5:1.00]
; HASWELL-NEXT:    vbroadcastss {{.*#+}} xmm2 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0] sched: [6:0.50]
; HASWELL-NEXT:    vfnmadd213ps {{.*#+}} xmm0 = -(xmm1 * xmm0) + xmm2 sched: [5:0.50]
; HASWELL-NEXT:    vfmadd132ps {{.*#+}} xmm0 = (xmm0 * xmm1) + xmm1 sched: [5:0.50]
; HASWELL-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm0 # sched: [11:0.50]
; HASWELL-NEXT:    retq # sched: [7:1.00]
;
; HASWELL-NO-FMA-LABEL: v4f32_one_step2:
; HASWELL-NO-FMA:       # %bb.0:
; HASWELL-NO-FMA-NEXT:    vrcpps %xmm0, %xmm1 # sched: [5:1.00]
; HASWELL-NO-FMA-NEXT:    vmulps %xmm1, %xmm0, %xmm0 # sched: [5:0.50]
; HASWELL-NO-FMA-NEXT:    vbroadcastss {{.*#+}} xmm2 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0] sched: [6:0.50]
; HASWELL-NO-FMA-NEXT:    vsubps %xmm0, %xmm2, %xmm0 # sched: [3:1.00]
; HASWELL-NO-FMA-NEXT:    vmulps %xmm0, %xmm1, %xmm0 # sched: [5:0.50]
; HASWELL-NO-FMA-NEXT:    vaddps %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
; HASWELL-NO-FMA-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm0 # sched: [11:0.50]
; HASWELL-NO-FMA-NEXT:    retq # sched: [7:1.00]
;
; KNL-LABEL: v4f32_one_step2:
; KNL:       # %bb.0:
; KNL-NEXT:    vrcpps %xmm0, %xmm1 # sched: [5:1.00]
; KNL-NEXT:    vbroadcastss {{.*#+}} xmm2 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0] sched: [6:0.50]
; KNL-NEXT:    vfnmadd213ps {{.*#+}} xmm0 = -(xmm1 * xmm0) + xmm2 sched: [5:0.50]
; KNL-NEXT:    vfmadd132ps {{.*#+}} xmm0 = (xmm0 * xmm1) + xmm1 sched: [5:0.50]
; KNL-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm0 # sched: [11:0.50]
; KNL-NEXT:    retq # sched: [7:1.00]
;
; SKX-LABEL: v4f32_one_step2:
; SKX:       # %bb.0:
; SKX-NEXT:    vrcpps %xmm0, %xmm1 # sched: [4:1.00]
; SKX-NEXT:    vfnmadd213ps {{.*#+}} xmm0 = -(xmm1 * xmm0) + mem sched: [10:0.50]
; SKX-NEXT:    vfmadd132ps {{.*#+}} xmm0 = (xmm0 * xmm1) + xmm1 sched: [4:0.50]
; SKX-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm0 # sched: [10:0.50]
; SKX-NEXT:    retq # sched: [7:1.00]
  %div = fdiv fast <4 x float> <float 1.0, float 2.0, float 3.0, float 4.0>, %x
  ret <4 x float> %div
}

define <4 x float> @v4f32_one_step_2_divs(<4 x float> %x) #1 {
; SSE-LABEL: v4f32_one_step_2_divs:
; SSE:       # %bb.0:
; SSE-NEXT:    rcpps %xmm0, %xmm1
; SSE-NEXT:    mulps %xmm1, %xmm0
; SSE-NEXT:    movaps {{.*#+}} xmm2 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0]
; SSE-NEXT:    subps %xmm0, %xmm2
; SSE-NEXT:    mulps %xmm1, %xmm2
; SSE-NEXT:    addps %xmm1, %xmm2
; SSE-NEXT:    movaps {{.*#+}} xmm0 = [1.0E+0,2.0E+0,3.0E+0,4.0E+0]
; SSE-NEXT:    mulps %xmm2, %xmm0
; SSE-NEXT:    mulps %xmm2, %xmm0
; SSE-NEXT:    retq
;
; AVX-RECIP-LABEL: v4f32_one_step_2_divs:
; AVX-RECIP:       # %bb.0:
; AVX-RECIP-NEXT:    vrcpps %xmm0, %xmm1
; AVX-RECIP-NEXT:    vmulps %xmm1, %xmm0, %xmm0
; AVX-RECIP-NEXT:    vmovaps {{.*#+}} xmm2 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0]
; AVX-RECIP-NEXT:    vsubps %xmm0, %xmm2, %xmm0
; AVX-RECIP-NEXT:    vmulps %xmm0, %xmm1, %xmm0
; AVX-RECIP-NEXT:    vaddps %xmm0, %xmm1, %xmm0
; AVX-RECIP-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm1
; AVX-RECIP-NEXT:    vmulps %xmm0, %xmm1, %xmm0
; AVX-RECIP-NEXT:    retq
;
; FMA-RECIP-LABEL: v4f32_one_step_2_divs:
; FMA-RECIP:       # %bb.0:
; FMA-RECIP-NEXT:    vrcpps %xmm0, %xmm1
; FMA-RECIP-NEXT:    vfnmadd213ps {{.*#+}} xmm0 = -(xmm1 * xmm0) + mem
; FMA-RECIP-NEXT:    vfmadd132ps {{.*#+}} xmm0 = (xmm0 * xmm1) + xmm1
; FMA-RECIP-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm1
; FMA-RECIP-NEXT:    vmulps %xmm0, %xmm1, %xmm0
; FMA-RECIP-NEXT:    retq
;
; BDVER2-LABEL: v4f32_one_step_2_divs:
; BDVER2:       # %bb.0:
; BDVER2-NEXT:    vrcpps %xmm0, %xmm1 # sched: [5:1.00]
; BDVER2-NEXT:    vfnmaddps {{.*}}(%rip), %xmm1, %xmm0, %xmm0 # sched: [10:0.50]
; BDVER2-NEXT:    vfmaddps %xmm1, %xmm0, %xmm1, %xmm0 # sched: [5:0.50]
; BDVER2-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm1 # sched: [10:1.00]
; BDVER2-NEXT:    vmulps %xmm0, %xmm1, %xmm0 # sched: [5:1.00]
; BDVER2-NEXT:    retq # sched: [5:1.00]
;
; BTVER2-LABEL: v4f32_one_step_2_divs:
; BTVER2:       # %bb.0:
; BTVER2-NEXT:    vmovaps {{.*#+}} xmm2 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0] sched: [5:1.00]
; BTVER2-NEXT:    vrcpps %xmm0, %xmm1 # sched: [2:1.00]
; BTVER2-NEXT:    vmulps %xmm1, %xmm0, %xmm0 # sched: [2:1.00]
; BTVER2-NEXT:    vsubps %xmm0, %xmm2, %xmm0 # sched: [3:1.00]
; BTVER2-NEXT:    vmulps %xmm0, %xmm1, %xmm0 # sched: [2:1.00]
; BTVER2-NEXT:    vaddps %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
; BTVER2-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm1 # sched: [7:1.00]
; BTVER2-NEXT:    vmulps %xmm0, %xmm1, %xmm0 # sched: [2:1.00]
; BTVER2-NEXT:    retq # sched: [4:1.00]
;
; SANDY-LABEL: v4f32_one_step_2_divs:
; SANDY:       # %bb.0:
; SANDY-NEXT:    vrcpps %xmm0, %xmm1 # sched: [5:1.00]
; SANDY-NEXT:    vmulps %xmm1, %xmm0, %xmm0 # sched: [5:1.00]
; SANDY-NEXT:    vmovaps {{.*#+}} xmm2 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0] sched: [6:0.50]
; SANDY-NEXT:    vsubps %xmm0, %xmm2, %xmm0 # sched: [3:1.00]
; SANDY-NEXT:    vmulps %xmm0, %xmm1, %xmm0 # sched: [5:1.00]
; SANDY-NEXT:    vaddps %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
; SANDY-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm1 # sched: [11:1.00]
; SANDY-NEXT:    vmulps %xmm0, %xmm1, %xmm0 # sched: [5:1.00]
; SANDY-NEXT:    retq # sched: [1:1.00]
;
; HASWELL-LABEL: v4f32_one_step_2_divs:
; HASWELL:       # %bb.0:
; HASWELL-NEXT:    vrcpps %xmm0, %xmm1 # sched: [5:1.00]
; HASWELL-NEXT:    vbroadcastss {{.*#+}} xmm2 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0] sched: [6:0.50]
; HASWELL-NEXT:    vfnmadd213ps {{.*#+}} xmm0 = -(xmm1 * xmm0) + xmm2 sched: [5:0.50]
; HASWELL-NEXT:    vfmadd132ps {{.*#+}} xmm0 = (xmm0 * xmm1) + xmm1 sched: [5:0.50]
; HASWELL-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm1 # sched: [11:0.50]
; HASWELL-NEXT:    vmulps %xmm0, %xmm1, %xmm0 # sched: [5:0.50]
; HASWELL-NEXT:    retq # sched: [7:1.00]
;
; HASWELL-NO-FMA-LABEL: v4f32_one_step_2_divs:
; HASWELL-NO-FMA:       # %bb.0:
; HASWELL-NO-FMA-NEXT:    vrcpps %xmm0, %xmm1 # sched: [5:1.00]
; HASWELL-NO-FMA-NEXT:    vmulps %xmm1, %xmm0, %xmm0 # sched: [5:0.50]
; HASWELL-NO-FMA-NEXT:    vbroadcastss {{.*#+}} xmm2 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0] sched: [6:0.50]
; HASWELL-NO-FMA-NEXT:    vsubps %xmm0, %xmm2, %xmm0 # sched: [3:1.00]
; HASWELL-NO-FMA-NEXT:    vmulps %xmm0, %xmm1, %xmm0 # sched: [5:0.50]
; HASWELL-NO-FMA-NEXT:    vaddps %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
; HASWELL-NO-FMA-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm1 # sched: [11:0.50]
; HASWELL-NO-FMA-NEXT:    vmulps %xmm0, %xmm1, %xmm0 # sched: [5:0.50]
; HASWELL-NO-FMA-NEXT:    retq # sched: [7:1.00]
;
; KNL-LABEL: v4f32_one_step_2_divs:
; KNL:       # %bb.0:
; KNL-NEXT:    vrcpps %xmm0, %xmm1 # sched: [5:1.00]
; KNL-NEXT:    vbroadcastss {{.*#+}} xmm2 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0] sched: [6:0.50]
; KNL-NEXT:    vfnmadd213ps {{.*#+}} xmm0 = -(xmm1 * xmm0) + xmm2 sched: [5:0.50]
; KNL-NEXT:    vfmadd132ps {{.*#+}} xmm0 = (xmm0 * xmm1) + xmm1 sched: [5:0.50]
; KNL-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm1 # sched: [11:0.50]
; KNL-NEXT:    vmulps %xmm0, %xmm1, %xmm0 # sched: [5:0.50]
; KNL-NEXT:    retq # sched: [7:1.00]
;
; SKX-LABEL: v4f32_one_step_2_divs:
; SKX:       # %bb.0:
; SKX-NEXT:    vrcpps %xmm0, %xmm1 # sched: [4:1.00]
; SKX-NEXT:    vfnmadd213ps {{.*#+}} xmm0 = -(xmm1 * xmm0) + mem sched: [10:0.50]
; SKX-NEXT:    vfmadd132ps {{.*#+}} xmm0 = (xmm0 * xmm1) + xmm1 sched: [4:0.50]
; SKX-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm1 # sched: [10:0.50]
; SKX-NEXT:    vmulps %xmm0, %xmm1, %xmm0 # sched: [4:0.50]
; SKX-NEXT:    retq # sched: [7:1.00]
  %div = fdiv fast <4 x float> <float 1.0, float 2.0, float 3.0, float 4.0>, %x
  %div2 = fdiv fast <4 x float> %div, %x
  ret <4 x float> %div2
}

define <4 x float> @v4f32_two_step2(<4 x float> %x) #2 {
; SSE-LABEL: v4f32_two_step2:
; SSE:       # %bb.0:
; SSE-NEXT:    rcpps %xmm0, %xmm2
; SSE-NEXT:    movaps %xmm0, %xmm3
; SSE-NEXT:    mulps %xmm2, %xmm3
; SSE-NEXT:    movaps {{.*#+}} xmm1 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0]
; SSE-NEXT:    movaps %xmm1, %xmm4
; SSE-NEXT:    subps %xmm3, %xmm4
; SSE-NEXT:    mulps %xmm2, %xmm4
; SSE-NEXT:    addps %xmm2, %xmm4
; SSE-NEXT:    mulps %xmm4, %xmm0
; SSE-NEXT:    subps %xmm0, %xmm1
; SSE-NEXT:    mulps %xmm4, %xmm1
; SSE-NEXT:    addps %xmm4, %xmm1
; SSE-NEXT:    mulps {{.*}}(%rip), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm0
; SSE-NEXT:    retq
;
; AVX-RECIP-LABEL: v4f32_two_step2:
; AVX-RECIP:       # %bb.0:
; AVX-RECIP-NEXT:    vrcpps %xmm0, %xmm1
; AVX-RECIP-NEXT:    vmulps %xmm1, %xmm0, %xmm2
; AVX-RECIP-NEXT:    vmovaps {{.*#+}} xmm3 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0]
; AVX-RECIP-NEXT:    vsubps %xmm2, %xmm3, %xmm2
; AVX-RECIP-NEXT:    vmulps %xmm2, %xmm1, %xmm2
; AVX-RECIP-NEXT:    vaddps %xmm2, %xmm1, %xmm1
; AVX-RECIP-NEXT:    vmulps %xmm1, %xmm0, %xmm0
; AVX-RECIP-NEXT:    vsubps %xmm0, %xmm3, %xmm0
; AVX-RECIP-NEXT:    vmulps %xmm0, %xmm1, %xmm0
; AVX-RECIP-NEXT:    vaddps %xmm0, %xmm1, %xmm0
; AVX-RECIP-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm0
; AVX-RECIP-NEXT:    retq
;
; FMA-RECIP-LABEL: v4f32_two_step2:
; FMA-RECIP:       # %bb.0:
; FMA-RECIP-NEXT:    vrcpps %xmm0, %xmm1
; FMA-RECIP-NEXT:    vmovaps {{.*#+}} xmm2 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0]
