1 | // REQUIRES: arm-registered-target |
2 | // REQUIRES: aarch64-registered-target |
3 | // RUN: %clang_cc1 -triple thumbv7-apple-darwin9 \ |
4 | // RUN: -target-abi aapcs \ |
5 | // RUN: -target-cpu cortex-a8 \ |
6 | // RUN: -mfloat-abi hard \ |
7 | // RUN: -ffreestanding \ |
8 | // RUN: -emit-llvm -w -o - %s | FileCheck %s |
9 | |
10 | // RUN: %clang_cc1 -triple armv7-unknown-nacl-gnueabi \ |
11 | // RUN: -target-cpu cortex-a8 \ |
12 | // RUN: -mfloat-abi hard \ |
13 | // RUN: -ffreestanding \ |
14 | // RUN: -emit-llvm -w -o - %s | FileCheck %s |
15 | |
16 | // RUN: %clang_cc1 -triple arm64-apple-darwin9 -target-feature +neon \ |
17 | // RUN: -ffreestanding \ |
18 | // RUN: -emit-llvm -w -o - %s | FileCheck -check-prefix=CHECK64 %s |
19 | |
20 | #include <arm_neon.h> |
21 | |
22 | struct homogeneous_struct { |
23 | float f[2]; |
24 | float f3; |
25 | float f4; |
26 | }; |
27 | // CHECK: define arm_aapcs_vfpcc %struct.homogeneous_struct @test_struct(%struct.homogeneous_struct %{{.*}}) |
28 | // CHECK64: define %struct.homogeneous_struct @test_struct([4 x float] %{{.*}}) |
29 | extern struct homogeneous_struct struct_callee(struct homogeneous_struct); |
30 | struct homogeneous_struct test_struct(struct homogeneous_struct arg) { |
31 | return struct_callee(arg); |
32 | } |
33 | |
34 | // CHECK: define arm_aapcs_vfpcc void @test_struct_variadic(%struct.homogeneous_struct* {{.*}}, ...) |
35 | struct homogeneous_struct test_struct_variadic(struct homogeneous_struct arg, ...) { |
36 | return struct_callee(arg); |
37 | } |
38 | |
39 | struct nested_array { |
40 | double d[4]; |
41 | }; |
42 | // CHECK: define arm_aapcs_vfpcc void @test_array(%struct.nested_array %{{.*}}) |
43 | // CHECK64: define void @test_array([4 x double] %{{.*}}) |
44 | extern void array_callee(struct nested_array); |
45 | void test_array(struct nested_array arg) { |
46 | array_callee(arg); |
47 | } |
48 | |
49 | extern void complex_callee(__complex__ double); |
50 | // CHECK: define arm_aapcs_vfpcc void @test_complex({ double, double } %{{.*}}) |
51 | // CHECK64: define void @test_complex([2 x double] %cd.coerce) |
52 | void test_complex(__complex__ double cd) { |
53 | complex_callee(cd); |
54 | } |
55 | |
56 | // Long double is the same as double on AAPCS, it should be homogeneous. |
57 | extern void complex_long_callee(__complex__ long double); |
58 | // CHECK: define arm_aapcs_vfpcc void @test_complex_long({ double, double } %{{.*}}) |
59 | void test_complex_long(__complex__ long double cd) { |
60 | complex_callee(cd); |
61 | } |
62 | |
63 | // Structs with more than 4 elements of the base type are not treated |
64 | // as homogeneous aggregates. Test that. |
65 | |
66 | struct big_struct { |
67 | float f1; |
68 | float f[2]; |
69 | float f3; |
70 | float f4; |
71 | }; |
72 | // CHECK: define arm_aapcs_vfpcc void @test_big([5 x i32] %{{.*}}) |
73 | // CHECK64: define void @test_big(%struct.big_struct* %{{.*}}) |
74 | // CHECK64: call void @llvm.memcpy |
75 | // CHECK64: call void @big_callee(%struct.big_struct* |
76 | extern void big_callee(struct big_struct); |
77 | void test_big(struct big_struct arg) { |
78 | big_callee(arg); |
79 | } |
80 | |
81 | // Make sure that aggregates with multiple base types are not treated as |
82 | // homogeneous aggregates. |
83 | |
84 | struct heterogeneous_struct { |
85 | float f1; |
86 | int i2; |
87 | }; |
88 | // CHECK: define arm_aapcs_vfpcc void @test_hetero([2 x i32] %{{.*}}) |
89 | // CHECK64: define void @test_hetero(i64 %{{.*}}) |
90 | extern void hetero_callee(struct heterogeneous_struct); |
91 | void test_hetero(struct heterogeneous_struct arg) { |
92 | hetero_callee(arg); |
93 | } |
94 | |
95 | // Neon multi-vector types are homogeneous aggregates. |
96 | // CHECK: define arm_aapcs_vfpcc <16 x i8> @f0(%struct.int8x16x4_t %{{.*}}) |
97 | // CHECK64: define <16 x i8> @f0([4 x <16 x i8>] %{{.*}}) |
98 | int8x16_t f0(int8x16x4_t v4) { |
99 | return vaddq_s8(v4.val[0], v4.val[3]); |
100 | } |
101 | |
102 | // ...and it doesn't matter whether the vectors are exactly the same, as long |
