[X86] Add tests for folding `fmul`/`fdiv` by Pow2 to `add`/`sub` of exp; NFC

[RKSimon]

llvm/test/CodeGen/X86/fold-int-pow2-with-fmul-or-fdiv.ll

5

These prefixes are broken, they should be more like this:

; RUN: llc < %s -mtriple=x86_64-unknown-unknown | FileCheck %s --check-prefixes=CHECK-SSE
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefixes=CHECK-AVX,CHECK-AVX2
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512f | FileCheck %s --check-prefixes=CHECK-AVX,CHECK-AVX512F

[goldstein.w.n]

Fix checks

[goldstein.w.n]

Add FMA test

[goldstein.w.n]

Rebase

[goldstein.w.n]

Rebase

[arsenm]

llvm/test/CodeGen/X86/fold-int-pow2-with-fmul-or-fdiv.ll

1265

No i32 cases?

[goldstein.w.n]

Add i32 tests

[RKSimon]

llvm/test/CodeGen/X86/fold-int-pow2-with-fmul-or-fdiv.ll

12

add nounwind to silence the cfi noise

[goldstein.w.n]

Remove nounwind tags / rebase

[goldstein.w.n]

Rebase

[goldstein.w.n]

Rebase

[goldstein.w.n]

Rebase

[RKSimon]

llvm/test/CodeGen/X86/fold-int-pow2-with-fmul-or-fdiv.ll

5

; RUN: llc < %s -mtriple=x86_64-unknown-unknown | FileCheck %s --check-prefixes=CHECK-SSE
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefixes=CHECK-AVX,CHECK-AVX2
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512f | FileCheck %s --check-prefixes=CHECK-AVX,CHECK-AVX512F,,CHECK-NO-FASTFMA
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=skx -fp-contract=fast | FileCheck %s --check-prefixes=CHECK-AVX,CHECK-AVX512F,CHECK-FMA

[goldstein.w.n]

Update checks

[goldstein.w.n]

Rebase

[RKSimon]

LGTM

[goldstein.w.n]

Rebase + ldexp

[goldstein.w.n]