; FMA-RECIP-NEXT:    vmovaps %xmm1, %xmm3
; FMA-RECIP-NEXT:    vfnmadd213ps {{.*#+}} xmm3 = -(xmm0 * xmm3) + xmm2
; FMA-RECIP-NEXT:    vfmadd132ps {{.*#+}} xmm3 = (xmm3 * xmm1) + xmm1
; FMA-RECIP-NEXT:    vfnmadd213ps {{.*#+}} xmm0 = -(xmm3 * xmm0) + xmm2
; FMA-RECIP-NEXT:    vfmadd132ps {{.*#+}} xmm0 = (xmm0 * xmm3) + xmm3
; FMA-RECIP-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm0
; FMA-RECIP-NEXT:    retq
;
; BDVER2-LABEL: v4f32_two_step2:
; BDVER2:       # %bb.0:
; BDVER2-NEXT:    vrcpps %xmm0, %xmm1 # sched: [5:1.00]
; BDVER2-NEXT:    vmovaps {{.*#+}} xmm2 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0] sched: [5:0.50]
; BDVER2-NEXT:    vfnmaddps %xmm2, %xmm1, %xmm0, %xmm3 # sched: [5:0.50]
; BDVER2-NEXT:    vfmaddps %xmm1, %xmm3, %xmm1, %xmm1 # sched: [5:0.50]
; BDVER2-NEXT:    vfnmaddps %xmm2, %xmm1, %xmm0, %xmm0 # sched: [5:0.50]
; BDVER2-NEXT:    vfmaddps %xmm1, %xmm0, %xmm1, %xmm0 # sched: [5:0.50]
; BDVER2-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm0 # sched: [10:1.00]
; BDVER2-NEXT:    retq # sched: [5:1.00]
;
; BTVER2-LABEL: v4f32_two_step2:
; BTVER2:       # %bb.0:
; BTVER2-NEXT:    vmovaps {{.*#+}} xmm3 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0] sched: [5:1.00]
; BTVER2-NEXT:    vrcpps %xmm0, %xmm1 # sched: [2:1.00]
; BTVER2-NEXT:    vmulps %xmm1, %xmm0, %xmm2 # sched: [2:1.00]
; BTVER2-NEXT:    vsubps %xmm2, %xmm3, %xmm2 # sched: [3:1.00]
; BTVER2-NEXT:    vmulps %xmm2, %xmm1, %xmm2 # sched: [2:1.00]
; BTVER2-NEXT:    vaddps %xmm2, %xmm1, %xmm1 # sched: [3:1.00]
; BTVER2-NEXT:    vmulps %xmm1, %xmm0, %xmm0 # sched: [2:1.00]
; BTVER2-NEXT:    vsubps %xmm0, %xmm3, %xmm0 # sched: [3:1.00]
; BTVER2-NEXT:    vmulps %xmm0, %xmm1, %xmm0 # sched: [2:1.00]
; BTVER2-NEXT:    vaddps %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
; BTVER2-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm0 # sched: [7:1.00]
; BTVER2-NEXT:    retq # sched: [4:1.00]
;
; SANDY-LABEL: v4f32_two_step2:
; SANDY:       # %bb.0:
; SANDY-NEXT:    vrcpps %xmm0, %xmm1 # sched: [5:1.00]
; SANDY-NEXT:    vmulps %xmm1, %xmm0, %xmm2 # sched: [5:1.00]
; SANDY-NEXT:    vmovaps {{.*#+}} xmm3 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0] sched: [6:0.50]
; SANDY-NEXT:    vsubps %xmm2, %xmm3, %xmm2 # sched: [3:1.00]
; SANDY-NEXT:    vmulps %xmm2, %xmm1, %xmm2 # sched: [5:1.00]
; SANDY-NEXT:    vaddps %xmm2, %xmm1, %xmm1 # sched: [3:1.00]
; SANDY-NEXT:    vmulps %xmm1, %xmm0, %xmm0 # sched: [5:1.00]
; SANDY-NEXT:    vsubps %xmm0, %xmm3, %xmm0 # sched: [3:1.00]
; SANDY-NEXT:    vmulps %xmm0, %xmm1, %xmm0 # sched: [5:1.00]
; SANDY-NEXT:    vaddps %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
; SANDY-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm0 # sched: [11:1.00]
; SANDY-NEXT:    retq # sched: [1:1.00]
;
; HASWELL-LABEL: v4f32_two_step2:
; HASWELL:       # %bb.0:
; HASWELL-NEXT:    vrcpps %xmm0, %xmm1 # sched: [5:1.00]
; HASWELL-NEXT:    vbroadcastss {{.*#+}} xmm2 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0] sched: [6:0.50]
; HASWELL-NEXT:    vmovaps %xmm1, %xmm3 # sched: [1:1.00]
; HASWELL-NEXT:    vfnmadd213ps {{.*#+}} xmm3 = -(xmm0 * xmm3) + xmm2 sched: [5:0.50]
; HASWELL-NEXT:    vfmadd132ps {{.*#+}} xmm3 = (xmm3 * xmm1) + xmm1 sched: [5:0.50]
; HASWELL-NEXT:    vfnmadd213ps {{.*#+}} xmm0 = -(xmm3 * xmm0) + xmm2 sched: [5:0.50]
; HASWELL-NEXT:    vfmadd132ps {{.*#+}} xmm0 = (xmm0 * xmm3) + xmm3 sched: [5:0.50]
; HASWELL-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm0 # sched: [11:0.50]
; HASWELL-NEXT:    retq # sched: [7:1.00]
;
; HASWELL-NO-FMA-LABEL: v4f32_two_step2:
; HASWELL-NO-FMA:       # %bb.0:
; HASWELL-NO-FMA-NEXT:    vrcpps %xmm0, %xmm1 # sched: [5:1.00]
; HASWELL-NO-FMA-NEXT:    vmulps %xmm1, %xmm0, %xmm2 # sched: [5:0.50]
; HASWELL-NO-FMA-NEXT:    vbroadcastss {{.*#+}} xmm3 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0] sched: [6:0.50]
; HASWELL-NO-FMA-NEXT:    vsubps %xmm2, %xmm3, %xmm2 # sched: [3:1.00]
; HASWELL-NO-FMA-NEXT:    vmulps %xmm2, %xmm1, %xmm2 # sched: [5:0.50]
; HASWELL-NO-FMA-NEXT:    vaddps %xmm2, %xmm1, %xmm1 # sched: [3:1.00]
; HASWELL-NO-FMA-NEXT:    vmulps %xmm1, %xmm0, %xmm0 # sched: [5:0.50]
; HASWELL-NO-FMA-NEXT:    vsubps %xmm0, %xmm3, %xmm0 # sched: [3:1.00]
; HASWELL-NO-FMA-NEXT:    vmulps %xmm0, %xmm1, %xmm0 # sched: [5:0.50]
; HASWELL-NO-FMA-NEXT:    vaddps %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
; HASWELL-NO-FMA-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm0 # sched: [11:0.50]
; HASWELL-NO-FMA-NEXT:    retq # sched: [7:1.00]
;
; KNL-LABEL: v4f32_two_step2:
; KNL:       # %bb.0:
; KNL-NEXT:    vrcpps %xmm0, %xmm1 # sched: [5:1.00]
; KNL-NEXT:    vbroadcastss {{.*#+}} xmm2 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0] sched: [6:0.50]
; KNL-NEXT:    vmovaps %xmm1, %xmm3 # sched: [1:1.00]
; KNL-NEXT:    vfnmadd213ps {{.*#+}} xmm3 = -(xmm0 * xmm3) + xmm2 sched: [5:0.50]
; KNL-NEXT:    vfmadd132ps {{.*#+}} xmm3 = (xmm3 * xmm1) + xmm1 sched: [5:0.50]
; KNL-NEXT:    vfnmadd213ps {{.*#+}} xmm0 = -(xmm3 * xmm0) + xmm2 sched: [5:0.50]
; KNL-NEXT:    vfmadd132ps {{.*#+}} xmm0 = (xmm0 * xmm3) + xmm3 sched: [5:0.50]
; KNL-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm0 # sched: [11:0.50]
; KNL-NEXT:    retq # sched: [7:1.00]
;
; SKX-LABEL: v4f32_two_step2:
; SKX:       # %bb.0:
; SKX-NEXT:    vrcpps %xmm0, %xmm1 # sched: [4:1.00]
; SKX-NEXT:    vbroadcastss {{.*#+}} xmm2 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0] sched: [6:0.50]
; SKX-NEXT:    vmovaps %xmm1, %xmm3 # sched: [1:0.33]
; SKX-NEXT:    vfnmadd213ps {{.*#+}} xmm3 = -(xmm0 * xmm3) + xmm2 sched: [4:0.50]
; SKX-NEXT:    vfmadd132ps {{.*#+}} xmm3 = (xmm3 * xmm1) + xmm1 sched: [4:0.50]
; SKX-NEXT:    vfnmadd213ps {{.*#+}} xmm0 = -(xmm3 * xmm0) + xmm2 sched: [4:0.50]
; SKX-NEXT:    vfmadd132ps {{.*#+}} xmm0 = (xmm0 * xmm3) + xmm3 sched: [4:0.50]
; SKX-NEXT:    vmulps {{.*}}(%rip), %xmm0, %xmm0 # sched: [10:0.50]
; SKX-NEXT:    retq # sched: [7:1.00]
  %div = fdiv fast <4 x float> <float 1.0, float 2.0, float 3.0, float 4.0>, %x
  ret <4 x float> %div
}

define <8 x float> @v8f32_one_step2(<8 x float> %x) #1 {
; SSE-LABEL: v8f32_one_step2:
; SSE:       # %bb.0:
; SSE-NEXT:    rcpps %xmm1, %xmm4
; SSE-NEXT:    mulps %xmm4, %xmm1
; SSE-NEXT:    movaps {{.*#+}} xmm2 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0]
; SSE-NEXT:    movaps %xmm2, %xmm3
; SSE-NEXT:    subps %xmm1, %xmm3
; SSE-NEXT:    mulps %xmm4, %xmm3
; SSE-NEXT:    addps %xmm4, %xmm3
; SSE-NEXT:    rcpps %xmm0, %xmm1
; SSE-NEXT:    mulps %xmm1, %xmm0
; SSE-NEXT:    subps %xmm0, %xmm2
; SSE-NEXT:    mulps %xmm1, %xmm2
; SSE-NEXT:    addps %xmm1, %xmm2
; SSE-NEXT:    mulps {{.*}}(%rip), %xmm2
; SSE-NEXT:    mulps {{.*}}(%rip), %xmm3
; SSE-NEXT:    movaps %xmm2, %xmm0
; SSE-NEXT:    movaps %xmm3, %xmm1
; SSE-NEXT:    retq
;
; AVX-RECIP-LABEL: v8f32_one_step2:
; AVX-RECIP:       # %bb.0:
; AVX-RECIP-NEXT:    vrcpps %ymm0, %ymm1
; AVX-RECIP-NEXT:    vmulps %ymm1, %ymm0, %ymm0
; AVX-RECIP-NEXT:    vmovaps {{.*#+}} ymm2 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0]
; AVX-RECIP-NEXT:    vsubps %ymm0, %ymm2, %ymm0
; AVX-RECIP-NEXT:    vmulps %ymm0, %ymm1, %ymm0
; AVX-RECIP-NEXT:    vaddps %ymm0, %ymm1, %ymm0
; AVX-RECIP-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0
; AVX-RECIP-NEXT:    retq
;
; FMA-RECIP-LABEL: v8f32_one_step2:
; FMA-RECIP:       # %bb.0:
; FMA-RECIP-NEXT:    vrcpps %ymm0, %ymm1
; FMA-RECIP-NEXT:    vfnmadd213ps {{.*#+}} ymm0 = -(ymm1 * ymm0) + mem
; FMA-RECIP-NEXT:    vfmadd132ps {{.*#+}} ymm0 = (ymm0 * ymm1) + ymm1
; FMA-RECIP-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0
; FMA-RECIP-NEXT:    retq
;
; BDVER2-LABEL: v8f32_one_step2:
; BDVER2:       # %bb.0:
; BDVER2-NEXT:    vrcpps %ymm0, %ymm1 # sched: [5:2.00]
; BDVER2-NEXT:    vfnmaddps {{.*}}(%rip), %ymm1, %ymm0, %ymm0 # sched: [10:1.00]
; BDVER2-NEXT:    vfmaddps %ymm1, %ymm0, %ymm1, %ymm0 # sched: [5:0.50]
; BDVER2-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [10:2.00]
; BDVER2-NEXT:    retq # sched: [5:1.00]
;
; BTVER2-LABEL: v8f32_one_step2:
; BTVER2:       # %bb.0:
; BTVER2-NEXT:    vmovaps {{.*#+}} ymm2 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0] sched: [5:1.00]
; BTVER2-NEXT:    vrcpps %ymm0, %ymm1 # sched: [2:2.00]
; BTVER2-NEXT:    vmulps %ymm1, %ymm0, %ymm0 # sched: [2:2.00]
; BTVER2-NEXT:    vsubps %ymm0, %ymm2, %ymm0 # sched: [3:2.00]
; BTVER2-NEXT:    vmulps %ymm0, %ymm1, %ymm0 # sched: [2:2.00]
; BTVER2-NEXT:    vaddps %ymm0, %ymm1, %ymm0 # sched: [3:2.00]
; BTVER2-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [7:2.00]
; BTVER2-NEXT:    retq # sched: [4:1.00]
;
; SANDY-LABEL: v8f32_one_step2:
; SANDY:       # %bb.0:
; SANDY-NEXT:    vrcpps %ymm0, %ymm1 # sched: [7:2.00]
; SANDY-NEXT:    vmulps %ymm1, %ymm0, %ymm0 # sched: [5:1.00]
; SANDY-NEXT:    vmovaps {{.*#+}} ymm2 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0] sched: [7:0.50]
; SANDY-NEXT:    vsubps %ymm0, %ymm2, %ymm0 # sched: [3:1.00]
; SANDY-NEXT:    vmulps %ymm0, %ymm1, %ymm0 # sched: [5:1.00]
; SANDY-NEXT:    vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
; SANDY-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [12:1.00]
; SANDY-NEXT:    retq # sched: [1:1.00]
;
; HASWELL-LABEL: v8f32_one_step2:
; HASWELL:       # %bb.0:
; HASWELL-NEXT:    vrcpps %ymm0, %ymm1 # sched: [11:2.00]
; HASWELL-NEXT:    vbroadcastss {{.*#+}} ymm2 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0] sched: [7:0.50]
; HASWELL-NEXT:    vfnmadd213ps {{.*#+}} ymm0 = -(ymm1 * ymm0) + ymm2 sched: [5:0.50]
