1; RUN: llc < %s -mtriple=i686-apple-darwin -mattr=+avx | FileCheck -check-prefix=X86 %s
2; RUN: llc < %s -mtriple=i386-pc-win32 -mattr=+avx | FileCheck -check-prefix=X86 %s
3; RUN: llc < %s -mtriple=x86_64-win32 -mattr=+avx | FileCheck -check-prefix=WIN64 %s
4; RUN: llc < %s -mtriple=x86_64-apple-darwin -mattr=+avx | FileCheck -check-prefix=X64 %s
5
6declare <16 x float> @func_float16_ptr(<16 x float>, <16 x float> *)
7declare <16 x float> @func_float16(<16 x float>, <16 x float>)
8declare i32 @func_int(i32, i32)
9
10; WIN64-LABEL: testf16_inp
11; WIN64: vaddps  {{.*}}, {{%ymm[0-1]}}
12; WIN64: vaddps  {{.*}}, {{%ymm[0-1]}}
13; WIN64: leaq    {{.*}}(%rsp), %rcx
14; WIN64: call
15; WIN64: ret
16
17; X86-LABEL: testf16_inp
18; X86: vaddps  {{.*}}, {{%ymm[0-1]}}
19; X86: vaddps  {{.*}}, {{%ymm[0-1]}}
20; Push is not deemed profitable if we're realigning the stack.
21; X86: {{pushl|movl}}   %eax
22; X86: call
23; X86: ret
24
25; X64-LABEL: testf16_inp
26; X64: vaddps  {{.*}}, {{%ymm[0-1]}}
27; X64: vaddps  {{.*}}, {{%ymm[0-1]}}
28; X64: movq    %rsp, %rdi
29; X64: call
30; X64: ret
31
32;test calling conventions - input parameters
33define <16 x float> @testf16_inp(<16 x float> %a, <16 x float> %b) nounwind {
34  %y = alloca <16 x float>, align 16
35  %x = fadd <16 x float> %a, %b
36  %1 = call intel_ocl_bicc <16 x float> @func_float16_ptr(<16 x float> %x, <16 x float>* %y)
37  %2 = load <16 x float>, <16 x float>* %y, align 16
38  %3 = fadd <16 x float> %2, %1
39  ret <16 x float> %3
40}
41
42;test calling conventions - preserved registers
43
44; preserved ymm6-ymm15
45; WIN64-LABEL: testf16_regs
46; WIN64: call
47; WIN64: vaddps  {{%ymm[6-7]}}, {{%ymm[0-1]}}, {{%ymm[0-1]}}
48; WIN64: vaddps  {{%ymm[6-7]}}, {{%ymm[0-1]}}, {{%ymm[0-1]}}
49; WIN64: ret
50
51; preserved ymm8-ymm15
52; X64-LABEL: testf16_regs
53; X64: call
54; X64: vaddps  {{%ymm[0-1]}}, {{%ymm[8-9]}}, {{%ymm[0-1]}}
55; X64: vaddps  {{%ymm[0-1]}}, {{%ymm[8-9]}}, {{%ymm[0-1]}}
56; X64: ret
57
58define <16 x float> @testf16_regs(<16 x float> %a, <16 x float> %b) nounwind {
59  %y = alloca <16 x float>, align 16
60  %x = fadd <16 x float> %a, %b
61  %1 = call intel_ocl_bicc <16 x float> @func_float16_ptr(<16 x float> %x, <16 x float>* %y)
62  %2 = load <16 x float>, <16 x float>* %y, align 16
63  %3 = fadd <16 x float> %1, %b
64  %4 = fadd <16 x float> %2, %3
65  ret <16 x float> %4
66}
67
68; test calling conventions - prolog and epilog
69; WIN64-LABEL: test_prolog_epilog
70; WIN64: vmovups {{%ymm([6-9]|1[0-5])}}, {{.*(%rbp).*}}     # 32-byte Spill
71; WIN64: vmovups {{%ymm([6-9]|1[0-5])}}, {{.*(%rbp).*}}     # 32-byte Spill