Rebase

llvm/test/CodeGen/X86/fold-int-pow2-with-fmul-or-fdiv.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 3
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown | FileCheck %s --check-prefixes=CHECK-SSE
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefixes=CHECK-AVX,CHECK-AVX2
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512f | FileCheck %s --check-prefixes=CHECK-AVX,CHECK-AVX512F,CHECK-NO-FASTFMA
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=skx -fp-contract=fast | FileCheck %s --check-prefixes=CHECK-AVX,CHECK-AVX512F,CHECK-FMA
+
+declare i16 @llvm.umax.i16(i16, i16)
+declare i64 @llvm.umin.i64(i64, i64)
+
+declare <4 x float> @llvm.ldexp.v4f32.v4i32(<4 x float>, <4 x i32>)
+
+define <4 x float> @fmul_pow2_4xfloat(<4 x i32> %i) {
+; CHECK-SSE-LABEL: fmul_pow2_4xfloat:
+; CHECK-SSE:       # %bb.0:
+; CHECK-SSE-NEXT:    pslld $23, %xmm0
+; CHECK-SSE-NEXT:    paddd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; CHECK-SSE-NEXT:    cvttps2dq %xmm0, %xmm0
+; CHECK-SSE-NEXT:    movdqa {{.*#+}} xmm1 = [65535,0,65535,0,65535,0,65535,0]
+; CHECK-SSE-NEXT:    pand %xmm0, %xmm1
+; CHECK-SSE-NEXT:    por {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1
+; CHECK-SSE-NEXT:    psrld $16, %xmm0
+; CHECK-SSE-NEXT:    por {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; CHECK-SSE-NEXT:    subps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; CHECK-SSE-NEXT:    addps %xmm1, %xmm0
+; CHECK-SSE-NEXT:    mulps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; CHECK-SSE-NEXT:    retq
+;
+; CHECK-AVX2-LABEL: fmul_pow2_4xfloat:
+; CHECK-AVX2:       # %bb.0:
+; CHECK-AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [1,1,1,1]
+; CHECK-AVX2-NEXT:    vpsllvd %xmm0, %xmm1, %xmm0
+; CHECK-AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [1258291200,1258291200,1258291200,1258291200]
+; CHECK-AVX2-NEXT:    vpblendw {{.*#+}} xmm1 = xmm0[0],xmm1[1],xmm0[2],xmm1[3],xmm0[4],xmm1[5],xmm0[6],xmm1[7]
+; CHECK-AVX2-NEXT:    vpsrld $16, %xmm0, %xmm0
+; CHECK-AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm2 = [1392508928,1392508928,1392508928,1392508928]
+; CHECK-AVX2-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2],xmm2[3],xmm0[4],xmm2[5],xmm0[6],xmm2[7]
+; CHECK-AVX2-NEXT:    vbroadcastss {{.*#+}} xmm2 = [5.49764202E+11,5.49764202E+11,5.49764202E+11,5.49764202E+11]
+; CHECK-AVX2-NEXT:    vsubps %xmm2, %xmm0, %xmm0
+; CHECK-AVX2-NEXT:    vaddps %xmm0, %xmm1, %xmm0
+; CHECK-AVX2-NEXT:    vbroadcastss {{.*#+}} xmm1 = [9.0E+0,9.0E+0,9.0E+0,9.0E+0]
+; CHECK-AVX2-NEXT:    vmulps %xmm1, %xmm0, %xmm0
+; CHECK-AVX2-NEXT:    retq
+;
+; CHECK-NO-FASTFMA-LABEL: fmul_pow2_4xfloat:
+; CHECK-NO-FASTFMA:       # %bb.0:
+; CHECK-NO-FASTFMA-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [1,1,1,1]
+; CHECK-NO-FASTFMA-NEXT:    vpsllvd %xmm0, %xmm1, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vcvtudq2ps %zmm0, %zmm0
+; CHECK-NO-FASTFMA-NEXT:    vbroadcastss {{.*#+}} xmm1 = [9.0E+0,9.0E+0,9.0E+0,9.0E+0]
+; CHECK-NO-FASTFMA-NEXT:    vmulps %xmm1, %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vzeroupper
+; CHECK-NO-FASTFMA-NEXT:    retq
+;
+; CHECK-FMA-LABEL: fmul_pow2_4xfloat:
+; CHECK-FMA:       # %bb.0:
+; CHECK-FMA-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [1,1,1,1]
+; CHECK-FMA-NEXT:    vpsllvd %xmm0, %xmm1, %xmm0
+; CHECK-FMA-NEXT:    vcvtudq2ps %xmm0, %xmm0
+; CHECK-FMA-NEXT:    vmulps {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to4}, %xmm0, %xmm0
+; CHECK-FMA-NEXT:    retq
+  %p2 = shl <4 x i32> <i32 1, i32 1, i32 1, i32 1>, %i
+  %p2_f = uitofp <4 x i32> %p2 to <4 x float>
+  %r = fmul <4 x float> <float 9.000000e+00, float 9.000000e+00, float 9.000000e+00, float 9.000000e+00>, %p2_f
+  ret <4 x float> %r
+}
+
+define <4 x float> @fmul_pow2_ldexp_4xfloat(<4 x i32> %i) {
+; CHECK-SSE-LABEL: fmul_pow2_ldexp_4xfloat:
+; CHECK-SSE:       # %bb.0:
+; CHECK-SSE-NEXT:    subq $56, %rsp
+; CHECK-SSE-NEXT:    .cfi_def_cfa_offset 64
+; CHECK-SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; CHECK-SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[3,3,3,3]
+; CHECK-SSE-NEXT:    movd %xmm1, %edi
+; CHECK-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; CHECK-SSE-NEXT:    callq ldexpf@PLT
+; CHECK-SSE-NEXT:    movaps %xmm0, (%rsp) # 16-byte Spill
+; CHECK-SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; CHECK-SSE-NEXT:    # xmm0 = mem[2,3,2,3]
+; CHECK-SSE-NEXT:    movd %xmm0, %edi
+; CHECK-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; CHECK-SSE-NEXT:    callq ldexpf@PLT
+; CHECK-SSE-NEXT:    unpcklps (%rsp), %xmm0 # 16-byte Folded Reload
+; CHECK-SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; CHECK-SSE-NEXT:    movaps %xmm0, (%rsp) # 16-byte Spill
+; CHECK-SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; CHECK-SSE-NEXT:    movd %xmm0, %edi
+; CHECK-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; CHECK-SSE-NEXT:    callq ldexpf@PLT
+; CHECK-SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; CHECK-SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; CHECK-SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; CHECK-SSE-NEXT:    movd %xmm0, %edi
+; CHECK-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; CHECK-SSE-NEXT:    callq ldexpf@PLT
+; CHECK-SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; CHECK-SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; CHECK-SSE-NEXT:    unpcklpd (%rsp), %xmm1 # 16-byte Folded Reload
+; CHECK-SSE-NEXT:    # xmm1 = xmm1[0],mem[0]
+; CHECK-SSE-NEXT:    movaps %xmm1, %xmm0
+; CHECK-SSE-NEXT:    addq $56, %rsp
+; CHECK-SSE-NEXT:    .cfi_def_cfa_offset 8
+; CHECK-SSE-NEXT:    retq
+;
+; CHECK-AVX-LABEL: fmul_pow2_ldexp_4xfloat:
+; CHECK-AVX:       # %bb.0:
+; CHECK-AVX-NEXT:    subq $40, %rsp
+; CHECK-AVX-NEXT:    .cfi_def_cfa_offset 48
+; CHECK-AVX-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; CHECK-AVX-NEXT:    vextractps $1, %xmm0, %edi
+; CHECK-AVX-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; CHECK-AVX-NEXT:    callq ldexpf@PLT
+; CHECK-AVX-NEXT:    vmovaps %xmm0, (%rsp) # 16-byte Spill
+; CHECK-AVX-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; CHECK-AVX-NEXT:    vmovd %xmm0, %edi
+; CHECK-AVX-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; CHECK-AVX-NEXT:    callq ldexpf@PLT
+; CHECK-AVX-NEXT:    vinsertps $16, (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
+; CHECK-AVX-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[2,3]
+; CHECK-AVX-NEXT:    vmovaps %xmm0, (%rsp) # 16-byte Spill
+; CHECK-AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; CHECK-AVX-NEXT:    vextractps $2, %xmm0, %edi
+; CHECK-AVX-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; CHECK-AVX-NEXT:    callq ldexpf@PLT
+; CHECK-AVX-NEXT:    vmovaps (%rsp), %xmm1 # 16-byte Reload
+; CHECK-AVX-NEXT:    vinsertps {{.*#+}} xmm0 = xmm1[0,1],xmm0[0],xmm1[3]
+; CHECK-AVX-NEXT:    vmovaps %xmm0, (%rsp) # 16-byte Spill
+; CHECK-AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; CHECK-AVX-NEXT:    vextractps $3, %xmm0, %edi
+; CHECK-AVX-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; CHECK-AVX-NEXT:    callq ldexpf@PLT
+; CHECK-AVX-NEXT:    vmovaps (%rsp), %xmm1 # 16-byte Reload
+; CHECK-AVX-NEXT:    vinsertps {{.*#+}} xmm0 = xmm1[0,1,2],xmm0[0]
+; CHECK-AVX-NEXT:    addq $40, %rsp
+; CHECK-AVX-NEXT:    .cfi_def_cfa_offset 8
+; CHECK-AVX-NEXT:    retq
+  %r = call <4 x float> @llvm.ldexp.v4f32.v4i32(<4 x float> <float 9.000000e+00, float 9.000000e+00, float 9.000000e+00, float 9.000000e+00>, <4 x i32> %i)
+  ret <4 x float> %r
+}
+
+define <4 x float> @fdiv_pow2_4xfloat(<4 x i32> %i) {
+; CHECK-SSE-LABEL: fdiv_pow2_4xfloat:
+; CHECK-SSE:       # %bb.0:
+; CHECK-SSE-NEXT:    pslld $23, %xmm0
+; CHECK-SSE-NEXT:    paddd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; CHECK-SSE-NEXT:    cvttps2dq %xmm0, %xmm1
+; CHECK-SSE-NEXT:    movdqa {{.*#+}} xmm0 = [65535,0,65535,0,65535,0,65535,0]
+; CHECK-SSE-NEXT:    pand %xmm1, %xmm0
+; CHECK-SSE-NEXT:    por {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; CHECK-SSE-NEXT:    psrld $16, %xmm1
+; CHECK-SSE-NEXT:    por {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1
+; CHECK-SSE-NEXT:    subps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1
+; CHECK-SSE-NEXT:    addps %xmm0, %xmm1
+; CHECK-SSE-NEXT:    movaps {{.*#+}} xmm0 = [9.0E+0,9.0E+0,9.0E+0,9.0E+0]
+; CHECK-SSE-NEXT:    divps %xmm1, %xmm0
+; CHECK-SSE-NEXT:    retq
+;
+; CHECK-AVX2-LABEL: fdiv_pow2_4xfloat:
+; CHECK-AVX2:       # %bb.0:
+; CHECK-AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [1,1,1,1]
+; CHECK-AVX2-NEXT:    vpsllvd %xmm0, %xmm1, %xmm0
+; CHECK-AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [1258291200,1258291200,1258291200,1258291200]
+; CHECK-AVX2-NEXT:    vpblendw {{.*#+}} xmm1 = xmm0[0],xmm1[1],xmm0[2],xmm1[3],xmm0[4],xmm1[5],xmm0[6],xmm1[7]
+; CHECK-AVX2-NEXT:    vpsrld $16, %xmm0, %xmm0
+; CHECK-AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm2 = [1392508928,1392508928,1392508928,1392508928]
+; CHECK-AVX2-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2],xmm2[3],xmm0[4],xmm2[5],xmm0[6],xmm2[7]
+; CHECK-AVX2-NEXT:    vbroadcastss {{.*#+}} xmm2 = [5.49764202E+11,5.49764202E+11,5.49764202E+11,5.49764202E+11]
+; CHECK-AVX2-NEXT:    vsubps %xmm2, %xmm0, %xmm0
+; CHECK-AVX2-NEXT:    vaddps %xmm0, %xmm1, %xmm0
+; CHECK-AVX2-NEXT:    vbroadcastss {{.*#+}} xmm1 = [9.0E+0,9.0E+0,9.0E+0,9.0E+0]
+; CHECK-AVX2-NEXT:    vdivps %xmm0, %xmm1, %xmm0
+; CHECK-AVX2-NEXT:    retq
+;
+; CHECK-NO-FASTFMA-LABEL: fdiv_pow2_4xfloat:
+; CHECK-NO-FASTFMA:       # %bb.0:
+; CHECK-NO-FASTFMA-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [1,1,1,1]
+; CHECK-NO-FASTFMA-NEXT:    vpsllvd %xmm0, %xmm1, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vcvtudq2ps %zmm0, %zmm0
+; CHECK-NO-FASTFMA-NEXT:    vbroadcastss {{.*#+}} xmm1 = [9.0E+0,9.0E+0,9.0E+0,9.0E+0]
+; CHECK-NO-FASTFMA-NEXT:    vdivps %xmm0, %xmm1, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vzeroupper
+; CHECK-NO-FASTFMA-NEXT:    retq
+;
+; CHECK-FMA-LABEL: fdiv_pow2_4xfloat:
+; CHECK-FMA:       # %bb.0:
+; CHECK-FMA-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [1,1,1,1]
+; CHECK-FMA-NEXT:    vpsllvd %xmm0, %xmm1, %xmm0
+; CHECK-FMA-NEXT:    vcvtudq2ps %xmm0, %xmm0
+; CHECK-FMA-NEXT:    vbroadcastss {{.*#+}} xmm1 = [9.0E+0,9.0E+0,9.0E+0,9.0E+0]
+; CHECK-FMA-NEXT:    vdivps %xmm0, %xmm1, %xmm0
+; CHECK-FMA-NEXT:    retq
+  %p2 = shl <4 x i32> <i32 1, i32 1, i32 1, i32 1>, %i
+  %p2_f = uitofp <4 x i32> %p2 to <4 x float>
+  %r = fdiv <4 x float> <float 9.000000e+00, float 9.000000e+00, float 9.000000e+00, float 9.000000e+00>, %p2_f
+  ret <4 x float> %r
+}
+
+declare <8 x half> @llvm.ldexp.v8f16.v8i16(<8 x half>, <8 x i16>)
+
+define <8 x half> @fmul_pow2_8xhalf(<8 x i16> %i) {
+; CHECK-SSE-LABEL: fmul_pow2_8xhalf:
+; CHECK-SSE:       # %bb.0:
+; CHECK-SSE-NEXT:    subq $88, %rsp
+; CHECK-SSE-NEXT:    .cfi_def_cfa_offset 96
+; CHECK-SSE-NEXT:    movdqa %xmm0, %xmm1
+; CHECK-SSE-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4,4,5,5,6,6,7,7]
+; CHECK-SSE-NEXT:    pslld $23, %xmm1
+; CHECK-SSE-NEXT:    movdqa {{.*#+}} xmm2 = [1065353216,1065353216,1065353216,1065353216]
+; CHECK-SSE-NEXT:    paddd %xmm2, %xmm1
+; CHECK-SSE-NEXT:    cvttps2dq %xmm1, %xmm1
+; CHECK-SSE-NEXT:    movaps %xmm1, (%rsp) # 16-byte Spill
+; CHECK-SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3]
+; CHECK-SSE-NEXT:    pslld $23, %xmm0
+; CHECK-SSE-NEXT:    paddd %xmm2, %xmm0
+; CHECK-SSE-NEXT:    cvttps2dq %xmm0, %xmm0
+; CHECK-SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; CHECK-SSE-NEXT:    pextrw $0, %xmm0, %eax
+; CHECK-SSE-NEXT:    xorps %xmm0, %xmm0
+; CHECK-SSE-NEXT:    cvtsi2ss %eax, %xmm0
+; CHECK-SSE-NEXT:    callq __truncsfhf2@PLT
+; CHECK-SSE-NEXT:    movss %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
+; CHECK-SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; CHECK-SSE-NEXT:    pextrw $2, %xmm0, %eax
+; CHECK-SSE-NEXT:    xorps %xmm0, %xmm0
+; CHECK-SSE-NEXT:    cvtsi2ss %eax, %xmm0
+; CHECK-SSE-NEXT:    callq __truncsfhf2@PLT
+; CHECK-SSE-NEXT:    movss %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
+; CHECK-SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; CHECK-SSE-NEXT:    pextrw $4, %xmm0, %eax
+; CHECK-SSE-NEXT:    xorps %xmm0, %xmm0
+; CHECK-SSE-NEXT:    cvtsi2ss %eax, %xmm0
+; CHECK-SSE-NEXT:    callq __truncsfhf2@PLT
+; CHECK-SSE-NEXT:    movss %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
+; CHECK-SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; CHECK-SSE-NEXT:    pextrw $6, %xmm0, %eax
+; CHECK-SSE-NEXT:    xorps %xmm0, %xmm0
+; CHECK-SSE-NEXT:    cvtsi2ss %eax, %xmm0
+; CHECK-SSE-NEXT:    callq __truncsfhf2@PLT
+; CHECK-SSE-NEXT:    movss %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
+; CHECK-SSE-NEXT:    movdqa (%rsp), %xmm0 # 16-byte Reload
+; CHECK-SSE-NEXT:    pextrw $0, %xmm0, %eax
+; CHECK-SSE-NEXT:    xorps %xmm0, %xmm0
+; CHECK-SSE-NEXT:    cvtsi2ss %eax, %xmm0
+; CHECK-SSE-NEXT:    callq __truncsfhf2@PLT
+; CHECK-SSE-NEXT:    movss %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
+; CHECK-SSE-NEXT:    movdqa (%rsp), %xmm0 # 16-byte Reload
+; CHECK-SSE-NEXT:    pextrw $2, %xmm0, %eax
+; CHECK-SSE-NEXT:    xorps %xmm0, %xmm0
+; CHECK-SSE-NEXT:    cvtsi2ss %eax, %xmm0
+; CHECK-SSE-NEXT:    callq __truncsfhf2@PLT
+; CHECK-SSE-NEXT:    movss %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
+; CHECK-SSE-NEXT:    movdqa (%rsp), %xmm0 # 16-byte Reload
+; CHECK-SSE-NEXT:    pextrw $4, %xmm0, %eax
+; CHECK-SSE-NEXT:    xorps %xmm0, %xmm0
+; CHECK-SSE-NEXT:    cvtsi2ss %eax, %xmm0
+; CHECK-SSE-NEXT:    callq __truncsfhf2@PLT
+; CHECK-SSE-NEXT:    movss %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
+; CHECK-SSE-NEXT:    movdqa (%rsp), %xmm0 # 16-byte Reload
+; CHECK-SSE-NEXT:    pextrw $6, %xmm0, %eax
+; CHECK-SSE-NEXT:    xorps %xmm0, %xmm0
+; CHECK-SSE-NEXT:    cvtsi2ss %eax, %xmm0
+; CHECK-SSE-NEXT:    callq __truncsfhf2@PLT
+; CHECK-SSE-NEXT:    callq __extendhfsf2@PLT
+; CHECK-SSE-NEXT:    mulss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; CHECK-SSE-NEXT:    callq __truncsfhf2@PLT
+; CHECK-SSE-NEXT:    movaps %xmm0, (%rsp) # 16-byte Spill
+; CHECK-SSE-NEXT:    movss {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 4-byte Reload
+; CHECK-SSE-NEXT:    # xmm0 = mem[0],zero,zero,zero
+; CHECK-SSE-NEXT:    callq __extendhfsf2@PLT
+; CHECK-SSE-NEXT:    mulss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; CHECK-SSE-NEXT:    callq __truncsfhf2@PLT
+; CHECK-SSE-NEXT:    punpcklwd (%rsp), %xmm0 # 16-byte Folded Reload
+; CHECK-SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; CHECK-SSE-NEXT:    movdqa %xmm0, (%rsp) # 16-byte Spill
+; CHECK-SSE-NEXT:    movss {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 4-byte Reload
+; CHECK-SSE-NEXT:    # xmm0 = mem[0],zero,zero,zero
+; CHECK-SSE-NEXT:    callq __extendhfsf2@PLT
+; CHECK-SSE-NEXT:    mulss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; CHECK-SSE-NEXT:    callq __truncsfhf2@PLT
+; CHECK-SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; CHECK-SSE-NEXT:    movss {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 4-byte Reload
+; CHECK-SSE-NEXT:    # xmm0 = mem[0],zero,zero,zero
+; CHECK-SSE-NEXT:    callq __extendhfsf2@PLT
+; CHECK-SSE-NEXT:    mulss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; CHECK-SSE-NEXT:    callq __truncsfhf2@PLT
+; CHECK-SSE-NEXT:    punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; CHECK-SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; CHECK-SSE-NEXT:    punpckldq (%rsp), %xmm0 # 16-byte Folded Reload
+; CHECK-SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; CHECK-SSE-NEXT:    movdqa %xmm0, (%rsp) # 16-byte Spill
+; CHECK-SSE-NEXT:    movss {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 4-byte Reload
+; CHECK-SSE-NEXT:    # xmm0 = mem[0],zero,zero,zero
+; CHECK-SSE-NEXT:    callq __extendhfsf2@PLT
+; CHECK-SSE-NEXT:    mulss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; CHECK-SSE-NEXT:    callq __truncsfhf2@PLT
+; CHECK-SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; CHECK-SSE-NEXT:    movss {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 4-byte Reload
+; CHECK-SSE-NEXT:    # xmm0 = mem[0],zero,zero,zero
+; CHECK-SSE-NEXT:    callq __extendhfsf2@PLT
+; CHECK-SSE-NEXT:    mulss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; CHECK-SSE-NEXT:    callq __truncsfhf2@PLT
+; CHECK-SSE-NEXT:    punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; CHECK-SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; CHECK-SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; CHECK-SSE-NEXT:    movss {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 4-byte Reload
+; CHECK-SSE-NEXT:    # xmm0 = mem[0],zero,zero,zero
+; CHECK-SSE-NEXT:    callq __extendhfsf2@PLT
+; CHECK-SSE-NEXT:    mulss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; CHECK-SSE-NEXT:    callq __truncsfhf2@PLT
+; CHECK-SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; CHECK-SSE-NEXT:    movss {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 4-byte Reload
+; CHECK-SSE-NEXT:    # xmm0 = mem[0],zero,zero,zero
+; CHECK-SSE-NEXT:    callq __extendhfsf2@PLT
+; CHECK-SSE-NEXT:    mulss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; CHECK-SSE-NEXT:    callq __truncsfhf2@PLT
+; CHECK-SSE-NEXT:    punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; CHECK-SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; CHECK-SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; CHECK-SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; CHECK-SSE-NEXT:    punpcklqdq (%rsp), %xmm0 # 16-byte Folded Reload
+; CHECK-SSE-NEXT:    # xmm0 = xmm0[0],mem[0]
+; CHECK-SSE-NEXT:    addq $88, %rsp
+; CHECK-SSE-NEXT:    .cfi_def_cfa_offset 8
+; CHECK-SSE-NEXT:    retq
+;
+; CHECK-AVX2-LABEL: fmul_pow2_8xhalf:
+; CHECK-AVX2:       # %bb.0:
+; CHECK-AVX2-NEXT:    subq $120, %rsp
+; CHECK-AVX2-NEXT:    .cfi_def_cfa_offset 128
+; CHECK-AVX2-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; CHECK-AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [1,1,1,1,1,1,1,1]
+; CHECK-AVX2-NEXT:    vpsllvd %ymm0, %ymm1, %ymm0
+; CHECK-AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,4,5,8,9,12,13,u,u,u,u,u,u,u,u,16,17,20,21,24,25,28,29,u,u,u,u,u,u,u,u]
+; CHECK-AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
+; CHECK-AVX2-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; CHECK-AVX2-NEXT:    vpextrw $0, %xmm0, %eax
+; CHECK-AVX2-NEXT:    vcvtsi2ss %eax, %xmm2, %xmm0
+; CHECK-AVX2-NEXT:    vzeroupper
+; CHECK-AVX2-NEXT:    callq __truncsfhf2@PLT
+; CHECK-AVX2-NEXT:    vmovss %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
+; CHECK-AVX2-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; CHECK-AVX2-NEXT:    vpextrw $1, %xmm0, %eax
+; CHECK-AVX2-NEXT:    vcvtsi2ss %eax, %xmm2, %xmm0
+; CHECK-AVX2-NEXT:    vzeroupper
+; CHECK-AVX2-NEXT:    callq __truncsfhf2@PLT
+; CHECK-AVX2-NEXT:    vmovss %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
+; CHECK-AVX2-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; CHECK-AVX2-NEXT:    vpextrw $2, %xmm0, %eax
+; CHECK-AVX2-NEXT:    vcvtsi2ss %eax, %xmm2, %xmm0
+; CHECK-AVX2-NEXT:    vzeroupper
+; CHECK-AVX2-NEXT:    callq __truncsfhf2@PLT
+; CHECK-AVX2-NEXT:    vmovss %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
+; CHECK-AVX2-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; CHECK-AVX2-NEXT:    vpextrw $3, %xmm0, %eax
+; CHECK-AVX2-NEXT:    vcvtsi2ss %eax, %xmm2, %xmm0
+; CHECK-AVX2-NEXT:    vzeroupper
+; CHECK-AVX2-NEXT:    callq __truncsfhf2@PLT
+; CHECK-AVX2-NEXT:    vmovss %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
+; CHECK-AVX2-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; CHECK-AVX2-NEXT:    vpextrw $4, %xmm0, %eax
+; CHECK-AVX2-NEXT:    vcvtsi2ss %eax, %xmm2, %xmm0
+; CHECK-AVX2-NEXT:    vzeroupper
+; CHECK-AVX2-NEXT:    callq __truncsfhf2@PLT
+; CHECK-AVX2-NEXT:    vmovss %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
+; CHECK-AVX2-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; CHECK-AVX2-NEXT:    vpextrw $5, %xmm0, %eax
+; CHECK-AVX2-NEXT:    vcvtsi2ss %eax, %xmm2, %xmm0
+; CHECK-AVX2-NEXT:    vzeroupper
+; CHECK-AVX2-NEXT:    callq __truncsfhf2@PLT
+; CHECK-AVX2-NEXT:    vmovss %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
+; CHECK-AVX2-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; CHECK-AVX2-NEXT:    vpextrw $6, %xmm0, %eax
+; CHECK-AVX2-NEXT:    vcvtsi2ss %eax, %xmm2, %xmm0
+; CHECK-AVX2-NEXT:    vzeroupper
+; CHECK-AVX2-NEXT:    callq __truncsfhf2@PLT