; HASWELL-NEXT:    vfmadd132ps {{.*#+}} ymm0 = (ymm0 * ymm1) + ymm1 sched: [5:0.50]
; HASWELL-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [12:0.50]
; HASWELL-NEXT:    retq # sched: [7:1.00]
;
; HASWELL-NO-FMA-LABEL: v8f32_one_step2:
; HASWELL-NO-FMA:       # %bb.0:
; HASWELL-NO-FMA-NEXT:    vrcpps %ymm0, %ymm1 # sched: [11:2.00]
; HASWELL-NO-FMA-NEXT:    vmulps %ymm1, %ymm0, %ymm0 # sched: [5:0.50]
; HASWELL-NO-FMA-NEXT:    vbroadcastss {{.*#+}} ymm2 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0] sched: [7:0.50]
; HASWELL-NO-FMA-NEXT:    vsubps %ymm0, %ymm2, %ymm0 # sched: [3:1.00]
; HASWELL-NO-FMA-NEXT:    vmulps %ymm0, %ymm1, %ymm0 # sched: [5:0.50]
; HASWELL-NO-FMA-NEXT:    vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
; HASWELL-NO-FMA-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [12:0.50]
; HASWELL-NO-FMA-NEXT:    retq # sched: [7:1.00]
;
; KNL-LABEL: v8f32_one_step2:
; KNL:       # %bb.0:
; KNL-NEXT:    vrcpps %ymm0, %ymm1 # sched: [11:2.00]
; KNL-NEXT:    vbroadcastss {{.*#+}} ymm2 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0] sched: [7:0.50]
; KNL-NEXT:    vfnmadd213ps {{.*#+}} ymm0 = -(ymm1 * ymm0) + ymm2 sched: [5:0.50]
; KNL-NEXT:    vfmadd132ps {{.*#+}} ymm0 = (ymm0 * ymm1) + ymm1 sched: [5:0.50]
; KNL-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [12:0.50]
; KNL-NEXT:    retq # sched: [7:1.00]
;
; SKX-LABEL: v8f32_one_step2:
; SKX:       # %bb.0:
; SKX-NEXT:    vrcpps %ymm0, %ymm1 # sched: [4:1.00]
; SKX-NEXT:    vfnmadd213ps {{.*#+}} ymm0 = -(ymm1 * ymm0) + mem sched: [11:0.50]
; SKX-NEXT:    vfmadd132ps {{.*#+}} ymm0 = (ymm0 * ymm1) + ymm1 sched: [4:0.50]
; SKX-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [11:0.50]
; SKX-NEXT:    retq # sched: [7:1.00]
  %div = fdiv fast <8 x float> <float 1.0, float 2.0, float 3.0, float 4.0, float 5.0, float 6.0, float 7.0, float 8.0>, %x
  ret <8 x float> %div
}

define <8 x float> @v8f32_one_step_2_divs(<8 x float> %x) #1 {
; SSE-LABEL: v8f32_one_step_2_divs:
; SSE:       # %bb.0:
; SSE-NEXT:    rcpps %xmm0, %xmm2
; SSE-NEXT:    mulps %xmm2, %xmm0
; SSE-NEXT:    movaps {{.*#+}} xmm3 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0]
; SSE-NEXT:    movaps %xmm3, %xmm4
; SSE-NEXT:    subps %xmm0, %xmm4
; SSE-NEXT:    mulps %xmm2, %xmm4
; SSE-NEXT:    addps %xmm2, %xmm4
; SSE-NEXT:    rcpps %xmm1, %xmm0
; SSE-NEXT:    mulps %xmm0, %xmm1
; SSE-NEXT:    subps %xmm1, %xmm3
; SSE-NEXT:    mulps %xmm0, %xmm3
; SSE-NEXT:    addps %xmm0, %xmm3
; SSE-NEXT:    movaps {{.*#+}} xmm1 = [5.0E+0,6.0E+0,7.0E+0,8.0E+0]
; SSE-NEXT:    mulps %xmm3, %xmm1
; SSE-NEXT:    movaps {{.*#+}} xmm0 = [1.0E+0,2.0E+0,3.0E+0,4.0E+0]
; SSE-NEXT:    mulps %xmm4, %xmm0
; SSE-NEXT:    mulps %xmm4, %xmm0
; SSE-NEXT:    mulps %xmm3, %xmm1
; SSE-NEXT:    retq
;
; AVX-RECIP-LABEL: v8f32_one_step_2_divs:
; AVX-RECIP:       # %bb.0:
; AVX-RECIP-NEXT:    vrcpps %ymm0, %ymm1
; AVX-RECIP-NEXT:    vmulps %ymm1, %ymm0, %ymm0
; AVX-RECIP-NEXT:    vmovaps {{.*#+}} ymm2 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0]
; AVX-RECIP-NEXT:    vsubps %ymm0, %ymm2, %ymm0
; AVX-RECIP-NEXT:    vmulps %ymm0, %ymm1, %ymm0
; AVX-RECIP-NEXT:    vaddps %ymm0, %ymm1, %ymm0
; AVX-RECIP-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm1
; AVX-RECIP-NEXT:    vmulps %ymm0, %ymm1, %ymm0
; AVX-RECIP-NEXT:    retq
;
; FMA-RECIP-LABEL: v8f32_one_step_2_divs:
; FMA-RECIP:       # %bb.0:
; FMA-RECIP-NEXT:    vrcpps %ymm0, %ymm1
; FMA-RECIP-NEXT:    vfnmadd213ps {{.*#+}} ymm0 = -(ymm1 * ymm0) + mem
; FMA-RECIP-NEXT:    vfmadd132ps {{.*#+}} ymm0 = (ymm0 * ymm1) + ymm1
; FMA-RECIP-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm1
; FMA-RECIP-NEXT:    vmulps %ymm0, %ymm1, %ymm0
; FMA-RECIP-NEXT:    retq
;
; BDVER2-LABEL: v8f32_one_step_2_divs:
; BDVER2:       # %bb.0:
; BDVER2-NEXT:    vrcpps %ymm0, %ymm1 # sched: [5:2.00]
; BDVER2-NEXT:    vfnmaddps {{.*}}(%rip), %ymm1, %ymm0, %ymm0 # sched: [10:1.00]
; BDVER2-NEXT:    vfmaddps %ymm1, %ymm0, %ymm1, %ymm0 # sched: [5:0.50]
; BDVER2-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm1 # sched: [10:2.00]
; BDVER2-NEXT:    vmulps %ymm0, %ymm1, %ymm0 # sched: [5:2.00]
; BDVER2-NEXT:    retq # sched: [5:1.00]
;
; BTVER2-LABEL: v8f32_one_step_2_divs:
; BTVER2:       # %bb.0:
; BTVER2-NEXT:    vmovaps {{.*#+}} ymm2 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0] sched: [5:1.00]
; BTVER2-NEXT:    vrcpps %ymm0, %ymm1 # sched: [2:2.00]
; BTVER2-NEXT:    vmulps %ymm1, %ymm0, %ymm0 # sched: [2:2.00]
; BTVER2-NEXT:    vsubps %ymm0, %ymm2, %ymm0 # sched: [3:2.00]
; BTVER2-NEXT:    vmulps %ymm0, %ymm1, %ymm0 # sched: [2:2.00]
; BTVER2-NEXT:    vaddps %ymm0, %ymm1, %ymm0 # sched: [3:2.00]
; BTVER2-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm1 # sched: [7:2.00]
; BTVER2-NEXT:    vmulps %ymm0, %ymm1, %ymm0 # sched: [2:2.00]
; BTVER2-NEXT:    retq # sched: [4:1.00]
;
; SANDY-LABEL: v8f32_one_step_2_divs:
; SANDY:       # %bb.0:
; SANDY-NEXT:    vrcpps %ymm0, %ymm1 # sched: [7:2.00]
; SANDY-NEXT:    vmulps %ymm1, %ymm0, %ymm0 # sched: [5:1.00]
; SANDY-NEXT:    vmovaps {{.*#+}} ymm2 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0] sched: [7:0.50]
; SANDY-NEXT:    vsubps %ymm0, %ymm2, %ymm0 # sched: [3:1.00]
; SANDY-NEXT:    vmulps %ymm0, %ymm1, %ymm0 # sched: [5:1.00]
; SANDY-NEXT:    vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
; SANDY-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm1 # sched: [12:1.00]
; SANDY-NEXT:    vmulps %ymm0, %ymm1, %ymm0 # sched: [5:1.00]
; SANDY-NEXT:    retq # sched: [1:1.00]
;
; HASWELL-LABEL: v8f32_one_step_2_divs:
; HASWELL:       # %bb.0:
; HASWELL-NEXT:    vrcpps %ymm0, %ymm1 # sched: [11:2.00]
; HASWELL-NEXT:    vbroadcastss {{.*#+}} ymm2 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0] sched: [7:0.50]
; HASWELL-NEXT:    vfnmadd213ps {{.*#+}} ymm0 = -(ymm1 * ymm0) + ymm2 sched: [5:0.50]
; HASWELL-NEXT:    vfmadd132ps {{.*#+}} ymm0 = (ymm0 * ymm1) + ymm1 sched: [5:0.50]
; HASWELL-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm1 # sched: [12:0.50]
; HASWELL-NEXT:    vmulps %ymm0, %ymm1, %ymm0 # sched: [5:0.50]
; HASWELL-NEXT:    retq # sched: [7:1.00]
;
; HASWELL-NO-FMA-LABEL: v8f32_one_step_2_divs:
; HASWELL-NO-FMA:       # %bb.0:
; HASWELL-NO-FMA-NEXT:    vrcpps %ymm0, %ymm1 # sched: [11:2.00]
; HASWELL-NO-FMA-NEXT:    vmulps %ymm1, %ymm0, %ymm0 # sched: [5:0.50]
; HASWELL-NO-FMA-NEXT:    vbroadcastss {{.*#+}} ymm2 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0] sched: [7:0.50]
; HASWELL-NO-FMA-NEXT:    vsubps %ymm0, %ymm2, %ymm0 # sched: [3:1.00]
; HASWELL-NO-FMA-NEXT:    vmulps %ymm0, %ymm1, %ymm0 # sched: [5:0.50]
; HASWELL-NO-FMA-NEXT:    vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
; HASWELL-NO-FMA-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm1 # sched: [12:0.50]
; HASWELL-NO-FMA-NEXT:    vmulps %ymm0, %ymm1, %ymm0 # sched: [5:0.50]
; HASWELL-NO-FMA-NEXT:    retq # sched: [7:1.00]
;
; KNL-LABEL: v8f32_one_step_2_divs:
; KNL:       # %bb.0:
; KNL-NEXT:    vrcpps %ymm0, %ymm1 # sched: [11:2.00]
; KNL-NEXT:    vbroadcastss {{.*#+}} ymm2 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0] sched: [7:0.50]
; KNL-NEXT:    vfnmadd213ps {{.*#+}} ymm0 = -(ymm1 * ymm0) + ymm2 sched: [5:0.50]
; KNL-NEXT:    vfmadd132ps {{.*#+}} ymm0 = (ymm0 * ymm1) + ymm1 sched: [5:0.50]
; KNL-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm1 # sched: [12:0.50]
; KNL-NEXT:    vmulps %ymm0, %ymm1, %ymm0 # sched: [5:0.50]
; KNL-NEXT:    retq # sched: [7:1.00]
;
; SKX-LABEL: v8f32_one_step_2_divs:
; SKX:       # %bb.0:
; SKX-NEXT:    vrcpps %ymm0, %ymm1 # sched: [4:1.00]
; SKX-NEXT:    vfnmadd213ps {{.*#+}} ymm0 = -(ymm1 * ymm0) + mem sched: [11:0.50]
; SKX-NEXT:    vfmadd132ps {{.*#+}} ymm0 = (ymm0 * ymm1) + ymm1 sched: [4:0.50]
; SKX-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm1 # sched: [11:0.50]
; SKX-NEXT:    vmulps %ymm0, %ymm1, %ymm0 # sched: [4:0.50]
; SKX-NEXT:    retq # sched: [7:1.00]
  %div = fdiv fast <8 x float> <float 1.0, float 2.0, float 3.0, float 4.0, float 5.0, float 6.0, float 7.0, float 8.0>, %x
  %div2 = fdiv fast <8 x float> %div, %x
  ret <8 x float> %div2
}

define <8 x float> @v8f32_two_step2(<8 x float> %x) #2 {
; SSE-LABEL: v8f32_two_step2:
; SSE:       # %bb.0:
; SSE-NEXT:    movaps %xmm0, %xmm2
; SSE-NEXT:    rcpps %xmm1, %xmm3
; SSE-NEXT:    movaps %xmm1, %xmm4
; SSE-NEXT:    mulps %xmm3, %xmm4
; SSE-NEXT:    movaps {{.*#+}} xmm0 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0]
; SSE-NEXT:    movaps %xmm0, %xmm5
; SSE-NEXT:    subps %xmm4, %xmm5
; SSE-NEXT:    mulps %xmm3, %xmm5
; SSE-NEXT:    addps %xmm3, %xmm5
; SSE-NEXT:    mulps %xmm5, %xmm1
; SSE-NEXT:    movaps %xmm0, %xmm3
; SSE-NEXT:    subps %xmm1, %xmm3
; SSE-NEXT:    mulps %xmm5, %xmm3
; SSE-NEXT:    addps %xmm5, %xmm3
; SSE-NEXT:    rcpps %xmm2, %xmm1
; SSE-NEXT:    movaps %xmm2, %xmm4
; SSE-NEXT:    mulps %xmm1, %xmm4
; SSE-NEXT:    movaps %xmm0, %xmm5
; SSE-NEXT:    subps %xmm4, %xmm5
; SSE-NEXT:    mulps %xmm1, %xmm5
; SSE-NEXT:    addps %xmm1, %xmm5
; SSE-NEXT:    mulps %xmm5, %xmm2
; SSE-NEXT:    subps %xmm2, %xmm0
; SSE-NEXT:    mulps %xmm5, %xmm0
; SSE-NEXT:    addps %xmm5, %xmm0
; SSE-NEXT:    mulps {{.*}}(%rip), %xmm0
; SSE-NEXT:    mulps {{.*}}(%rip), %xmm3
; SSE-NEXT:    movaps %xmm3, %xmm1
; SSE-NEXT:    retq
;
; AVX-RECIP-LABEL: v8f32_two_step2:
; AVX-RECIP:       # %bb.0:
; AVX-RECIP-NEXT:    vrcpps %ymm0, %ymm1
; AVX-RECIP-NEXT:    vmulps %ymm1, %ymm0, %ymm2
; AVX-RECIP-NEXT:    vmovaps {{.*#+}} ymm3 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0]
; AVX-RECIP-NEXT:    vsubps %ymm2, %ymm3, %ymm2