72; WIN64: vmovups {{%ymm([6-9]|1[0-5])}}, {{.*(%rbp).*}}     # 32-byte Spill
73; WIN64: vmovups {{%ymm([6-9]|1[0-5])}}, {{.*(%rbp).*}}     # 32-byte Spill
74; WIN64: vmovups {{%ymm([6-9]|1[0-5])}}, {{.*(%rbp).*}}     # 32-byte Spill
75; WIN64: vmovups {{%ymm([6-9]|1[0-5])}}, {{.*(%rbp).*}}     # 32-byte Spill
76; WIN64: vmovups {{%ymm([6-9]|1[0-5])}}, {{.*(%rbp).*}}     # 32-byte Spill
77; WIN64: vmovups {{%ymm([6-9]|1[0-5])}}, {{.*(%rbp).*}}     # 32-byte Spill
78; WIN64: vmovups {{%ymm([6-9]|1[0-5])}}, {{.*(%rbp).*}}     # 32-byte Spill
79; WIN64: vmovups {{%ymm([6-9]|1[0-5])}}, {{.*(%rbp).*}}     # 32-byte Spill
80; WIN64: call
81; WIN64: vmovups {{.*(%rbp).*}}, {{%ymm([6-9]|1[0-5])}}     # 32-byte Reload
82; WIN64: vmovups {{.*(%rbp).*}}, {{%ymm([6-9]|1[0-5])}}     # 32-byte Reload
83; WIN64: vmovups {{.*(%rbp).*}}, {{%ymm([6-9]|1[0-5])}}     # 32-byte Reload
84; WIN64: vmovups {{.*(%rbp).*}}, {{%ymm([6-9]|1[0-5])}}     # 32-byte Reload
85; WIN64: vmovups {{.*(%rbp).*}}, {{%ymm([6-9]|1[0-5])}}     # 32-byte Reload
86; WIN64: vmovups {{.*(%rbp).*}}, {{%ymm([6-9]|1[0-5])}}     # 32-byte Reload
87; WIN64: vmovups {{.*(%rbp).*}}, {{%ymm([6-9]|1[0-5])}}     # 32-byte Reload
88; WIN64: vmovups {{.*(%rbp).*}}, {{%ymm([6-9]|1[0-5])}}     # 32-byte Reload
89; WIN64: vmovups {{.*(%rbp).*}}, {{%ymm([6-9]|1[0-5])}}     # 32-byte Reload
90; WIN64: vmovups {{.*(%rbp).*}}, {{%ymm([6-9]|1[0-5])}}     # 32-byte Reload
91
92; X64-LABEL: test_prolog_epilog
93; X64: vmovups {{%ymm([8-9]|1[0-5])}}, {{.*}}(%rsp)  ## 32-byte Spill
94; X64: vmovups {{%ymm([8-9]|1[0-5])}}, {{.*}}(%rsp)  ## 32-byte Spill
95; X64: vmovups {{%ymm([8-9]|1[0-5])}}, {{.*}}(%rsp)  ## 32-byte Spill
96; X64: vmovups {{%ymm([8-9]|1[0-5])}}, {{.*}}(%rsp)  ## 32-byte Spill
97; X64: vmovups {{%ymm([8-9]|1[0-5])}}, {{.*}}(%rsp)  ## 32-byte Spill
98; X64: vmovups {{%ymm([8-9]|1[0-5])}}, {{.*}}(%rsp)  ## 32-byte Spill
99; X64: vmovups {{%ymm([8-9]|1[0-5])}}, {{.*}}(%rsp)  ## 32-byte Spill
100; X64: vmovups {{%ymm([8-9]|1[0-5])}}, {{.*}}(%rsp)  ## 32-byte Spill
101; X64: call
102; X64: vmovups {{.*}}(%rsp), {{%ymm([8-9]|1[0-5])}} ## 32-byte Reload
103; X64: vmovups {{.*}}(%rsp), {{%ymm([8-9]|1[0-5])}} ## 32-byte Reload
104; X64: vmovups {{.*}}(%rsp), {{%ymm([8-9]|1[0-5])}} ## 32-byte Reload
105; X64: vmovups {{.*}}(%rsp), {{%ymm([8-9]|1[0-5])}} ## 32-byte Reload
106; X64: vmovups {{.*}}(%rsp), {{%ymm([8-9]|1[0-5])}} ## 32-byte Reload
107; X64: vmovups {{.*}}(%rsp), {{%ymm([8-9]|1[0-5])}} ## 32-byte Reload
108; X64: vmovups {{.*}}(%rsp), {{%ymm([8-9]|1[0-5])}} ## 32-byte Reload
109; X64: vmovups {{.*}}(%rsp), {{%ymm([8-9]|1[0-5])}} ## 32-byte Reload