+; CHECK-AVX2-NEXT:    vmovss %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
+; CHECK-AVX2-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; CHECK-AVX2-NEXT:    vpextrw $7, %xmm0, %eax
+; CHECK-AVX2-NEXT:    vcvtsi2ss %eax, %xmm2, %xmm0
+; CHECK-AVX2-NEXT:    vzeroupper
+; CHECK-AVX2-NEXT:    callq __truncsfhf2@PLT
+; CHECK-AVX2-NEXT:    callq __extendhfsf2@PLT
+; CHECK-AVX2-NEXT:    vmulss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; CHECK-AVX2-NEXT:    callq __truncsfhf2@PLT
+; CHECK-AVX2-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; CHECK-AVX2-NEXT:    vmovss {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 4-byte Reload
+; CHECK-AVX2-NEXT:    # xmm0 = mem[0],zero,zero,zero
+; CHECK-AVX2-NEXT:    callq __extendhfsf2@PLT
+; CHECK-AVX2-NEXT:    vmulss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; CHECK-AVX2-NEXT:    callq __truncsfhf2@PLT
+; CHECK-AVX2-NEXT:    vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; CHECK-AVX2-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; CHECK-AVX2-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; CHECK-AVX2-NEXT:    vmovss {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 4-byte Reload
+; CHECK-AVX2-NEXT:    # xmm0 = mem[0],zero,zero,zero
+; CHECK-AVX2-NEXT:    callq __extendhfsf2@PLT
+; CHECK-AVX2-NEXT:    vmulss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; CHECK-AVX2-NEXT:    callq __truncsfhf2@PLT
+; CHECK-AVX2-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; CHECK-AVX2-NEXT:    vmovss {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 4-byte Reload
+; CHECK-AVX2-NEXT:    # xmm0 = mem[0],zero,zero,zero
+; CHECK-AVX2-NEXT:    callq __extendhfsf2@PLT
+; CHECK-AVX2-NEXT:    vmulss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; CHECK-AVX2-NEXT:    callq __truncsfhf2@PLT
+; CHECK-AVX2-NEXT:    vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; CHECK-AVX2-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; CHECK-AVX2-NEXT:    vpunpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; CHECK-AVX2-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; CHECK-AVX2-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; CHECK-AVX2-NEXT:    vmovss {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 4-byte Reload
+; CHECK-AVX2-NEXT:    # xmm0 = mem[0],zero,zero,zero
+; CHECK-AVX2-NEXT:    callq __extendhfsf2@PLT
+; CHECK-AVX2-NEXT:    vmulss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; CHECK-AVX2-NEXT:    callq __truncsfhf2@PLT
+; CHECK-AVX2-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; CHECK-AVX2-NEXT:    vmovss {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 4-byte Reload
+; CHECK-AVX2-NEXT:    # xmm0 = mem[0],zero,zero,zero
+; CHECK-AVX2-NEXT:    callq __extendhfsf2@PLT
+; CHECK-AVX2-NEXT:    vmulss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; CHECK-AVX2-NEXT:    callq __truncsfhf2@PLT
+; CHECK-AVX2-NEXT:    vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; CHECK-AVX2-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; CHECK-AVX2-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; CHECK-AVX2-NEXT:    vmovss {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 4-byte Reload
+; CHECK-AVX2-NEXT:    # xmm0 = mem[0],zero,zero,zero
+; CHECK-AVX2-NEXT:    callq __extendhfsf2@PLT
+; CHECK-AVX2-NEXT:    vmulss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; CHECK-AVX2-NEXT:    callq __truncsfhf2@PLT
+; CHECK-AVX2-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; CHECK-AVX2-NEXT:    vmovss {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 4-byte Reload
+; CHECK-AVX2-NEXT:    # xmm0 = mem[0],zero,zero,zero
+; CHECK-AVX2-NEXT:    callq __extendhfsf2@PLT
+; CHECK-AVX2-NEXT:    vmulss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; CHECK-AVX2-NEXT:    callq __truncsfhf2@PLT
+; CHECK-AVX2-NEXT:    vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; CHECK-AVX2-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; CHECK-AVX2-NEXT:    vpunpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; CHECK-AVX2-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; CHECK-AVX2-NEXT:    vpunpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; CHECK-AVX2-NEXT:    # xmm0 = xmm0[0],mem[0]
+; CHECK-AVX2-NEXT:    addq $120, %rsp
+; CHECK-AVX2-NEXT:    .cfi_def_cfa_offset 8
+; CHECK-AVX2-NEXT:    retq
+;
+; CHECK-NO-FASTFMA-LABEL: fmul_pow2_8xhalf:
+; CHECK-NO-FASTFMA:       # %bb.0:
+; CHECK-NO-FASTFMA-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; CHECK-NO-FASTFMA-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [1,1,1,1,1,1,1,1]
+; CHECK-NO-FASTFMA-NEXT:    vpsllvd %ymm0, %ymm1, %ymm1
+; CHECK-NO-FASTFMA-NEXT:    vpmovdw %zmm1, %ymm0
+; CHECK-NO-FASTFMA-NEXT:    vpextrw $7, %xmm0, %eax
+; CHECK-NO-FASTFMA-NEXT:    vcvtsi2ss %eax, %xmm2, %xmm2
+; CHECK-NO-FASTFMA-NEXT:    vcvtps2ph $4, %xmm2, %xmm2
+; CHECK-NO-FASTFMA-NEXT:    vmovd %xmm2, %eax
+; CHECK-NO-FASTFMA-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm2
+; CHECK-NO-FASTFMA-NEXT:    vpextrw $6, %xmm0, %eax
+; CHECK-NO-FASTFMA-NEXT:    vcvtsi2ss %eax, %xmm3, %xmm3
+; CHECK-NO-FASTFMA-NEXT:    vcvtps2ph $4, %xmm3, %xmm3
+; CHECK-NO-FASTFMA-NEXT:    vmovd %xmm3, %eax
+; CHECK-NO-FASTFMA-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm3
+; CHECK-NO-FASTFMA-NEXT:    vpunpcklwd {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
+; CHECK-NO-FASTFMA-NEXT:    vpextrw $5, %xmm0, %eax
+; CHECK-NO-FASTFMA-NEXT:    vcvtsi2ss %eax, %xmm4, %xmm3
+; CHECK-NO-FASTFMA-NEXT:    vcvtps2ph $4, %xmm3, %xmm3
+; CHECK-NO-FASTFMA-NEXT:    vmovd %xmm3, %eax
+; CHECK-NO-FASTFMA-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm3
+; CHECK-NO-FASTFMA-NEXT:    vpextrw $4, %xmm0, %eax
+; CHECK-NO-FASTFMA-NEXT:    vcvtsi2ss %eax, %xmm4, %xmm4
+; CHECK-NO-FASTFMA-NEXT:    vcvtps2ph $4, %xmm4, %xmm4
+; CHECK-NO-FASTFMA-NEXT:    vmovd %xmm4, %eax
+; CHECK-NO-FASTFMA-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm4
+; CHECK-NO-FASTFMA-NEXT:    vpunpcklwd {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
+; CHECK-NO-FASTFMA-NEXT:    vpunpckldq {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
+; CHECK-NO-FASTFMA-NEXT:    vpextrw $3, %xmm0, %eax
+; CHECK-NO-FASTFMA-NEXT:    vcvtsi2ss %eax, %xmm5, %xmm3
+; CHECK-NO-FASTFMA-NEXT:    vcvtps2ph $4, %xmm3, %xmm3
+; CHECK-NO-FASTFMA-NEXT:    vmovd %xmm3, %eax
+; CHECK-NO-FASTFMA-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm3
+; CHECK-NO-FASTFMA-NEXT:    vpextrw $2, %xmm0, %eax
+; CHECK-NO-FASTFMA-NEXT:    vcvtsi2ss %eax, %xmm5, %xmm4
+; CHECK-NO-FASTFMA-NEXT:    vcvtps2ph $4, %xmm4, %xmm4
+; CHECK-NO-FASTFMA-NEXT:    vmovd %xmm4, %eax
+; CHECK-NO-FASTFMA-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm4
+; CHECK-NO-FASTFMA-NEXT:    vpunpcklwd {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
+; CHECK-NO-FASTFMA-NEXT:    vpextrw $0, %xmm1, %eax
+; CHECK-NO-FASTFMA-NEXT:    vcvtsi2ss %eax, %xmm5, %xmm1
+; CHECK-NO-FASTFMA-NEXT:    vcvtps2ph $4, %xmm1, %xmm1
+; CHECK-NO-FASTFMA-NEXT:    vmovd %xmm1, %eax
+; CHECK-NO-FASTFMA-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm1
+; CHECK-NO-FASTFMA-NEXT:    vpextrw $1, %xmm0, %eax
+; CHECK-NO-FASTFMA-NEXT:    vcvtsi2ss %eax, %xmm5, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vcvtps2ph $4, %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vmovd %xmm0, %eax
+; CHECK-NO-FASTFMA-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; CHECK-NO-FASTFMA-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
+; CHECK-NO-FASTFMA-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm2[0]
+; CHECK-NO-FASTFMA-NEXT:    vcvtph2ps %xmm0, %ymm0
+; CHECK-NO-FASTFMA-NEXT:    vbroadcastss {{.*#+}} ymm1 = [8.192E+3,8.192E+3,8.192E+3,8.192E+3,8.192E+3,8.192E+3,8.192E+3,8.192E+3]
+; CHECK-NO-FASTFMA-NEXT:    vmulps %ymm1, %ymm0, %ymm0
+; CHECK-NO-FASTFMA-NEXT:    vcvtps2ph $4, %ymm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vzeroupper
+; CHECK-NO-FASTFMA-NEXT:    retq
+;
+; CHECK-FMA-LABEL: fmul_pow2_8xhalf:
+; CHECK-FMA:       # %bb.0:
+; CHECK-FMA-NEXT:    vpbroadcastw {{.*#+}} xmm1 = [1,1,1,1,1,1,1,1]
+; CHECK-FMA-NEXT:    vpsllvw %xmm0, %xmm1, %xmm0
+; CHECK-FMA-NEXT:    vpextrw $7, %xmm0, %eax
+; CHECK-FMA-NEXT:    vcvtsi2ss %eax, %xmm2, %xmm1
+; CHECK-FMA-NEXT:    vcvtps2ph $4, %xmm1, %xmm1
+; CHECK-FMA-NEXT:    vmovd %xmm1, %eax
+; CHECK-FMA-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm1
+; CHECK-FMA-NEXT:    vpextrw $6, %xmm0, %eax
+; CHECK-FMA-NEXT:    vcvtsi2ss %eax, %xmm2, %xmm2
+; CHECK-FMA-NEXT:    vcvtps2ph $4, %xmm2, %xmm2
+; CHECK-FMA-NEXT:    vmovd %xmm2, %eax
+; CHECK-FMA-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm2
+; CHECK-FMA-NEXT:    vpextrw $5, %xmm0, %eax
+; CHECK-FMA-NEXT:    vcvtsi2ss %eax, %xmm3, %xmm3
+; CHECK-FMA-NEXT:    vcvtps2ph $4, %xmm3, %xmm3
+; CHECK-FMA-NEXT:    vmovd %xmm3, %eax
+; CHECK-FMA-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm3
+; CHECK-FMA-NEXT:    vpextrw $4, %xmm0, %eax
+; CHECK-FMA-NEXT:    vcvtsi2ss %eax, %xmm4, %xmm4
+; CHECK-FMA-NEXT:    vpunpcklwd {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
+; CHECK-FMA-NEXT:    vcvtps2ph $4, %xmm4, %xmm2
+; CHECK-FMA-NEXT:    vmovd %xmm2, %eax
+; CHECK-FMA-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm2
+; CHECK-FMA-NEXT:    vpextrw $3, %xmm0, %eax
+; CHECK-FMA-NEXT:    vcvtsi2ss %eax, %xmm5, %xmm4
+; CHECK-FMA-NEXT:    vcvtps2ph $4, %xmm4, %xmm4
+; CHECK-FMA-NEXT:    vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3]
+; CHECK-FMA-NEXT:    vmovd %xmm4, %eax
+; CHECK-FMA-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm3
+; CHECK-FMA-NEXT:    vpextrw $2, %xmm0, %eax
+; CHECK-FMA-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; CHECK-FMA-NEXT:    vcvtsi2ss %eax, %xmm5, %xmm2
+; CHECK-FMA-NEXT:    vcvtps2ph $4, %xmm2, %xmm2
+; CHECK-FMA-NEXT:    vmovd %xmm2, %eax
+; CHECK-FMA-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm2
+; CHECK-FMA-NEXT:    vpextrw $1, %xmm0, %eax
+; CHECK-FMA-NEXT:    vcvtsi2ss %eax, %xmm5, %xmm4
+; CHECK-FMA-NEXT:    vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3]
+; CHECK-FMA-NEXT:    vcvtps2ph $4, %xmm4, %xmm3
+; CHECK-FMA-NEXT:    vmovd %xmm3, %eax
+; CHECK-FMA-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm3
+; CHECK-FMA-NEXT:    vpextrw $0, %xmm0, %eax
+; CHECK-FMA-NEXT:    vcvtsi2ss %eax, %xmm5, %xmm0
+; CHECK-FMA-NEXT:    vcvtps2ph $4, %xmm0, %xmm0
+; CHECK-FMA-NEXT:    vmovd %xmm0, %eax
+; CHECK-FMA-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0
+; CHECK-FMA-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3]
+; CHECK-FMA-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; CHECK-FMA-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; CHECK-FMA-NEXT:    vcvtph2ps %xmm0, %ymm0
+; CHECK-FMA-NEXT:    vmulps {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to8}, %ymm0, %ymm0
+; CHECK-FMA-NEXT:    vcvtps2ph $4, %ymm0, %xmm0
+; CHECK-FMA-NEXT:    vzeroupper
+; CHECK-FMA-NEXT:    retq
+  %p2 = shl <8 x i16> <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>, %i
+  %p2_f = uitofp <8 x i16> %p2 to <8 x half>
+  %r = fmul <8 x half> <half 0xH7000, half 0xH7000, half 0xH7000, half 0xH7000, half 0xH7000, half 0xH7000, half 0xH7000, half 0xH7000>, %p2_f
+  ret <8 x half> %r
+}
+
+define <8 x half> @fmul_pow2_ldexp_8xhalf(<8 x i16> %i) {
+; CHECK-SSE-LABEL: fmul_pow2_ldexp_8xhalf:
+; CHECK-SSE:       # %bb.0:
+; CHECK-SSE-NEXT:    subq $72, %rsp
+; CHECK-SSE-NEXT:    .cfi_def_cfa_offset 80
+; CHECK-SSE-NEXT:    movdqa %xmm0, (%rsp) # 16-byte Spill
+; CHECK-SSE-NEXT:    pextrw $7, %xmm0, %edi
+; CHECK-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; CHECK-SSE-NEXT:    callq ldexpf@PLT
+; CHECK-SSE-NEXT:    callq __truncsfhf2@PLT
+; CHECK-SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; CHECK-SSE-NEXT:    movdqa (%rsp), %xmm0 # 16-byte Reload
+; CHECK-SSE-NEXT:    pextrw $6, %xmm0, %edi
+; CHECK-SSE-NEXT:    movd {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; CHECK-SSE-NEXT:    callq ldexpf@PLT
+; CHECK-SSE-NEXT:    callq __truncsfhf2@PLT
+; CHECK-SSE-NEXT:    punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; CHECK-SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; CHECK-SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; CHECK-SSE-NEXT:    movdqa (%rsp), %xmm0 # 16-byte Reload
+; CHECK-SSE-NEXT:    pextrw $5, %xmm0, %edi
+; CHECK-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; CHECK-SSE-NEXT:    callq ldexpf@PLT
+; CHECK-SSE-NEXT:    callq __truncsfhf2@PLT
+; CHECK-SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; CHECK-SSE-NEXT:    movdqa (%rsp), %xmm0 # 16-byte Reload
+; CHECK-SSE-NEXT:    pextrw $4, %xmm0, %edi
+; CHECK-SSE-NEXT:    movd {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; CHECK-SSE-NEXT:    callq ldexpf@PLT
+; CHECK-SSE-NEXT:    callq __truncsfhf2@PLT
+; CHECK-SSE-NEXT:    punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; CHECK-SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; CHECK-SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; CHECK-SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; CHECK-SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; CHECK-SSE-NEXT:    movdqa (%rsp), %xmm0 # 16-byte Reload
+; CHECK-SSE-NEXT:    pextrw $3, %xmm0, %edi
+; CHECK-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; CHECK-SSE-NEXT:    callq ldexpf@PLT
+; CHECK-SSE-NEXT:    callq __truncsfhf2@PLT
+; CHECK-SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; CHECK-SSE-NEXT:    movdqa (%rsp), %xmm0 # 16-byte Reload
+; CHECK-SSE-NEXT:    pextrw $2, %xmm0, %edi
+; CHECK-SSE-NEXT:    movd {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; CHECK-SSE-NEXT:    callq ldexpf@PLT
+; CHECK-SSE-NEXT:    callq __truncsfhf2@PLT
+; CHECK-SSE-NEXT:    punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; CHECK-SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; CHECK-SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; CHECK-SSE-NEXT:    movdqa (%rsp), %xmm0 # 16-byte Reload
+; CHECK-SSE-NEXT:    pextrw $1, %xmm0, %edi
+; CHECK-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; CHECK-SSE-NEXT:    callq ldexpf@PLT
+; CHECK-SSE-NEXT:    callq __truncsfhf2@PLT
+; CHECK-SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; CHECK-SSE-NEXT:    movdqa (%rsp), %xmm0 # 16-byte Reload
+; CHECK-SSE-NEXT:    movd %xmm0, %eax
+; CHECK-SSE-NEXT:    movzwl %ax, %edi
+; CHECK-SSE-NEXT:    movd {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; CHECK-SSE-NEXT:    callq ldexpf@PLT
+; CHECK-SSE-NEXT:    callq __truncsfhf2@PLT
+; CHECK-SSE-NEXT:    punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; CHECK-SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; CHECK-SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; CHECK-SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; CHECK-SSE-NEXT:    punpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; CHECK-SSE-NEXT:    # xmm0 = xmm0[0],mem[0]
+; CHECK-SSE-NEXT:    addq $72, %rsp
+; CHECK-SSE-NEXT:    .cfi_def_cfa_offset 8
+; CHECK-SSE-NEXT:    retq
+;
+; CHECK-AVX2-LABEL: fmul_pow2_ldexp_8xhalf:
+; CHECK-AVX2:       # %bb.0:
+; CHECK-AVX2-NEXT:    subq $72, %rsp
+; CHECK-AVX2-NEXT:    .cfi_def_cfa_offset 80
+; CHECK-AVX2-NEXT:    vmovdqa %xmm0, (%rsp) # 16-byte Spill
+; CHECK-AVX2-NEXT:    vpextrw $7, %xmm0, %edi
+; CHECK-AVX2-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; CHECK-AVX2-NEXT:    callq ldexpf@PLT
+; CHECK-AVX2-NEXT:    callq __truncsfhf2@PLT
+; CHECK-AVX2-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; CHECK-AVX2-NEXT:    vmovdqa (%rsp), %xmm0 # 16-byte Reload
+; CHECK-AVX2-NEXT:    vpextrw $6, %xmm0, %edi
+; CHECK-AVX2-NEXT:    vmovd {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; CHECK-AVX2-NEXT:    callq ldexpf@PLT
+; CHECK-AVX2-NEXT:    callq __truncsfhf2@PLT
+; CHECK-AVX2-NEXT:    vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; CHECK-AVX2-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; CHECK-AVX2-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; CHECK-AVX2-NEXT:    vmovdqa (%rsp), %xmm0 # 16-byte Reload
+; CHECK-AVX2-NEXT:    vpextrw $5, %xmm0, %edi
+; CHECK-AVX2-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; CHECK-AVX2-NEXT:    callq ldexpf@PLT
+; CHECK-AVX2-NEXT:    callq __truncsfhf2@PLT
+; CHECK-AVX2-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; CHECK-AVX2-NEXT:    vmovdqa (%rsp), %xmm0 # 16-byte Reload
+; CHECK-AVX2-NEXT:    vpextrw $4, %xmm0, %edi
+; CHECK-AVX2-NEXT:    vmovd {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; CHECK-AVX2-NEXT:    callq ldexpf@PLT
+; CHECK-AVX2-NEXT:    callq __truncsfhf2@PLT
+; CHECK-AVX2-NEXT:    vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; CHECK-AVX2-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; CHECK-AVX2-NEXT:    vpunpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; CHECK-AVX2-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; CHECK-AVX2-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; CHECK-AVX2-NEXT:    vmovdqa (%rsp), %xmm0 # 16-byte Reload
+; CHECK-AVX2-NEXT:    vpextrw $3, %xmm0, %edi
+; CHECK-AVX2-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; CHECK-AVX2-NEXT:    callq ldexpf@PLT
+; CHECK-AVX2-NEXT:    callq __truncsfhf2@PLT
+; CHECK-AVX2-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; CHECK-AVX2-NEXT:    vmovdqa (%rsp), %xmm0 # 16-byte Reload
+; CHECK-AVX2-NEXT:    vpextrw $2, %xmm0, %edi
+; CHECK-AVX2-NEXT:    vmovd {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; CHECK-AVX2-NEXT:    callq ldexpf@PLT
+; CHECK-AVX2-NEXT:    callq __truncsfhf2@PLT
+; CHECK-AVX2-NEXT:    vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; CHECK-AVX2-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; CHECK-AVX2-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; CHECK-AVX2-NEXT:    vmovdqa (%rsp), %xmm0 # 16-byte Reload
+; CHECK-AVX2-NEXT:    vpextrw $1, %xmm0, %edi
+; CHECK-AVX2-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; CHECK-AVX2-NEXT:    callq ldexpf@PLT
+; CHECK-AVX2-NEXT:    callq __truncsfhf2@PLT
+; CHECK-AVX2-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; CHECK-AVX2-NEXT:    vmovdqa (%rsp), %xmm0 # 16-byte Reload
+; CHECK-AVX2-NEXT:    vmovd %xmm0, %eax
+; CHECK-AVX2-NEXT:    movzwl %ax, %edi
+; CHECK-AVX2-NEXT:    vmovd {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; CHECK-AVX2-NEXT:    callq ldexpf@PLT
+; CHECK-AVX2-NEXT:    callq __truncsfhf2@PLT
+; CHECK-AVX2-NEXT:    vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; CHECK-AVX2-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; CHECK-AVX2-NEXT:    vpunpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; CHECK-AVX2-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; CHECK-AVX2-NEXT:    vpunpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; CHECK-AVX2-NEXT:    # xmm0 = xmm0[0],mem[0]
+; CHECK-AVX2-NEXT:    addq $72, %rsp
+; CHECK-AVX2-NEXT:    .cfi_def_cfa_offset 8
+; CHECK-AVX2-NEXT:    retq
+;
+; CHECK-AVX512F-LABEL: fmul_pow2_ldexp_8xhalf:
+; CHECK-AVX512F:       # %bb.0:
+; CHECK-AVX512F-NEXT:    subq $72, %rsp
+; CHECK-AVX512F-NEXT:    .cfi_def_cfa_offset 80
+; CHECK-AVX512F-NEXT:    vmovdqa %xmm0, (%rsp) # 16-byte Spill
+; CHECK-AVX512F-NEXT:    vpextrw $7, %xmm0, %edi
+; CHECK-AVX512F-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; CHECK-AVX512F-NEXT:    callq ldexpf@PLT
+; CHECK-AVX512F-NEXT:    vcvtps2ph $4, %xmm0, %xmm0
+; CHECK-AVX512F-NEXT:    vmovd %xmm0, %eax
+; CHECK-AVX512F-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0
+; CHECK-AVX512F-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; CHECK-AVX512F-NEXT:    vmovdqa (%rsp), %xmm0 # 16-byte Reload
+; CHECK-AVX512F-NEXT:    vpextrw $6, %xmm0, %edi
+; CHECK-AVX512F-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; CHECK-AVX512F-NEXT:    callq ldexpf@PLT
+; CHECK-AVX512F-NEXT:    vcvtps2ph $4, %xmm0, %xmm0
+; CHECK-AVX512F-NEXT:    vmovd %xmm0, %eax
+; CHECK-AVX512F-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0
+; CHECK-AVX512F-NEXT:    vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; CHECK-AVX512F-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; CHECK-AVX512F-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; CHECK-AVX512F-NEXT:    vmovdqa (%rsp), %xmm0 # 16-byte Reload
+; CHECK-AVX512F-NEXT:    vpextrw $5, %xmm0, %edi
+; CHECK-AVX512F-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; CHECK-AVX512F-NEXT:    callq ldexpf@PLT
+; CHECK-AVX512F-NEXT:    vcvtps2ph $4, %xmm0, %xmm0
+; CHECK-AVX512F-NEXT:    vmovd %xmm0, %eax
+; CHECK-AVX512F-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0
