; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=i386-pc-win32 -mcpu=nehalem | FileCheck -check-prefix=WIN32 %s
; RUN: llc < %s -mtriple=x86_64-win32 -mcpu=nehalem | FileCheck -check-prefix=WIN64 %s
; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=nehalem | FileCheck -check-prefix=NOT_WIN %s

declare <16 x float> @func_float16_ptr(<16 x float>, ptr)
declare <16 x float> @func_float16(<16 x float>, <16 x float>)

;test calling conventions - input parameters
define <16 x float> @testf16_inp(<16 x float> %a, <16 x float> %b) nounwind {
; WIN32-LABEL: testf16_inp:
; WIN32:       # %bb.0:
; WIN32-NEXT:    pushl %ebp
; WIN32-NEXT:    movl %esp, %ebp
; WIN32-NEXT:    andl $-16, %esp
; WIN32-NEXT:    subl $80, %esp
; WIN32-NEXT:    movups 72(%ebp), %xmm4
; WIN32-NEXT:    movups 8(%ebp), %xmm3
; WIN32-NEXT:    addps %xmm4, %xmm3
; WIN32-NEXT:    movups 56(%ebp), %xmm4
; WIN32-NEXT:    movups 40(%ebp), %xmm5
; WIN32-NEXT:    movups 24(%ebp), %xmm6
; WIN32-NEXT:    movl %esp, %eax
; WIN32-NEXT:    addps %xmm6, %xmm0
; WIN32-NEXT:    addps %xmm5, %xmm1
; WIN32-NEXT:    addps %xmm4, %xmm2
; WIN32-NEXT:    pushl %eax
; WIN32-NEXT:    calll _func_float16_ptr
; WIN32-NEXT:    addl $4, %esp
; WIN32-NEXT:    addps (%esp), %xmm0
; WIN32-NEXT:    addps {{[0-9]+}}(%esp), %xmm1
; WIN32-NEXT:    addps {{[0-9]+}}(%esp), %xmm2
; WIN32-NEXT:    addps {{[0-9]+}}(%esp), %xmm3
; WIN32-NEXT:    movl %ebp, %esp
; WIN32-NEXT:    popl %ebp
; WIN32-NEXT:    retl
;
; WIN64-LABEL: testf16_inp:
; WIN64:       # %bb.0:
; WIN64-NEXT:    subq $104, %rsp
; WIN64-NEXT:    movaps (%r9), %xmm3
; WIN64-NEXT:    movaps (%r8), %xmm2
; WIN64-NEXT:    movaps (%rdx), %xmm1
; WIN64-NEXT:    movaps (%rcx), %xmm0
; WIN64-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; WIN64-NEXT:    movq {{[0-9]+}}(%rsp), %rcx
; WIN64-NEXT:    movq {{[0-9]+}}(%rsp), %rdx
; WIN64-NEXT:    movq {{[0-9]+}}(%rsp), %r8
; WIN64-NEXT:    addps (%r8), %xmm0
; WIN64-NEXT:    addps (%rdx), %xmm1
; WIN64-NEXT:    addps (%rcx), %xmm2
; WIN64-NEXT:    addps (%rax), %xmm3
; WIN64-NEXT:    leaq {{[0-9]+}}(%rsp), %rcx
; WIN64-NEXT:    callq func_float16_ptr
; WIN64-NEXT:    addps {{[0-9]+}}(%rsp), %xmm0
; WIN64-NEXT:    addps {{[0-9]+}}(%rsp), %xmm1
; WIN64-NEXT:    addps {{[0-9]+}}(%rsp), %xmm2
; WIN64-NEXT:    addps {{[0-9]+}}(%rsp), %xmm3
; WIN64-NEXT:    addq $104, %rsp
; WIN64-NEXT:    retq
;
; NOT_WIN-LABEL: testf16_inp:
; NOT_WIN:       ## %bb.0:
; NOT_WIN-NEXT:    subq $72, %rsp
; NOT_WIN-NEXT:    addps %xmm4, %xmm0
; NOT_WIN-NEXT:    addps %xmm5, %xmm1
; NOT_WIN-NEXT:    addps %xmm6, %xmm2
; NOT_WIN-NEXT:    addps %xmm7, %xmm3
; NOT_WIN-NEXT:    movq %rsp, %rdi
; NOT_WIN-NEXT:    callq _func_float16_ptr
; NOT_WIN-NEXT:    addps (%rsp), %xmm0
; NOT_WIN-NEXT:    addps {{[0-9]+}}(%rsp), %xmm1
; NOT_WIN-NEXT:    addps {{[0-9]+}}(%rsp), %xmm2
; NOT_WIN-NEXT:    addps {{[0-9]+}}(%rsp), %xmm3
; NOT_WIN-NEXT:    addq $72, %rsp
; NOT_WIN-NEXT:    retq
  %y = alloca <16 x float>, align 16
  %x = fadd <16 x float> %a, %b
  %1 = call intel_ocl_bicc <16 x float> @func_float16_ptr(<16 x float> %x, ptr %y)
  %2 = load <16 x float>, ptr %y, align 16
  %3 = fadd <16 x float> %2, %1
  ret <16 x float> %3
}

; test calling conventions - preserved registers

; preserves xmm6-xmm15 on windows, xmm8-xmm15 on other plateforms.