110define intel_ocl_bicc <16 x float> @test_prolog_epilog(<16 x float> %a, <16 x float> %b) nounwind {
111   %c = call <16 x float> @func_float16(<16 x float> %a, <16 x float> %b)
112   ret <16 x float> %c
113}
114
115; test functions with integer parameters
116; pass parameters on stack for 32-bit platform
117; X86-LABEL: test_int
118; X86: pushl {{.*}}
119; X86: pushl {{.*}}
120; X86: call
121; X86: addl {{.*}}, %eax
122
123; pass parameters in registers for 64-bit platform
124; X64-LABEL: test_int
125; X64: movl {{.*}}, %esi
126; X64: leal {{.*}}, %edi
127; X64: call
128; X64: addl {{.*}}, %eax
129define i32 @test_int(i32 %a, i32 %b) nounwind {
130    %c1 = add i32 %a, %b
131	%c2 = call intel_ocl_bicc i32 @func_int(i32 %c1, i32 %a)
132    %c = add i32 %c2, %b
133	ret i32 %c
134}
135
136; WIN64-LABEL: test_float4
137; WIN64-NOT: vzeroupper
138; WIN64: call
139; WIN64-NOT: vzeroupper
140; WIN64: call
141; WIN64: ret
142
143; X64-LABEL: test_float4
144; X64-NOT: vzeroupper
145; X64: call
146; X64-NOT: vzeroupper
147; X64: call
148; X64: ret
149
150; X86-LABEL: test_float4
151; X86: vzeroupper
152; X86: call
153; X86: vzeroupper
154; X86: call
155; X86: ret
156
157declare <4 x float> @func_float4(<4 x float>, <4 x float>, <4 x float>)
158
159define <8 x float> @test_float4(<8 x float> %a, <8 x float> %b, <8 x float> %c) nounwind readnone {
160entry:
161  %0 = shufflevector <8 x float> %a, <8 x float> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
162  %1 = shufflevector <8 x float> %b, <8 x float> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
163  %2 = shufflevector <8 x float> %c, <8 x float> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
164  %call.i = tail call intel_ocl_bicc <4 x float> @func_float4(<4 x float> %0, <4 x float> %1, <4 x float> %2) nounwind
165  %3 = shufflevector <4 x float> %call.i, <4 x float> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef>
166  %4 = shufflevector <8 x float> %a, <8 x float> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
167  %5 = shufflevector <8 x float> %b, <8 x float> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
168  %6 = shufflevector <8 x float> %c, <8 x float> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
169  %call.i2 = tail call intel_ocl_bicc <4 x float> @func_float4(<4 x float> %4, <4 x float> %5, <4 x float> %6) nounwind
170  %7 = shufflevector <4 x float> %call.i2, <4 x float> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef>
171  %8 = shufflevector <8 x float> %3, <8 x float> %7, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 8, i32 9, i32 10, i32 11>
172  ret <8 x float> %8
173}
174