+; CHECK-AVX512F-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; CHECK-AVX512F-NEXT:    vmovdqa (%rsp), %xmm0 # 16-byte Reload
+; CHECK-AVX512F-NEXT:    vpextrw $4, %xmm0, %edi
+; CHECK-AVX512F-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; CHECK-AVX512F-NEXT:    callq ldexpf@PLT
+; CHECK-AVX512F-NEXT:    vcvtps2ph $4, %xmm0, %xmm0
+; CHECK-AVX512F-NEXT:    vmovd %xmm0, %eax
+; CHECK-AVX512F-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0
+; CHECK-AVX512F-NEXT:    vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; CHECK-AVX512F-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; CHECK-AVX512F-NEXT:    vpunpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; CHECK-AVX512F-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; CHECK-AVX512F-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; CHECK-AVX512F-NEXT:    vmovdqa (%rsp), %xmm0 # 16-byte Reload
+; CHECK-AVX512F-NEXT:    vpextrw $3, %xmm0, %edi
+; CHECK-AVX512F-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; CHECK-AVX512F-NEXT:    callq ldexpf@PLT
+; CHECK-AVX512F-NEXT:    vcvtps2ph $4, %xmm0, %xmm0
+; CHECK-AVX512F-NEXT:    vmovd %xmm0, %eax
+; CHECK-AVX512F-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0
+; CHECK-AVX512F-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; CHECK-AVX512F-NEXT:    vmovdqa (%rsp), %xmm0 # 16-byte Reload
+; CHECK-AVX512F-NEXT:    vpextrw $2, %xmm0, %edi
+; CHECK-AVX512F-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; CHECK-AVX512F-NEXT:    callq ldexpf@PLT
+; CHECK-AVX512F-NEXT:    vcvtps2ph $4, %xmm0, %xmm0
+; CHECK-AVX512F-NEXT:    vmovd %xmm0, %eax
+; CHECK-AVX512F-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0
+; CHECK-AVX512F-NEXT:    vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; CHECK-AVX512F-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; CHECK-AVX512F-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; CHECK-AVX512F-NEXT:    vmovdqa (%rsp), %xmm0 # 16-byte Reload
+; CHECK-AVX512F-NEXT:    vpextrw $1, %xmm0, %edi
+; CHECK-AVX512F-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; CHECK-AVX512F-NEXT:    callq ldexpf@PLT
+; CHECK-AVX512F-NEXT:    vcvtps2ph $4, %xmm0, %xmm0
+; CHECK-AVX512F-NEXT:    vmovd %xmm0, %eax
+; CHECK-AVX512F-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0
+; CHECK-AVX512F-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; CHECK-AVX512F-NEXT:    vmovdqa (%rsp), %xmm0 # 16-byte Reload
+; CHECK-AVX512F-NEXT:    vmovd %xmm0, %eax
+; CHECK-AVX512F-NEXT:    movzwl %ax, %edi
+; CHECK-AVX512F-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; CHECK-AVX512F-NEXT:    callq ldexpf@PLT
+; CHECK-AVX512F-NEXT:    vcvtps2ph $4, %xmm0, %xmm0
+; CHECK-AVX512F-NEXT:    vmovd %xmm0, %eax
+; CHECK-AVX512F-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0
+; CHECK-AVX512F-NEXT:    vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; CHECK-AVX512F-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; CHECK-AVX512F-NEXT:    vpunpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; CHECK-AVX512F-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; CHECK-AVX512F-NEXT:    vpunpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; CHECK-AVX512F-NEXT:    # xmm0 = xmm0[0],mem[0]
+; CHECK-AVX512F-NEXT:    addq $72, %rsp
+; CHECK-AVX512F-NEXT:    .cfi_def_cfa_offset 8
+; CHECK-AVX512F-NEXT:    retq
+  %r = call <8 x half> @llvm.ldexp.v8f16.v8i16(<8 x half> <half 0xH7000, half 0xH7000, half 0xH7000, half 0xH7000, half 0xH7000, half 0xH7000, half 0xH7000, half 0xH7000>, <8 x i16> %i)
+  ret <8 x half> %r
+}
+
+define <8 x half> @fdiv_pow2_8xhalf(<8 x i16> %i) {
+; CHECK-SSE-LABEL: fdiv_pow2_8xhalf:
+; CHECK-SSE:       # %bb.0:
+; CHECK-SSE-NEXT:    subq $88, %rsp
+; CHECK-SSE-NEXT:    .cfi_def_cfa_offset 96
+; CHECK-SSE-NEXT:    movdqa %xmm0, %xmm1
+; CHECK-SSE-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4,4,5,5,6,6,7,7]
+; CHECK-SSE-NEXT:    pslld $23, %xmm1
+; CHECK-SSE-NEXT:    movdqa {{.*#+}} xmm2 = [1065353216,1065353216,1065353216,1065353216]
+; CHECK-SSE-NEXT:    paddd %xmm2, %xmm1
+; CHECK-SSE-NEXT:    cvttps2dq %xmm1, %xmm1
+; CHECK-SSE-NEXT:    movaps %xmm1, (%rsp) # 16-byte Spill
+; CHECK-SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3]
+; CHECK-SSE-NEXT:    pslld $23, %xmm0
+; CHECK-SSE-NEXT:    paddd %xmm2, %xmm0
+; CHECK-SSE-NEXT:    cvttps2dq %xmm0, %xmm0
+; CHECK-SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; CHECK-SSE-NEXT:    pextrw $0, %xmm0, %eax
+; CHECK-SSE-NEXT:    xorps %xmm0, %xmm0
+; CHECK-SSE-NEXT:    cvtsi2ss %eax, %xmm0
+; CHECK-SSE-NEXT:    callq __truncsfhf2@PLT
+; CHECK-SSE-NEXT:    movss %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
+; CHECK-SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; CHECK-SSE-NEXT:    pextrw $2, %xmm0, %eax
+; CHECK-SSE-NEXT:    xorps %xmm0, %xmm0
+; CHECK-SSE-NEXT:    cvtsi2ss %eax, %xmm0
+; CHECK-SSE-NEXT:    callq __truncsfhf2@PLT
+; CHECK-SSE-NEXT:    movss %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
+; CHECK-SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; CHECK-SSE-NEXT:    pextrw $4, %xmm0, %eax
+; CHECK-SSE-NEXT:    xorps %xmm0, %xmm0
+; CHECK-SSE-NEXT:    cvtsi2ss %eax, %xmm0
+; CHECK-SSE-NEXT:    callq __truncsfhf2@PLT
+; CHECK-SSE-NEXT:    movss %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
+; CHECK-SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; CHECK-SSE-NEXT:    pextrw $6, %xmm0, %eax
+; CHECK-SSE-NEXT:    xorps %xmm0, %xmm0
+; CHECK-SSE-NEXT:    cvtsi2ss %eax, %xmm0
+; CHECK-SSE-NEXT:    callq __truncsfhf2@PLT
+; CHECK-SSE-NEXT:    movss %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
+; CHECK-SSE-NEXT:    movdqa (%rsp), %xmm0 # 16-byte Reload
+; CHECK-SSE-NEXT:    pextrw $0, %xmm0, %eax
+; CHECK-SSE-NEXT:    xorps %xmm0, %xmm0
+; CHECK-SSE-NEXT:    cvtsi2ss %eax, %xmm0
+; CHECK-SSE-NEXT:    callq __truncsfhf2@PLT
+; CHECK-SSE-NEXT:    movss %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
+; CHECK-SSE-NEXT:    movdqa (%rsp), %xmm0 # 16-byte Reload
+; CHECK-SSE-NEXT:    pextrw $2, %xmm0, %eax
+; CHECK-SSE-NEXT:    xorps %xmm0, %xmm0
+; CHECK-SSE-NEXT:    cvtsi2ss %eax, %xmm0
+; CHECK-SSE-NEXT:    callq __truncsfhf2@PLT
+; CHECK-SSE-NEXT:    movss %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
+; CHECK-SSE-NEXT:    movdqa (%rsp), %xmm0 # 16-byte Reload
+; CHECK-SSE-NEXT:    pextrw $4, %xmm0, %eax
+; CHECK-SSE-NEXT:    xorps %xmm0, %xmm0
+; CHECK-SSE-NEXT:    cvtsi2ss %eax, %xmm0
+; CHECK-SSE-NEXT:    callq __truncsfhf2@PLT
+; CHECK-SSE-NEXT:    movss %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
+; CHECK-SSE-NEXT:    movdqa (%rsp), %xmm0 # 16-byte Reload
+; CHECK-SSE-NEXT:    pextrw $6, %xmm0, %eax
+; CHECK-SSE-NEXT:    xorps %xmm0, %xmm0
+; CHECK-SSE-NEXT:    cvtsi2ss %eax, %xmm0
+; CHECK-SSE-NEXT:    callq __truncsfhf2@PLT
+; CHECK-SSE-NEXT:    callq __extendhfsf2@PLT
+; CHECK-SSE-NEXT:    movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; CHECK-SSE-NEXT:    divss %xmm0, %xmm1
+; CHECK-SSE-NEXT:    movaps %xmm1, %xmm0
+; CHECK-SSE-NEXT:    callq __truncsfhf2@PLT
+; CHECK-SSE-NEXT:    movaps %xmm0, (%rsp) # 16-byte Spill
+; CHECK-SSE-NEXT:    movss {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 4-byte Reload
+; CHECK-SSE-NEXT:    # xmm0 = mem[0],zero,zero,zero
+; CHECK-SSE-NEXT:    callq __extendhfsf2@PLT
+; CHECK-SSE-NEXT:    movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; CHECK-SSE-NEXT:    divss %xmm0, %xmm1
+; CHECK-SSE-NEXT:    movaps %xmm1, %xmm0
+; CHECK-SSE-NEXT:    callq __truncsfhf2@PLT
+; CHECK-SSE-NEXT:    punpcklwd (%rsp), %xmm0 # 16-byte Folded Reload
+; CHECK-SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; CHECK-SSE-NEXT:    movdqa %xmm0, (%rsp) # 16-byte Spill
+; CHECK-SSE-NEXT:    movss {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 4-byte Reload
+; CHECK-SSE-NEXT:    # xmm0 = mem[0],zero,zero,zero
+; CHECK-SSE-NEXT:    callq __extendhfsf2@PLT
+; CHECK-SSE-NEXT:    movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; CHECK-SSE-NEXT:    divss %xmm0, %xmm1
+; CHECK-SSE-NEXT:    movaps %xmm1, %xmm0
+; CHECK-SSE-NEXT:    callq __truncsfhf2@PLT
+; CHECK-SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; CHECK-SSE-NEXT:    movss {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 4-byte Reload
+; CHECK-SSE-NEXT:    # xmm0 = mem[0],zero,zero,zero
+; CHECK-SSE-NEXT:    callq __extendhfsf2@PLT
+; CHECK-SSE-NEXT:    movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; CHECK-SSE-NEXT:    divss %xmm0, %xmm1
+; CHECK-SSE-NEXT:    movaps %xmm1, %xmm0
+; CHECK-SSE-NEXT:    callq __truncsfhf2@PLT
+; CHECK-SSE-NEXT:    punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; CHECK-SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; CHECK-SSE-NEXT:    punpckldq (%rsp), %xmm0 # 16-byte Folded Reload
+; CHECK-SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; CHECK-SSE-NEXT:    movdqa %xmm0, (%rsp) # 16-byte Spill
+; CHECK-SSE-NEXT:    movss {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 4-byte Reload
+; CHECK-SSE-NEXT:    # xmm0 = mem[0],zero,zero,zero
+; CHECK-SSE-NEXT:    callq __extendhfsf2@PLT
+; CHECK-SSE-NEXT:    movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; CHECK-SSE-NEXT:    divss %xmm0, %xmm1
+; CHECK-SSE-NEXT:    movaps %xmm1, %xmm0
+; CHECK-SSE-NEXT:    callq __truncsfhf2@PLT
+; CHECK-SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; CHECK-SSE-NEXT:    movss {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 4-byte Reload
+; CHECK-SSE-NEXT:    # xmm0 = mem[0],zero,zero,zero
+; CHECK-SSE-NEXT:    callq __extendhfsf2@PLT
+; CHECK-SSE-NEXT:    movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; CHECK-SSE-NEXT:    divss %xmm0, %xmm1
+; CHECK-SSE-NEXT:    movaps %xmm1, %xmm0
+; CHECK-SSE-NEXT:    callq __truncsfhf2@PLT
+; CHECK-SSE-NEXT:    punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; CHECK-SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; CHECK-SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; CHECK-SSE-NEXT:    movss {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 4-byte Reload
+; CHECK-SSE-NEXT:    # xmm0 = mem[0],zero,zero,zero
+; CHECK-SSE-NEXT:    callq __extendhfsf2@PLT
+; CHECK-SSE-NEXT:    movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; CHECK-SSE-NEXT:    divss %xmm0, %xmm1
+; CHECK-SSE-NEXT:    movaps %xmm1, %xmm0
+; CHECK-SSE-NEXT:    callq __truncsfhf2@PLT
+; CHECK-SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; CHECK-SSE-NEXT:    movss {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 4-byte Reload
+; CHECK-SSE-NEXT:    # xmm0 = mem[0],zero,zero,zero
+; CHECK-SSE-NEXT:    callq __extendhfsf2@PLT
+; CHECK-SSE-NEXT:    movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; CHECK-SSE-NEXT:    divss %xmm0, %xmm1
+; CHECK-SSE-NEXT:    movaps %xmm1, %xmm0
+; CHECK-SSE-NEXT:    callq __truncsfhf2@PLT
+; CHECK-SSE-NEXT:    punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; CHECK-SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; CHECK-SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; CHECK-SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; CHECK-SSE-NEXT:    punpcklqdq (%rsp), %xmm0 # 16-byte Folded Reload
+; CHECK-SSE-NEXT:    # xmm0 = xmm0[0],mem[0]
+; CHECK-SSE-NEXT:    addq $88, %rsp
+; CHECK-SSE-NEXT:    .cfi_def_cfa_offset 8
+; CHECK-SSE-NEXT:    retq
+;
+; CHECK-AVX2-LABEL: fdiv_pow2_8xhalf:
+; CHECK-AVX2:       # %bb.0:
+; CHECK-AVX2-NEXT:    subq $120, %rsp
+; CHECK-AVX2-NEXT:    .cfi_def_cfa_offset 128
+; CHECK-AVX2-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; CHECK-AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [1,1,1,1,1,1,1,1]
+; CHECK-AVX2-NEXT:    vpsllvd %ymm0, %ymm1, %ymm0
+; CHECK-AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,4,5,8,9,12,13,u,u,u,u,u,u,u,u,16,17,20,21,24,25,28,29,u,u,u,u,u,u,u,u]
+; CHECK-AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
+; CHECK-AVX2-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; CHECK-AVX2-NEXT:    vpextrw $0, %xmm0, %eax
+; CHECK-AVX2-NEXT:    vcvtsi2ss %eax, %xmm2, %xmm0
+; CHECK-AVX2-NEXT:    vzeroupper
+; CHECK-AVX2-NEXT:    callq __truncsfhf2@PLT
+; CHECK-AVX2-NEXT:    vmovss %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
+; CHECK-AVX2-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; CHECK-AVX2-NEXT:    vpextrw $1, %xmm0, %eax
+; CHECK-AVX2-NEXT:    vcvtsi2ss %eax, %xmm2, %xmm0
+; CHECK-AVX2-NEXT:    vzeroupper
+; CHECK-AVX2-NEXT:    callq __truncsfhf2@PLT
+; CHECK-AVX2-NEXT:    vmovss %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
+; CHECK-AVX2-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; CHECK-AVX2-NEXT:    vpextrw $2, %xmm0, %eax
+; CHECK-AVX2-NEXT:    vcvtsi2ss %eax, %xmm2, %xmm0
+; CHECK-AVX2-NEXT:    vzeroupper
+; CHECK-AVX2-NEXT:    callq __truncsfhf2@PLT
+; CHECK-AVX2-NEXT:    vmovss %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
+; CHECK-AVX2-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; CHECK-AVX2-NEXT:    vpextrw $3, %xmm0, %eax
+; CHECK-AVX2-NEXT:    vcvtsi2ss %eax, %xmm2, %xmm0
+; CHECK-AVX2-NEXT:    vzeroupper
+; CHECK-AVX2-NEXT:    callq __truncsfhf2@PLT
+; CHECK-AVX2-NEXT:    vmovss %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
+; CHECK-AVX2-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; CHECK-AVX2-NEXT:    vpextrw $4, %xmm0, %eax
+; CHECK-AVX2-NEXT:    vcvtsi2ss %eax, %xmm2, %xmm0
+; CHECK-AVX2-NEXT:    vzeroupper
+; CHECK-AVX2-NEXT:    callq __truncsfhf2@PLT
+; CHECK-AVX2-NEXT:    vmovss %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
+; CHECK-AVX2-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; CHECK-AVX2-NEXT:    vpextrw $5, %xmm0, %eax
+; CHECK-AVX2-NEXT:    vcvtsi2ss %eax, %xmm2, %xmm0
+; CHECK-AVX2-NEXT:    vzeroupper
+; CHECK-AVX2-NEXT:    callq __truncsfhf2@PLT
+; CHECK-AVX2-NEXT:    vmovss %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
+; CHECK-AVX2-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; CHECK-AVX2-NEXT:    vpextrw $6, %xmm0, %eax
+; CHECK-AVX2-NEXT:    vcvtsi2ss %eax, %xmm2, %xmm0
+; CHECK-AVX2-NEXT:    vzeroupper
+; CHECK-AVX2-NEXT:    callq __truncsfhf2@PLT
+; CHECK-AVX2-NEXT:    vmovss %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
+; CHECK-AVX2-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; CHECK-AVX2-NEXT:    vpextrw $7, %xmm0, %eax
+; CHECK-AVX2-NEXT:    vcvtsi2ss %eax, %xmm2, %xmm0
+; CHECK-AVX2-NEXT:    vzeroupper
+; CHECK-AVX2-NEXT:    callq __truncsfhf2@PLT
+; CHECK-AVX2-NEXT:    callq __extendhfsf2@PLT
+; CHECK-AVX2-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; CHECK-AVX2-NEXT:    vdivss %xmm0, %xmm1, %xmm0
+; CHECK-AVX2-NEXT:    callq __truncsfhf2@PLT
+; CHECK-AVX2-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; CHECK-AVX2-NEXT:    vmovss {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 4-byte Reload
+; CHECK-AVX2-NEXT:    # xmm0 = mem[0],zero,zero,zero
+; CHECK-AVX2-NEXT:    callq __extendhfsf2@PLT
+; CHECK-AVX2-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; CHECK-AVX2-NEXT:    vdivss %xmm0, %xmm1, %xmm0
+; CHECK-AVX2-NEXT:    callq __truncsfhf2@PLT
+; CHECK-AVX2-NEXT:    vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; CHECK-AVX2-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; CHECK-AVX2-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; CHECK-AVX2-NEXT:    vmovss {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 4-byte Reload
+; CHECK-AVX2-NEXT:    # xmm0 = mem[0],zero,zero,zero
+; CHECK-AVX2-NEXT:    callq __extendhfsf2@PLT
+; CHECK-AVX2-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; CHECK-AVX2-NEXT:    vdivss %xmm0, %xmm1, %xmm0
+; CHECK-AVX2-NEXT:    callq __truncsfhf2@PLT
+; CHECK-AVX2-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; CHECK-AVX2-NEXT:    vmovss {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 4-byte Reload
+; CHECK-AVX2-NEXT:    # xmm0 = mem[0],zero,zero,zero
+; CHECK-AVX2-NEXT:    callq __extendhfsf2@PLT
+; CHECK-AVX2-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; CHECK-AVX2-NEXT:    vdivss %xmm0, %xmm1, %xmm0
+; CHECK-AVX2-NEXT:    callq __truncsfhf2@PLT
+; CHECK-AVX2-NEXT:    vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; CHECK-AVX2-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; CHECK-AVX2-NEXT:    vpunpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; CHECK-AVX2-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; CHECK-AVX2-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; CHECK-AVX2-NEXT:    vmovss {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 4-byte Reload
+; CHECK-AVX2-NEXT:    # xmm0 = mem[0],zero,zero,zero
+; CHECK-AVX2-NEXT:    callq __extendhfsf2@PLT
+; CHECK-AVX2-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; CHECK-AVX2-NEXT:    vdivss %xmm0, %xmm1, %xmm0
+; CHECK-AVX2-NEXT:    callq __truncsfhf2@PLT
+; CHECK-AVX2-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; CHECK-AVX2-NEXT:    vmovss {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 4-byte Reload
+; CHECK-AVX2-NEXT:    # xmm0 = mem[0],zero,zero,zero
+; CHECK-AVX2-NEXT:    callq __extendhfsf2@PLT
+; CHECK-AVX2-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; CHECK-AVX2-NEXT:    vdivss %xmm0, %xmm1, %xmm0
+; CHECK-AVX2-NEXT:    callq __truncsfhf2@PLT
+; CHECK-AVX2-NEXT:    vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; CHECK-AVX2-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; CHECK-AVX2-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; CHECK-AVX2-NEXT:    vmovss {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 4-byte Reload
+; CHECK-AVX2-NEXT:    # xmm0 = mem[0],zero,zero,zero
+; CHECK-AVX2-NEXT:    callq __extendhfsf2@PLT
+; CHECK-AVX2-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; CHECK-AVX2-NEXT:    vdivss %xmm0, %xmm1, %xmm0
+; CHECK-AVX2-NEXT:    callq __truncsfhf2@PLT
+; CHECK-AVX2-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; CHECK-AVX2-NEXT:    vmovss {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 4-byte Reload
+; CHECK-AVX2-NEXT:    # xmm0 = mem[0],zero,zero,zero
+; CHECK-AVX2-NEXT:    callq __extendhfsf2@PLT
+; CHECK-AVX2-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; CHECK-AVX2-NEXT:    vdivss %xmm0, %xmm1, %xmm0
+; CHECK-AVX2-NEXT:    callq __truncsfhf2@PLT
+; CHECK-AVX2-NEXT:    vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; CHECK-AVX2-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; CHECK-AVX2-NEXT:    vpunpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; CHECK-AVX2-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; CHECK-AVX2-NEXT:    vpunpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; CHECK-AVX2-NEXT:    # xmm0 = xmm0[0],mem[0]
+; CHECK-AVX2-NEXT:    addq $120, %rsp
+; CHECK-AVX2-NEXT:    .cfi_def_cfa_offset 8
+; CHECK-AVX2-NEXT:    retq
+;
+; CHECK-NO-FASTFMA-LABEL: fdiv_pow2_8xhalf:
+; CHECK-NO-FASTFMA:       # %bb.0:
+; CHECK-NO-FASTFMA-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; CHECK-NO-FASTFMA-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [1,1,1,1,1,1,1,1]
+; CHECK-NO-FASTFMA-NEXT:    vpsllvd %ymm0, %ymm1, %ymm1
+; CHECK-NO-FASTFMA-NEXT:    vpmovdw %zmm1, %ymm0
+; CHECK-NO-FASTFMA-NEXT:    vpextrw $7, %xmm0, %eax
+; CHECK-NO-FASTFMA-NEXT:    vcvtsi2ss %eax, %xmm2, %xmm2
+; CHECK-NO-FASTFMA-NEXT:    vcvtps2ph $4, %xmm2, %xmm2
+; CHECK-NO-FASTFMA-NEXT:    vmovd %xmm2, %eax
+; CHECK-NO-FASTFMA-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm2
+; CHECK-NO-FASTFMA-NEXT:    vpextrw $6, %xmm0, %eax
+; CHECK-NO-FASTFMA-NEXT:    vcvtsi2ss %eax, %xmm3, %xmm3
+; CHECK-NO-FASTFMA-NEXT:    vcvtps2ph $4, %xmm3, %xmm3
+; CHECK-NO-FASTFMA-NEXT:    vmovd %xmm3, %eax
+; CHECK-NO-FASTFMA-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm3
+; CHECK-NO-FASTFMA-NEXT:    vpunpcklwd {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
+; CHECK-NO-FASTFMA-NEXT:    vpextrw $5, %xmm0, %eax
+; CHECK-NO-FASTFMA-NEXT:    vcvtsi2ss %eax, %xmm4, %xmm3
+; CHECK-NO-FASTFMA-NEXT:    vcvtps2ph $4, %xmm3, %xmm3
+; CHECK-NO-FASTFMA-NEXT:    vmovd %xmm3, %eax
+; CHECK-NO-FASTFMA-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm3
+; CHECK-NO-FASTFMA-NEXT:    vpextrw $4, %xmm0, %eax
+; CHECK-NO-FASTFMA-NEXT:    vcvtsi2ss %eax, %xmm4, %xmm4
+; CHECK-NO-FASTFMA-NEXT:    vcvtps2ph $4, %xmm4, %xmm4
+; CHECK-NO-FASTFMA-NEXT:    vmovd %xmm4, %eax
+; CHECK-NO-FASTFMA-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm4
+; CHECK-NO-FASTFMA-NEXT:    vpunpcklwd {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
+; CHECK-NO-FASTFMA-NEXT:    vpunpckldq {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
+; CHECK-NO-FASTFMA-NEXT:    vpextrw $3, %xmm0, %eax
+; CHECK-NO-FASTFMA-NEXT:    vcvtsi2ss %eax, %xmm5, %xmm3
+; CHECK-NO-FASTFMA-NEXT:    vcvtps2ph $4, %xmm3, %xmm3