define <16 x float> @testf16_regs(<16 x float> %a, <16 x float> %b) nounwind {
; WIN32-LABEL: testf16_regs:
; WIN32:       # %bb.0:
; WIN32-NEXT:    pushl %ebp
; WIN32-NEXT:    movl %esp, %ebp
; WIN32-NEXT:    andl $-16, %esp
; WIN32-NEXT:    subl $80, %esp
; WIN32-NEXT:    movups 72(%ebp), %xmm6
; WIN32-NEXT:    movups 8(%ebp), %xmm3
; WIN32-NEXT:    movups 56(%ebp), %xmm7
; WIN32-NEXT:    movups 40(%ebp), %xmm5
; WIN32-NEXT:    movups 24(%ebp), %xmm4
; WIN32-NEXT:    movl %esp, %eax
; WIN32-NEXT:    addps %xmm4, %xmm0
; WIN32-NEXT:    addps %xmm5, %xmm1
; WIN32-NEXT:    addps %xmm7, %xmm2
; WIN32-NEXT:    addps %xmm6, %xmm3
; WIN32-NEXT:    pushl %eax
; WIN32-NEXT:    calll _func_float16_ptr
; WIN32-NEXT:    addl $4, %esp
; WIN32-NEXT:    movups 72(%ebp), %xmm4
; WIN32-NEXT:    addps %xmm4, %xmm3
; WIN32-NEXT:    movups 56(%ebp), %xmm4
; WIN32-NEXT:    addps %xmm4, %xmm2
; WIN32-NEXT:    movups 40(%ebp), %xmm4
; WIN32-NEXT:    addps %xmm4, %xmm1
; WIN32-NEXT:    movups 24(%ebp), %xmm4
; WIN32-NEXT:    addps %xmm4, %xmm0
; WIN32-NEXT:    addps (%esp), %xmm0
; WIN32-NEXT:    addps {{[0-9]+}}(%esp), %xmm1
; WIN32-NEXT:    addps {{[0-9]+}}(%esp), %xmm2
; WIN32-NEXT:    addps {{[0-9]+}}(%esp), %xmm3
; WIN32-NEXT:    movl %ebp, %esp
; WIN32-NEXT:    popl %ebp
; WIN32-NEXT:    retl
;
; WIN64-LABEL: testf16_regs:
; WIN64:       # %bb.0:
; WIN64-NEXT:    subq $168, %rsp
; WIN64-NEXT:    movaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; WIN64-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; WIN64-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; WIN64-NEXT:    movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; WIN64-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; WIN64-NEXT:    movaps (%rax), %xmm6
; WIN64-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; WIN64-NEXT:    movaps (%rax), %xmm7
; WIN64-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; WIN64-NEXT:    movaps (%rax), %xmm8
; WIN64-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; WIN64-NEXT:    movaps (%rax), %xmm9
; WIN64-NEXT:    movaps (%rcx), %xmm0
; WIN64-NEXT:    addps %xmm9, %xmm0
; WIN64-NEXT:    movaps (%rdx), %xmm1
; WIN64-NEXT:    addps %xmm8, %xmm1
; WIN64-NEXT:    movaps (%r8), %xmm2
; WIN64-NEXT:    addps %xmm7, %xmm2
; WIN64-NEXT:    movaps (%r9), %xmm3
; WIN64-NEXT:    addps %xmm6, %xmm3
; WIN64-NEXT:    leaq {{[0-9]+}}(%rsp), %rcx
; WIN64-NEXT:    callq func_float16_ptr
; WIN64-NEXT:    addps %xmm6, %xmm3
; WIN64-NEXT:    addps %xmm7, %xmm2
; WIN64-NEXT:    addps %xmm8, %xmm1
; WIN64-NEXT:    addps %xmm9, %xmm0
; WIN64-NEXT:    addps {{[0-9]+}}(%rsp), %xmm0
; WIN64-NEXT:    addps {{[0-9]+}}(%rsp), %xmm1
; WIN64-NEXT:    addps {{[0-9]+}}(%rsp), %xmm2