; AVX-RECIP-NEXT:    vmulps %ymm2, %ymm1, %ymm2
; AVX-RECIP-NEXT:    vaddps %ymm2, %ymm1, %ymm1
; AVX-RECIP-NEXT:    vmulps %ymm1, %ymm0, %ymm0
; AVX-RECIP-NEXT:    vsubps %ymm0, %ymm3, %ymm0
; AVX-RECIP-NEXT:    vmulps %ymm0, %ymm1, %ymm0
; AVX-RECIP-NEXT:    vaddps %ymm0, %ymm1, %ymm0
; AVX-RECIP-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0
; AVX-RECIP-NEXT:    retq
;
; FMA-RECIP-LABEL: v8f32_two_step2:
; FMA-RECIP:       # %bb.0:
; FMA-RECIP-NEXT:    vrcpps %ymm0, %ymm1
; FMA-RECIP-NEXT:    vmovaps {{.*#+}} ymm2 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0]
; FMA-RECIP-NEXT:    vmovaps %ymm1, %ymm3
; FMA-RECIP-NEXT:    vfnmadd213ps {{.*#+}} ymm3 = -(ymm0 * ymm3) + ymm2
; FMA-RECIP-NEXT:    vfmadd132ps {{.*#+}} ymm3 = (ymm3 * ymm1) + ymm1
; FMA-RECIP-NEXT:    vfnmadd213ps {{.*#+}} ymm0 = -(ymm3 * ymm0) + ymm2
; FMA-RECIP-NEXT:    vfmadd132ps {{.*#+}} ymm0 = (ymm0 * ymm3) + ymm3
; FMA-RECIP-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0
; FMA-RECIP-NEXT:    retq
;
; BDVER2-LABEL: v8f32_two_step2:
; BDVER2:       # %bb.0:
; BDVER2-NEXT:    vrcpps %ymm0, %ymm1 # sched: [5:2.00]
; BDVER2-NEXT:    vmovaps {{.*#+}} ymm2 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0] sched: [5:0.50]
; BDVER2-NEXT:    vfnmaddps %ymm2, %ymm1, %ymm0, %ymm3 # sched: [5:0.50]
; BDVER2-NEXT:    vfmaddps %ymm1, %ymm3, %ymm1, %ymm1 # sched: [5:0.50]
; BDVER2-NEXT:    vfnmaddps %ymm2, %ymm1, %ymm0, %ymm0 # sched: [5:0.50]
; BDVER2-NEXT:    vfmaddps %ymm1, %ymm0, %ymm1, %ymm0 # sched: [5:0.50]
; BDVER2-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [10:2.00]
; BDVER2-NEXT:    retq # sched: [5:1.00]
;
; BTVER2-LABEL: v8f32_two_step2:
; BTVER2:       # %bb.0:
; BTVER2-NEXT:    vmovaps {{.*#+}} ymm3 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0] sched: [5:1.00]
; BTVER2-NEXT:    vrcpps %ymm0, %ymm1 # sched: [2:2.00]
; BTVER2-NEXT:    vmulps %ymm1, %ymm0, %ymm2 # sched: [2:2.00]
; BTVER2-NEXT:    vsubps %ymm2, %ymm3, %ymm2 # sched: [3:2.00]
; BTVER2-NEXT:    vmulps %ymm2, %ymm1, %ymm2 # sched: [2:2.00]
; BTVER2-NEXT:    vaddps %ymm2, %ymm1, %ymm1 # sched: [3:2.00]
; BTVER2-NEXT:    vmulps %ymm1, %ymm0, %ymm0 # sched: [2:2.00]
; BTVER2-NEXT:    vsubps %ymm0, %ymm3, %ymm0 # sched: [3:2.00]
; BTVER2-NEXT:    vmulps %ymm0, %ymm1, %ymm0 # sched: [2:2.00]
; BTVER2-NEXT:    vaddps %ymm0, %ymm1, %ymm0 # sched: [3:2.00]
; BTVER2-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [7:2.00]
; BTVER2-NEXT:    retq # sched: [4:1.00]
;
; SANDY-LABEL: v8f32_two_step2:
; SANDY:       # %bb.0:
; SANDY-NEXT:    vrcpps %ymm0, %ymm1 # sched: [7:2.00]
; SANDY-NEXT:    vmulps %ymm1, %ymm0, %ymm2 # sched: [5:1.00]
; SANDY-NEXT:    vmovaps {{.*#+}} ymm3 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0] sched: [7:0.50]
; SANDY-NEXT:    vsubps %ymm2, %ymm3, %ymm2 # sched: [3:1.00]
; SANDY-NEXT:    vmulps %ymm2, %ymm1, %ymm2 # sched: [5:1.00]
; SANDY-NEXT:    vaddps %ymm2, %ymm1, %ymm1 # sched: [3:1.00]
; SANDY-NEXT:    vmulps %ymm1, %ymm0, %ymm0 # sched: [5:1.00]
; SANDY-NEXT:    vsubps %ymm0, %ymm3, %ymm0 # sched: [3:1.00]
; SANDY-NEXT:    vmulps %ymm0, %ymm1, %ymm0 # sched: [5:1.00]
; SANDY-NEXT:    vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
; SANDY-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [12:1.00]
; SANDY-NEXT:    retq # sched: [1:1.00]
;
; HASWELL-LABEL: v8f32_two_step2:
; HASWELL:       # %bb.0:
; HASWELL-NEXT:    vrcpps %ymm0, %ymm1 # sched: [11:2.00]
; HASWELL-NEXT:    vbroadcastss {{.*#+}} ymm2 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0] sched: [7:0.50]
; HASWELL-NEXT:    vmovaps %ymm1, %ymm3 # sched: [1:1.00]
; HASWELL-NEXT:    vfnmadd213ps {{.*#+}} ymm3 = -(ymm0 * ymm3) + ymm2 sched: [5:0.50]
; HASWELL-NEXT:    vfmadd132ps {{.*#+}} ymm3 = (ymm3 * ymm1) + ymm1 sched: [5:0.50]
; HASWELL-NEXT:    vfnmadd213ps {{.*#+}} ymm0 = -(ymm3 * ymm0) + ymm2 sched: [5:0.50]
; HASWELL-NEXT:    vfmadd132ps {{.*#+}} ymm0 = (ymm0 * ymm3) + ymm3 sched: [5:0.50]
; HASWELL-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [12:0.50]
; HASWELL-NEXT:    retq # sched: [7:1.00]
;
; HASWELL-NO-FMA-LABEL: v8f32_two_step2:
; HASWELL-NO-FMA:       # %bb.0:
; HASWELL-NO-FMA-NEXT:    vrcpps %ymm0, %ymm1 # sched: [11:2.00]
; HASWELL-NO-FMA-NEXT:    vmulps %ymm1, %ymm0, %ymm2 # sched: [5:0.50]
; HASWELL-NO-FMA-NEXT:    vbroadcastss {{.*#+}} ymm3 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0] sched: [7:0.50]
; HASWELL-NO-FMA-NEXT:    vsubps %ymm2, %ymm3, %ymm2 # sched: [3:1.00]
; HASWELL-NO-FMA-NEXT:    vmulps %ymm2, %ymm1, %ymm2 # sched: [5:0.50]
; HASWELL-NO-FMA-NEXT:    vaddps %ymm2, %ymm1, %ymm1 # sched: [3:1.00]
; HASWELL-NO-FMA-NEXT:    vmulps %ymm1, %ymm0, %ymm0 # sched: [5:0.50]
; HASWELL-NO-FMA-NEXT:    vsubps %ymm0, %ymm3, %ymm0 # sched: [3:1.00]
; HASWELL-NO-FMA-NEXT:    vmulps %ymm0, %ymm1, %ymm0 # sched: [5:0.50]
; HASWELL-NO-FMA-NEXT:    vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
; HASWELL-NO-FMA-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [12:0.50]
; HASWELL-NO-FMA-NEXT:    retq # sched: [7:1.00]
;
; KNL-LABEL: v8f32_two_step2:
; KNL:       # %bb.0:
; KNL-NEXT:    vrcpps %ymm0, %ymm1 # sched: [11:2.00]
; KNL-NEXT:    vbroadcastss {{.*#+}} ymm2 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0] sched: [7:0.50]
; KNL-NEXT:    vmovaps %ymm1, %ymm3 # sched: [1:1.00]
; KNL-NEXT:    vfnmadd213ps {{.*#+}} ymm3 = -(ymm0 * ymm3) + ymm2 sched: [5:0.50]
; KNL-NEXT:    vfmadd132ps {{.*#+}} ymm3 = (ymm3 * ymm1) + ymm1 sched: [5:0.50]
; KNL-NEXT:    vfnmadd213ps {{.*#+}} ymm0 = -(ymm3 * ymm0) + ymm2 sched: [5:0.50]
; KNL-NEXT:    vfmadd132ps {{.*#+}} ymm0 = (ymm0 * ymm3) + ymm3 sched: [5:0.50]
; KNL-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [12:0.50]
; KNL-NEXT:    retq # sched: [7:1.00]
;
; SKX-LABEL: v8f32_two_step2:
; SKX:       # %bb.0:
; SKX-NEXT:    vrcpps %ymm0, %ymm1 # sched: [4:1.00]
; SKX-NEXT:    vbroadcastss {{.*#+}} ymm2 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0] sched: [7:0.50]
; SKX-NEXT:    vmovaps %ymm1, %ymm3 # sched: [1:0.33]
; SKX-NEXT:    vfnmadd213ps {{.*#+}} ymm3 = -(ymm0 * ymm3) + ymm2 sched: [4:0.50]
; SKX-NEXT:    vfmadd132ps {{.*#+}} ymm3 = (ymm3 * ymm1) + ymm1 sched: [4:0.50]
; SKX-NEXT:    vfnmadd213ps {{.*#+}} ymm0 = -(ymm3 * ymm0) + ymm2 sched: [4:0.50]
; SKX-NEXT:    vfmadd132ps {{.*#+}} ymm0 = (ymm0 * ymm3) + ymm3 sched: [4:0.50]
; SKX-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [11:0.50]
; SKX-NEXT:    retq # sched: [7:1.00]
  %div = fdiv fast <8 x float> <float 1.0, float 2.0, float 3.0, float 4.0, float 5.0, float 6.0, float 7.0, float 8.0>, %x
  ret <8 x float> %div
}

define <8 x float> @v8f32_no_step(<8 x float> %x) #3 {
; SSE-LABEL: v8f32_no_step:
; SSE:       # %bb.0:
; SSE-NEXT:    rcpps %xmm0, %xmm0
; SSE-NEXT:    rcpps %xmm1, %xmm1
; SSE-NEXT:    retq
;
; AVX-RECIP-LABEL: v8f32_no_step:
; AVX-RECIP:       # %bb.0:
; AVX-RECIP-NEXT:    vrcpps %ymm0, %ymm0
; AVX-RECIP-NEXT:    retq
;
; FMA-RECIP-LABEL: v8f32_no_step:
; FMA-RECIP:       # %bb.0:
; FMA-RECIP-NEXT:    vrcpps %ymm0, %ymm0
; FMA-RECIP-NEXT:    retq
;
; BDVER2-LABEL: v8f32_no_step:
; BDVER2:       # %bb.0:
; BDVER2-NEXT:    vrcpps %ymm0, %ymm0 # sched: [5:2.00]
; BDVER2-NEXT:    retq # sched: [5:1.00]
;
; BTVER2-LABEL: v8f32_no_step:
; BTVER2:       # %bb.0:
; BTVER2-NEXT:    vrcpps %ymm0, %ymm0 # sched: [2:2.00]
; BTVER2-NEXT:    retq # sched: [4:1.00]
;
; SANDY-LABEL: v8f32_no_step:
; SANDY:       # %bb.0:
; SANDY-NEXT:    vrcpps %ymm0, %ymm0 # sched: [7:2.00]
; SANDY-NEXT:    retq # sched: [1:1.00]
;
; HASWELL-LABEL: v8f32_no_step:
; HASWELL:       # %bb.0:
; HASWELL-NEXT:    vrcpps %ymm0, %ymm0 # sched: [11:2.00]
; HASWELL-NEXT:    retq # sched: [7:1.00]
;
; HASWELL-NO-FMA-LABEL: v8f32_no_step:
; HASWELL-NO-FMA:       # %bb.0:
; HASWELL-NO-FMA-NEXT:    vrcpps %ymm0, %ymm0 # sched: [11:2.00]
; HASWELL-NO-FMA-NEXT:    retq # sched: [7:1.00]
;
; KNL-LABEL: v8f32_no_step:
; KNL:       # %bb.0:
; KNL-NEXT:    vrcpps %ymm0, %ymm0 # sched: [11:2.00]
; KNL-NEXT:    retq # sched: [7:1.00]
;
; SKX-LABEL: v8f32_no_step:
; SKX:       # %bb.0:
; SKX-NEXT:    vrcpps %ymm0, %ymm0 # sched: [4:1.00]
; SKX-NEXT:    retq # sched: [7:1.00]
  %div = fdiv fast <8 x float> <float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0>, %x
  ret <8 x float> %div
}

define <8 x float> @v8f32_no_step2(<8 x float> %x) #3 {
; SSE-LABEL: v8f32_no_step2:
; SSE:       # %bb.0:
; SSE-NEXT:    rcpps %xmm1, %xmm1
; SSE-NEXT:    rcpps %xmm0, %xmm0
; SSE-NEXT:    mulps {{.*}}(%rip), %xmm0
; SSE-NEXT:    mulps {{.*}}(%rip), %xmm1
; SSE-NEXT:    retq
;
; AVX-RECIP-LABEL: v8f32_no_step2:
; AVX-RECIP:       # %bb.0:
; AVX-RECIP-NEXT:    vrcpps %ymm0, %ymm0
; AVX-RECIP-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0
; AVX-RECIP-NEXT:    retq
;
; FMA-RECIP-LABEL: v8f32_no_step2:
; FMA-RECIP:       # %bb.0:
; FMA-RECIP-NEXT:    vrcpps %ymm0, %ymm0
; FMA-RECIP-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0
; FMA-RECIP-NEXT:    retq
;
; BDVER2-LABEL: v8f32_no_step2:
; BDVER2:       # %bb.0:
; BDVER2-NEXT:    vrcpps %ymm0, %ymm0 # sched: [5:2.00]
; BDVER2-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [10:2.00]
; BDVER2-NEXT:    retq # sched: [5:1.00]
;
; BTVER2-LABEL: v8f32_no_step2:
; BTVER2:       # %bb.0:
; BTVER2-NEXT:    vrcpps %ymm0, %ymm0 # sched: [2:2.00]
; BTVER2-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [7:2.00]
; BTVER2-NEXT:    retq # sched: [4:1.00]
;
; SANDY-LABEL: v8f32_no_step2:
; SANDY:       # %bb.0:
; SANDY-NEXT:    vrcpps %ymm0, %ymm0 # sched: [7:2.00]