+; CHECK-NO-FASTFMA-NEXT:    vmovd %xmm3, %eax
+; CHECK-NO-FASTFMA-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm3
+; CHECK-NO-FASTFMA-NEXT:    vpextrw $2, %xmm0, %eax
+; CHECK-NO-FASTFMA-NEXT:    vcvtsi2ss %eax, %xmm5, %xmm4
+; CHECK-NO-FASTFMA-NEXT:    vcvtps2ph $4, %xmm4, %xmm4
+; CHECK-NO-FASTFMA-NEXT:    vmovd %xmm4, %eax
+; CHECK-NO-FASTFMA-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm4
+; CHECK-NO-FASTFMA-NEXT:    vpunpcklwd {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
+; CHECK-NO-FASTFMA-NEXT:    vpextrw $0, %xmm1, %eax
+; CHECK-NO-FASTFMA-NEXT:    vcvtsi2ss %eax, %xmm5, %xmm1
+; CHECK-NO-FASTFMA-NEXT:    vcvtps2ph $4, %xmm1, %xmm1
+; CHECK-NO-FASTFMA-NEXT:    vmovd %xmm1, %eax
+; CHECK-NO-FASTFMA-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm1
+; CHECK-NO-FASTFMA-NEXT:    vpextrw $1, %xmm0, %eax
+; CHECK-NO-FASTFMA-NEXT:    vcvtsi2ss %eax, %xmm5, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vcvtps2ph $4, %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vmovd %xmm0, %eax
+; CHECK-NO-FASTFMA-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; CHECK-NO-FASTFMA-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
+; CHECK-NO-FASTFMA-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm2[0]
+; CHECK-NO-FASTFMA-NEXT:    vcvtph2ps %xmm0, %ymm0
+; CHECK-NO-FASTFMA-NEXT:    vbroadcastss {{.*#+}} ymm1 = [8.192E+3,8.192E+3,8.192E+3,8.192E+3,8.192E+3,8.192E+3,8.192E+3,8.192E+3]
+; CHECK-NO-FASTFMA-NEXT:    vdivps %ymm0, %ymm1, %ymm0
+; CHECK-NO-FASTFMA-NEXT:    vcvtps2ph $4, %ymm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vzeroupper
+; CHECK-NO-FASTFMA-NEXT:    retq
+;
+; CHECK-FMA-LABEL: fdiv_pow2_8xhalf:
+; CHECK-FMA:       # %bb.0:
+; CHECK-FMA-NEXT:    vpbroadcastw {{.*#+}} xmm1 = [1,1,1,1,1,1,1,1]
+; CHECK-FMA-NEXT:    vpsllvw %xmm0, %xmm1, %xmm0
+; CHECK-FMA-NEXT:    vpextrw $7, %xmm0, %eax
+; CHECK-FMA-NEXT:    vcvtsi2ss %eax, %xmm2, %xmm1
+; CHECK-FMA-NEXT:    vcvtps2ph $4, %xmm1, %xmm1
+; CHECK-FMA-NEXT:    vmovd %xmm1, %eax
+; CHECK-FMA-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm1
+; CHECK-FMA-NEXT:    vpextrw $6, %xmm0, %eax
+; CHECK-FMA-NEXT:    vcvtsi2ss %eax, %xmm2, %xmm2
+; CHECK-FMA-NEXT:    vcvtps2ph $4, %xmm2, %xmm2
+; CHECK-FMA-NEXT:    vmovd %xmm2, %eax
+; CHECK-FMA-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm2
+; CHECK-FMA-NEXT:    vpextrw $5, %xmm0, %eax
+; CHECK-FMA-NEXT:    vcvtsi2ss %eax, %xmm3, %xmm3
+; CHECK-FMA-NEXT:    vcvtps2ph $4, %xmm3, %xmm3
+; CHECK-FMA-NEXT:    vmovd %xmm3, %eax
+; CHECK-FMA-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm3
+; CHECK-FMA-NEXT:    vpextrw $4, %xmm0, %eax
+; CHECK-FMA-NEXT:    vcvtsi2ss %eax, %xmm4, %xmm4
+; CHECK-FMA-NEXT:    vcvtps2ph $4, %xmm4, %xmm4
+; CHECK-FMA-NEXT:    vmovd %xmm4, %eax
+; CHECK-FMA-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm4
+; CHECK-FMA-NEXT:    vpextrw $3, %xmm0, %eax
+; CHECK-FMA-NEXT:    vcvtsi2ss %eax, %xmm5, %xmm5
+; CHECK-FMA-NEXT:    vcvtps2ph $4, %xmm5, %xmm5
+; CHECK-FMA-NEXT:    vmovd %xmm5, %eax
+; CHECK-FMA-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm5
+; CHECK-FMA-NEXT:    vpextrw $2, %xmm0, %eax
+; CHECK-FMA-NEXT:    vcvtsi2ss %eax, %xmm6, %xmm6
+; CHECK-FMA-NEXT:    vcvtps2ph $4, %xmm6, %xmm6
+; CHECK-FMA-NEXT:    vpunpcklwd {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
+; CHECK-FMA-NEXT:    vmovd %xmm6, %eax
+; CHECK-FMA-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm2
+; CHECK-FMA-NEXT:    vpextrw $1, %xmm0, %eax
+; CHECK-FMA-NEXT:    vpunpcklwd {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
+; CHECK-FMA-NEXT:    vcvtsi2ss %eax, %xmm7, %xmm4
+; CHECK-FMA-NEXT:    vcvtps2ph $4, %xmm4, %xmm4
+; CHECK-FMA-NEXT:    vmovd %xmm4, %eax
+; CHECK-FMA-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm4
+; CHECK-FMA-NEXT:    vpextrw $0, %xmm0, %eax
+; CHECK-FMA-NEXT:    vcvtsi2ss %eax, %xmm7, %xmm0
+; CHECK-FMA-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
+; CHECK-FMA-NEXT:    vcvtps2ph $4, %xmm0, %xmm0
+; CHECK-FMA-NEXT:    vmovd %xmm0, %eax
+; CHECK-FMA-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0
+; CHECK-FMA-NEXT:    vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm5[0],xmm2[1],xmm5[1],xmm2[2],xmm5[2],xmm2[3],xmm5[3]
+; CHECK-FMA-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1],xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; CHECK-FMA-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; CHECK-FMA-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; CHECK-FMA-NEXT:    vcvtph2ps %xmm0, %ymm0
+; CHECK-FMA-NEXT:    vbroadcastss {{.*#+}} ymm1 = [8.192E+3,8.192E+3,8.192E+3,8.192E+3,8.192E+3,8.192E+3,8.192E+3,8.192E+3]
+; CHECK-FMA-NEXT:    vdivps %ymm0, %ymm1, %ymm0
+; CHECK-FMA-NEXT:    vcvtps2ph $4, %ymm0, %xmm0
+; CHECK-FMA-NEXT:    vzeroupper
+; CHECK-FMA-NEXT:    retq
+  %p2 = shl <8 x i16> <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>, %i
+  %p2_f = uitofp <8 x i16> %p2 to <8 x half>
+  %r = fdiv <8 x half> <half 0xH7000, half 0xH7000, half 0xH7000, half 0xH7000, half 0xH7000, half 0xH7000, half 0xH7000, half 0xH7000>, %p2_f
+  ret <8 x half> %r
+}
+
+define double @fmul_pow_shl_cnt(i64 %cnt) nounwind {
+; CHECK-SSE-LABEL: fmul_pow_shl_cnt:
+; CHECK-SSE:       # %bb.0:
+; CHECK-SSE-NEXT:    movq %rdi, %rcx
+; CHECK-SSE-NEXT:    movl $1, %eax
+; CHECK-SSE-NEXT:    # kill: def $cl killed $cl killed $rcx
+; CHECK-SSE-NEXT:    shlq %cl, %rax
+; CHECK-SSE-NEXT:    movq %rax, %xmm1
+; CHECK-SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],mem[0],xmm1[1],mem[1]
+; CHECK-SSE-NEXT:    subpd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1
+; CHECK-SSE-NEXT:    movapd %xmm1, %xmm0
+; CHECK-SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
+; CHECK-SSE-NEXT:    addsd %xmm1, %xmm0
+; CHECK-SSE-NEXT:    mulsd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; CHECK-SSE-NEXT:    retq
+;
+; CHECK-AVX2-LABEL: fmul_pow_shl_cnt:
+; CHECK-AVX2:       # %bb.0:
+; CHECK-AVX2-NEXT:    movq %rdi, %rcx
+; CHECK-AVX2-NEXT:    movl $1, %eax
+; CHECK-AVX2-NEXT:    # kill: def $cl killed $cl killed $rcx
+; CHECK-AVX2-NEXT:    shlq %cl, %rax
+; CHECK-AVX2-NEXT:    vmovq %rax, %xmm0
+; CHECK-AVX2-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; CHECK-AVX2-NEXT:    vsubpd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; CHECK-AVX2-NEXT:    vshufpd {{.*#+}} xmm1 = xmm0[1,0]
+; CHECK-AVX2-NEXT:    vaddsd %xmm0, %xmm1, %xmm0
+; CHECK-AVX2-NEXT:    vmulsd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; CHECK-AVX2-NEXT:    retq
+;
+; CHECK-NO-FASTFMA-LABEL: fmul_pow_shl_cnt:
+; CHECK-NO-FASTFMA:       # %bb.0:
+; CHECK-NO-FASTFMA-NEXT:    movq %rdi, %rcx
+; CHECK-NO-FASTFMA-NEXT:    movl $1, %eax
+; CHECK-NO-FASTFMA-NEXT:    # kill: def $cl killed $cl killed $rcx
+; CHECK-NO-FASTFMA-NEXT:    shlq %cl, %rax
+; CHECK-NO-FASTFMA-NEXT:    vcvtusi2sd %rax, %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vmulsd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    retq
+;
+; CHECK-FMA-LABEL: fmul_pow_shl_cnt:
+; CHECK-FMA:       # %bb.0:
+; CHECK-FMA-NEXT:    movl $1, %eax
+; CHECK-FMA-NEXT:    shlxq %rdi, %rax, %rax
+; CHECK-FMA-NEXT:    vcvtusi2sd %rax, %xmm0, %xmm0
+; CHECK-FMA-NEXT:    vmulsd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; CHECK-FMA-NEXT:    retq
+  %shl = shl nuw i64 1, %cnt
+  %conv = uitofp i64 %shl to double
+  %mul = fmul double 9.000000e+00, %conv
+  ret double %mul
+}
+
+define double @fmul_pow_shl_cnt2(i64 %cnt) nounwind {
+; CHECK-SSE-LABEL: fmul_pow_shl_cnt2:
+; CHECK-SSE:       # %bb.0:
+; CHECK-SSE-NEXT:    movq %rdi, %rcx
+; CHECK-SSE-NEXT:    movl $2, %eax
+; CHECK-SSE-NEXT:    # kill: def $cl killed $cl killed $rcx
+; CHECK-SSE-NEXT:    shlq %cl, %rax
+; CHECK-SSE-NEXT:    movq %rax, %xmm1
+; CHECK-SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],mem[0],xmm1[1],mem[1]
+; CHECK-SSE-NEXT:    subpd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1
+; CHECK-SSE-NEXT:    movapd %xmm1, %xmm0
+; CHECK-SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
+; CHECK-SSE-NEXT:    addsd %xmm1, %xmm0
+; CHECK-SSE-NEXT:    mulsd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; CHECK-SSE-NEXT:    retq
+;
+; CHECK-AVX2-LABEL: fmul_pow_shl_cnt2:
+; CHECK-AVX2:       # %bb.0:
+; CHECK-AVX2-NEXT:    movq %rdi, %rcx
+; CHECK-AVX2-NEXT:    movl $2, %eax
+; CHECK-AVX2-NEXT:    # kill: def $cl killed $cl killed $rcx
+; CHECK-AVX2-NEXT:    shlq %cl, %rax
+; CHECK-AVX2-NEXT:    vmovq %rax, %xmm0
+; CHECK-AVX2-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; CHECK-AVX2-NEXT:    vsubpd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; CHECK-AVX2-NEXT:    vshufpd {{.*#+}} xmm1 = xmm0[1,0]
+; CHECK-AVX2-NEXT:    vaddsd %xmm0, %xmm1, %xmm0
+; CHECK-AVX2-NEXT:    vmulsd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; CHECK-AVX2-NEXT:    retq
+;
+; CHECK-NO-FASTFMA-LABEL: fmul_pow_shl_cnt2:
+; CHECK-NO-FASTFMA:       # %bb.0:
+; CHECK-NO-FASTFMA-NEXT:    movq %rdi, %rcx
+; CHECK-NO-FASTFMA-NEXT:    movl $2, %eax
+; CHECK-NO-FASTFMA-NEXT:    # kill: def $cl killed $cl killed $rcx
+; CHECK-NO-FASTFMA-NEXT:    shlq %cl, %rax
+; CHECK-NO-FASTFMA-NEXT:    vcvtusi2sd %rax, %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vmulsd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    retq
+;
+; CHECK-FMA-LABEL: fmul_pow_shl_cnt2:
+; CHECK-FMA:       # %bb.0:
+; CHECK-FMA-NEXT:    movl $2, %eax
+; CHECK-FMA-NEXT:    shlxq %rdi, %rax, %rax
+; CHECK-FMA-NEXT:    vcvtusi2sd %rax, %xmm0, %xmm0
+; CHECK-FMA-NEXT:    vmulsd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; CHECK-FMA-NEXT:    retq
+  %shl = shl nuw i64 2, %cnt
+  %conv = uitofp i64 %shl to double
+  %mul = fmul double -9.000000e+00, %conv
+  ret double %mul
+}
+
+define float @fmul_pow_select(i32 %cnt, i1 %c) nounwind {
+; CHECK-SSE-LABEL: fmul_pow_select:
+; CHECK-SSE:       # %bb.0:
+; CHECK-SSE-NEXT:    movl %edi, %ecx
+; CHECK-SSE-NEXT:    andl $1, %esi
+; CHECK-SSE-NEXT:    movl $2, %eax
+; CHECK-SSE-NEXT:    subl %esi, %eax
+; CHECK-SSE-NEXT:    # kill: def $cl killed $cl killed $ecx
+; CHECK-SSE-NEXT:    shll %cl, %eax
+; CHECK-SSE-NEXT:    cvtsi2ss %rax, %xmm0
+; CHECK-SSE-NEXT:    mulss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; CHECK-SSE-NEXT:    retq
+;
+; CHECK-AVX2-LABEL: fmul_pow_select:
+; CHECK-AVX2:       # %bb.0:
+; CHECK-AVX2-NEXT:    movl %edi, %ecx
+; CHECK-AVX2-NEXT:    andl $1, %esi
+; CHECK-AVX2-NEXT:    movl $2, %eax
+; CHECK-AVX2-NEXT:    subl %esi, %eax
+; CHECK-AVX2-NEXT:    # kill: def $cl killed $cl killed $ecx
+; CHECK-AVX2-NEXT:    shll %cl, %eax
+; CHECK-AVX2-NEXT:    vcvtsi2ss %rax, %xmm0, %xmm0
+; CHECK-AVX2-NEXT:    vmulss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; CHECK-AVX2-NEXT:    retq
+;
+; CHECK-NO-FASTFMA-LABEL: fmul_pow_select:
+; CHECK-NO-FASTFMA:       # %bb.0:
+; CHECK-NO-FASTFMA-NEXT:    movl %edi, %ecx
+; CHECK-NO-FASTFMA-NEXT:    andl $1, %esi
+; CHECK-NO-FASTFMA-NEXT:    movl $2, %eax
+; CHECK-NO-FASTFMA-NEXT:    subl %esi, %eax
+; CHECK-NO-FASTFMA-NEXT:    # kill: def $cl killed $cl killed $ecx
+; CHECK-NO-FASTFMA-NEXT:    shll %cl, %eax
+; CHECK-NO-FASTFMA-NEXT:    vcvtusi2ss %eax, %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vmulss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    retq
+;
+; CHECK-FMA-LABEL: fmul_pow_select:
+; CHECK-FMA:       # %bb.0:
+; CHECK-FMA-NEXT:    andl $1, %esi
+; CHECK-FMA-NEXT:    movl $2, %eax
+; CHECK-FMA-NEXT:    subl %esi, %eax
+; CHECK-FMA-NEXT:    shlxl %edi, %eax, %eax
+; CHECK-FMA-NEXT:    vcvtusi2ss %eax, %xmm0, %xmm0
+; CHECK-FMA-NEXT:    vmulss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; CHECK-FMA-NEXT:    retq
+  %shl2 = shl nuw i32 2, %cnt
+  %shl1 = shl nuw i32 1, %cnt
+  %shl = select i1 %c, i32 %shl1, i32 %shl2
+  %conv = uitofp i32 %shl to float
+  %mul = fmul float 9.000000e+00, %conv
+  ret float %mul
+}
+
+define float @fmul_fly_pow_mul_min_pow2(i64 %cnt) nounwind {
+; CHECK-SSE-LABEL: fmul_fly_pow_mul_min_pow2:
+; CHECK-SSE:       # %bb.0:
+; CHECK-SSE-NEXT:    movq %rdi, %rcx
+; CHECK-SSE-NEXT:    movl $8, %eax
+; CHECK-SSE-NEXT:    # kill: def $cl killed $cl killed $rcx
+; CHECK-SSE-NEXT:    shlq %cl, %rax
+; CHECK-SSE-NEXT:    cmpq $8192, %rax # imm = 0x2000
+; CHECK-SSE-NEXT:    movl $8192, %ecx # imm = 0x2000
+; CHECK-SSE-NEXT:    cmovbq %rax, %rcx
+; CHECK-SSE-NEXT:    cvtsi2ss %rcx, %xmm0
+; CHECK-SSE-NEXT:    mulss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; CHECK-SSE-NEXT:    retq
+;
+; CHECK-AVX2-LABEL: fmul_fly_pow_mul_min_pow2:
+; CHECK-AVX2:       # %bb.0:
+; CHECK-AVX2-NEXT:    movq %rdi, %rcx
+; CHECK-AVX2-NEXT:    movl $8, %eax
+; CHECK-AVX2-NEXT:    # kill: def $cl killed $cl killed $rcx
+; CHECK-AVX2-NEXT:    shlq %cl, %rax
+; CHECK-AVX2-NEXT:    cmpq $8192, %rax # imm = 0x2000
+; CHECK-AVX2-NEXT:    movl $8192, %ecx # imm = 0x2000
+; CHECK-AVX2-NEXT:    cmovbq %rax, %rcx
+; CHECK-AVX2-NEXT:    vcvtsi2ss %rcx, %xmm0, %xmm0
+; CHECK-AVX2-NEXT:    vmulss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; CHECK-AVX2-NEXT:    retq
+;
+; CHECK-NO-FASTFMA-LABEL: fmul_fly_pow_mul_min_pow2:
+; CHECK-NO-FASTFMA:       # %bb.0:
+; CHECK-NO-FASTFMA-NEXT:    movq %rdi, %rcx
+; CHECK-NO-FASTFMA-NEXT:    movl $8, %eax
+; CHECK-NO-FASTFMA-NEXT:    # kill: def $cl killed $cl killed $rcx
+; CHECK-NO-FASTFMA-NEXT:    shlq %cl, %rax
+; CHECK-NO-FASTFMA-NEXT:    cmpq $8192, %rax # imm = 0x2000
+; CHECK-NO-FASTFMA-NEXT:    movl $8192, %ecx # imm = 0x2000
+; CHECK-NO-FASTFMA-NEXT:    cmovbq %rax, %rcx
+; CHECK-NO-FASTFMA-NEXT:    vcvtsi2ss %rcx, %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vmulss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    retq
+;
+; CHECK-FMA-LABEL: fmul_fly_pow_mul_min_pow2:
+; CHECK-FMA:       # %bb.0:
+; CHECK-FMA-NEXT:    movl $8, %eax
+; CHECK-FMA-NEXT:    shlxq %rdi, %rax, %rax
+; CHECK-FMA-NEXT:    cmpq $8192, %rax # imm = 0x2000
+; CHECK-FMA-NEXT:    movl $8192, %ecx # imm = 0x2000
+; CHECK-FMA-NEXT:    cmovbq %rax, %rcx
+; CHECK-FMA-NEXT:    vcvtsi2ss %rcx, %xmm0, %xmm0
+; CHECK-FMA-NEXT:    vmulss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; CHECK-FMA-NEXT:    retq
+  %shl8 = shl nuw i64 8, %cnt
+  %shl = call i64 @llvm.umin.i64(i64 %shl8, i64 8192)
+  %conv = uitofp i64 %shl to float
+  %mul = fmul float 9.000000e+00, %conv
+  ret float %mul
+}
+
+define double @fmul_pow_mul_max_pow2(i16 %cnt) nounwind {
+; CHECK-SSE-LABEL: fmul_pow_mul_max_pow2:
+; CHECK-SSE:       # %bb.0:
+; CHECK-SSE-NEXT:    movl %edi, %ecx
+; CHECK-SSE-NEXT:    movl $2, %eax
+; CHECK-SSE-NEXT:    shll %cl, %eax
+; CHECK-SSE-NEXT:    movl $1, %edx
+; CHECK-SSE-NEXT:    # kill: def $cl killed $cl killed $ecx
+; CHECK-SSE-NEXT:    shll %cl, %edx
+; CHECK-SSE-NEXT:    cmpw %ax, %dx
+; CHECK-SSE-NEXT:    cmovbel %eax, %edx
+; CHECK-SSE-NEXT:    movzwl %dx, %eax
+; CHECK-SSE-NEXT:    cvtsi2sd %eax, %xmm0
+; CHECK-SSE-NEXT:    mulsd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; CHECK-SSE-NEXT:    retq
+;
+; CHECK-AVX2-LABEL: fmul_pow_mul_max_pow2:
+; CHECK-AVX2:       # %bb.0:
+; CHECK-AVX2-NEXT:    movl %edi, %ecx
+; CHECK-AVX2-NEXT:    movl $2, %eax
+; CHECK-AVX2-NEXT:    shll %cl, %eax
+; CHECK-AVX2-NEXT:    movl $1, %edx
+; CHECK-AVX2-NEXT:    # kill: def $cl killed $cl killed $ecx
+; CHECK-AVX2-NEXT:    shll %cl, %edx
+; CHECK-AVX2-NEXT:    cmpw %ax, %dx
+; CHECK-AVX2-NEXT:    cmovbel %eax, %edx
+; CHECK-AVX2-NEXT:    movzwl %dx, %eax
+; CHECK-AVX2-NEXT:    vcvtsi2sd %eax, %xmm0, %xmm0
+; CHECK-AVX2-NEXT:    vmulsd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; CHECK-AVX2-NEXT:    retq
+;
+; CHECK-NO-FASTFMA-LABEL: fmul_pow_mul_max_pow2:
+; CHECK-NO-FASTFMA:       # %bb.0:
+; CHECK-NO-FASTFMA-NEXT:    movl %edi, %ecx
+; CHECK-NO-FASTFMA-NEXT:    movl $2, %eax
+; CHECK-NO-FASTFMA-NEXT:    shll %cl, %eax
+; CHECK-NO-FASTFMA-NEXT:    movl $1, %edx
+; CHECK-NO-FASTFMA-NEXT:    # kill: def $cl killed $cl killed $ecx
+; CHECK-NO-FASTFMA-NEXT:    shll %cl, %edx
+; CHECK-NO-FASTFMA-NEXT:    cmpw %ax, %dx
+; CHECK-NO-FASTFMA-NEXT:    cmovbel %eax, %edx
+; CHECK-NO-FASTFMA-NEXT:    movzwl %dx, %eax
+; CHECK-NO-FASTFMA-NEXT:    vcvtsi2sd %eax, %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vmulsd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    retq
+;
+; CHECK-FMA-LABEL: fmul_pow_mul_max_pow2:
+; CHECK-FMA:       # %bb.0:
+; CHECK-FMA-NEXT:    movl $2, %eax
+; CHECK-FMA-NEXT:    shlxl %edi, %eax, %eax
+; CHECK-FMA-NEXT:    movl $1, %ecx
+; CHECK-FMA-NEXT:    shlxl %edi, %ecx, %ecx
+; CHECK-FMA-NEXT:    cmpw %ax, %cx
+; CHECK-FMA-NEXT:    cmoval %ecx, %eax
+; CHECK-FMA-NEXT:    movzwl %ax, %eax
+; CHECK-FMA-NEXT:    vcvtsi2sd %eax, %xmm0, %xmm0
+; CHECK-FMA-NEXT:    vmulsd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; CHECK-FMA-NEXT:    retq
+  %shl2 = shl nuw i16 2, %cnt
+  %shl1 = shl nuw i16 1, %cnt
+  %shl = call i16 @llvm.umax.i16(i16 %shl1, i16 %shl2)
+  %conv = uitofp i16 %shl to double
+  %mul = fmul double 3.000000e+00, %conv
+  ret double %mul
+}
+
+define double @fmul_pow_shl_cnt_fail_maybe_non_pow2(i64 %v, i64 %cnt) nounwind {
+; CHECK-SSE-LABEL: fmul_pow_shl_cnt_fail_maybe_non_pow2:
+; CHECK-SSE:       # %bb.0:
+; CHECK-SSE-NEXT:    movq %rsi, %rcx
+; CHECK-SSE-NEXT:    # kill: def $cl killed $cl killed $rcx
+; CHECK-SSE-NEXT:    shlq %cl, %rdi
+; CHECK-SSE-NEXT:    movq %rdi, %xmm1
+; CHECK-SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],mem[0],xmm1[1],mem[1]
+; CHECK-SSE-NEXT:    subpd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1
+; CHECK-SSE-NEXT:    movapd %xmm1, %xmm0
+; CHECK-SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
+; CHECK-SSE-NEXT:    addsd %xmm1, %xmm0
+; CHECK-SSE-NEXT:    mulsd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; CHECK-SSE-NEXT:    retq
+;
+; CHECK-AVX2-LABEL: fmul_pow_shl_cnt_fail_maybe_non_pow2:
+; CHECK-AVX2:       # %bb.0:
+; CHECK-AVX2-NEXT:    movq %rsi, %rcx
+; CHECK-AVX2-NEXT:    # kill: def $cl killed $cl killed $rcx
+; CHECK-AVX2-NEXT:    shlq %cl, %rdi
+; CHECK-AVX2-NEXT:    vmovq %rdi, %xmm0
+; CHECK-AVX2-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; CHECK-AVX2-NEXT:    vsubpd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; CHECK-AVX2-NEXT:    vshufpd {{.*#+}} xmm1 = xmm0[1,0]
+; CHECK-AVX2-NEXT:    vaddsd %xmm0, %xmm1, %xmm0
+; CHECK-AVX2-NEXT:    vmulsd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; CHECK-AVX2-NEXT:    retq
+;
+; CHECK-NO-FASTFMA-LABEL: fmul_pow_shl_cnt_fail_maybe_non_pow2:
+; CHECK-NO-FASTFMA:       # %bb.0:
+; CHECK-NO-FASTFMA-NEXT:    movq %rsi, %rcx
+; CHECK-NO-FASTFMA-NEXT:    # kill: def $cl killed $cl killed $rcx
+; CHECK-NO-FASTFMA-NEXT:    shlq %cl, %rdi
+; CHECK-NO-FASTFMA-NEXT:    vcvtusi2sd %rdi, %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vmulsd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    retq
+;
+; CHECK-FMA-LABEL: fmul_pow_shl_cnt_fail_maybe_non_pow2:
+; CHECK-FMA:       # %bb.0:
+; CHECK-FMA-NEXT:    shlxq %rsi, %rdi, %rax
+; CHECK-FMA-NEXT:    vcvtusi2sd %rax, %xmm0, %xmm0
+; CHECK-FMA-NEXT:    vmulsd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; CHECK-FMA-NEXT:    retq
+  %shl = shl nuw i64 %v, %cnt
+  %conv = uitofp i64 %shl to double
+  %mul = fmul double 9.000000e+00, %conv
+  ret double %mul
+}
+
+define <2 x float> @fmul_pow_shl_cnt_vec_fail_expensive_cast(<2 x i64> %cnt) nounwind {
+; CHECK-SSE-LABEL: fmul_pow_shl_cnt_vec_fail_expensive_cast:
+; CHECK-SSE:       # %bb.0:
+; CHECK-SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[2,3,2,3]
+; CHECK-SSE-NEXT:    movdqa {{.*#+}} xmm3 = [2,2]
+; CHECK-SSE-NEXT:    movdqa %xmm3, %xmm1
+; CHECK-SSE-NEXT:    psllq %xmm2, %xmm1
+; CHECK-SSE-NEXT:    psllq %xmm0, %xmm3
+; CHECK-SSE-NEXT:    movq %xmm3, %rax
+; CHECK-SSE-NEXT:    testq %rax, %rax
+; CHECK-SSE-NEXT:    js .LBB12_1
+; CHECK-SSE-NEXT:  # %bb.2:
+; CHECK-SSE-NEXT:    xorps %xmm0, %xmm0
+; CHECK-SSE-NEXT:    cvtsi2ss %rax, %xmm0
+; CHECK-SSE-NEXT:    jmp .LBB12_3
+; CHECK-SSE-NEXT:  .LBB12_1:
+; CHECK-SSE-NEXT:    movq %rax, %rcx
+; CHECK-SSE-NEXT:    shrq %rcx
+; CHECK-SSE-NEXT:    andl $1, %eax
+; CHECK-SSE-NEXT:    orq %rcx, %rax
+; CHECK-SSE-NEXT:    xorps %xmm0, %xmm0
+; CHECK-SSE-NEXT:    cvtsi2ss %rax, %xmm0
+; CHECK-SSE-NEXT:    addss %xmm0, %xmm0
+; CHECK-SSE-NEXT:  .LBB12_3:
+; CHECK-SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; CHECK-SSE-NEXT:    movq %xmm1, %rax
+; CHECK-SSE-NEXT:    testq %rax, %rax
+; CHECK-SSE-NEXT:    js .LBB12_4
+; CHECK-SSE-NEXT:  # %bb.5:
+; CHECK-SSE-NEXT:    xorps %xmm1, %xmm1
+; CHECK-SSE-NEXT:    cvtsi2ss %rax, %xmm1
+; CHECK-SSE-NEXT:    jmp .LBB12_6
+; CHECK-SSE-NEXT:  .LBB12_4:
+; CHECK-SSE-NEXT:    movq %rax, %rcx