103 | // as they have the same size. |
104 | |
105 | struct neon_struct { |
106 | int8x8x2_t v12; |
107 | int32x2_t v3; |
108 | int16x4_t v4; |
109 | }; |
110 | // CHECK: define arm_aapcs_vfpcc void @test_neon(%struct.neon_struct %{{.*}}) |
111 | // CHECK64: define void @test_neon([4 x <8 x i8>] %{{.*}}) |
112 | extern void neon_callee(struct neon_struct); |
113 | void test_neon(struct neon_struct arg) { |
114 | neon_callee(arg); |
115 | } |
116 | |
117 | // CHECK-LABEL: define arm_aapcs_vfpcc void @f33(%struct.s33* byval align 4 %s) |
118 | struct s33 { char buf[32*32]; }; |
119 | void f33(struct s33 s) { } |
120 | |
121 | typedef struct { long long x; int y; } struct_long_long_int; |
122 | // CHECK: define arm_aapcs_vfpcc void @test_vfp_stack_gpr_split_1(double %a, double %b, double %c, double %d, double %e, double %f, double %g, double %h, double %i, i32 %j, i64 %k, i32 %l) |
123 | void test_vfp_stack_gpr_split_1(double a, double b, double c, double d, double e, double f, double g, double h, double i, int j, long long k, int l) {} |
124 | |
125 | // CHECK: define arm_aapcs_vfpcc void @test_vfp_stack_gpr_split_2(double %a, double %b, double %c, double %d, double %e, double %f, double %g, double %h, double %i, i32 %j, [2 x i64] %k.coerce) |
126 | void test_vfp_stack_gpr_split_2(double a, double b, double c, double d, double e, double f, double g, double h, double i, int j, struct_long_long_int k) {} |
127 | |
128 | // CHECK: define arm_aapcs_vfpcc void @test_vfp_stack_gpr_split_3(%struct.struct_long_long_int* noalias sret %agg.result, double %a, double %b, double %c, double %d, double %e, double %f, double %g, double %h, double %i, [2 x i64] %k.coerce) |
129 | struct_long_long_int test_vfp_stack_gpr_split_3(double a, double b, double c, double d, double e, double f, double g, double h, double i, struct_long_long_int k) {} |
130 | |
131 | typedef struct { int a; int b:4; int c; } struct_int_bitfield_int; |
132 | // CHECK: define arm_aapcs_vfpcc void @test_test_vfp_stack_gpr_split_bitfield(double %a, double %b, double %c, double %d, double %e, double %f, double %g, double %h, double %i, i32 %j, i32 %k, [3 x i32] %l.coerce) |
133 | void test_test_vfp_stack_gpr_split_bitfield(double a, double b, double c, double d, double e, double f, double g, double h, double i, int j, int k, struct_int_bitfield_int l) {} |
134 | |
135 | // Note: this struct requires internal padding |
136 | typedef struct { int x; long long y; } struct_int_long_long; |
137 | // CHECK: define arm_aapcs_vfpcc void @test_vfp_stack_gpr_split_4(double %a, double %b, double %c, double %d, double %e, double %f, double %g, double %h, double %i, i32 %j, [2 x i64] %k.coerce) |
138 | void test_vfp_stack_gpr_split_4(double a, double b, double c, double d, double e, double f, double g, double h, double i, int j, struct_int_long_long k) {} |
139 | |
140 | // This very large struct (passed byval) uses up the GPRs, so no padding is needed |
141 | typedef struct { int x[17]; } struct_seventeen_ints; |
142 | typedef struct { int x[4]; } struct_four_ints; |
143 | // CHECK: define arm_aapcs_vfpcc void @test_vfp_stack_gpr_split_5(%struct.struct_seventeen_ints* byval align 4 %a, double %b, double %c, double %d, double %e, double %f, double %g, double %h, double %i, double %j, [4 x i32] %k.coerce) |
144 | void test_vfp_stack_gpr_split_5(struct_seventeen_ints a, double b, double c, double d, double e, double f, double g, double h, double i, double j, struct_four_ints k) {} |
145 | |
146 | // Here, parameter k would need padding to prevent it from being split, but it |
147 | // is passed ByVal (due to being > 64 bytes), so the backend handles this instead. |
148 | void test_vfp_stack_gpr_split_6(double a, double b, double c, double d, double e, double f, double g, double h, double i, int j, struct_seventeen_ints k) {} |
149 | // CHECK: define arm_aapcs_vfpcc void @test_vfp_stack_gpr_split_6(double %a, double %b, double %c, double %d, double %e, double %f, double %g, double %h, double %i, i32 %j, %struct.struct_seventeen_ints* byval align 4 %k) |
150 | |