; SANDY-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [12:1.00]
; SANDY-NEXT:    retq # sched: [1:1.00]
;
; HASWELL-LABEL: v8f32_no_step2:
; HASWELL:       # %bb.0:
; HASWELL-NEXT:    vrcpps %ymm0, %ymm0 # sched: [11:2.00]
; HASWELL-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [12:0.50]
; HASWELL-NEXT:    retq # sched: [7:1.00]
;
; HASWELL-NO-FMA-LABEL: v8f32_no_step2:
; HASWELL-NO-FMA:       # %bb.0:
; HASWELL-NO-FMA-NEXT:    vrcpps %ymm0, %ymm0 # sched: [11:2.00]
; HASWELL-NO-FMA-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [12:0.50]
; HASWELL-NO-FMA-NEXT:    retq # sched: [7:1.00]
;
; KNL-LABEL: v8f32_no_step2:
; KNL:       # %bb.0:
; KNL-NEXT:    vrcpps %ymm0, %ymm0 # sched: [11:2.00]
; KNL-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [12:0.50]
; KNL-NEXT:    retq # sched: [7:1.00]
;
; SKX-LABEL: v8f32_no_step2:
; SKX:       # %bb.0:
; SKX-NEXT:    vrcpps %ymm0, %ymm0 # sched: [4:1.00]
; SKX-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [11:0.50]
; SKX-NEXT:    retq # sched: [7:1.00]
  %div = fdiv fast <8 x float> <float 1.0, float 2.0, float 3.0, float 4.0, float 5.0, float 6.0, float 7.0, float 8.0>, %x
  ret <8 x float> %div
}

define <16 x float> @v16f32_one_step2(<16 x float> %x) #1 {
; SSE-LABEL: v16f32_one_step2:
; SSE:       # %bb.0:
; SSE-NEXT:    movaps %xmm3, %xmm4
; SSE-NEXT:    movaps %xmm2, %xmm5
; SSE-NEXT:    movaps %xmm0, %xmm6
; SSE-NEXT:    rcpps %xmm3, %xmm2
; SSE-NEXT:    mulps %xmm2, %xmm4
; SSE-NEXT:    movaps {{.*#+}} xmm0 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0]
; SSE-NEXT:    movaps %xmm0, %xmm3
; SSE-NEXT:    subps %xmm4, %xmm3
; SSE-NEXT:    mulps %xmm2, %xmm3
; SSE-NEXT:    addps %xmm2, %xmm3
; SSE-NEXT:    rcpps %xmm5, %xmm4
; SSE-NEXT:    mulps %xmm4, %xmm5
; SSE-NEXT:    movaps %xmm0, %xmm2
; SSE-NEXT:    subps %xmm5, %xmm2
; SSE-NEXT:    mulps %xmm4, %xmm2
; SSE-NEXT:    addps %xmm4, %xmm2
; SSE-NEXT:    rcpps %xmm1, %xmm5
; SSE-NEXT:    mulps %xmm5, %xmm1
; SSE-NEXT:    movaps %xmm0, %xmm4
; SSE-NEXT:    subps %xmm1, %xmm4
; SSE-NEXT:    mulps %xmm5, %xmm4
; SSE-NEXT:    addps %xmm5, %xmm4
; SSE-NEXT:    rcpps %xmm6, %xmm1
; SSE-NEXT:    mulps %xmm1, %xmm6
; SSE-NEXT:    subps %xmm6, %xmm0
; SSE-NEXT:    mulps %xmm1, %xmm0
; SSE-NEXT:    addps %xmm1, %xmm0
; SSE-NEXT:    mulps {{.*}}(%rip), %xmm0
; SSE-NEXT:    mulps {{.*}}(%rip), %xmm4
; SSE-NEXT:    mulps {{.*}}(%rip), %xmm2
; SSE-NEXT:    mulps {{.*}}(%rip), %xmm3
; SSE-NEXT:    movaps %xmm4, %xmm1
; SSE-NEXT:    retq
;
; AVX-RECIP-LABEL: v16f32_one_step2:
; AVX-RECIP:       # %bb.0:
; AVX-RECIP-NEXT:    vrcpps %ymm1, %ymm2
; AVX-RECIP-NEXT:    vmulps %ymm2, %ymm1, %ymm1
; AVX-RECIP-NEXT:    vmovaps {{.*#+}} ymm3 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0]
; AVX-RECIP-NEXT:    vsubps %ymm1, %ymm3, %ymm1
; AVX-RECIP-NEXT:    vmulps %ymm1, %ymm2, %ymm1
; AVX-RECIP-NEXT:    vaddps %ymm1, %ymm2, %ymm1
; AVX-RECIP-NEXT:    vrcpps %ymm0, %ymm2
; AVX-RECIP-NEXT:    vmulps %ymm2, %ymm0, %ymm0
; AVX-RECIP-NEXT:    vsubps %ymm0, %ymm3, %ymm0
; AVX-RECIP-NEXT:    vmulps %ymm0, %ymm2, %ymm0
; AVX-RECIP-NEXT:    vaddps %ymm0, %ymm2, %ymm0
; AVX-RECIP-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0
; AVX-RECIP-NEXT:    vmulps {{.*}}(%rip), %ymm1, %ymm1
; AVX-RECIP-NEXT:    retq
;
; FMA-RECIP-LABEL: v16f32_one_step2:
; FMA-RECIP:       # %bb.0:
; FMA-RECIP-NEXT:    vrcpps %ymm1, %ymm2
; FMA-RECIP-NEXT:    vmovaps {{.*#+}} ymm3 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0]
; FMA-RECIP-NEXT:    vfnmadd213ps {{.*#+}} ymm1 = -(ymm2 * ymm1) + ymm3
; FMA-RECIP-NEXT:    vfmadd132ps {{.*#+}} ymm1 = (ymm1 * ymm2) + ymm2
; FMA-RECIP-NEXT:    vrcpps %ymm0, %ymm2
; FMA-RECIP-NEXT:    vfnmadd213ps {{.*#+}} ymm0 = -(ymm2 * ymm0) + ymm3
; FMA-RECIP-NEXT:    vfmadd132ps {{.*#+}} ymm0 = (ymm0 * ymm2) + ymm2
; FMA-RECIP-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0
; FMA-RECIP-NEXT:    vmulps {{.*}}(%rip), %ymm1, %ymm1
; FMA-RECIP-NEXT:    retq
;
; BDVER2-LABEL: v16f32_one_step2:
; BDVER2:       # %bb.0:
; BDVER2-NEXT:    vrcpps %ymm1, %ymm2 # sched: [5:2.00]
; BDVER2-NEXT:    vmovaps {{.*#+}} ymm3 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0] sched: [5:0.50]
; BDVER2-NEXT:    vrcpps %ymm0, %ymm4 # sched: [5:2.00]
; BDVER2-NEXT:    vfnmaddps %ymm3, %ymm2, %ymm1, %ymm1 # sched: [5:0.50]
; BDVER2-NEXT:    vfnmaddps %ymm3, %ymm4, %ymm0, %ymm0 # sched: [5:0.50]
; BDVER2-NEXT:    vfmaddps %ymm2, %ymm1, %ymm2, %ymm1 # sched: [5:0.50]
; BDVER2-NEXT:    vfmaddps %ymm4, %ymm0, %ymm4, %ymm0 # sched: [5:0.50]
; BDVER2-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [10:2.00]
; BDVER2-NEXT:    vmulps {{.*}}(%rip), %ymm1, %ymm1 # sched: [10:2.00]
; BDVER2-NEXT:    retq # sched: [5:1.00]
;
; BTVER2-LABEL: v16f32_one_step2:
; BTVER2:       # %bb.0:
; BTVER2-NEXT:    vmovaps {{.*#+}} ymm3 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0] sched: [5:1.00]
; BTVER2-NEXT:    vrcpps %ymm1, %ymm2 # sched: [2:2.00]
; BTVER2-NEXT:    vrcpps %ymm0, %ymm4 # sched: [2:2.00]
; BTVER2-NEXT:    vmulps %ymm2, %ymm1, %ymm1 # sched: [2:2.00]
; BTVER2-NEXT:    vmulps %ymm4, %ymm0, %ymm0 # sched: [2:2.00]
; BTVER2-NEXT:    vsubps %ymm1, %ymm3, %ymm1 # sched: [3:2.00]
; BTVER2-NEXT:    vsubps %ymm0, %ymm3, %ymm0 # sched: [3:2.00]
; BTVER2-NEXT:    vmulps %ymm1, %ymm2, %ymm1 # sched: [2:2.00]
; BTVER2-NEXT:    vmulps %ymm0, %ymm4, %ymm0 # sched: [2:2.00]
; BTVER2-NEXT:    vaddps %ymm1, %ymm2, %ymm1 # sched: [3:2.00]
; BTVER2-NEXT:    vaddps %ymm0, %ymm4, %ymm0 # sched: [3:2.00]
; BTVER2-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [7:2.00]
; BTVER2-NEXT:    vmulps {{.*}}(%rip), %ymm1, %ymm1 # sched: [7:2.00]
; BTVER2-NEXT:    retq # sched: [4:1.00]
;
; SANDY-LABEL: v16f32_one_step2:
; SANDY:       # %bb.0:
; SANDY-NEXT:    vrcpps %ymm1, %ymm2 # sched: [7:2.00]
; SANDY-NEXT:    vmulps %ymm2, %ymm1, %ymm1 # sched: [5:1.00]
; SANDY-NEXT:    vmovaps {{.*#+}} ymm3 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0] sched: [7:0.50]
; SANDY-NEXT:    vsubps %ymm1, %ymm3, %ymm1 # sched: [3:1.00]
; SANDY-NEXT:    vmulps %ymm1, %ymm2, %ymm1 # sched: [5:1.00]
; SANDY-NEXT:    vaddps %ymm1, %ymm2, %ymm1 # sched: [3:1.00]
; SANDY-NEXT:    vrcpps %ymm0, %ymm2 # sched: [7:2.00]
; SANDY-NEXT:    vmulps %ymm2, %ymm0, %ymm0 # sched: [5:1.00]
; SANDY-NEXT:    vsubps %ymm0, %ymm3, %ymm0 # sched: [3:1.00]
; SANDY-NEXT:    vmulps %ymm0, %ymm2, %ymm0 # sched: [5:1.00]
; SANDY-NEXT:    vaddps %ymm0, %ymm2, %ymm0 # sched: [3:1.00]
; SANDY-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [12:1.00]
; SANDY-NEXT:    vmulps {{.*}}(%rip), %ymm1, %ymm1 # sched: [12:1.00]
; SANDY-NEXT:    retq # sched: [1:1.00]
;
; HASWELL-LABEL: v16f32_one_step2:
; HASWELL:       # %bb.0:
; HASWELL-NEXT:    vrcpps %ymm1, %ymm2 # sched: [11:2.00]
; HASWELL-NEXT:    vbroadcastss {{.*#+}} ymm3 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0] sched: [7:0.50]
; HASWELL-NEXT:    vrcpps %ymm0, %ymm4 # sched: [11:2.00]
; HASWELL-NEXT:    vfnmadd213ps {{.*#+}} ymm1 = -(ymm2 * ymm1) + ymm3 sched: [5:0.50]
; HASWELL-NEXT:    vfmadd132ps {{.*#+}} ymm1 = (ymm1 * ymm2) + ymm2 sched: [5:0.50]
; HASWELL-NEXT:    vfnmadd213ps {{.*#+}} ymm0 = -(ymm4 * ymm0) + ymm3 sched: [5:0.50]
; HASWELL-NEXT:    vfmadd132ps {{.*#+}} ymm0 = (ymm0 * ymm4) + ymm4 sched: [5:0.50]
; HASWELL-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [12:0.50]
; HASWELL-NEXT:    vmulps {{.*}}(%rip), %ymm1, %ymm1 # sched: [12:0.50]
; HASWELL-NEXT:    retq # sched: [7:1.00]
;
; HASWELL-NO-FMA-LABEL: v16f32_one_step2:
; HASWELL-NO-FMA:       # %bb.0:
; HASWELL-NO-FMA-NEXT:    vrcpps %ymm1, %ymm2 # sched: [11:2.00]
; HASWELL-NO-FMA-NEXT:    vmulps %ymm2, %ymm1, %ymm1 # sched: [5:0.50]
; HASWELL-NO-FMA-NEXT:    vbroadcastss {{.*#+}} ymm3 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0] sched: [7:0.50]
; HASWELL-NO-FMA-NEXT:    vsubps %ymm1, %ymm3, %ymm1 # sched: [3:1.00]
; HASWELL-NO-FMA-NEXT:    vmulps %ymm1, %ymm2, %ymm1 # sched: [5:0.50]
; HASWELL-NO-FMA-NEXT:    vaddps %ymm1, %ymm2, %ymm1 # sched: [3:1.00]
; HASWELL-NO-FMA-NEXT:    vrcpps %ymm0, %ymm2 # sched: [11:2.00]
; HASWELL-NO-FMA-NEXT:    vmulps %ymm2, %ymm0, %ymm0 # sched: [5:0.50]
; HASWELL-NO-FMA-NEXT:    vsubps %ymm0, %ymm3, %ymm0 # sched: [3:1.00]
; HASWELL-NO-FMA-NEXT:    vmulps %ymm0, %ymm2, %ymm0 # sched: [5:0.50]
; HASWELL-NO-FMA-NEXT:    vaddps %ymm0, %ymm2, %ymm0 # sched: [3:1.00]
; HASWELL-NO-FMA-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [12:0.50]
; HASWELL-NO-FMA-NEXT:    vmulps {{.*}}(%rip), %ymm1, %ymm1 # sched: [12:0.50]
; HASWELL-NO-FMA-NEXT:    retq # sched: [7:1.00]
;
; KNL-LABEL: v16f32_one_step2:
; KNL:       # %bb.0:
; KNL-NEXT:    vrcp14ps %zmm0, %zmm1 # sched: [11:2.00]
; KNL-NEXT:    vfnmadd213ps {{.*#+}} zmm0 = -(zmm1 * zmm0) + mem sched: [12:0.50]
; KNL-NEXT:    vfmadd132ps {{.*#+}} zmm0 = (zmm0 * zmm1) + zmm1 sched: [5:0.50]
; KNL-NEXT:    vmulps {{.*}}(%rip), %zmm0, %zmm0 # sched: [12:0.50]
; KNL-NEXT:    retq # sched: [7:1.00]
;
; SKX-LABEL: v16f32_one_step2:
; SKX:       # %bb.0:
; SKX-NEXT:    vrcp14ps %zmm0, %zmm1 # sched: [4:2.00]
; SKX-NEXT:    vfnmadd213ps {{.*#+}} zmm0 = -(zmm1 * zmm0) + mem sched: [11:0.50]
; SKX-NEXT:    vfmadd132ps {{.*#+}} zmm0 = (zmm0 * zmm1) + zmm1 sched: [4:0.50]
; SKX-NEXT:    vmulps {{.*}}(%rip), %zmm0, %zmm0 # sched: [11:0.50]
; SKX-NEXT:    retq # sched: [7:1.00]
  %div = fdiv fast <16 x float> <float 1.0, float 2.0, float 3.0, float 4.0, float 5.0, float 6.0, float 7.0, float 8.0, float 9.0, float 10.0, float 11.0, float 12.0, float 13.0, float 14.0, float 15.0, float 16.0>, %x
  ret <16 x float> %div
}