+; CHECK-SSE-NEXT:    shrq %rcx
+; CHECK-SSE-NEXT:    andl $1, %eax
+; CHECK-SSE-NEXT:    orq %rcx, %rax
+; CHECK-SSE-NEXT:    xorps %xmm1, %xmm1
+; CHECK-SSE-NEXT:    cvtsi2ss %rax, %xmm1
+; CHECK-SSE-NEXT:    addss %xmm1, %xmm1
+; CHECK-SSE-NEXT:  .LBB12_6:
+; CHECK-SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; CHECK-SSE-NEXT:    mulps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; CHECK-SSE-NEXT:    retq
+;
+; CHECK-AVX2-LABEL: fmul_pow_shl_cnt_vec_fail_expensive_cast:
+; CHECK-AVX2:       # %bb.0:
+; CHECK-AVX2-NEXT:    vpbroadcastq {{.*#+}} xmm1 = [2,2]
+; CHECK-AVX2-NEXT:    vpsllvq %xmm0, %xmm1, %xmm0
+; CHECK-AVX2-NEXT:    vpsrlq $1, %xmm0, %xmm1
+; CHECK-AVX2-NEXT:    vblendvpd %xmm0, %xmm1, %xmm0, %xmm1
+; CHECK-AVX2-NEXT:    vpextrq $1, %xmm1, %rax
+; CHECK-AVX2-NEXT:    vcvtsi2ss %rax, %xmm2, %xmm2
+; CHECK-AVX2-NEXT:    vmovq %xmm1, %rax
+; CHECK-AVX2-NEXT:    vcvtsi2ss %rax, %xmm3, %xmm1
+; CHECK-AVX2-NEXT:    vinsertps {{.*#+}} xmm1 = xmm1[0],xmm2[0],zero,zero
+; CHECK-AVX2-NEXT:    vaddps %xmm1, %xmm1, %xmm2
+; CHECK-AVX2-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; CHECK-AVX2-NEXT:    vpcmpgtq %xmm0, %xmm3, %xmm0
+; CHECK-AVX2-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[1,3,2,3]
+; CHECK-AVX2-NEXT:    vblendvps %xmm0, %xmm2, %xmm1, %xmm0
+; CHECK-AVX2-NEXT:    vbroadcastss {{.*#+}} xmm1 = [1.5E+1,1.5E+1,1.5E+1,1.5E+1]
+; CHECK-AVX2-NEXT:    vmulps %xmm1, %xmm0, %xmm0
+; CHECK-AVX2-NEXT:    retq
+;
+; CHECK-NO-FASTFMA-LABEL: fmul_pow_shl_cnt_vec_fail_expensive_cast:
+; CHECK-NO-FASTFMA:       # %bb.0:
+; CHECK-NO-FASTFMA-NEXT:    vpbroadcastq {{.*#+}} xmm1 = [2,2]
+; CHECK-NO-FASTFMA-NEXT:    vpsllvq %xmm0, %xmm1, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vpextrq $1, %xmm0, %rax
+; CHECK-NO-FASTFMA-NEXT:    vcvtusi2ss %rax, %xmm2, %xmm1
+; CHECK-NO-FASTFMA-NEXT:    vmovq %xmm0, %rax
+; CHECK-NO-FASTFMA-NEXT:    vcvtusi2ss %rax, %xmm2, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vinsertps {{.*#+}} xmm0 = xmm0[0],xmm1[0],zero,zero
+; CHECK-NO-FASTFMA-NEXT:    vbroadcastss {{.*#+}} xmm1 = [1.5E+1,1.5E+1,1.5E+1,1.5E+1]
+; CHECK-NO-FASTFMA-NEXT:    vmulps %xmm1, %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    retq
+;
+; CHECK-FMA-LABEL: fmul_pow_shl_cnt_vec_fail_expensive_cast:
+; CHECK-FMA:       # %bb.0:
+; CHECK-FMA-NEXT:    vpbroadcastq {{.*#+}} xmm1 = [2,2]
+; CHECK-FMA-NEXT:    vpsllvq %xmm0, %xmm1, %xmm0
+; CHECK-FMA-NEXT:    vcvtuqq2ps %xmm0, %xmm0
+; CHECK-FMA-NEXT:    vmulps {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to4}, %xmm0, %xmm0
+; CHECK-FMA-NEXT:    retq
+  %shl = shl nsw nuw <2 x i64> <i64 2, i64 2>, %cnt
+  %conv = uitofp <2 x i64> %shl to <2 x float>
+  %mul = fmul <2 x float> <float 15.000000e+00, float 15.000000e+00>, %conv
+  ret <2 x float> %mul
+}
+
+define <2 x double> @fmul_pow_shl_cnt_vec(<2 x i64> %cnt) nounwind {
+; CHECK-SSE-LABEL: fmul_pow_shl_cnt_vec:
+; CHECK-SSE:       # %bb.0:
+; CHECK-SSE-NEXT:    movdqa {{.*#+}} xmm1 = [2,2]
+; CHECK-SSE-NEXT:    movdqa %xmm1, %xmm2
+; CHECK-SSE-NEXT:    psllq %xmm0, %xmm2
+; CHECK-SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; CHECK-SSE-NEXT:    psllq %xmm0, %xmm1
+; CHECK-SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; CHECK-SSE-NEXT:    movapd {{.*#+}} xmm0 = [4294967295,4294967295]
+; CHECK-SSE-NEXT:    andpd %xmm1, %xmm0
+; CHECK-SSE-NEXT:    orpd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; CHECK-SSE-NEXT:    psrlq $32, %xmm1
+; CHECK-SSE-NEXT:    por {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1
+; CHECK-SSE-NEXT:    subpd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1
+; CHECK-SSE-NEXT:    addpd %xmm0, %xmm1
+; CHECK-SSE-NEXT:    mulpd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1
+; CHECK-SSE-NEXT:    movapd %xmm1, %xmm0
+; CHECK-SSE-NEXT:    retq
+;
+; CHECK-AVX2-LABEL: fmul_pow_shl_cnt_vec:
+; CHECK-AVX2:       # %bb.0:
+; CHECK-AVX2-NEXT:    vpbroadcastq {{.*#+}} xmm1 = [2,2]
+; CHECK-AVX2-NEXT:    vpsllvq %xmm0, %xmm1, %xmm0
+; CHECK-AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; CHECK-AVX2-NEXT:    vpblendd {{.*#+}} xmm1 = xmm0[0],xmm1[1],xmm0[2],xmm1[3]
+; CHECK-AVX2-NEXT:    vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
+; CHECK-AVX2-NEXT:    vpsrlq $32, %xmm0, %xmm0
+; CHECK-AVX2-NEXT:    vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; CHECK-AVX2-NEXT:    vsubpd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; CHECK-AVX2-NEXT:    vaddpd %xmm0, %xmm1, %xmm0
+; CHECK-AVX2-NEXT:    vmulpd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; CHECK-AVX2-NEXT:    retq
+;
+; CHECK-NO-FASTFMA-LABEL: fmul_pow_shl_cnt_vec:
+; CHECK-NO-FASTFMA:       # %bb.0:
+; CHECK-NO-FASTFMA-NEXT:    vpbroadcastq {{.*#+}} xmm1 = [2,2]
+; CHECK-NO-FASTFMA-NEXT:    vpsllvq %xmm0, %xmm1, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; CHECK-NO-FASTFMA-NEXT:    vpblendd {{.*#+}} xmm1 = xmm0[0],xmm1[1],xmm0[2],xmm1[3]
+; CHECK-NO-FASTFMA-NEXT:    vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
+; CHECK-NO-FASTFMA-NEXT:    vpsrlq $32, %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vsubpd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vaddpd %xmm0, %xmm1, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vmulpd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    retq
+;
+; CHECK-FMA-LABEL: fmul_pow_shl_cnt_vec:
+; CHECK-FMA:       # %bb.0:
+; CHECK-FMA-NEXT:    vpbroadcastq {{.*#+}} xmm1 = [2,2]
+; CHECK-FMA-NEXT:    vpsllvq %xmm0, %xmm1, %xmm0
+; CHECK-FMA-NEXT:    vcvtuqq2pd %xmm0, %xmm0
+; CHECK-FMA-NEXT:    vmulpd {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to2}, %xmm0, %xmm0
+; CHECK-FMA-NEXT:    retq
+  %shl = shl nsw nuw <2 x i64> <i64 2, i64 2>, %cnt
+  %conv = uitofp <2 x i64> %shl to <2 x double>
+  %mul = fmul <2 x double> <double 15.000000e+00, double 15.000000e+00>, %conv
+  ret <2 x double> %mul
+}
+
+define <4 x float> @fmul_pow_shl_cnt_vec_preserve_fma(<4 x i32> %cnt, <4 x float> %add) nounwind {
+; CHECK-SSE-LABEL: fmul_pow_shl_cnt_vec_preserve_fma:
+; CHECK-SSE:       # %bb.0:
+; CHECK-SSE-NEXT:    pslld $23, %xmm0
+; CHECK-SSE-NEXT:    paddd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; CHECK-SSE-NEXT:    cvttps2dq %xmm0, %xmm0
+; CHECK-SSE-NEXT:    movdqa {{.*#+}} xmm2 = [2,2,2,2]
+; CHECK-SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[1,1,3,3]
+; CHECK-SSE-NEXT:    pmuludq %xmm2, %xmm0
+; CHECK-SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; CHECK-SSE-NEXT:    pmuludq %xmm2, %xmm3
+; CHECK-SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,2,2,3]
+; CHECK-SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; CHECK-SSE-NEXT:    movdqa {{.*#+}} xmm2 = [65535,65535,65535,65535]
+; CHECK-SSE-NEXT:    pand %xmm0, %xmm2
+; CHECK-SSE-NEXT:    por {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2
+; CHECK-SSE-NEXT:    psrld $16, %xmm0
+; CHECK-SSE-NEXT:    por {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; CHECK-SSE-NEXT:    subps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; CHECK-SSE-NEXT:    addps %xmm2, %xmm0
+; CHECK-SSE-NEXT:    mulps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; CHECK-SSE-NEXT:    addps %xmm1, %xmm0
+; CHECK-SSE-NEXT:    retq
+;
+; CHECK-AVX2-LABEL: fmul_pow_shl_cnt_vec_preserve_fma:
+; CHECK-AVX2:       # %bb.0:
+; CHECK-AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm2 = [2,2,2,2]
+; CHECK-AVX2-NEXT:    vpsllvd %xmm0, %xmm2, %xmm0
+; CHECK-AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm2 = [1258291200,1258291200,1258291200,1258291200]
+; CHECK-AVX2-NEXT:    vpblendw {{.*#+}} xmm2 = xmm0[0],xmm2[1],xmm0[2],xmm2[3],xmm0[4],xmm2[5],xmm0[6],xmm2[7]
+; CHECK-AVX2-NEXT:    vpsrld $16, %xmm0, %xmm0
+; CHECK-AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm3 = [1392508928,1392508928,1392508928,1392508928]
+; CHECK-AVX2-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0],xmm3[1],xmm0[2],xmm3[3],xmm0[4],xmm3[5],xmm0[6],xmm3[7]
+; CHECK-AVX2-NEXT:    vbroadcastss {{.*#+}} xmm3 = [5.49764202E+11,5.49764202E+11,5.49764202E+11,5.49764202E+11]
+; CHECK-AVX2-NEXT:    vsubps %xmm3, %xmm0, %xmm0
+; CHECK-AVX2-NEXT:    vaddps %xmm0, %xmm2, %xmm0
+; CHECK-AVX2-NEXT:    vbroadcastss {{.*#+}} xmm2 = [5.0E+0,5.0E+0,5.0E+0,5.0E+0]
+; CHECK-AVX2-NEXT:    vmulps %xmm2, %xmm0, %xmm0
+; CHECK-AVX2-NEXT:    vaddps %xmm1, %xmm0, %xmm0
+; CHECK-AVX2-NEXT:    retq
+;
+; CHECK-NO-FASTFMA-LABEL: fmul_pow_shl_cnt_vec_preserve_fma:
+; CHECK-NO-FASTFMA:       # %bb.0:
+; CHECK-NO-FASTFMA-NEXT:    vpbroadcastd {{.*#+}} xmm2 = [2,2,2,2]
+; CHECK-NO-FASTFMA-NEXT:    vpsllvd %xmm0, %xmm2, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vcvtudq2ps %zmm0, %zmm0
+; CHECK-NO-FASTFMA-NEXT:    vbroadcastss {{.*#+}} xmm2 = [5.0E+0,5.0E+0,5.0E+0,5.0E+0]
+; CHECK-NO-FASTFMA-NEXT:    vmulps %xmm2, %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vaddps %xmm1, %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vzeroupper
+; CHECK-NO-FASTFMA-NEXT:    retq
+;
+; CHECK-FMA-LABEL: fmul_pow_shl_cnt_vec_preserve_fma:
+; CHECK-FMA:       # %bb.0:
+; CHECK-FMA-NEXT:    vpbroadcastd {{.*#+}} xmm2 = [2,2,2,2]
+; CHECK-FMA-NEXT:    vpsllvd %xmm0, %xmm2, %xmm0
+; CHECK-FMA-NEXT:    vcvtudq2ps %xmm0, %xmm0
+; CHECK-FMA-NEXT:    vfmadd132ps {{.*#+}} xmm0 = (xmm0 * mem) + xmm1
+; CHECK-FMA-NEXT:    retq
+  %shl = shl nsw nuw <4 x i32> <i32 2, i32 2, i32 2, i32 2>, %cnt
+  %conv = uitofp <4 x i32> %shl to <4 x float>
+  %mul = fmul <4 x float> <float 5.000000e+00, float 5.000000e+00, float 5.000000e+00, float 5.000000e+00>, %conv
+  %res = fadd <4 x float> %mul, %add
+  ret <4 x float> %res
+}
+
+define <2 x double> @fmul_pow_shl_cnt_vec_non_splat_todo(<2 x i64> %cnt) nounwind {
+; CHECK-SSE-LABEL: fmul_pow_shl_cnt_vec_non_splat_todo:
+; CHECK-SSE:       # %bb.0:
+; CHECK-SSE-NEXT:    movdqa {{.*#+}} xmm1 = [2,2]
+; CHECK-SSE-NEXT:    movdqa %xmm1, %xmm2
+; CHECK-SSE-NEXT:    psllq %xmm0, %xmm2
+; CHECK-SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; CHECK-SSE-NEXT:    psllq %xmm0, %xmm1
+; CHECK-SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; CHECK-SSE-NEXT:    movapd {{.*#+}} xmm0 = [4294967295,4294967295]
+; CHECK-SSE-NEXT:    andpd %xmm1, %xmm0
+; CHECK-SSE-NEXT:    orpd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; CHECK-SSE-NEXT:    psrlq $32, %xmm1
+; CHECK-SSE-NEXT:    por {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1
+; CHECK-SSE-NEXT:    subpd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1
+; CHECK-SSE-NEXT:    addpd %xmm0, %xmm1
+; CHECK-SSE-NEXT:    mulpd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1
+; CHECK-SSE-NEXT:    movapd %xmm1, %xmm0
+; CHECK-SSE-NEXT:    retq
+;
+; CHECK-AVX2-LABEL: fmul_pow_shl_cnt_vec_non_splat_todo:
+; CHECK-AVX2:       # %bb.0:
+; CHECK-AVX2-NEXT:    vpbroadcastq {{.*#+}} xmm1 = [2,2]
+; CHECK-AVX2-NEXT:    vpsllvq %xmm0, %xmm1, %xmm0
+; CHECK-AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; CHECK-AVX2-NEXT:    vpblendd {{.*#+}} xmm1 = xmm0[0],xmm1[1],xmm0[2],xmm1[3]
+; CHECK-AVX2-NEXT:    vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
+; CHECK-AVX2-NEXT:    vpsrlq $32, %xmm0, %xmm0
+; CHECK-AVX2-NEXT:    vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; CHECK-AVX2-NEXT:    vsubpd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; CHECK-AVX2-NEXT:    vaddpd %xmm0, %xmm1, %xmm0
+; CHECK-AVX2-NEXT:    vmulpd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; CHECK-AVX2-NEXT:    retq
+;
+; CHECK-NO-FASTFMA-LABEL: fmul_pow_shl_cnt_vec_non_splat_todo:
+; CHECK-NO-FASTFMA:       # %bb.0:
+; CHECK-NO-FASTFMA-NEXT:    vpbroadcastq {{.*#+}} xmm1 = [2,2]
+; CHECK-NO-FASTFMA-NEXT:    vpsllvq %xmm0, %xmm1, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; CHECK-NO-FASTFMA-NEXT:    vpblendd {{.*#+}} xmm1 = xmm0[0],xmm1[1],xmm0[2],xmm1[3]
+; CHECK-NO-FASTFMA-NEXT:    vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
+; CHECK-NO-FASTFMA-NEXT:    vpsrlq $32, %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vsubpd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vaddpd %xmm0, %xmm1, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vmulpd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    retq
+;
+; CHECK-FMA-LABEL: fmul_pow_shl_cnt_vec_non_splat_todo:
+; CHECK-FMA:       # %bb.0:
+; CHECK-FMA-NEXT:    vpbroadcastq {{.*#+}} xmm1 = [2,2]
+; CHECK-FMA-NEXT:    vpsllvq %xmm0, %xmm1, %xmm0
+; CHECK-FMA-NEXT:    vcvtuqq2pd %xmm0, %xmm0
+; CHECK-FMA-NEXT:    vmulpd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; CHECK-FMA-NEXT:    retq
+  %shl = shl nsw nuw <2 x i64> <i64 2, i64 2>, %cnt
+  %conv = uitofp <2 x i64> %shl to <2 x double>
+  %mul = fmul <2 x double> <double 15.000000e+00, double 14.000000e+00>, %conv
+  ret <2 x double> %mul
+}
+
+define <2 x double> @fmul_pow_shl_cnt_vec_non_splat2_todo(<2 x i64> %cnt) nounwind {
+; CHECK-SSE-LABEL: fmul_pow_shl_cnt_vec_non_splat2_todo:
+; CHECK-SSE:       # %bb.0:
+; CHECK-SSE-NEXT:    movdqa {{.*#+}} xmm1 = [2,1]
+; CHECK-SSE-NEXT:    movdqa %xmm1, %xmm2
+; CHECK-SSE-NEXT:    psllq %xmm0, %xmm2
+; CHECK-SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; CHECK-SSE-NEXT:    psllq %xmm0, %xmm1
+; CHECK-SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; CHECK-SSE-NEXT:    movapd {{.*#+}} xmm0 = [4294967295,4294967295]
+; CHECK-SSE-NEXT:    andpd %xmm1, %xmm0
+; CHECK-SSE-NEXT:    orpd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; CHECK-SSE-NEXT:    psrlq $32, %xmm1
+; CHECK-SSE-NEXT:    por {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1
+; CHECK-SSE-NEXT:    subpd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1
+; CHECK-SSE-NEXT:    addpd %xmm0, %xmm1
+; CHECK-SSE-NEXT:    mulpd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1
+; CHECK-SSE-NEXT:    movapd %xmm1, %xmm0
+; CHECK-SSE-NEXT:    retq
+;
+; CHECK-AVX2-LABEL: fmul_pow_shl_cnt_vec_non_splat2_todo:
+; CHECK-AVX2:       # %bb.0:
+; CHECK-AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [2,1]
+; CHECK-AVX2-NEXT:    vpsllvq %xmm0, %xmm1, %xmm0
+; CHECK-AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; CHECK-AVX2-NEXT:    vpblendd {{.*#+}} xmm1 = xmm0[0],xmm1[1],xmm0[2],xmm1[3]
+; CHECK-AVX2-NEXT:    vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
+; CHECK-AVX2-NEXT:    vpsrlq $32, %xmm0, %xmm0
+; CHECK-AVX2-NEXT:    vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; CHECK-AVX2-NEXT:    vsubpd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; CHECK-AVX2-NEXT:    vaddpd %xmm0, %xmm1, %xmm0
+; CHECK-AVX2-NEXT:    vmulpd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; CHECK-AVX2-NEXT:    retq
+;
+; CHECK-NO-FASTFMA-LABEL: fmul_pow_shl_cnt_vec_non_splat2_todo:
+; CHECK-NO-FASTFMA:       # %bb.0:
+; CHECK-NO-FASTFMA-NEXT:    vmovdqa {{.*#+}} xmm1 = [2,1]
+; CHECK-NO-FASTFMA-NEXT:    vpsllvq %xmm0, %xmm1, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; CHECK-NO-FASTFMA-NEXT:    vpblendd {{.*#+}} xmm1 = xmm0[0],xmm1[1],xmm0[2],xmm1[3]
+; CHECK-NO-FASTFMA-NEXT:    vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
+; CHECK-NO-FASTFMA-NEXT:    vpsrlq $32, %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vsubpd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vaddpd %xmm0, %xmm1, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vmulpd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    retq
+;
+; CHECK-FMA-LABEL: fmul_pow_shl_cnt_vec_non_splat2_todo:
+; CHECK-FMA:       # %bb.0:
+; CHECK-FMA-NEXT:    vmovdqa {{.*#+}} xmm1 = [2,1]
+; CHECK-FMA-NEXT:    vpsllvq %xmm0, %xmm1, %xmm0
+; CHECK-FMA-NEXT:    vcvtuqq2pd %xmm0, %xmm0
+; CHECK-FMA-NEXT:    vmulpd {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to2}, %xmm0, %xmm0
+; CHECK-FMA-NEXT:    retq
+  %shl = shl nsw nuw <2 x i64> <i64 2, i64 1>, %cnt
+  %conv = uitofp <2 x i64> %shl to <2 x double>
+  %mul = fmul <2 x double> <double 15.000000e+00, double 15.000000e+00>, %conv
+  ret <2 x double> %mul
+}
+
+define <2 x half> @fmul_pow_shl_cnt_vec_fail_to_large(<2 x i16> %cnt) nounwind {
+; CHECK-SSE-LABEL: fmul_pow_shl_cnt_vec_fail_to_large:
+; CHECK-SSE:       # %bb.0:
+; CHECK-SSE-NEXT:    subq $40, %rsp
+; CHECK-SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3]
+; CHECK-SSE-NEXT:    pslld $23, %xmm0
+; CHECK-SSE-NEXT:    paddd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; CHECK-SSE-NEXT:    cvttps2dq %xmm0, %xmm0
+; CHECK-SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
+; CHECK-SSE-NEXT:    pmullw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; CHECK-SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; CHECK-SSE-NEXT:    pextrw $1, %xmm0, %eax
+; CHECK-SSE-NEXT:    xorps %xmm0, %xmm0
+; CHECK-SSE-NEXT:    cvtsi2ss %eax, %xmm0
+; CHECK-SSE-NEXT:    callq __truncsfhf2@PLT
+; CHECK-SSE-NEXT:    movss %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
+; CHECK-SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; CHECK-SSE-NEXT:    pextrw $0, %xmm0, %eax
+; CHECK-SSE-NEXT:    xorps %xmm0, %xmm0
+; CHECK-SSE-NEXT:    cvtsi2ss %eax, %xmm0
+; CHECK-SSE-NEXT:    callq __truncsfhf2@PLT
+; CHECK-SSE-NEXT:    callq __extendhfsf2@PLT
+; CHECK-SSE-NEXT:    mulss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; CHECK-SSE-NEXT:    callq __truncsfhf2@PLT
+; CHECK-SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; CHECK-SSE-NEXT:    movss {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 4-byte Reload
+; CHECK-SSE-NEXT:    # xmm0 = mem[0],zero,zero,zero
+; CHECK-SSE-NEXT:    callq __extendhfsf2@PLT
+; CHECK-SSE-NEXT:    mulss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; CHECK-SSE-NEXT:    callq __truncsfhf2@PLT
+; CHECK-SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; CHECK-SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; CHECK-SSE-NEXT:    movdqa %xmm1, %xmm0
+; CHECK-SSE-NEXT:    addq $40, %rsp
+; CHECK-SSE-NEXT:    retq
+;
+; CHECK-AVX2-LABEL: fmul_pow_shl_cnt_vec_fail_to_large:
+; CHECK-AVX2:       # %bb.0:
+; CHECK-AVX2-NEXT:    subq $40, %rsp
+; CHECK-AVX2-NEXT:    vpmovzxwd {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero
+; CHECK-AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [2,2,2,2]
+; CHECK-AVX2-NEXT:    vpsllvd %xmm0, %xmm1, %xmm0
+; CHECK-AVX2-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; CHECK-AVX2-NEXT:    vpextrw $2, %xmm0, %eax
+; CHECK-AVX2-NEXT:    vcvtsi2ss %eax, %xmm2, %xmm0
+; CHECK-AVX2-NEXT:    callq __truncsfhf2@PLT
+; CHECK-AVX2-NEXT:    vmovss %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
+; CHECK-AVX2-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; CHECK-AVX2-NEXT:    vpextrw $0, %xmm0, %eax
+; CHECK-AVX2-NEXT:    vcvtsi2ss %eax, %xmm2, %xmm0
+; CHECK-AVX2-NEXT:    callq __truncsfhf2@PLT
+; CHECK-AVX2-NEXT:    callq __extendhfsf2@PLT
+; CHECK-AVX2-NEXT:    vmulss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; CHECK-AVX2-NEXT:    callq __truncsfhf2@PLT
+; CHECK-AVX2-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; CHECK-AVX2-NEXT:    vmovss {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 4-byte Reload
+; CHECK-AVX2-NEXT:    # xmm0 = mem[0],zero,zero,zero
+; CHECK-AVX2-NEXT:    callq __extendhfsf2@PLT
+; CHECK-AVX2-NEXT:    vmulss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; CHECK-AVX2-NEXT:    callq __truncsfhf2@PLT
+; CHECK-AVX2-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; CHECK-AVX2-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; CHECK-AVX2-NEXT:    addq $40, %rsp
+; CHECK-AVX2-NEXT:    retq
+;
+; CHECK-NO-FASTFMA-LABEL: fmul_pow_shl_cnt_vec_fail_to_large:
+; CHECK-NO-FASTFMA:       # %bb.0:
+; CHECK-NO-FASTFMA-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; CHECK-NO-FASTFMA-NEXT:    vmovdqa {{.*#+}} ymm1 = [2,2,0,0,0,0,0,0]
+; CHECK-NO-FASTFMA-NEXT:    vpsllvd %ymm0, %ymm1, %ymm0
+; CHECK-NO-FASTFMA-NEXT:    vpmovdw %zmm0, %ymm1
+; CHECK-NO-FASTFMA-NEXT:    vpextrw $0, %xmm0, %eax
+; CHECK-NO-FASTFMA-NEXT:    vcvtsi2ss %eax, %xmm2, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vcvtps2ph $4, %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vmovd %xmm0, %eax
+; CHECK-NO-FASTFMA-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vpextrw $1, %xmm1, %eax
+; CHECK-NO-FASTFMA-NEXT:    vcvtsi2ss %eax, %xmm2, %xmm1
+; CHECK-NO-FASTFMA-NEXT:    vcvtps2ph $4, %xmm1, %xmm1
+; CHECK-NO-FASTFMA-NEXT:    vmovd %xmm1, %eax
+; CHECK-NO-FASTFMA-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm1
+; CHECK-NO-FASTFMA-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; CHECK-NO-FASTFMA-NEXT:    vmovaps {{.*#+}} xmm1 = [16,0,0,0]
+; CHECK-NO-FASTFMA-NEXT:    xorl %eax, %eax
+; CHECK-NO-FASTFMA-NEXT:    vcvtsi2ss %eax, %xmm2, %xmm2
+; CHECK-NO-FASTFMA-NEXT:    vcvtps2ph $4, %xmm2, %xmm2