; WIN64-NEXT:    addps {{[0-9]+}}(%rsp), %xmm3
; WIN64-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
; WIN64-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
; WIN64-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; WIN64-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
; WIN64-NEXT:    addq $168, %rsp
; WIN64-NEXT:    retq
;
; NOT_WIN-LABEL: testf16_regs:
; NOT_WIN:       ## %bb.0:
; NOT_WIN-NEXT:    subq $72, %rsp
; NOT_WIN-NEXT:    movaps %xmm7, %xmm9
; NOT_WIN-NEXT:    movaps %xmm6, %xmm10
; NOT_WIN-NEXT:    movaps %xmm5, %xmm11
; NOT_WIN-NEXT:    movaps %xmm4, %xmm8
; NOT_WIN-NEXT:    addps %xmm4, %xmm0
; NOT_WIN-NEXT:    addps %xmm5, %xmm1
; NOT_WIN-NEXT:    addps %xmm6, %xmm2
; NOT_WIN-NEXT:    addps %xmm7, %xmm3
; NOT_WIN-NEXT:    movq %rsp, %rdi
; NOT_WIN-NEXT:    callq _func_float16_ptr
; NOT_WIN-NEXT:    addps %xmm9, %xmm3
; NOT_WIN-NEXT:    addps %xmm10, %xmm2
; NOT_WIN-NEXT:    addps %xmm11, %xmm1
; NOT_WIN-NEXT:    addps %xmm8, %xmm0
; NOT_WIN-NEXT:    addps (%rsp), %xmm0
; NOT_WIN-NEXT:    addps {{[0-9]+}}(%rsp), %xmm1
; NOT_WIN-NEXT:    addps {{[0-9]+}}(%rsp), %xmm2
; NOT_WIN-NEXT:    addps {{[0-9]+}}(%rsp), %xmm3
; NOT_WIN-NEXT:    addq $72, %rsp
; NOT_WIN-NEXT:    retq
  %y = alloca <16 x float>, align 16
  %x = fadd <16 x float> %a, %b
  %1 = call intel_ocl_bicc <16 x float> @func_float16_ptr(<16 x float> %x, ptr %y)
  %2 = load <16 x float>, ptr %y, align 16
  %3 = fadd <16 x float> %1, %b
  %4 = fadd <16 x float> %2, %3
  ret <16 x float> %4
}

; test calling conventions - prolog and epilog
define intel_ocl_bicc <16 x float> @test_prolog_epilog(<16 x float> %a, <16 x float> %b) nounwind {
; WIN32-LABEL: test_prolog_epilog:
; WIN32:       # %bb.0:
; WIN32-NEXT:    pushl %ebp
; WIN32-NEXT:    movl %esp, %ebp
; WIN32-NEXT:    andl $-16, %esp
; WIN32-NEXT:    subl $96, %esp
; WIN32-NEXT:    movups 8(%ebp), %xmm4
; WIN32-NEXT:    movups 24(%ebp), %xmm5
; WIN32-NEXT:    movups 40(%ebp), %xmm6
; WIN32-NEXT:    movups 56(%ebp), %xmm7
; WIN32-NEXT:    movups %xmm7, {{[0-9]+}}(%esp)
; WIN32-NEXT:    movups %xmm6, {{[0-9]+}}(%esp)
; WIN32-NEXT:    movups %xmm5, {{[0-9]+}}(%esp)
; WIN32-NEXT:    movups %xmm4, {{[0-9]+}}(%esp)
; WIN32-NEXT:    movups %xmm3, (%esp)
; WIN32-NEXT:    calll _func_float16
; WIN32-NEXT:    movl %ebp, %esp
; WIN32-NEXT:    popl %ebp
; WIN32-NEXT:    retl
;
; WIN64-LABEL: test_prolog_epilog:
; WIN64:       # %bb.0:
; WIN64-NEXT:    subq $232, %rsp
; WIN64-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; WIN64-NEXT:    movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; WIN64-NEXT:    movaps (%r9), %xmm4
; WIN64-NEXT:    movaps (%rdx), %xmm5
; WIN64-NEXT:    movaps (%r8), %xmm6