define <16 x float> @v16f32_one_step_2_divs(<16 x float> %x) #1 {
; SSE-LABEL: v16f32_one_step_2_divs:
; SSE:       # %bb.0:
; SSE-NEXT:    rcpps %xmm0, %xmm6
; SSE-NEXT:    mulps %xmm6, %xmm0
; SSE-NEXT:    movaps {{.*#+}} xmm4 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0]
; SSE-NEXT:    movaps %xmm4, %xmm5
; SSE-NEXT:    subps %xmm0, %xmm5
; SSE-NEXT:    mulps %xmm6, %xmm5
; SSE-NEXT:    addps %xmm6, %xmm5
; SSE-NEXT:    rcpps %xmm1, %xmm0
; SSE-NEXT:    mulps %xmm0, %xmm1
; SSE-NEXT:    movaps %xmm4, %xmm6
; SSE-NEXT:    subps %xmm1, %xmm6
; SSE-NEXT:    mulps %xmm0, %xmm6
; SSE-NEXT:    addps %xmm0, %xmm6
; SSE-NEXT:    rcpps %xmm2, %xmm0
; SSE-NEXT:    mulps %xmm0, %xmm2
; SSE-NEXT:    movaps %xmm4, %xmm7
; SSE-NEXT:    subps %xmm2, %xmm7
; SSE-NEXT:    mulps %xmm0, %xmm7
; SSE-NEXT:    addps %xmm0, %xmm7
; SSE-NEXT:    rcpps %xmm3, %xmm0
; SSE-NEXT:    mulps %xmm0, %xmm3
; SSE-NEXT:    subps %xmm3, %xmm4
; SSE-NEXT:    mulps %xmm0, %xmm4
; SSE-NEXT:    addps %xmm0, %xmm4
; SSE-NEXT:    movaps {{.*#+}} xmm3 = [1.3E+1,1.4E+1,1.5E+1,1.6E+1]
; SSE-NEXT:    mulps %xmm4, %xmm3
; SSE-NEXT:    movaps {{.*#+}} xmm2 = [9.0E+0,1.0E+1,1.1E+1,1.2E+1]
; SSE-NEXT:    mulps %xmm7, %xmm2
; SSE-NEXT:    movaps {{.*#+}} xmm1 = [5.0E+0,6.0E+0,7.0E+0,8.0E+0]
; SSE-NEXT:    mulps %xmm6, %xmm1
; SSE-NEXT:    movaps {{.*#+}} xmm0 = [1.0E+0,2.0E+0,3.0E+0,4.0E+0]
; SSE-NEXT:    mulps %xmm5, %xmm0
; SSE-NEXT:    mulps %xmm5, %xmm0
; SSE-NEXT:    mulps %xmm6, %xmm1
; SSE-NEXT:    mulps %xmm7, %xmm2
; SSE-NEXT:    mulps %xmm4, %xmm3
; SSE-NEXT:    retq
;
; AVX-RECIP-LABEL: v16f32_one_step_2_divs:
; AVX-RECIP:       # %bb.0:
; AVX-RECIP-NEXT:    vrcpps %ymm0, %ymm2
; AVX-RECIP-NEXT:    vmulps %ymm2, %ymm0, %ymm0
; AVX-RECIP-NEXT:    vmovaps {{.*#+}} ymm3 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0]
; AVX-RECIP-NEXT:    vsubps %ymm0, %ymm3, %ymm0
; AVX-RECIP-NEXT:    vmulps %ymm0, %ymm2, %ymm0
; AVX-RECIP-NEXT:    vaddps %ymm0, %ymm2, %ymm0
; AVX-RECIP-NEXT:    vrcpps %ymm1, %ymm2
; AVX-RECIP-NEXT:    vmulps %ymm2, %ymm1, %ymm1
; AVX-RECIP-NEXT:    vsubps %ymm1, %ymm3, %ymm1
; AVX-RECIP-NEXT:    vmulps %ymm1, %ymm2, %ymm1
; AVX-RECIP-NEXT:    vaddps %ymm1, %ymm2, %ymm1
; AVX-RECIP-NEXT:    vmulps {{.*}}(%rip), %ymm1, %ymm2
; AVX-RECIP-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm3
; AVX-RECIP-NEXT:    vmulps %ymm0, %ymm3, %ymm0
; AVX-RECIP-NEXT:    vmulps %ymm1, %ymm2, %ymm1
; AVX-RECIP-NEXT:    retq
;
; FMA-RECIP-LABEL: v16f32_one_step_2_divs:
; FMA-RECIP:       # %bb.0:
; FMA-RECIP-NEXT:    vrcpps %ymm0, %ymm2
; FMA-RECIP-NEXT:    vmovaps {{.*#+}} ymm3 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0]
; FMA-RECIP-NEXT:    vfnmadd213ps {{.*#+}} ymm0 = -(ymm2 * ymm0) + ymm3
; FMA-RECIP-NEXT:    vfmadd132ps {{.*#+}} ymm0 = (ymm0 * ymm2) + ymm2
; FMA-RECIP-NEXT:    vrcpps %ymm1, %ymm2
; FMA-RECIP-NEXT:    vfnmadd213ps {{.*#+}} ymm1 = -(ymm2 * ymm1) + ymm3
; FMA-RECIP-NEXT:    vfmadd132ps {{.*#+}} ymm1 = (ymm1 * ymm2) + ymm2
; FMA-RECIP-NEXT:    vmulps {{.*}}(%rip), %ymm1, %ymm2
; FMA-RECIP-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm3
; FMA-RECIP-NEXT:    vmulps %ymm0, %ymm3, %ymm0
; FMA-RECIP-NEXT:    vmulps %ymm1, %ymm2, %ymm1
; FMA-RECIP-NEXT:    retq
;
; BDVER2-LABEL: v16f32_one_step_2_divs:
; BDVER2:       # %bb.0:
; BDVER2-NEXT:    vrcpps %ymm0, %ymm2 # sched: [5:2.00]
; BDVER2-NEXT:    vmovaps {{.*#+}} ymm3 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0] sched: [5:0.50]
; BDVER2-NEXT:    vfnmaddps %ymm3, %ymm2, %ymm0, %ymm0 # sched: [5:0.50]
; BDVER2-NEXT:    vfmaddps %ymm2, %ymm0, %ymm2, %ymm0 # sched: [5:0.50]
; BDVER2-NEXT:    vrcpps %ymm1, %ymm2 # sched: [5:2.00]
; BDVER2-NEXT:    vfnmaddps %ymm3, %ymm2, %ymm1, %ymm1 # sched: [5:0.50]
; BDVER2-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm3 # sched: [10:2.00]
; BDVER2-NEXT:    vfmaddps %ymm2, %ymm1, %ymm2, %ymm1 # sched: [5:0.50]
; BDVER2-NEXT:    vmulps {{.*}}(%rip), %ymm1, %ymm2 # sched: [10:2.00]
; BDVER2-NEXT:    vmulps %ymm0, %ymm3, %ymm0 # sched: [5:2.00]
; BDVER2-NEXT:    vmulps %ymm1, %ymm2, %ymm1 # sched: [5:2.00]
; BDVER2-NEXT:    retq # sched: [5:1.00]
;
; BTVER2-LABEL: v16f32_one_step_2_divs:
; BTVER2:       # %bb.0:
; BTVER2-NEXT:    vmovaps {{.*#+}} ymm3 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0] sched: [5:1.00]
; BTVER2-NEXT:    vrcpps %ymm0, %ymm2 # sched: [2:2.00]
; BTVER2-NEXT:    vmulps %ymm2, %ymm0, %ymm0 # sched: [2:2.00]
; BTVER2-NEXT:    vsubps %ymm0, %ymm3, %ymm0 # sched: [3:2.00]
; BTVER2-NEXT:    vmulps %ymm0, %ymm2, %ymm0 # sched: [2:2.00]
; BTVER2-NEXT:    vaddps %ymm0, %ymm2, %ymm0 # sched: [3:2.00]
; BTVER2-NEXT:    vrcpps %ymm1, %ymm2 # sched: [2:2.00]
; BTVER2-NEXT:    vmulps %ymm2, %ymm1, %ymm1 # sched: [2:2.00]
; BTVER2-NEXT:    vsubps %ymm1, %ymm3, %ymm1 # sched: [3:2.00]
; BTVER2-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm3 # sched: [7:2.00]
; BTVER2-NEXT:    vmulps %ymm1, %ymm2, %ymm1 # sched: [2:2.00]
; BTVER2-NEXT:    vaddps %ymm1, %ymm2, %ymm1 # sched: [3:2.00]
; BTVER2-NEXT:    vmulps {{.*}}(%rip), %ymm1, %ymm2 # sched: [7:2.00]
; BTVER2-NEXT:    vmulps %ymm0, %ymm3, %ymm0 # sched: [2:2.00]
; BTVER2-NEXT:    vmulps %ymm1, %ymm2, %ymm1 # sched: [2:2.00]
; BTVER2-NEXT:    retq # sched: [4:1.00]
;
; SANDY-LABEL: v16f32_one_step_2_divs:
; SANDY:       # %bb.0:
; SANDY-NEXT:    vrcpps %ymm0, %ymm2 # sched: [7:2.00]
; SANDY-NEXT:    vmulps %ymm2, %ymm0, %ymm0 # sched: [5:1.00]
; SANDY-NEXT:    vmovaps {{.*#+}} ymm3 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0] sched: [7:0.50]
; SANDY-NEXT:    vsubps %ymm0, %ymm3, %ymm0 # sched: [3:1.00]
; SANDY-NEXT:    vrcpps %ymm1, %ymm4 # sched: [7:2.00]
; SANDY-NEXT:    vmulps %ymm0, %ymm2, %ymm0 # sched: [5:1.00]
; SANDY-NEXT:    vaddps %ymm0, %ymm2, %ymm0 # sched: [3:1.00]
; SANDY-NEXT:    vmulps %ymm4, %ymm1, %ymm1 # sched: [5:1.00]
; SANDY-NEXT:    vsubps %ymm1, %ymm3, %ymm1 # sched: [3:1.00]
; SANDY-NEXT:    vmulps %ymm1, %ymm4, %ymm1 # sched: [5:1.00]
; SANDY-NEXT:    vaddps %ymm1, %ymm4, %ymm1 # sched: [3:1.00]
; SANDY-NEXT:    vmulps {{.*}}(%rip), %ymm1, %ymm2 # sched: [12:1.00]
; SANDY-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm3 # sched: [12:1.00]
; SANDY-NEXT:    vmulps %ymm0, %ymm3, %ymm0 # sched: [5:1.00]
; SANDY-NEXT:    vmulps %ymm1, %ymm2, %ymm1 # sched: [5:1.00]
; SANDY-NEXT:    retq # sched: [1:1.00]
;
; HASWELL-LABEL: v16f32_one_step_2_divs:
; HASWELL:       # %bb.0:
; HASWELL-NEXT:    vrcpps %ymm0, %ymm2 # sched: [11:2.00]
; HASWELL-NEXT:    vbroadcastss {{.*#+}} ymm3 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0] sched: [7:0.50]
; HASWELL-NEXT:    vfnmadd213ps {{.*#+}} ymm0 = -(ymm2 * ymm0) + ymm3 sched: [5:0.50]
; HASWELL-NEXT:    vfmadd132ps {{.*#+}} ymm0 = (ymm0 * ymm2) + ymm2 sched: [5:0.50]
; HASWELL-NEXT:    vrcpps %ymm1, %ymm2 # sched: [11:2.00]
; HASWELL-NEXT:    vfnmadd213ps {{.*#+}} ymm1 = -(ymm2 * ymm1) + ymm3 sched: [5:0.50]
; HASWELL-NEXT:    vfmadd132ps {{.*#+}} ymm1 = (ymm1 * ymm2) + ymm2 sched: [5:0.50]
; HASWELL-NEXT:    vmulps {{.*}}(%rip), %ymm1, %ymm2 # sched: [12:0.50]
; HASWELL-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm3 # sched: [12:0.50]
; HASWELL-NEXT:    vmulps %ymm0, %ymm3, %ymm0 # sched: [5:0.50]
; HASWELL-NEXT:    vmulps %ymm1, %ymm2, %ymm1 # sched: [5:0.50]
; HASWELL-NEXT:    retq # sched: [7:1.00]
;
; HASWELL-NO-FMA-LABEL: v16f32_one_step_2_divs:
; HASWELL-NO-FMA:       # %bb.0:
; HASWELL-NO-FMA-NEXT:    vrcpps %ymm0, %ymm2 # sched: [11:2.00]
; HASWELL-NO-FMA-NEXT:    vmulps %ymm2, %ymm0, %ymm0 # sched: [5:0.50]
; HASWELL-NO-FMA-NEXT:    vbroadcastss {{.*#+}} ymm3 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0] sched: [7:0.50]
; HASWELL-NO-FMA-NEXT:    vsubps %ymm0, %ymm3, %ymm0 # sched: [3:1.00]
; HASWELL-NO-FMA-NEXT:    vrcpps %ymm1, %ymm4 # sched: [11:2.00]
; HASWELL-NO-FMA-NEXT:    vmulps %ymm0, %ymm2, %ymm0 # sched: [5:0.50]
; HASWELL-NO-FMA-NEXT:    vaddps %ymm0, %ymm2, %ymm0 # sched: [3:1.00]
; HASWELL-NO-FMA-NEXT:    vmulps %ymm4, %ymm1, %ymm1 # sched: [5:0.50]
; HASWELL-NO-FMA-NEXT:    vsubps %ymm1, %ymm3, %ymm1 # sched: [3:1.00]
; HASWELL-NO-FMA-NEXT:    vmulps %ymm1, %ymm4, %ymm1 # sched: [5:0.50]
; HASWELL-NO-FMA-NEXT:    vaddps %ymm1, %ymm4, %ymm1 # sched: [3:1.00]
; HASWELL-NO-FMA-NEXT:    vmulps {{.*}}(%rip), %ymm1, %ymm2 # sched: [12:0.50]
; HASWELL-NO-FMA-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm3 # sched: [12:0.50]
; HASWELL-NO-FMA-NEXT:    vmulps %ymm0, %ymm3, %ymm0 # sched: [5:0.50]
; HASWELL-NO-FMA-NEXT:    vmulps %ymm1, %ymm2, %ymm1 # sched: [5:0.50]
; HASWELL-NO-FMA-NEXT:    retq # sched: [7:1.00]
;
; KNL-LABEL: v16f32_one_step_2_divs:
; KNL:       # %bb.0:
; KNL-NEXT:    vrcp14ps %zmm0, %zmm1 # sched: [11:2.00]
; KNL-NEXT:    vfnmadd213ps {{.*#+}} zmm0 = -(zmm1 * zmm0) + mem sched: [12:0.50]
; KNL-NEXT:    vfmadd132ps {{.*#+}} zmm0 = (zmm0 * zmm1) + zmm1 sched: [5:0.50]
; KNL-NEXT:    vmulps {{.*}}(%rip), %zmm0, %zmm1 # sched: [12:0.50]
; KNL-NEXT:    vmulps %zmm0, %zmm1, %zmm0 # sched: [5:0.50]
; KNL-NEXT:    retq # sched: [7:1.00]
;
; SKX-LABEL: v16f32_one_step_2_divs:
; SKX:       # %bb.0:
; SKX-NEXT:    vrcp14ps %zmm0, %zmm1 # sched: [4:2.00]
; SKX-NEXT:    vfnmadd213ps {{.*#+}} zmm0 = -(zmm1 * zmm0) + mem sched: [11:0.50]
; SKX-NEXT:    vfmadd132ps {{.*#+}} zmm0 = (zmm0 * zmm1) + zmm1 sched: [4:0.50]
; SKX-NEXT:    vmulps {{.*}}(%rip), %zmm0, %zmm1 # sched: [11:0.50]
; SKX-NEXT:    vmulps %zmm0, %zmm1, %zmm0 # sched: [4:0.50]
; SKX-NEXT:    retq # sched: [7:1.00]