+; CHECK-NO-FASTFMA-NEXT:    vmovd %xmm2, %eax
+; CHECK-NO-FASTFMA-NEXT:    vmovd %eax, %xmm2
+; CHECK-NO-FASTFMA-NEXT:    vpbroadcastw %xmm2, %xmm2
+; CHECK-NO-FASTFMA-NEXT:    vpermt2ps %zmm0, %zmm1, %zmm2
+; CHECK-NO-FASTFMA-NEXT:    vcvtph2ps %xmm2, %ymm0
+; CHECK-NO-FASTFMA-NEXT:    vbroadcastss {{.*#+}} ymm1 = [1.5E+1,1.5E+1,1.5E+1,1.5E+1,1.5E+1,1.5E+1,1.5E+1,1.5E+1]
+; CHECK-NO-FASTFMA-NEXT:    vmulps %ymm1, %ymm0, %ymm0
+; CHECK-NO-FASTFMA-NEXT:    vcvtps2ph $4, %ymm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vzeroupper
+; CHECK-NO-FASTFMA-NEXT:    retq
+;
+; CHECK-FMA-LABEL: fmul_pow_shl_cnt_vec_fail_to_large:
+; CHECK-FMA:       # %bb.0:
+; CHECK-FMA-NEXT:    vpbroadcastw {{.*#+}} xmm1 = [2,2,2,2,2,2,2,2]
+; CHECK-FMA-NEXT:    vpsllvw %xmm0, %xmm1, %xmm0
+; CHECK-FMA-NEXT:    vpextrw $7, %xmm0, %eax
+; CHECK-FMA-NEXT:    vcvtsi2ss %eax, %xmm2, %xmm1
+; CHECK-FMA-NEXT:    vcvtps2ph $4, %xmm1, %xmm1
+; CHECK-FMA-NEXT:    vmovd %xmm1, %eax
+; CHECK-FMA-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm1
+; CHECK-FMA-NEXT:    vpextrw $6, %xmm0, %eax
+; CHECK-FMA-NEXT:    vcvtsi2ss %eax, %xmm2, %xmm2
+; CHECK-FMA-NEXT:    vcvtps2ph $4, %xmm2, %xmm2
+; CHECK-FMA-NEXT:    vmovd %xmm2, %eax
+; CHECK-FMA-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm2
+; CHECK-FMA-NEXT:    vpextrw $5, %xmm0, %eax
+; CHECK-FMA-NEXT:    vcvtsi2ss %eax, %xmm3, %xmm3
+; CHECK-FMA-NEXT:    vcvtps2ph $4, %xmm3, %xmm3
+; CHECK-FMA-NEXT:    vmovd %xmm3, %eax
+; CHECK-FMA-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm3
+; CHECK-FMA-NEXT:    vpextrw $4, %xmm0, %eax
+; CHECK-FMA-NEXT:    vcvtsi2ss %eax, %xmm4, %xmm4
+; CHECK-FMA-NEXT:    vpunpcklwd {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
+; CHECK-FMA-NEXT:    vcvtps2ph $4, %xmm4, %xmm2
+; CHECK-FMA-NEXT:    vmovd %xmm2, %eax
+; CHECK-FMA-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm2
+; CHECK-FMA-NEXT:    vpextrw $3, %xmm0, %eax
+; CHECK-FMA-NEXT:    vcvtsi2ss %eax, %xmm5, %xmm4
+; CHECK-FMA-NEXT:    vcvtps2ph $4, %xmm4, %xmm4
+; CHECK-FMA-NEXT:    vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3]
+; CHECK-FMA-NEXT:    vmovd %xmm4, %eax
+; CHECK-FMA-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm3
+; CHECK-FMA-NEXT:    vpextrw $2, %xmm0, %eax
+; CHECK-FMA-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; CHECK-FMA-NEXT:    vcvtsi2ss %eax, %xmm5, %xmm2
+; CHECK-FMA-NEXT:    vcvtps2ph $4, %xmm2, %xmm2
+; CHECK-FMA-NEXT:    vmovd %xmm2, %eax
+; CHECK-FMA-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm2
+; CHECK-FMA-NEXT:    vpextrw $1, %xmm0, %eax
+; CHECK-FMA-NEXT:    vcvtsi2ss %eax, %xmm5, %xmm4
+; CHECK-FMA-NEXT:    vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3]
+; CHECK-FMA-NEXT:    vcvtps2ph $4, %xmm4, %xmm3
+; CHECK-FMA-NEXT:    vmovd %xmm3, %eax
+; CHECK-FMA-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm3
+; CHECK-FMA-NEXT:    vpextrw $0, %xmm0, %eax
+; CHECK-FMA-NEXT:    vcvtsi2ss %eax, %xmm5, %xmm0
+; CHECK-FMA-NEXT:    vcvtps2ph $4, %xmm0, %xmm0
+; CHECK-FMA-NEXT:    vmovd %xmm0, %eax
+; CHECK-FMA-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0
+; CHECK-FMA-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3]
+; CHECK-FMA-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; CHECK-FMA-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; CHECK-FMA-NEXT:    vcvtph2ps %xmm0, %ymm0
+; CHECK-FMA-NEXT:    vmulps {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to8}, %ymm0, %ymm0
+; CHECK-FMA-NEXT:    vcvtps2ph $4, %ymm0, %xmm0
+; CHECK-FMA-NEXT:    vzeroupper
+; CHECK-FMA-NEXT:    retq
+  %shl = shl nsw nuw <2 x i16> <i16 2, i16 2>, %cnt
+  %conv = uitofp <2 x i16> %shl to <2 x half>
+  %mul = fmul <2 x half> <half 15.000000e+00, half 15.000000e+00>, %conv
+  ret <2 x half> %mul
+}
+
+define double @fmul_pow_shl_cnt_fail_maybe_bad_exp(i64 %cnt) nounwind {
+; CHECK-SSE-LABEL: fmul_pow_shl_cnt_fail_maybe_bad_exp:
+; CHECK-SSE:       # %bb.0:
+; CHECK-SSE-NEXT:    movq %rdi, %rcx
+; CHECK-SSE-NEXT:    movl $1, %eax
+; CHECK-SSE-NEXT:    # kill: def $cl killed $cl killed $rcx
+; CHECK-SSE-NEXT:    shlq %cl, %rax
+; CHECK-SSE-NEXT:    movq %rax, %xmm1
+; CHECK-SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],mem[0],xmm1[1],mem[1]
+; CHECK-SSE-NEXT:    subpd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1
+; CHECK-SSE-NEXT:    movapd %xmm1, %xmm0
+; CHECK-SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
+; CHECK-SSE-NEXT:    addsd %xmm1, %xmm0
+; CHECK-SSE-NEXT:    mulsd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; CHECK-SSE-NEXT:    retq
+;
+; CHECK-AVX2-LABEL: fmul_pow_shl_cnt_fail_maybe_bad_exp:
+; CHECK-AVX2:       # %bb.0:
+; CHECK-AVX2-NEXT:    movq %rdi, %rcx
+; CHECK-AVX2-NEXT:    movl $1, %eax
+; CHECK-AVX2-NEXT:    # kill: def $cl killed $cl killed $rcx
+; CHECK-AVX2-NEXT:    shlq %cl, %rax
+; CHECK-AVX2-NEXT:    vmovq %rax, %xmm0
+; CHECK-AVX2-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; CHECK-AVX2-NEXT:    vsubpd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; CHECK-AVX2-NEXT:    vshufpd {{.*#+}} xmm1 = xmm0[1,0]
+; CHECK-AVX2-NEXT:    vaddsd %xmm0, %xmm1, %xmm0
+; CHECK-AVX2-NEXT:    vmulsd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; CHECK-AVX2-NEXT:    retq
+;
+; CHECK-NO-FASTFMA-LABEL: fmul_pow_shl_cnt_fail_maybe_bad_exp:
+; CHECK-NO-FASTFMA:       # %bb.0:
+; CHECK-NO-FASTFMA-NEXT:    movq %rdi, %rcx
+; CHECK-NO-FASTFMA-NEXT:    movl $1, %eax
+; CHECK-NO-FASTFMA-NEXT:    # kill: def $cl killed $cl killed $rcx
+; CHECK-NO-FASTFMA-NEXT:    shlq %cl, %rax
+; CHECK-NO-FASTFMA-NEXT:    vcvtusi2sd %rax, %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vmulsd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    retq
+;
+; CHECK-FMA-LABEL: fmul_pow_shl_cnt_fail_maybe_bad_exp:
+; CHECK-FMA:       # %bb.0:
+; CHECK-FMA-NEXT:    movl $1, %eax
+; CHECK-FMA-NEXT:    shlxq %rdi, %rax, %rax
+; CHECK-FMA-NEXT:    vcvtusi2sd %rax, %xmm0, %xmm0
+; CHECK-FMA-NEXT:    vmulsd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; CHECK-FMA-NEXT:    retq
+  %shl = shl nuw i64 1, %cnt
+  %conv = uitofp i64 %shl to double
+  %mul = fmul double 9.745314e+288, %conv
+  ret double %mul
+}
+
+define double @fmul_pow_shl_cnt_safe(i16 %cnt) nounwind {
+; CHECK-SSE-LABEL: fmul_pow_shl_cnt_safe:
+; CHECK-SSE:       # %bb.0:
+; CHECK-SSE-NEXT:    movl %edi, %ecx
+; CHECK-SSE-NEXT:    movl $1, %eax
+; CHECK-SSE-NEXT:    # kill: def $cl killed $cl killed $ecx
+; CHECK-SSE-NEXT:    shll %cl, %eax
+; CHECK-SSE-NEXT:    movzwl %ax, %eax
+; CHECK-SSE-NEXT:    cvtsi2sd %eax, %xmm0
+; CHECK-SSE-NEXT:    mulsd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; CHECK-SSE-NEXT:    retq
+;
+; CHECK-AVX2-LABEL: fmul_pow_shl_cnt_safe:
+; CHECK-AVX2:       # %bb.0:
+; CHECK-AVX2-NEXT:    movl %edi, %ecx
+; CHECK-AVX2-NEXT:    movl $1, %eax
+; CHECK-AVX2-NEXT:    # kill: def $cl killed $cl killed $ecx
+; CHECK-AVX2-NEXT:    shll %cl, %eax
+; CHECK-AVX2-NEXT:    movzwl %ax, %eax
+; CHECK-AVX2-NEXT:    vcvtsi2sd %eax, %xmm0, %xmm0
+; CHECK-AVX2-NEXT:    vmulsd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; CHECK-AVX2-NEXT:    retq
+;
+; CHECK-NO-FASTFMA-LABEL: fmul_pow_shl_cnt_safe:
+; CHECK-NO-FASTFMA:       # %bb.0:
+; CHECK-NO-FASTFMA-NEXT:    movl %edi, %ecx
+; CHECK-NO-FASTFMA-NEXT:    movl $1, %eax
+; CHECK-NO-FASTFMA-NEXT:    # kill: def $cl killed $cl killed $ecx
+; CHECK-NO-FASTFMA-NEXT:    shll %cl, %eax
+; CHECK-NO-FASTFMA-NEXT:    movzwl %ax, %eax
+; CHECK-NO-FASTFMA-NEXT:    vcvtsi2sd %eax, %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vmulsd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    retq
+;
+; CHECK-FMA-LABEL: fmul_pow_shl_cnt_safe:
+; CHECK-FMA:       # %bb.0:
+; CHECK-FMA-NEXT:    movl $1, %eax
+; CHECK-FMA-NEXT:    shlxl %edi, %eax, %eax
+; CHECK-FMA-NEXT:    movzwl %ax, %eax
+; CHECK-FMA-NEXT:    vcvtsi2sd %eax, %xmm0, %xmm0
+; CHECK-FMA-NEXT:    vmulsd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; CHECK-FMA-NEXT:    retq
+  %shl = shl nuw i16 1, %cnt
+  %conv = uitofp i16 %shl to double
+  %mul = fmul double 9.745314e+288, %conv
+  ret double %mul
+}
+
+define <2 x double> @fdiv_pow_shl_cnt_vec(<2 x i64> %cnt) nounwind {
+; CHECK-SSE-LABEL: fdiv_pow_shl_cnt_vec:
+; CHECK-SSE:       # %bb.0:
+; CHECK-SSE-NEXT:    movdqa {{.*#+}} xmm1 = [1,1]
+; CHECK-SSE-NEXT:    movdqa %xmm1, %xmm2
+; CHECK-SSE-NEXT:    psllq %xmm0, %xmm2
+; CHECK-SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; CHECK-SSE-NEXT:    psllq %xmm0, %xmm1
+; CHECK-SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; CHECK-SSE-NEXT:    movapd {{.*#+}} xmm0 = [4294967295,4294967295]
+; CHECK-SSE-NEXT:    andpd %xmm1, %xmm0
+; CHECK-SSE-NEXT:    orpd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; CHECK-SSE-NEXT:    psrlq $32, %xmm1
+; CHECK-SSE-NEXT:    por {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1
+; CHECK-SSE-NEXT:    subpd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1
+; CHECK-SSE-NEXT:    addpd %xmm0, %xmm1
+; CHECK-SSE-NEXT:    movapd {{.*#+}} xmm0 = [1.0E+0,1.0E+0]
+; CHECK-SSE-NEXT:    divpd %xmm1, %xmm0
+; CHECK-SSE-NEXT:    retq
+;
+; CHECK-AVX2-LABEL: fdiv_pow_shl_cnt_vec:
+; CHECK-AVX2:       # %bb.0:
+; CHECK-AVX2-NEXT:    vpbroadcastq {{.*#+}} xmm1 = [1,1]
+; CHECK-AVX2-NEXT:    vpsllvq %xmm0, %xmm1, %xmm0
+; CHECK-AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; CHECK-AVX2-NEXT:    vpblendd {{.*#+}} xmm1 = xmm0[0],xmm1[1],xmm0[2],xmm1[3]
+; CHECK-AVX2-NEXT:    vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
+; CHECK-AVX2-NEXT:    vpsrlq $32, %xmm0, %xmm0
+; CHECK-AVX2-NEXT:    vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; CHECK-AVX2-NEXT:    vsubpd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; CHECK-AVX2-NEXT:    vaddpd %xmm0, %xmm1, %xmm0
+; CHECK-AVX2-NEXT:    vmovddup {{.*#+}} xmm1 = [1.0E+0,1.0E+0]
+; CHECK-AVX2-NEXT:    # xmm1 = mem[0,0]
+; CHECK-AVX2-NEXT:    vdivpd %xmm0, %xmm1, %xmm0
+; CHECK-AVX2-NEXT:    retq
+;
+; CHECK-NO-FASTFMA-LABEL: fdiv_pow_shl_cnt_vec:
+; CHECK-NO-FASTFMA:       # %bb.0:
+; CHECK-NO-FASTFMA-NEXT:    vpbroadcastq {{.*#+}} xmm1 = [1,1]
+; CHECK-NO-FASTFMA-NEXT:    vpsllvq %xmm0, %xmm1, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; CHECK-NO-FASTFMA-NEXT:    vpblendd {{.*#+}} xmm1 = xmm0[0],xmm1[1],xmm0[2],xmm1[3]
+; CHECK-NO-FASTFMA-NEXT:    vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
+; CHECK-NO-FASTFMA-NEXT:    vpsrlq $32, %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vsubpd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vaddpd %xmm0, %xmm1, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vmovddup {{.*#+}} xmm1 = [1.0E+0,1.0E+0]
+; CHECK-NO-FASTFMA-NEXT:    # xmm1 = mem[0,0]
+; CHECK-NO-FASTFMA-NEXT:    vdivpd %xmm0, %xmm1, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    retq
+;
+; CHECK-FMA-LABEL: fdiv_pow_shl_cnt_vec:
+; CHECK-FMA:       # %bb.0:
+; CHECK-FMA-NEXT:    vpbroadcastq {{.*#+}} xmm1 = [1,1]
+; CHECK-FMA-NEXT:    vpsllvq %xmm0, %xmm1, %xmm0
+; CHECK-FMA-NEXT:    vcvtuqq2pd %xmm0, %xmm0
+; CHECK-FMA-NEXT:    vmovddup {{.*#+}} xmm1 = [1.0E+0,1.0E+0]
+; CHECK-FMA-NEXT:    # xmm1 = mem[0,0]
+; CHECK-FMA-NEXT:    vdivpd %xmm0, %xmm1, %xmm0
+; CHECK-FMA-NEXT:    retq
+  %shl = shl nuw <2 x i64> <i64 1, i64 1>, %cnt
+  %conv = uitofp <2 x i64> %shl to <2 x double>
+  %mul = fdiv <2 x double> <double 1.000000e+00, double 1.000000e+00>, %conv
+  ret <2 x double> %mul
+}
+
+define <2 x float> @fdiv_pow_shl_cnt_vec_with_expensive_cast(<2 x i64> %cnt) nounwind {
+; CHECK-SSE-LABEL: fdiv_pow_shl_cnt_vec_with_expensive_cast:
+; CHECK-SSE:       # %bb.0:
+; CHECK-SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; CHECK-SSE-NEXT:    movdqa {{.*#+}} xmm3 = [1,1]
+; CHECK-SSE-NEXT:    movdqa %xmm3, %xmm2
+; CHECK-SSE-NEXT:    psllq %xmm1, %xmm2
+; CHECK-SSE-NEXT:    psllq %xmm0, %xmm3
+; CHECK-SSE-NEXT:    movq %xmm3, %rax
+; CHECK-SSE-NEXT:    testq %rax, %rax
+; CHECK-SSE-NEXT:    js .LBB21_1
+; CHECK-SSE-NEXT:  # %bb.2:
+; CHECK-SSE-NEXT:    xorps %xmm1, %xmm1
+; CHECK-SSE-NEXT:    cvtsi2ss %rax, %xmm1
+; CHECK-SSE-NEXT:    jmp .LBB21_3
+; CHECK-SSE-NEXT:  .LBB21_1:
+; CHECK-SSE-NEXT:    movq %rax, %rcx
+; CHECK-SSE-NEXT:    shrq %rcx
+; CHECK-SSE-NEXT:    andl $1, %eax
+; CHECK-SSE-NEXT:    orq %rcx, %rax
+; CHECK-SSE-NEXT:    xorps %xmm1, %xmm1
+; CHECK-SSE-NEXT:    cvtsi2ss %rax, %xmm1
+; CHECK-SSE-NEXT:    addss %xmm1, %xmm1
+; CHECK-SSE-NEXT:  .LBB21_3:
+; CHECK-SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[2,3,2,3]
+; CHECK-SSE-NEXT:    movq %xmm0, %rax
+; CHECK-SSE-NEXT:    testq %rax, %rax
+; CHECK-SSE-NEXT:    js .LBB21_4
+; CHECK-SSE-NEXT:  # %bb.5:
+; CHECK-SSE-NEXT:    xorps %xmm0, %xmm0
+; CHECK-SSE-NEXT:    cvtsi2ss %rax, %xmm0
+; CHECK-SSE-NEXT:    jmp .LBB21_6
+; CHECK-SSE-NEXT:  .LBB21_4:
+; CHECK-SSE-NEXT:    movq %rax, %rcx
+; CHECK-SSE-NEXT:    shrq %rcx
+; CHECK-SSE-NEXT:    andl $1, %eax
+; CHECK-SSE-NEXT:    orq %rcx, %rax
+; CHECK-SSE-NEXT:    xorps %xmm0, %xmm0
+; CHECK-SSE-NEXT:    cvtsi2ss %rax, %xmm0
+; CHECK-SSE-NEXT:    addss %xmm0, %xmm0
+; CHECK-SSE-NEXT:  .LBB21_6:
+; CHECK-SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; CHECK-SSE-NEXT:    movaps {{.*#+}} xmm0 = <1.0E+0,1.0E+0,u,u>
+; CHECK-SSE-NEXT:    divps %xmm1, %xmm0
+; CHECK-SSE-NEXT:    retq
+;
+; CHECK-AVX2-LABEL: fdiv_pow_shl_cnt_vec_with_expensive_cast:
+; CHECK-AVX2:       # %bb.0:
+; CHECK-AVX2-NEXT:    vpbroadcastq {{.*#+}} xmm1 = [1,1]
+; CHECK-AVX2-NEXT:    vpsllvq %xmm0, %xmm1, %xmm0
+; CHECK-AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm1
+; CHECK-AVX2-NEXT:    vpsrlq $1, %xmm0, %xmm2
+; CHECK-AVX2-NEXT:    vpor %xmm1, %xmm2, %xmm1
+; CHECK-AVX2-NEXT:    vblendvpd %xmm0, %xmm1, %xmm0, %xmm1
+; CHECK-AVX2-NEXT:    vpextrq $1, %xmm1, %rax
+; CHECK-AVX2-NEXT:    vcvtsi2ss %rax, %xmm3, %xmm2
+; CHECK-AVX2-NEXT:    vmovq %xmm1, %rax
+; CHECK-AVX2-NEXT:    vcvtsi2ss %rax, %xmm3, %xmm1
+; CHECK-AVX2-NEXT:    vinsertps {{.*#+}} xmm1 = xmm1[0],xmm2[0],zero,zero
+; CHECK-AVX2-NEXT:    vaddps %xmm1, %xmm1, %xmm2
+; CHECK-AVX2-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; CHECK-AVX2-NEXT:    vpcmpgtq %xmm0, %xmm3, %xmm0
+; CHECK-AVX2-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[1,3,2,3]
+; CHECK-AVX2-NEXT:    vblendvps %xmm0, %xmm2, %xmm1, %xmm0
+; CHECK-AVX2-NEXT:    vbroadcastss {{.*#+}} xmm1 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0]
+; CHECK-AVX2-NEXT:    vdivps %xmm0, %xmm1, %xmm0
+; CHECK-AVX2-NEXT:    retq
+;
+; CHECK-NO-FASTFMA-LABEL: fdiv_pow_shl_cnt_vec_with_expensive_cast:
+; CHECK-NO-FASTFMA:       # %bb.0:
+; CHECK-NO-FASTFMA-NEXT:    vpbroadcastq {{.*#+}} xmm1 = [1,1]
+; CHECK-NO-FASTFMA-NEXT:    vpsllvq %xmm0, %xmm1, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vpextrq $1, %xmm0, %rax
+; CHECK-NO-FASTFMA-NEXT:    vcvtusi2ss %rax, %xmm2, %xmm1
+; CHECK-NO-FASTFMA-NEXT:    vmovq %xmm0, %rax
+; CHECK-NO-FASTFMA-NEXT:    vcvtusi2ss %rax, %xmm2, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vinsertps {{.*#+}} xmm0 = xmm0[0],xmm1[0],zero,zero
+; CHECK-NO-FASTFMA-NEXT:    vbroadcastss {{.*#+}} xmm1 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0]
+; CHECK-NO-FASTFMA-NEXT:    vdivps %xmm0, %xmm1, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    retq
+;
+; CHECK-FMA-LABEL: fdiv_pow_shl_cnt_vec_with_expensive_cast:
+; CHECK-FMA:       # %bb.0:
+; CHECK-FMA-NEXT:    vpbroadcastq {{.*#+}} xmm1 = [1,1]
+; CHECK-FMA-NEXT:    vpsllvq %xmm0, %xmm1, %xmm0
+; CHECK-FMA-NEXT:    vcvtuqq2ps %xmm0, %xmm0
+; CHECK-FMA-NEXT:    vbroadcastss {{.*#+}} xmm1 = [1.0E+0,1.0E+0,1.0E+0,1.0E+0]
+; CHECK-FMA-NEXT:    vdivps %xmm0, %xmm1, %xmm0
+; CHECK-FMA-NEXT:    retq
+  %shl = shl nuw <2 x i64> <i64 1, i64 1>, %cnt
+  %conv = uitofp <2 x i64> %shl to <2 x float>
+  %mul = fdiv <2 x float> <float 1.000000e+00, float 1.000000e+00>, %conv
+  ret <2 x float> %mul
+}
+
+define float @fdiv_pow_shl_cnt_fail_maybe_z(i64 %cnt) nounwind {
+; CHECK-SSE-LABEL: fdiv_pow_shl_cnt_fail_maybe_z:
+; CHECK-SSE:       # %bb.0:
+; CHECK-SSE-NEXT:    movq %rdi, %rcx
+; CHECK-SSE-NEXT:    movl $8, %eax
+; CHECK-SSE-NEXT:    # kill: def $cl killed $cl killed $rcx
+; CHECK-SSE-NEXT:    shlq %cl, %rax
+; CHECK-SSE-NEXT:    testq %rax, %rax
+; CHECK-SSE-NEXT:    js .LBB22_1
+; CHECK-SSE-NEXT:  # %bb.2:
+; CHECK-SSE-NEXT:    cvtsi2ss %rax, %xmm1
+; CHECK-SSE-NEXT:    jmp .LBB22_3
+; CHECK-SSE-NEXT:  .LBB22_1:
+; CHECK-SSE-NEXT:    shrq %rax
+; CHECK-SSE-NEXT:    cvtsi2ss %rax, %xmm1
+; CHECK-SSE-NEXT:    addss %xmm1, %xmm1
+; CHECK-SSE-NEXT:  .LBB22_3:
+; CHECK-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; CHECK-SSE-NEXT:    divss %xmm1, %xmm0
+; CHECK-SSE-NEXT:    retq
+;
+; CHECK-AVX2-LABEL: fdiv_pow_shl_cnt_fail_maybe_z:
+; CHECK-AVX2:       # %bb.0:
+; CHECK-AVX2-NEXT:    movq %rdi, %rcx
+; CHECK-AVX2-NEXT:    movl $8, %eax
+; CHECK-AVX2-NEXT:    # kill: def $cl killed $cl killed $rcx
+; CHECK-AVX2-NEXT:    shlq %cl, %rax
+; CHECK-AVX2-NEXT:    testq %rax, %rax
+; CHECK-AVX2-NEXT:    js .LBB22_1
+; CHECK-AVX2-NEXT:  # %bb.2:
+; CHECK-AVX2-NEXT:    vcvtsi2ss %rax, %xmm0, %xmm0
+; CHECK-AVX2-NEXT:    jmp .LBB22_3
+; CHECK-AVX2-NEXT:  .LBB22_1:
+; CHECK-AVX2-NEXT:    shrq %rax
+; CHECK-AVX2-NEXT:    vcvtsi2ss %rax, %xmm0, %xmm0
+; CHECK-AVX2-NEXT:    vaddss %xmm0, %xmm0, %xmm0
+; CHECK-AVX2-NEXT:  .LBB22_3:
+; CHECK-AVX2-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; CHECK-AVX2-NEXT:    vdivss %xmm0, %xmm1, %xmm0
+; CHECK-AVX2-NEXT:    retq
+;
+; CHECK-NO-FASTFMA-LABEL: fdiv_pow_shl_cnt_fail_maybe_z:
+; CHECK-NO-FASTFMA:       # %bb.0:
+; CHECK-NO-FASTFMA-NEXT:    movq %rdi, %rcx
+; CHECK-NO-FASTFMA-NEXT:    movl $8, %eax
+; CHECK-NO-FASTFMA-NEXT:    # kill: def $cl killed $cl killed $rcx
+; CHECK-NO-FASTFMA-NEXT:    shlq %cl, %rax
+; CHECK-NO-FASTFMA-NEXT:    vcvtusi2ss %rax, %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; CHECK-NO-FASTFMA-NEXT:    vdivss %xmm0, %xmm1, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    retq
+;
+; CHECK-FMA-LABEL: fdiv_pow_shl_cnt_fail_maybe_z:
+; CHECK-FMA:       # %bb.0:
+; CHECK-FMA-NEXT:    movl $8, %eax
+; CHECK-FMA-NEXT:    shlxq %rdi, %rax, %rax
+; CHECK-FMA-NEXT:    vcvtusi2ss %rax, %xmm0, %xmm0
+; CHECK-FMA-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; CHECK-FMA-NEXT:    vdivss %xmm0, %xmm1, %xmm0
+; CHECK-FMA-NEXT:    retq
+  %shl = shl i64 8, %cnt
+  %conv = uitofp i64 %shl to float
+  %mul = fdiv float -9.000000e+00, %conv
+  ret float %mul
+}
+
+define float @fdiv_pow_shl_cnt_fail_neg_int(i64 %cnt) nounwind {
+; CHECK-SSE-LABEL: fdiv_pow_shl_cnt_fail_neg_int:
+; CHECK-SSE:       # %bb.0:
+; CHECK-SSE-NEXT:    movq %rdi, %rcx
+; CHECK-SSE-NEXT:    movl $8, %eax
+; CHECK-SSE-NEXT:    # kill: def $cl killed $cl killed $rcx
+; CHECK-SSE-NEXT:    shlq %cl, %rax
+; CHECK-SSE-NEXT:    cvtsi2ss %rax, %xmm1
+; CHECK-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; CHECK-SSE-NEXT:    divss %xmm1, %xmm0
+; CHECK-SSE-NEXT:    retq
+;
+; CHECK-AVX2-LABEL: fdiv_pow_shl_cnt_fail_neg_int:
+; CHECK-AVX2:       # %bb.0:
+; CHECK-AVX2-NEXT:    movq %rdi, %rcx
+; CHECK-AVX2-NEXT:    movl $8, %eax
+; CHECK-AVX2-NEXT:    # kill: def $cl killed $cl killed $rcx
+; CHECK-AVX2-NEXT:    shlq %cl, %rax
+; CHECK-AVX2-NEXT:    vcvtsi2ss %rax, %xmm0, %xmm0
+; CHECK-AVX2-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; CHECK-AVX2-NEXT:    vdivss %xmm0, %xmm1, %xmm0
+; CHECK-AVX2-NEXT:    retq
+;
+; CHECK-NO-FASTFMA-LABEL: fdiv_pow_shl_cnt_fail_neg_int:
+; CHECK-NO-FASTFMA:       # %bb.0:
+; CHECK-NO-FASTFMA-NEXT:    movq %rdi, %rcx
+; CHECK-NO-FASTFMA-NEXT:    movl $8, %eax
+; CHECK-NO-FASTFMA-NEXT:    # kill: def $cl killed $cl killed $rcx
+; CHECK-NO-FASTFMA-NEXT:    shlq %cl, %rax
+; CHECK-NO-FASTFMA-NEXT:    vcvtsi2ss %rax, %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; CHECK-NO-FASTFMA-NEXT:    vdivss %xmm0, %xmm1, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    retq
+;
+; CHECK-FMA-LABEL: fdiv_pow_shl_cnt_fail_neg_int:
+; CHECK-FMA:       # %bb.0:
+; CHECK-FMA-NEXT:    movl $8, %eax
+; CHECK-FMA-NEXT:    shlxq %rdi, %rax, %rax
+; CHECK-FMA-NEXT:    vcvtsi2ss %rax, %xmm0, %xmm0
+; CHECK-FMA-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; CHECK-FMA-NEXT:    vdivss %xmm0, %xmm1, %xmm0
+; CHECK-FMA-NEXT:    retq
+  %shl = shl i64 8, %cnt
+  %conv = sitofp i64 %shl to float
+  %mul = fdiv float -9.000000e+00, %conv
+  ret float %mul
+}
+
+define float @fdiv_pow_shl_cnt(i64 %cnt_in) nounwind {
+; CHECK-SSE-LABEL: fdiv_pow_shl_cnt:
+; CHECK-SSE:       # %bb.0:
+; CHECK-SSE-NEXT:    movq %rdi, %rcx
+; CHECK-SSE-NEXT:    andb $31, %cl
+; CHECK-SSE-NEXT:    movl $8, %eax