; WIN64-NEXT:    movaps (%rcx), %xmm7
; WIN64-NEXT:    movaps %xmm1, {{[0-9]+}}(%rsp)
; WIN64-NEXT:    movaps %xmm0, {{[0-9]+}}(%rsp)
; WIN64-NEXT:    movaps %xmm2, {{[0-9]+}}(%rsp)
; WIN64-NEXT:    movaps %xmm3, {{[0-9]+}}(%rsp)
; WIN64-NEXT:    movaps %xmm7, {{[0-9]+}}(%rsp)
; WIN64-NEXT:    movaps %xmm6, {{[0-9]+}}(%rsp)
; WIN64-NEXT:    movaps %xmm5, {{[0-9]+}}(%rsp)
; WIN64-NEXT:    leaq {{[0-9]+}}(%rsp), %rax
; WIN64-NEXT:    movq %rax, {{[0-9]+}}(%rsp)
; WIN64-NEXT:    leaq {{[0-9]+}}(%rsp), %rax
; WIN64-NEXT:    movq %rax, {{[0-9]+}}(%rsp)
; WIN64-NEXT:    movaps %xmm4, {{[0-9]+}}(%rsp)
; WIN64-NEXT:    leaq {{[0-9]+}}(%rsp), %rax
; WIN64-NEXT:    movq %rax, {{[0-9]+}}(%rsp)
; WIN64-NEXT:    leaq {{[0-9]+}}(%rsp), %rax
; WIN64-NEXT:    movq %rax, {{[0-9]+}}(%rsp)
; WIN64-NEXT:    leaq {{[0-9]+}}(%rsp), %rcx
; WIN64-NEXT:    leaq {{[0-9]+}}(%rsp), %rdx
; WIN64-NEXT:    leaq {{[0-9]+}}(%rsp), %r8
; WIN64-NEXT:    leaq {{[0-9]+}}(%rsp), %r9
; WIN64-NEXT:    callq func_float16
; WIN64-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
; WIN64-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
; WIN64-NEXT:    addq $232, %rsp
; WIN64-NEXT:    retq
;
; NOT_WIN-LABEL: test_prolog_epilog:
; NOT_WIN:       ## %bb.0:
; NOT_WIN-NEXT:    subq $136, %rsp
; NOT_WIN-NEXT:    movaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
; NOT_WIN-NEXT:    movaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
; NOT_WIN-NEXT:    movaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
; NOT_WIN-NEXT:    movaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
; NOT_WIN-NEXT:    movaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
; NOT_WIN-NEXT:    movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
; NOT_WIN-NEXT:    movaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
; NOT_WIN-NEXT:    movaps %xmm8, (%rsp) ## 16-byte Spill
; NOT_WIN-NEXT:    callq _func_float16
; NOT_WIN-NEXT:    movaps (%rsp), %xmm8 ## 16-byte Reload
; NOT_WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 ## 16-byte Reload
; NOT_WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 ## 16-byte Reload
; NOT_WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 ## 16-byte Reload
; NOT_WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 ## 16-byte Reload
; NOT_WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 ## 16-byte Reload
; NOT_WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 ## 16-byte Reload
; NOT_WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 ## 16-byte Reload
; NOT_WIN-NEXT:    addq $136, %rsp
; NOT_WIN-NEXT:    retq
   %c = call <16 x float> @func_float16(<16 x float> %a, <16 x float> %b)
   ret <16 x float> %c
}