  %div = fdiv fast <16 x float> <float 1.0, float 2.0, float 3.0, float 4.0, float 5.0, float 6.0, float 7.0, float 8.0, float 9.0, float 10.0, float 11.0, float 12.0, float 13.0, float 14.0, float 15.0, float 16.0>, %x
  %div2 = fdiv fast <16 x float> %div, %x
  ret <16 x float> %div2
}

define <16 x float> @v16f32_two_step2(<16 x float> %x) #2 {
; SSE-LABEL: v16f32_two_step2:
; SSE:       # %bb.0:
; SSE-NEXT:    movaps %xmm3, %xmm6
; SSE-NEXT:    movaps %xmm2, %xmm5
; SSE-NEXT:    movaps %xmm0, %xmm4
; SSE-NEXT:    rcpps %xmm3, %xmm2
; SSE-NEXT:    mulps %xmm2, %xmm3
; SSE-NEXT:    movaps {{.*#+}} xmm0 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0]
; SSE-NEXT:    movaps %xmm0, %xmm7
; SSE-NEXT:    subps %xmm3, %xmm7
; SSE-NEXT:    mulps %xmm2, %xmm7
; SSE-NEXT:    addps %xmm2, %xmm7
; SSE-NEXT:    mulps %xmm7, %xmm6
; SSE-NEXT:    movaps %xmm0, %xmm3
; SSE-NEXT:    subps %xmm6, %xmm3
; SSE-NEXT:    mulps %xmm7, %xmm3
; SSE-NEXT:    addps %xmm7, %xmm3
; SSE-NEXT:    rcpps %xmm5, %xmm2
; SSE-NEXT:    movaps %xmm5, %xmm6
; SSE-NEXT:    mulps %xmm2, %xmm6
; SSE-NEXT:    movaps %xmm0, %xmm7
; SSE-NEXT:    subps %xmm6, %xmm7
; SSE-NEXT:    mulps %xmm2, %xmm7
; SSE-NEXT:    addps %xmm2, %xmm7
; SSE-NEXT:    mulps %xmm7, %xmm5
; SSE-NEXT:    movaps %xmm0, %xmm2
; SSE-NEXT:    subps %xmm5, %xmm2
; SSE-NEXT:    mulps %xmm7, %xmm2
; SSE-NEXT:    addps %xmm7, %xmm2
; SSE-NEXT:    rcpps %xmm1, %xmm5
; SSE-NEXT:    movaps %xmm1, %xmm6
; SSE-NEXT:    mulps %xmm5, %xmm6
; SSE-NEXT:    movaps %xmm0, %xmm7
; SSE-NEXT:    subps %xmm6, %xmm7
; SSE-NEXT:    mulps %xmm5, %xmm7
; SSE-NEXT:    addps %xmm5, %xmm7
; SSE-NEXT:    mulps %xmm7, %xmm1
; SSE-NEXT:    movaps %xmm0, %xmm5
; SSE-NEXT:    subps %xmm1, %xmm5
; SSE-NEXT:    mulps %xmm7, %xmm5
; SSE-NEXT:    addps %xmm7, %xmm5
; SSE-NEXT:    rcpps %xmm4, %xmm1
; SSE-NEXT:    movaps %xmm4, %xmm6
; SSE-NEXT:    mulps %xmm1, %xmm6
; SSE-NEXT:    movaps %xmm0, %xmm7
; SSE-NEXT:    subps %xmm6, %xmm7
; SSE-NEXT:    mulps %xmm1, %xmm7
; SSE-NEXT:    addps %xmm1, %xmm7
; SSE-NEXT:    mulps %xmm7, %xmm4
; SSE-NEXT:    subps %xmm4, %xmm0
; SSE-NEXT:    mulps %xmm7, %xmm0
; SSE-NEXT:    addps %xmm7, %xmm0
; SSE-NEXT:    mulps {{.*}}(%rip), %xmm0
; SSE-NEXT:    mulps {{.*}}(%rip), %xmm5
; SSE-NEXT:    mulps {{.*}}(%rip), %xmm2
; SSE-NEXT:    mulps {{.*}}(%rip), %xmm3
; SSE-NEXT:    movaps %xmm5, %xmm1
; SSE-NEXT:    retq
;
; AVX-RECIP-LABEL: v16f32_two_step2:
; AVX-RECIP:       # %bb.0:
; AVX-RECIP-NEXT:    vrcpps %ymm1, %ymm2
; AVX-RECIP-NEXT:    vmulps %ymm2, %ymm1, %ymm3
; AVX-RECIP-NEXT:    vmovaps {{.*#+}} ymm4 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0]
; AVX-RECIP-NEXT:    vsubps %ymm3, %ymm4, %ymm3
; AVX-RECIP-NEXT:    vmulps %ymm3, %ymm2, %ymm3
; AVX-RECIP-NEXT:    vaddps %ymm3, %ymm2, %ymm2
; AVX-RECIP-NEXT:    vmulps %ymm2, %ymm1, %ymm1
; AVX-RECIP-NEXT:    vsubps %ymm1, %ymm4, %ymm1
; AVX-RECIP-NEXT:    vmulps %ymm1, %ymm2, %ymm1
; AVX-RECIP-NEXT:    vaddps %ymm1, %ymm2, %ymm1
; AVX-RECIP-NEXT:    vrcpps %ymm0, %ymm2
; AVX-RECIP-NEXT:    vmulps %ymm2, %ymm0, %ymm3
; AVX-RECIP-NEXT:    vsubps %ymm3, %ymm4, %ymm3
; AVX-RECIP-NEXT:    vmulps %ymm3, %ymm2, %ymm3
; AVX-RECIP-NEXT:    vaddps %ymm3, %ymm2, %ymm2
; AVX-RECIP-NEXT:    vmulps %ymm2, %ymm0, %ymm0
; AVX-RECIP-NEXT:    vsubps %ymm0, %ymm4, %ymm0
; AVX-RECIP-NEXT:    vmulps %ymm0, %ymm2, %ymm0
; AVX-RECIP-NEXT:    vaddps %ymm0, %ymm2, %ymm0
; AVX-RECIP-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0
; AVX-RECIP-NEXT:    vmulps {{.*}}(%rip), %ymm1, %ymm1
; AVX-RECIP-NEXT:    retq
;
; FMA-RECIP-LABEL: v16f32_two_step2:
; FMA-RECIP:       # %bb.0:
; FMA-RECIP-NEXT:    vrcpps %ymm1, %ymm2
; FMA-RECIP-NEXT:    vmovaps {{.*#+}} ymm3 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0]
; FMA-RECIP-NEXT:    vmovaps %ymm2, %ymm4
; FMA-RECIP-NEXT:    vfnmadd213ps {{.*#+}} ymm4 = -(ymm1 * ymm4) + ymm3
; FMA-RECIP-NEXT:    vfmadd132ps {{.*#+}} ymm4 = (ymm4 * ymm2) + ymm2
; FMA-RECIP-NEXT:    vfnmadd213ps {{.*#+}} ymm1 = -(ymm4 * ymm1) + ymm3
; FMA-RECIP-NEXT:    vfmadd132ps {{.*#+}} ymm1 = (ymm1 * ymm4) + ymm4
; FMA-RECIP-NEXT:    vrcpps %ymm0, %ymm2
; FMA-RECIP-NEXT:    vmovaps %ymm2, %ymm4
; FMA-RECIP-NEXT:    vfnmadd213ps {{.*#+}} ymm4 = -(ymm0 * ymm4) + ymm3
; FMA-RECIP-NEXT:    vfmadd132ps {{.*#+}} ymm4 = (ymm4 * ymm2) + ymm2
; FMA-RECIP-NEXT:    vfnmadd213ps {{.*#+}} ymm0 = -(ymm4 * ymm0) + ymm3
; FMA-RECIP-NEXT:    vfmadd132ps {{.*#+}} ymm0 = (ymm0 * ymm4) + ymm4
; FMA-RECIP-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0
; FMA-RECIP-NEXT:    vmulps {{.*}}(%rip), %ymm1, %ymm1
; FMA-RECIP-NEXT:    retq
;
; BDVER2-LABEL: v16f32_two_step2:
; BDVER2:       # %bb.0:
; BDVER2-NEXT:    vrcpps %ymm1, %ymm2 # sched: [5:2.00]
; BDVER2-NEXT:    vmovaps {{.*#+}} ymm3 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0] sched: [5:0.50]
; BDVER2-NEXT:    vfnmaddps %ymm3, %ymm2, %ymm1, %ymm4 # sched: [5:0.50]
; BDVER2-NEXT:    vfmaddps %ymm2, %ymm4, %ymm2, %ymm2 # sched: [5:0.50]
; BDVER2-NEXT:    vfnmaddps %ymm3, %ymm2, %ymm1, %ymm1 # sched: [5:0.50]
; BDVER2-NEXT:    vfmaddps %ymm2, %ymm1, %ymm2, %ymm1 # sched: [5:0.50]
; BDVER2-NEXT:    vrcpps %ymm0, %ymm2 # sched: [5:2.00]
; BDVER2-NEXT:    vmulps {{.*}}(%rip), %ymm1, %ymm1 # sched: [10:2.00]
; BDVER2-NEXT:    vfnmaddps %ymm3, %ymm2, %ymm0, %ymm4 # sched: [5:0.50]
; BDVER2-NEXT:    vfmaddps %ymm2, %ymm4, %ymm2, %ymm2 # sched: [5:0.50]
; BDVER2-NEXT:    vfnmaddps %ymm3, %ymm2, %ymm0, %ymm0 # sched: [5:0.50]
; BDVER2-NEXT:    vfmaddps %ymm2, %ymm0, %ymm2, %ymm0 # sched: [5:0.50]
; BDVER2-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [10:2.00]
; BDVER2-NEXT:    retq # sched: [5:1.00]
;
; BTVER2-LABEL: v16f32_two_step2:
; BTVER2:       # %bb.0:
; BTVER2-NEXT:    vmovaps {{.*#+}} ymm4 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0] sched: [5:1.00]
; BTVER2-NEXT:    vrcpps %ymm1, %ymm2 # sched: [2:2.00]
; BTVER2-NEXT:    vmulps %ymm2, %ymm1, %ymm3 # sched: [2:2.00]
; BTVER2-NEXT:    vsubps %ymm3, %ymm4, %ymm3 # sched: [3:2.00]
; BTVER2-NEXT:    vmulps %ymm3, %ymm2, %ymm3 # sched: [2:2.00]
; BTVER2-NEXT:    vaddps %ymm3, %ymm2, %ymm2 # sched: [3:2.00]
; BTVER2-NEXT:    vmulps %ymm2, %ymm1, %ymm1 # sched: [2:2.00]
; BTVER2-NEXT:    vsubps %ymm1, %ymm4, %ymm1 # sched: [3:2.00]
; BTVER2-NEXT:    vmulps %ymm1, %ymm2, %ymm1 # sched: [2:2.00]
; BTVER2-NEXT:    vaddps %ymm1, %ymm2, %ymm1 # sched: [3:2.00]
; BTVER2-NEXT:    vrcpps %ymm0, %ymm2 # sched: [2:2.00]
; BTVER2-NEXT:    vmulps {{.*}}(%rip), %ymm1, %ymm1 # sched: [7:2.00]
; BTVER2-NEXT:    vmulps %ymm2, %ymm0, %ymm3 # sched: [2:2.00]
; BTVER2-NEXT:    vsubps %ymm3, %ymm4, %ymm3 # sched: [3:2.00]
; BTVER2-NEXT:    vmulps %ymm3, %ymm2, %ymm3 # sched: [2:2.00]
; BTVER2-NEXT:    vaddps %ymm3, %ymm2, %ymm2 # sched: [3:2.00]
; BTVER2-NEXT:    vmulps %ymm2, %ymm0, %ymm0 # sched: [2:2.00]
; BTVER2-NEXT:    vsubps %ymm0, %ymm4, %ymm0 # sched: [3:2.00]
; BTVER2-NEXT:    vmulps %ymm0, %ymm2, %ymm0 # sched: [2:2.00]
; BTVER2-NEXT:    vaddps %ymm0, %ymm2, %ymm0 # sched: [3:2.00]
; BTVER2-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [7:2.00]
; BTVER2-NEXT:    retq # sched: [4:1.00]
;
; SANDY-LABEL: v16f32_two_step2:
; SANDY:       # %bb.0:
; SANDY-NEXT:    vrcpps %ymm1, %ymm2 # sched: [7:2.00]
; SANDY-NEXT:    vmulps %ymm2, %ymm1, %ymm3 # sched: [5:1.00]
; SANDY-NEXT:    vmovaps {{.*#+}} ymm4 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0] sched: [7:0.50]
; SANDY-NEXT:    vsubps %ymm3, %ymm4, %ymm3 # sched: [3:1.00]
; SANDY-NEXT:    vmulps %ymm3, %ymm2, %ymm3 # sched: [5:1.00]
; SANDY-NEXT:    vaddps %ymm3, %ymm2, %ymm2 # sched: [3:1.00]
; SANDY-NEXT:    vmulps %ymm2, %ymm1, %ymm1 # sched: [5:1.00]
; SANDY-NEXT:    vsubps %ymm1, %ymm4, %ymm1 # sched: [3:1.00]
; SANDY-NEXT:    vmulps %ymm1, %ymm2, %ymm1 # sched: [5:1.00]
; SANDY-NEXT:    vaddps %ymm1, %ymm2, %ymm1 # sched: [3:1.00]
; SANDY-NEXT:    vrcpps %ymm0, %ymm2 # sched: [7:2.00]
; SANDY-NEXT:    vmulps %ymm2, %ymm0, %ymm3 # sched: [5:1.00]
; SANDY-NEXT:    vsubps %ymm3, %ymm4, %ymm3 # sched: [3:1.00]
; SANDY-NEXT:    vmulps %ymm3, %ymm2, %ymm3 # sched: [5:1.00]
; SANDY-NEXT:    vaddps %ymm3, %ymm2, %ymm2 # sched: [3:1.00]
; SANDY-NEXT:    vmulps %ymm2, %ymm0, %ymm0 # sched: [5:1.00]
; SANDY-NEXT:    vsubps %ymm0, %ymm4, %ymm0 # sched: [3:1.00]
; SANDY-NEXT:    vmulps %ymm0, %ymm2, %ymm0 # sched: [5:1.00]
; SANDY-NEXT:    vaddps %ymm0, %ymm2, %ymm0 # sched: [3:1.00]
; SANDY-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [12:1.00]
; SANDY-NEXT:    vmulps {{.*}}(%rip), %ymm1, %ymm1 # sched: [12:1.00]
; SANDY-NEXT:    retq # sched: [1:1.00]
;
; HASWELL-LABEL: v16f32_two_step2:
; HASWELL:       # %bb.0:
; HASWELL-NEXT:    vrcpps %ymm1, %ymm2 # sched: [11:2.00]
; HASWELL-NEXT:    vbroadcastss {{.*#+}} ymm3 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0] sched: [7:0.50]
; HASWELL-NEXT:    vmovaps %ymm2, %ymm4 # sched: [1:1.00]
; HASWELL-NEXT:    vfnmadd213ps {{.*#+}} ymm4 = -(ymm1 * ymm4) + ymm3 sched: [5:0.50]
; HASWELL-NEXT:    vfmadd132ps {{.*#+}} ymm4 = (ymm4 * ymm2) + ymm2 sched: [5:0.50]
; HASWELL-NEXT:    vfnmadd213ps {{.*#+}} ymm1 = -(ymm4 * ymm1) + ymm3 sched: [5:0.50]
; HASWELL-NEXT:    vfmadd132ps {{.*#+}} ymm1 = (ymm1 * ymm4) + ymm4 sched: [5:0.50]
; HASWELL-NEXT:    vrcpps %ymm0, %ymm2 # sched: [11:2.00]
; HASWELL-NEXT:    vmovaps %ymm2, %ymm4 # sched: [1:1.00]
; HASWELL-NEXT:    vfnmadd213ps {{.*#+}} ymm4 = -(ymm0 * ymm4) + ymm3 sched: [5:0.50]
; HASWELL-NEXT:    vfmadd132ps {{.*#+}} ymm4 = (ymm4 * ymm2) + ymm2 sched: [5:0.50]
; HASWELL-NEXT:    vfnmadd213ps {{.*#+}} ymm0 = -(ymm4 * ymm0) + ymm3 sched: [5:0.50]
; HASWELL-NEXT:    vfmadd132ps {{.*#+}} ymm0 = (ymm0 * ymm4) + ymm4 sched: [5:0.50]