+; CHECK-SSE-NEXT:    # kill: def $cl killed $cl killed $rcx
+; CHECK-SSE-NEXT:    shlq %cl, %rax
+; CHECK-SSE-NEXT:    cvtsi2ss %rax, %xmm1
+; CHECK-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; CHECK-SSE-NEXT:    divss %xmm1, %xmm0
+; CHECK-SSE-NEXT:    retq
+;
+; CHECK-AVX2-LABEL: fdiv_pow_shl_cnt:
+; CHECK-AVX2:       # %bb.0:
+; CHECK-AVX2-NEXT:    movq %rdi, %rcx
+; CHECK-AVX2-NEXT:    andb $31, %cl
+; CHECK-AVX2-NEXT:    movl $8, %eax
+; CHECK-AVX2-NEXT:    # kill: def $cl killed $cl killed $rcx
+; CHECK-AVX2-NEXT:    shlq %cl, %rax
+; CHECK-AVX2-NEXT:    vcvtsi2ss %rax, %xmm0, %xmm0
+; CHECK-AVX2-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; CHECK-AVX2-NEXT:    vdivss %xmm0, %xmm1, %xmm0
+; CHECK-AVX2-NEXT:    retq
+;
+; CHECK-NO-FASTFMA-LABEL: fdiv_pow_shl_cnt:
+; CHECK-NO-FASTFMA:       # %bb.0:
+; CHECK-NO-FASTFMA-NEXT:    movq %rdi, %rcx
+; CHECK-NO-FASTFMA-NEXT:    andb $31, %cl
+; CHECK-NO-FASTFMA-NEXT:    movl $8, %eax
+; CHECK-NO-FASTFMA-NEXT:    # kill: def $cl killed $cl killed $rcx
+; CHECK-NO-FASTFMA-NEXT:    shlq %cl, %rax
+; CHECK-NO-FASTFMA-NEXT:    vcvtsi2ss %rax, %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; CHECK-NO-FASTFMA-NEXT:    vdivss %xmm0, %xmm1, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    retq
+;
+; CHECK-FMA-LABEL: fdiv_pow_shl_cnt:
+; CHECK-FMA:       # %bb.0:
+; CHECK-FMA-NEXT:    andb $31, %dil
+; CHECK-FMA-NEXT:    movl $8, %eax
+; CHECK-FMA-NEXT:    shlxq %rdi, %rax, %rax
+; CHECK-FMA-NEXT:    vcvtsi2ss %rax, %xmm0, %xmm0
+; CHECK-FMA-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; CHECK-FMA-NEXT:    vdivss %xmm0, %xmm1, %xmm0
+; CHECK-FMA-NEXT:    retq
+  %cnt = and i64 %cnt_in, 31
+  %shl = shl i64 8, %cnt
+  %conv = sitofp i64 %shl to float
+  %mul = fdiv float -0.500000e+00, %conv
+  ret float %mul
+}
+
+define half @fdiv_pow_shl_cnt_fail_out_of_bounds(i32 %cnt) nounwind {
+; CHECK-SSE-LABEL: fdiv_pow_shl_cnt_fail_out_of_bounds:
+; CHECK-SSE:       # %bb.0:
+; CHECK-SSE-NEXT:    pushq %rax
+; CHECK-SSE-NEXT:    movl %edi, %ecx
+; CHECK-SSE-NEXT:    movl $1, %eax
+; CHECK-SSE-NEXT:    # kill: def $cl killed $cl killed $ecx
+; CHECK-SSE-NEXT:    shll %cl, %eax
+; CHECK-SSE-NEXT:    cvtsi2ss %rax, %xmm0
+; CHECK-SSE-NEXT:    callq __truncsfhf2@PLT
+; CHECK-SSE-NEXT:    callq __extendhfsf2@PLT
+; CHECK-SSE-NEXT:    movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; CHECK-SSE-NEXT:    divss %xmm0, %xmm1
+; CHECK-SSE-NEXT:    movaps %xmm1, %xmm0
+; CHECK-SSE-NEXT:    callq __truncsfhf2@PLT
+; CHECK-SSE-NEXT:    popq %rax
+; CHECK-SSE-NEXT:    retq
+;
+; CHECK-AVX2-LABEL: fdiv_pow_shl_cnt_fail_out_of_bounds:
+; CHECK-AVX2:       # %bb.0:
+; CHECK-AVX2-NEXT:    pushq %rax
+; CHECK-AVX2-NEXT:    movl %edi, %ecx
+; CHECK-AVX2-NEXT:    movl $1, %eax
+; CHECK-AVX2-NEXT:    # kill: def $cl killed $cl killed $ecx
+; CHECK-AVX2-NEXT:    shll %cl, %eax
+; CHECK-AVX2-NEXT:    vcvtsi2ss %rax, %xmm0, %xmm0
+; CHECK-AVX2-NEXT:    callq __truncsfhf2@PLT
+; CHECK-AVX2-NEXT:    callq __extendhfsf2@PLT
+; CHECK-AVX2-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; CHECK-AVX2-NEXT:    vdivss %xmm0, %xmm1, %xmm0
+; CHECK-AVX2-NEXT:    callq __truncsfhf2@PLT
+; CHECK-AVX2-NEXT:    popq %rax
+; CHECK-AVX2-NEXT:    retq
+;
+; CHECK-NO-FASTFMA-LABEL: fdiv_pow_shl_cnt_fail_out_of_bounds:
+; CHECK-NO-FASTFMA:       # %bb.0:
+; CHECK-NO-FASTFMA-NEXT:    movl %edi, %ecx
+; CHECK-NO-FASTFMA-NEXT:    movl $1, %eax
+; CHECK-NO-FASTFMA-NEXT:    # kill: def $cl killed $cl killed $ecx
+; CHECK-NO-FASTFMA-NEXT:    shll %cl, %eax
+; CHECK-NO-FASTFMA-NEXT:    vcvtusi2ss %eax, %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vcvtps2ph $4, %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vpmovzxwq {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero
+; CHECK-NO-FASTFMA-NEXT:    vcvtph2ps %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; CHECK-NO-FASTFMA-NEXT:    vdivss %xmm0, %xmm1, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vcvtps2ph $4, %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vmovd %xmm0, %eax
+; CHECK-NO-FASTFMA-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    retq
+;
+; CHECK-FMA-LABEL: fdiv_pow_shl_cnt_fail_out_of_bounds:
+; CHECK-FMA:       # %bb.0:
+; CHECK-FMA-NEXT:    movl $1, %eax
+; CHECK-FMA-NEXT:    shlxl %edi, %eax, %eax
+; CHECK-FMA-NEXT:    vcvtusi2ss %eax, %xmm0, %xmm0
+; CHECK-FMA-NEXT:    vcvtps2ph $4, %xmm0, %xmm0
+; CHECK-FMA-NEXT:    vpmovzxwq {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero
+; CHECK-FMA-NEXT:    vcvtph2ps %xmm0, %xmm0
+; CHECK-FMA-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; CHECK-FMA-NEXT:    vdivss %xmm0, %xmm1, %xmm0
+; CHECK-FMA-NEXT:    vcvtps2ph $4, %xmm0, %xmm0
+; CHECK-FMA-NEXT:    vmovd %xmm0, %eax
+; CHECK-FMA-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0
+; CHECK-FMA-NEXT:    retq
+  %shl = shl nuw i32 1, %cnt
+  %conv = uitofp i32 %shl to half
+  %mul = fdiv half 0xH7000, %conv
+  ret half %mul
+}
+
+define half @fdiv_pow_shl_cnt_in_bounds(i16 %cnt) nounwind {
+; CHECK-SSE-LABEL: fdiv_pow_shl_cnt_in_bounds:
+; CHECK-SSE:       # %bb.0:
+; CHECK-SSE-NEXT:    pushq %rax
+; CHECK-SSE-NEXT:    movl %edi, %ecx
+; CHECK-SSE-NEXT:    movl $1, %eax
+; CHECK-SSE-NEXT:    # kill: def $cl killed $cl killed $ecx
+; CHECK-SSE-NEXT:    shll %cl, %eax
+; CHECK-SSE-NEXT:    movzwl %ax, %eax
+; CHECK-SSE-NEXT:    cvtsi2ss %eax, %xmm0
+; CHECK-SSE-NEXT:    callq __truncsfhf2@PLT
+; CHECK-SSE-NEXT:    callq __extendhfsf2@PLT
+; CHECK-SSE-NEXT:    movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; CHECK-SSE-NEXT:    divss %xmm0, %xmm1
+; CHECK-SSE-NEXT:    movaps %xmm1, %xmm0
+; CHECK-SSE-NEXT:    callq __truncsfhf2@PLT
+; CHECK-SSE-NEXT:    popq %rax
+; CHECK-SSE-NEXT:    retq
+;
+; CHECK-AVX2-LABEL: fdiv_pow_shl_cnt_in_bounds:
+; CHECK-AVX2:       # %bb.0:
+; CHECK-AVX2-NEXT:    pushq %rax
+; CHECK-AVX2-NEXT:    movl %edi, %ecx
+; CHECK-AVX2-NEXT:    movl $1, %eax
+; CHECK-AVX2-NEXT:    # kill: def $cl killed $cl killed $ecx
+; CHECK-AVX2-NEXT:    shll %cl, %eax
+; CHECK-AVX2-NEXT:    movzwl %ax, %eax
+; CHECK-AVX2-NEXT:    vcvtsi2ss %eax, %xmm0, %xmm0
+; CHECK-AVX2-NEXT:    callq __truncsfhf2@PLT
+; CHECK-AVX2-NEXT:    callq __extendhfsf2@PLT
+; CHECK-AVX2-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; CHECK-AVX2-NEXT:    vdivss %xmm0, %xmm1, %xmm0
+; CHECK-AVX2-NEXT:    callq __truncsfhf2@PLT
+; CHECK-AVX2-NEXT:    popq %rax
+; CHECK-AVX2-NEXT:    retq
+;
+; CHECK-NO-FASTFMA-LABEL: fdiv_pow_shl_cnt_in_bounds:
+; CHECK-NO-FASTFMA:       # %bb.0:
+; CHECK-NO-FASTFMA-NEXT:    movl %edi, %ecx
+; CHECK-NO-FASTFMA-NEXT:    movl $1, %eax
+; CHECK-NO-FASTFMA-NEXT:    # kill: def $cl killed $cl killed $ecx
+; CHECK-NO-FASTFMA-NEXT:    shll %cl, %eax
+; CHECK-NO-FASTFMA-NEXT:    movzwl %ax, %eax
+; CHECK-NO-FASTFMA-NEXT:    vcvtsi2ss %eax, %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vcvtps2ph $4, %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vpmovzxwq {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero
+; CHECK-NO-FASTFMA-NEXT:    vcvtph2ps %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; CHECK-NO-FASTFMA-NEXT:    vdivss %xmm0, %xmm1, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vcvtps2ph $4, %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vmovd %xmm0, %eax
+; CHECK-NO-FASTFMA-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    retq
+;
+; CHECK-FMA-LABEL: fdiv_pow_shl_cnt_in_bounds:
+; CHECK-FMA:       # %bb.0:
+; CHECK-FMA-NEXT:    movl $1, %eax
+; CHECK-FMA-NEXT:    shlxl %edi, %eax, %eax
+; CHECK-FMA-NEXT:    movzwl %ax, %eax
+; CHECK-FMA-NEXT:    vcvtsi2ss %eax, %xmm0, %xmm0
+; CHECK-FMA-NEXT:    vcvtps2ph $4, %xmm0, %xmm0
+; CHECK-FMA-NEXT:    vpmovzxwq {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero
+; CHECK-FMA-NEXT:    vcvtph2ps %xmm0, %xmm0
+; CHECK-FMA-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; CHECK-FMA-NEXT:    vdivss %xmm0, %xmm1, %xmm0
+; CHECK-FMA-NEXT:    vcvtps2ph $4, %xmm0, %xmm0
+; CHECK-FMA-NEXT:    vmovd %xmm0, %eax
+; CHECK-FMA-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0
+; CHECK-FMA-NEXT:    retq
+  %shl = shl nuw i16 1, %cnt
+  %conv = uitofp i16 %shl to half
+  %mul = fdiv half 0xH7000, %conv
+  ret half %mul
+}
+
+define half @fdiv_pow_shl_cnt_in_bounds2(i16 %cnt) nounwind {
+; CHECK-SSE-LABEL: fdiv_pow_shl_cnt_in_bounds2:
+; CHECK-SSE:       # %bb.0:
+; CHECK-SSE-NEXT:    pushq %rax
+; CHECK-SSE-NEXT:    movl %edi, %ecx
+; CHECK-SSE-NEXT:    movl $1, %eax
+; CHECK-SSE-NEXT:    # kill: def $cl killed $cl killed $ecx
+; CHECK-SSE-NEXT:    shll %cl, %eax
+; CHECK-SSE-NEXT:    movzwl %ax, %eax
+; CHECK-SSE-NEXT:    cvtsi2ss %eax, %xmm0
+; CHECK-SSE-NEXT:    callq __truncsfhf2@PLT
+; CHECK-SSE-NEXT:    callq __extendhfsf2@PLT
+; CHECK-SSE-NEXT:    movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; CHECK-SSE-NEXT:    divss %xmm0, %xmm1
+; CHECK-SSE-NEXT:    movaps %xmm1, %xmm0
+; CHECK-SSE-NEXT:    callq __truncsfhf2@PLT
+; CHECK-SSE-NEXT:    popq %rax
+; CHECK-SSE-NEXT:    retq
+;
+; CHECK-AVX2-LABEL: fdiv_pow_shl_cnt_in_bounds2:
+; CHECK-AVX2:       # %bb.0:
+; CHECK-AVX2-NEXT:    pushq %rax
+; CHECK-AVX2-NEXT:    movl %edi, %ecx
+; CHECK-AVX2-NEXT:    movl $1, %eax
+; CHECK-AVX2-NEXT:    # kill: def $cl killed $cl killed $ecx
+; CHECK-AVX2-NEXT:    shll %cl, %eax
+; CHECK-AVX2-NEXT:    movzwl %ax, %eax
+; CHECK-AVX2-NEXT:    vcvtsi2ss %eax, %xmm0, %xmm0
+; CHECK-AVX2-NEXT:    callq __truncsfhf2@PLT
+; CHECK-AVX2-NEXT:    callq __extendhfsf2@PLT
+; CHECK-AVX2-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; CHECK-AVX2-NEXT:    vdivss %xmm0, %xmm1, %xmm0
+; CHECK-AVX2-NEXT:    callq __truncsfhf2@PLT
+; CHECK-AVX2-NEXT:    popq %rax
+; CHECK-AVX2-NEXT:    retq
+;
+; CHECK-NO-FASTFMA-LABEL: fdiv_pow_shl_cnt_in_bounds2:
+; CHECK-NO-FASTFMA:       # %bb.0:
+; CHECK-NO-FASTFMA-NEXT:    movl %edi, %ecx
+; CHECK-NO-FASTFMA-NEXT:    movl $1, %eax
+; CHECK-NO-FASTFMA-NEXT:    # kill: def $cl killed $cl killed $ecx
+; CHECK-NO-FASTFMA-NEXT:    shll %cl, %eax
+; CHECK-NO-FASTFMA-NEXT:    movzwl %ax, %eax
+; CHECK-NO-FASTFMA-NEXT:    vcvtsi2ss %eax, %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vcvtps2ph $4, %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vpmovzxwq {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero
+; CHECK-NO-FASTFMA-NEXT:    vcvtph2ps %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; CHECK-NO-FASTFMA-NEXT:    vdivss %xmm0, %xmm1, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vcvtps2ph $4, %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vmovd %xmm0, %eax
+; CHECK-NO-FASTFMA-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    retq
+;
+; CHECK-FMA-LABEL: fdiv_pow_shl_cnt_in_bounds2:
+; CHECK-FMA:       # %bb.0:
+; CHECK-FMA-NEXT:    movl $1, %eax
+; CHECK-FMA-NEXT:    shlxl %edi, %eax, %eax
+; CHECK-FMA-NEXT:    movzwl %ax, %eax
+; CHECK-FMA-NEXT:    vcvtsi2ss %eax, %xmm0, %xmm0
+; CHECK-FMA-NEXT:    vcvtps2ph $4, %xmm0, %xmm0
+; CHECK-FMA-NEXT:    vpmovzxwq {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero
+; CHECK-FMA-NEXT:    vcvtph2ps %xmm0, %xmm0
+; CHECK-FMA-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; CHECK-FMA-NEXT:    vdivss %xmm0, %xmm1, %xmm0
+; CHECK-FMA-NEXT:    vcvtps2ph $4, %xmm0, %xmm0
+; CHECK-FMA-NEXT:    vmovd %xmm0, %eax
+; CHECK-FMA-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0
+; CHECK-FMA-NEXT:    retq
+  %shl = shl nuw i16 1, %cnt
+  %conv = uitofp i16 %shl to half
+  %mul = fdiv half 0xH4800, %conv
+  ret half %mul
+}
+
+define half @fdiv_pow_shl_cnt_fail_out_of_bound2(i16 %cnt) nounwind {
+; CHECK-SSE-LABEL: fdiv_pow_shl_cnt_fail_out_of_bound2:
+; CHECK-SSE:       # %bb.0:
+; CHECK-SSE-NEXT:    pushq %rax
+; CHECK-SSE-NEXT:    movl %edi, %ecx
+; CHECK-SSE-NEXT:    movl $1, %eax
+; CHECK-SSE-NEXT:    # kill: def $cl killed $cl killed $ecx
+; CHECK-SSE-NEXT:    shll %cl, %eax
+; CHECK-SSE-NEXT:    movzwl %ax, %eax
+; CHECK-SSE-NEXT:    cvtsi2ss %eax, %xmm0
+; CHECK-SSE-NEXT:    callq __truncsfhf2@PLT
+; CHECK-SSE-NEXT:    callq __extendhfsf2@PLT
+; CHECK-SSE-NEXT:    movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; CHECK-SSE-NEXT:    divss %xmm0, %xmm1
+; CHECK-SSE-NEXT:    movaps %xmm1, %xmm0
+; CHECK-SSE-NEXT:    callq __truncsfhf2@PLT
+; CHECK-SSE-NEXT:    popq %rax
+; CHECK-SSE-NEXT:    retq
+;
+; CHECK-AVX2-LABEL: fdiv_pow_shl_cnt_fail_out_of_bound2:
+; CHECK-AVX2:       # %bb.0:
+; CHECK-AVX2-NEXT:    pushq %rax
+; CHECK-AVX2-NEXT:    movl %edi, %ecx
+; CHECK-AVX2-NEXT:    movl $1, %eax
+; CHECK-AVX2-NEXT:    # kill: def $cl killed $cl killed $ecx
+; CHECK-AVX2-NEXT:    shll %cl, %eax
+; CHECK-AVX2-NEXT:    movzwl %ax, %eax
+; CHECK-AVX2-NEXT:    vcvtsi2ss %eax, %xmm0, %xmm0
+; CHECK-AVX2-NEXT:    callq __truncsfhf2@PLT
+; CHECK-AVX2-NEXT:    callq __extendhfsf2@PLT
+; CHECK-AVX2-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; CHECK-AVX2-NEXT:    vdivss %xmm0, %xmm1, %xmm0
+; CHECK-AVX2-NEXT:    callq __truncsfhf2@PLT
+; CHECK-AVX2-NEXT:    popq %rax
+; CHECK-AVX2-NEXT:    retq
+;
+; CHECK-NO-FASTFMA-LABEL: fdiv_pow_shl_cnt_fail_out_of_bound2:
+; CHECK-NO-FASTFMA:       # %bb.0:
+; CHECK-NO-FASTFMA-NEXT:    movl %edi, %ecx
+; CHECK-NO-FASTFMA-NEXT:    movl $1, %eax
+; CHECK-NO-FASTFMA-NEXT:    # kill: def $cl killed $cl killed $ecx
+; CHECK-NO-FASTFMA-NEXT:    shll %cl, %eax
+; CHECK-NO-FASTFMA-NEXT:    movzwl %ax, %eax
+; CHECK-NO-FASTFMA-NEXT:    vcvtsi2ss %eax, %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vcvtps2ph $4, %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vpmovzxwq {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero
+; CHECK-NO-FASTFMA-NEXT:    vcvtph2ps %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; CHECK-NO-FASTFMA-NEXT:    vdivss %xmm0, %xmm1, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vcvtps2ph $4, %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vmovd %xmm0, %eax
+; CHECK-NO-FASTFMA-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    retq
+;
+; CHECK-FMA-LABEL: fdiv_pow_shl_cnt_fail_out_of_bound2:
+; CHECK-FMA:       # %bb.0:
+; CHECK-FMA-NEXT:    movl $1, %eax
+; CHECK-FMA-NEXT:    shlxl %edi, %eax, %eax
+; CHECK-FMA-NEXT:    movzwl %ax, %eax
+; CHECK-FMA-NEXT:    vcvtsi2ss %eax, %xmm0, %xmm0
+; CHECK-FMA-NEXT:    vcvtps2ph $4, %xmm0, %xmm0
+; CHECK-FMA-NEXT:    vpmovzxwq {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero
+; CHECK-FMA-NEXT:    vcvtph2ps %xmm0, %xmm0
+; CHECK-FMA-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; CHECK-FMA-NEXT:    vdivss %xmm0, %xmm1, %xmm0
+; CHECK-FMA-NEXT:    vcvtps2ph $4, %xmm0, %xmm0
+; CHECK-FMA-NEXT:    vmovd %xmm0, %eax
+; CHECK-FMA-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0
+; CHECK-FMA-NEXT:    retq
+  %shl = shl nuw i16 1, %cnt
+  %conv = uitofp i16 %shl to half
+  %mul = fdiv half 0xH4000, %conv
+  ret half %mul
+}
+
+define double @fdiv_pow_shl_cnt32_to_dbl_okay(i32 %cnt) nounwind {
+; CHECK-SSE-LABEL: fdiv_pow_shl_cnt32_to_dbl_okay:
+; CHECK-SSE:       # %bb.0:
+; CHECK-SSE-NEXT:    movl %edi, %ecx
+; CHECK-SSE-NEXT:    movl $1, %eax
+; CHECK-SSE-NEXT:    # kill: def $cl killed $cl killed $ecx
+; CHECK-SSE-NEXT:    shll %cl, %eax
+; CHECK-SSE-NEXT:    cvtsi2sd %rax, %xmm1
+; CHECK-SSE-NEXT:    movsd {{.*#+}} xmm0 = mem[0],zero
+; CHECK-SSE-NEXT:    divsd %xmm1, %xmm0
+; CHECK-SSE-NEXT:    retq
+;
+; CHECK-AVX2-LABEL: fdiv_pow_shl_cnt32_to_dbl_okay:
+; CHECK-AVX2:       # %bb.0:
+; CHECK-AVX2-NEXT:    movl %edi, %ecx
+; CHECK-AVX2-NEXT:    movl $1, %eax
+; CHECK-AVX2-NEXT:    # kill: def $cl killed $cl killed $ecx
+; CHECK-AVX2-NEXT:    shll %cl, %eax
+; CHECK-AVX2-NEXT:    vcvtsi2sd %rax, %xmm0, %xmm0
+; CHECK-AVX2-NEXT:    vmovsd {{.*#+}} xmm1 = mem[0],zero
+; CHECK-AVX2-NEXT:    vdivsd %xmm0, %xmm1, %xmm0
+; CHECK-AVX2-NEXT:    retq
+;
+; CHECK-NO-FASTFMA-LABEL: fdiv_pow_shl_cnt32_to_dbl_okay:
+; CHECK-NO-FASTFMA:       # %bb.0:
+; CHECK-NO-FASTFMA-NEXT:    movl %edi, %ecx
+; CHECK-NO-FASTFMA-NEXT:    movl $1, %eax
+; CHECK-NO-FASTFMA-NEXT:    # kill: def $cl killed $cl killed $ecx
+; CHECK-NO-FASTFMA-NEXT:    shll %cl, %eax
+; CHECK-NO-FASTFMA-NEXT:    vcvtusi2sd %eax, %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vmovsd {{.*#+}} xmm1 = mem[0],zero
+; CHECK-NO-FASTFMA-NEXT:    vdivsd %xmm0, %xmm1, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    retq
+;
+; CHECK-FMA-LABEL: fdiv_pow_shl_cnt32_to_dbl_okay:
+; CHECK-FMA:       # %bb.0:
+; CHECK-FMA-NEXT:    movl $1, %eax
+; CHECK-FMA-NEXT:    shlxl %edi, %eax, %eax
+; CHECK-FMA-NEXT:    vcvtusi2sd %eax, %xmm0, %xmm0
+; CHECK-FMA-NEXT:    vmovsd {{.*#+}} xmm1 = mem[0],zero
+; CHECK-FMA-NEXT:    vdivsd %xmm0, %xmm1, %xmm0
+; CHECK-FMA-NEXT:    retq
+  %shl = shl nuw i32 1, %cnt
+  %conv = uitofp i32 %shl to double
+  %mul = fdiv double 0x36A0000000000000, %conv
+  ret double %mul
+}
+
+define float @fdiv_pow_shl_cnt32_out_of_bounds2(i32 %cnt) nounwind {
+; CHECK-SSE-LABEL: fdiv_pow_shl_cnt32_out_of_bounds2:
+; CHECK-SSE:       # %bb.0:
+; CHECK-SSE-NEXT:    movl %edi, %ecx
+; CHECK-SSE-NEXT:    movl $1, %eax
+; CHECK-SSE-NEXT:    # kill: def $cl killed $cl killed $ecx
+; CHECK-SSE-NEXT:    shll %cl, %eax
+; CHECK-SSE-NEXT:    cvtsi2ss %rax, %xmm1
+; CHECK-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; CHECK-SSE-NEXT:    divss %xmm1, %xmm0
+; CHECK-SSE-NEXT:    retq
+;
+; CHECK-AVX2-LABEL: fdiv_pow_shl_cnt32_out_of_bounds2:
+; CHECK-AVX2:       # %bb.0:
+; CHECK-AVX2-NEXT:    movl %edi, %ecx
+; CHECK-AVX2-NEXT:    movl $1, %eax
+; CHECK-AVX2-NEXT:    # kill: def $cl killed $cl killed $ecx
+; CHECK-AVX2-NEXT:    shll %cl, %eax
+; CHECK-AVX2-NEXT:    vcvtsi2ss %rax, %xmm0, %xmm0
+; CHECK-AVX2-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; CHECK-AVX2-NEXT:    vdivss %xmm0, %xmm1, %xmm0
+; CHECK-AVX2-NEXT:    retq
+;
+; CHECK-NO-FASTFMA-LABEL: fdiv_pow_shl_cnt32_out_of_bounds2:
+; CHECK-NO-FASTFMA:       # %bb.0:
+; CHECK-NO-FASTFMA-NEXT:    movl %edi, %ecx
+; CHECK-NO-FASTFMA-NEXT:    movl $1, %eax
+; CHECK-NO-FASTFMA-NEXT:    # kill: def $cl killed $cl killed $ecx
+; CHECK-NO-FASTFMA-NEXT:    shll %cl, %eax
+; CHECK-NO-FASTFMA-NEXT:    vcvtusi2ss %eax, %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; CHECK-NO-FASTFMA-NEXT:    vdivss %xmm0, %xmm1, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    retq
+;
+; CHECK-FMA-LABEL: fdiv_pow_shl_cnt32_out_of_bounds2:
+; CHECK-FMA:       # %bb.0:
+; CHECK-FMA-NEXT:    movl $1, %eax
+; CHECK-FMA-NEXT:    shlxl %edi, %eax, %eax
+; CHECK-FMA-NEXT:    vcvtusi2ss %eax, %xmm0, %xmm0
+; CHECK-FMA-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; CHECK-FMA-NEXT:    vdivss %xmm0, %xmm1, %xmm0
+; CHECK-FMA-NEXT:    retq
+  %shl = shl nuw i32 1, %cnt
+  %conv = uitofp i32 %shl to float
+  %mul = fdiv float 0x3a1fffff00000000, %conv
+  ret float %mul
+}
+
+define float @fdiv_pow_shl_cnt32_okay(i32 %cnt) nounwind {
+; CHECK-SSE-LABEL: fdiv_pow_shl_cnt32_okay:
+; CHECK-SSE:       # %bb.0:
+; CHECK-SSE-NEXT:    movl %edi, %ecx
+; CHECK-SSE-NEXT:    movl $1, %eax
+; CHECK-SSE-NEXT:    # kill: def $cl killed $cl killed $ecx
+; CHECK-SSE-NEXT:    shll %cl, %eax
+; CHECK-SSE-NEXT:    cvtsi2ss %rax, %xmm1
+; CHECK-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; CHECK-SSE-NEXT:    divss %xmm1, %xmm0
+; CHECK-SSE-NEXT:    retq
+;
+; CHECK-AVX2-LABEL: fdiv_pow_shl_cnt32_okay:
+; CHECK-AVX2:       # %bb.0:
+; CHECK-AVX2-NEXT:    movl %edi, %ecx
+; CHECK-AVX2-NEXT:    movl $1, %eax
+; CHECK-AVX2-NEXT:    # kill: def $cl killed $cl killed $ecx
+; CHECK-AVX2-NEXT:    shll %cl, %eax
+; CHECK-AVX2-NEXT:    vcvtsi2ss %rax, %xmm0, %xmm0
+; CHECK-AVX2-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; CHECK-AVX2-NEXT:    vdivss %xmm0, %xmm1, %xmm0
+; CHECK-AVX2-NEXT:    retq
+;
+; CHECK-NO-FASTFMA-LABEL: fdiv_pow_shl_cnt32_okay:
+; CHECK-NO-FASTFMA:       # %bb.0:
+; CHECK-NO-FASTFMA-NEXT:    movl %edi, %ecx
+; CHECK-NO-FASTFMA-NEXT:    movl $1, %eax
+; CHECK-NO-FASTFMA-NEXT:    # kill: def $cl killed $cl killed $ecx
+; CHECK-NO-FASTFMA-NEXT:    shll %cl, %eax
+; CHECK-NO-FASTFMA-NEXT:    vcvtusi2ss %eax, %xmm0, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; CHECK-NO-FASTFMA-NEXT:    vdivss %xmm0, %xmm1, %xmm0
+; CHECK-NO-FASTFMA-NEXT:    retq
+;
+; CHECK-FMA-LABEL: fdiv_pow_shl_cnt32_okay:
+; CHECK-FMA:       # %bb.0:
+; CHECK-FMA-NEXT:    movl $1, %eax
+; CHECK-FMA-NEXT:    shlxl %edi, %eax, %eax
+; CHECK-FMA-NEXT:    vcvtusi2ss %eax, %xmm0, %xmm0
+; CHECK-FMA-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; CHECK-FMA-NEXT:    vdivss %xmm0, %xmm1, %xmm0
+; CHECK-FMA-NEXT:    retq
+  %shl = shl nuw i32 1, %cnt
+  %conv = uitofp i32 %shl to float
+  %mul = fdiv float 0x3a20000000000000, %conv
+  ret float %mul
+}