; HASWELL-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [12:0.50]
; HASWELL-NEXT:    vmulps {{.*}}(%rip), %ymm1, %ymm1 # sched: [12:0.50]
; HASWELL-NEXT:    retq # sched: [7:1.00]
;
; HASWELL-NO-FMA-LABEL: v16f32_two_step2:
; HASWELL-NO-FMA:       # %bb.0:
; HASWELL-NO-FMA-NEXT:    vrcpps %ymm1, %ymm2 # sched: [11:2.00]
; HASWELL-NO-FMA-NEXT:    vmulps %ymm2, %ymm1, %ymm3 # sched: [5:0.50]
; HASWELL-NO-FMA-NEXT:    vbroadcastss {{.*#+}} ymm4 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0] sched: [7:0.50]
; HASWELL-NO-FMA-NEXT:    vsubps %ymm3, %ymm4, %ymm3 # sched: [3:1.00]
; HASWELL-NO-FMA-NEXT:    vmulps %ymm3, %ymm2, %ymm3 # sched: [5:0.50]
; HASWELL-NO-FMA-NEXT:    vaddps %ymm3, %ymm2, %ymm2 # sched: [3:1.00]
; HASWELL-NO-FMA-NEXT:    vmulps %ymm2, %ymm1, %ymm1 # sched: [5:0.50]
; HASWELL-NO-FMA-NEXT:    vsubps %ymm1, %ymm4, %ymm1 # sched: [3:1.00]
; HASWELL-NO-FMA-NEXT:    vmulps %ymm1, %ymm2, %ymm1 # sched: [5:0.50]
; HASWELL-NO-FMA-NEXT:    vaddps %ymm1, %ymm2, %ymm1 # sched: [3:1.00]
; HASWELL-NO-FMA-NEXT:    vrcpps %ymm0, %ymm2 # sched: [11:2.00]
; HASWELL-NO-FMA-NEXT:    vmulps %ymm2, %ymm0, %ymm3 # sched: [5:0.50]
; HASWELL-NO-FMA-NEXT:    vsubps %ymm3, %ymm4, %ymm3 # sched: [3:1.00]
; HASWELL-NO-FMA-NEXT:    vmulps %ymm3, %ymm2, %ymm3 # sched: [5:0.50]
; HASWELL-NO-FMA-NEXT:    vaddps %ymm3, %ymm2, %ymm2 # sched: [3:1.00]
; HASWELL-NO-FMA-NEXT:    vmulps %ymm2, %ymm0, %ymm0 # sched: [5:0.50]
; HASWELL-NO-FMA-NEXT:    vsubps %ymm0, %ymm4, %ymm0 # sched: [3:1.00]
; HASWELL-NO-FMA-NEXT:    vmulps %ymm0, %ymm2, %ymm0 # sched: [5:0.50]
; HASWELL-NO-FMA-NEXT:    vaddps %ymm0, %ymm2, %ymm0 # sched: [3:1.00]
; HASWELL-NO-FMA-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [12:0.50]
; HASWELL-NO-FMA-NEXT:    vmulps {{.*}}(%rip), %ymm1, %ymm1 # sched: [12:0.50]
; HASWELL-NO-FMA-NEXT:    retq # sched: [7:1.00]
;
; KNL-LABEL: v16f32_two_step2:
; KNL:       # %bb.0:
; KNL-NEXT:    vrcp14ps %zmm0, %zmm1 # sched: [11:2.00]
; KNL-NEXT:    vbroadcastss {{.*#+}} zmm2 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0] sched: [10:1.00]
; KNL-NEXT:    vmovaps %zmm1, %zmm3 # sched: [1:1.00]
; KNL-NEXT:    vfnmadd213ps {{.*#+}} zmm3 = -(zmm0 * zmm3) + zmm2 sched: [5:0.50]
; KNL-NEXT:    vfmadd132ps {{.*#+}} zmm3 = (zmm3 * zmm1) + zmm1 sched: [5:0.50]
; KNL-NEXT:    vfnmadd213ps {{.*#+}} zmm0 = -(zmm3 * zmm0) + zmm2 sched: [5:0.50]
; KNL-NEXT:    vfmadd132ps {{.*#+}} zmm0 = (zmm0 * zmm3) + zmm3 sched: [5:0.50]
; KNL-NEXT:    vmulps {{.*}}(%rip), %zmm0, %zmm0 # sched: [12:0.50]
; KNL-NEXT:    retq # sched: [7:1.00]
;
; SKX-LABEL: v16f32_two_step2:
; SKX:       # %bb.0:
; SKX-NEXT:    vrcp14ps %zmm0, %zmm1 # sched: [4:2.00]
; SKX-NEXT:    vbroadcastss {{.*#+}} zmm2 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0,1.0E+0] sched: [8:0.50]
; SKX-NEXT:    vmovaps %zmm1, %zmm3 # sched: [1:0.33]
; SKX-NEXT:    vfnmadd213ps {{.*#+}} zmm3 = -(zmm0 * zmm3) + zmm2 sched: [4:0.50]
; SKX-NEXT:    vfmadd132ps {{.*#+}} zmm3 = (zmm3 * zmm1) + zmm1 sched: [4:0.50]
; SKX-NEXT:    vfnmadd213ps {{.*#+}} zmm0 = -(zmm3 * zmm0) + zmm2 sched: [4:0.50]
; SKX-NEXT:    vfmadd132ps {{.*#+}} zmm0 = (zmm0 * zmm3) + zmm3 sched: [4:0.50]
; SKX-NEXT:    vmulps {{.*}}(%rip), %zmm0, %zmm0 # sched: [11:0.50]
; SKX-NEXT:    retq # sched: [7:1.00]
  %div = fdiv fast <16 x float> <float 1.0, float 2.0, float 3.0, float 4.0, float 5.0, float 6.0, float 7.0, float 8.0, float 9.0, float 10.0, float 11.0, float 12.0, float 13.0, float 14.0, float 15.0, float 16.0>, %x
  ret <16 x float> %div
}

define <16 x float> @v16f32_no_step(<16 x float> %x) #3 {
; SSE-LABEL: v16f32_no_step:
; SSE:       # %bb.0:
; SSE-NEXT:    rcpps %xmm0, %xmm0
; SSE-NEXT:    rcpps %xmm1, %xmm1
; SSE-NEXT:    rcpps %xmm2, %xmm2
; SSE-NEXT:    rcpps %xmm3, %xmm3
; SSE-NEXT:    retq
;
; AVX-RECIP-LABEL: v16f32_no_step:
; AVX-RECIP:       # %bb.0:
; AVX-RECIP-NEXT:    vrcpps %ymm0, %ymm0
; AVX-RECIP-NEXT:    vrcpps %ymm1, %ymm1
; AVX-RECIP-NEXT:    retq
;
; FMA-RECIP-LABEL: v16f32_no_step:
; FMA-RECIP:       # %bb.0:
; FMA-RECIP-NEXT:    vrcpps %ymm0, %ymm0
; FMA-RECIP-NEXT:    vrcpps %ymm1, %ymm1
; FMA-RECIP-NEXT:    retq
;
; BDVER2-LABEL: v16f32_no_step:
; BDVER2:       # %bb.0:
; BDVER2-NEXT:    vrcpps %ymm0, %ymm0 # sched: [5:2.00]
; BDVER2-NEXT:    vrcpps %ymm1, %ymm1 # sched: [5:2.00]
; BDVER2-NEXT:    retq # sched: [5:1.00]
;
; BTVER2-LABEL: v16f32_no_step:
; BTVER2:       # %bb.0:
; BTVER2-NEXT:    vrcpps %ymm0, %ymm0 # sched: [2:2.00]
; BTVER2-NEXT:    vrcpps %ymm1, %ymm1 # sched: [2:2.00]
; BTVER2-NEXT:    retq # sched: [4:1.00]
;
; SANDY-LABEL: v16f32_no_step:
; SANDY:       # %bb.0:
; SANDY-NEXT:    vrcpps %ymm0, %ymm0 # sched: [7:2.00]
; SANDY-NEXT:    vrcpps %ymm1, %ymm1 # sched: [7:2.00]
; SANDY-NEXT:    retq # sched: [1:1.00]
;
; HASWELL-LABEL: v16f32_no_step:
; HASWELL:       # %bb.0:
; HASWELL-NEXT:    vrcpps %ymm0, %ymm0 # sched: [11:2.00]
; HASWELL-NEXT:    vrcpps %ymm1, %ymm1 # sched: [11:2.00]
; HASWELL-NEXT:    retq # sched: [7:1.00]
;
; HASWELL-NO-FMA-LABEL: v16f32_no_step:
; HASWELL-NO-FMA:       # %bb.0:
; HASWELL-NO-FMA-NEXT:    vrcpps %ymm0, %ymm0 # sched: [11:2.00]
; HASWELL-NO-FMA-NEXT:    vrcpps %ymm1, %ymm1 # sched: [11:2.00]
; HASWELL-NO-FMA-NEXT:    retq # sched: [7:1.00]
;
; KNL-LABEL: v16f32_no_step:
; KNL:       # %bb.0:
; KNL-NEXT:    vrcp14ps %zmm0, %zmm0 # sched: [11:2.00]
; KNL-NEXT:    retq # sched: [7:1.00]
;
; SKX-LABEL: v16f32_no_step:
; SKX:       # %bb.0:
; SKX-NEXT:    vrcp14ps %zmm0, %zmm0 # sched: [4:2.00]
; SKX-NEXT:    retq # sched: [7:1.00]
  %div = fdiv fast <16 x float> <float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0>, %x
  ret <16 x float> %div
}

define <16 x float> @v16f32_no_step2(<16 x float> %x) #3 {
; SSE-LABEL: v16f32_no_step2:
; SSE:       # %bb.0:
; SSE-NEXT:    rcpps %xmm3, %xmm3
; SSE-NEXT:    rcpps %xmm2, %xmm2
; SSE-NEXT:    rcpps %xmm1, %xmm1
; SSE-NEXT:    rcpps %xmm0, %xmm0
; SSE-NEXT:    mulps {{.*}}(%rip), %xmm0
; SSE-NEXT:    mulps {{.*}}(%rip), %xmm1
; SSE-NEXT:    mulps {{.*}}(%rip), %xmm2
; SSE-NEXT:    mulps {{.*}}(%rip), %xmm3
; SSE-NEXT:    retq
;
; AVX-RECIP-LABEL: v16f32_no_step2:
; AVX-RECIP:       # %bb.0:
; AVX-RECIP-NEXT:    vrcpps %ymm1, %ymm1
; AVX-RECIP-NEXT:    vrcpps %ymm0, %ymm0
; AVX-RECIP-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0
; AVX-RECIP-NEXT:    vmulps {{.*}}(%rip), %ymm1, %ymm1
; AVX-RECIP-NEXT:    retq
;
; FMA-RECIP-LABEL: v16f32_no_step2:
; FMA-RECIP:       # %bb.0:
; FMA-RECIP-NEXT:    vrcpps %ymm1, %ymm1
; FMA-RECIP-NEXT:    vrcpps %ymm0, %ymm0
; FMA-RECIP-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0
; FMA-RECIP-NEXT:    vmulps {{.*}}(%rip), %ymm1, %ymm1
; FMA-RECIP-NEXT:    retq
;
; BDVER2-LABEL: v16f32_no_step2:
; BDVER2:       # %bb.0:
; BDVER2-NEXT:    vrcpps %ymm1, %ymm1 # sched: [5:2.00]
; BDVER2-NEXT:    vrcpps %ymm0, %ymm0 # sched: [5:2.00]
; BDVER2-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [10:2.00]
; BDVER2-NEXT:    vmulps {{.*}}(%rip), %ymm1, %ymm1 # sched: [10:2.00]
; BDVER2-NEXT:    retq # sched: [5:1.00]
;
; BTVER2-LABEL: v16f32_no_step2:
; BTVER2:       # %bb.0:
; BTVER2-NEXT:    vrcpps %ymm1, %ymm1 # sched: [2:2.00]
; BTVER2-NEXT:    vrcpps %ymm0, %ymm0 # sched: [2:2.00]
; BTVER2-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [7:2.00]
; BTVER2-NEXT:    vmulps {{.*}}(%rip), %ymm1, %ymm1 # sched: [7:2.00]
; BTVER2-NEXT:    retq # sched: [4:1.00]
;
; SANDY-LABEL: v16f32_no_step2:
; SANDY:       # %bb.0:
; SANDY-NEXT:    vrcpps %ymm1, %ymm1 # sched: [7:2.00]
; SANDY-NEXT:    vrcpps %ymm0, %ymm0 # sched: [7:2.00]
; SANDY-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [12:1.00]
; SANDY-NEXT:    vmulps {{.*}}(%rip), %ymm1, %ymm1 # sched: [12:1.00]
; SANDY-NEXT:    retq # sched: [1:1.00]
;
; HASWELL-LABEL: v16f32_no_step2:
; HASWELL:       # %bb.0:
; HASWELL-NEXT:    vrcpps %ymm1, %ymm1 # sched: [11:2.00]
; HASWELL-NEXT:    vrcpps %ymm0, %ymm0 # sched: [11:2.00]
; HASWELL-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [12:0.50]
; HASWELL-NEXT:    vmulps {{.*}}(%rip), %ymm1, %ymm1 # sched: [12:0.50]
; HASWELL-NEXT:    retq # sched: [7:1.00]
;
; HASWELL-NO-FMA-LABEL: v16f32_no_step2:
; HASWELL-NO-FMA:       # %bb.0:
; HASWELL-NO-FMA-NEXT:    vrcpps %ymm1, %ymm1 # sched: [11:2.00]
; HASWELL-NO-FMA-NEXT:    vrcpps %ymm0, %ymm0 # sched: [11:2.00]
; HASWELL-NO-FMA-NEXT:    vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [12:0.50]
; HASWELL-NO-FMA-NEXT:    vmulps {{.*}}(%rip), %ymm1, %ymm1 # sched: [12:0.50]
; HASWELL-NO-FMA-NEXT:    retq # sched: [7:1.00]
;
; KNL-LABEL: v16f32_no_step2:
; KNL:       # %bb.0:
; KNL-NEXT:    vrcp14ps %zmm0, %zmm0 # sched: [11:2.00]
; KNL-NEXT:    vmulps {{.*}}(%rip), %zmm0, %zmm0 # sched: [12:0.50]
; KNL-NEXT:    retq # sched: [7:1.00]
;
; SKX-LABEL: v16f32_no_step2:
; SKX:       # %bb.0:
; SKX-NEXT:    vrcp14ps %zmm0, %zmm0 # sched: [4:2.00]
; SKX-NEXT:    vmulps {{.*}}(%rip), %zmm0, %zmm0 # sched: [11:0.50]
; SKX-NEXT:    retq # sched: [7:1.00]
  %div = fdiv fast <16 x float> <float 1.0, float 2.0, float 3.0, float 4.0, float 5.0, float 6.0, float 7.0, float 8.0, float 9.0, float 10.0, float 11.0, float 12.0, float 13.0, float 14.0, float 15.0, float 16.0>, %x
  ret <16 x float> %div
}

attributes #0 = { "unsafe-fp-math"="true" "reciprocal-estimates"="!divf,!vec-divf" }
attributes #1 = { "unsafe-fp-math"="true" "reciprocal-estimates"="divf,vec-divf" }
attributes #2 = { "unsafe-fp-math"="true" "reciprocal-estimates"="divf:2,vec-divf:2" }
attributes #3 = { "unsafe-fp-math"="true" "reciprocal-estimates"="divf:0,vec-divf:0" }