1 // RUN: %clang_cc1 -target-feature +altivec -target-feature +vsx \
2 // RUN:   -faltivec-src-compat=mixed -triple powerpc-unknown-unknown -S -emit-llvm %s -o - | FileCheck %s
3 // RUN: not %clang_cc1 -target-feature +altivec -target-feature +vsx \
4 // RUN:   -faltivec-src-compat=gcc -triple powerpc-unknown-unknown -S -emit-llvm %s -o - 2>&1 | FileCheck %s --check-prefix=ERROR
5 // RUN: %clang_cc1 -target-feature +altivec -target-feature +vsx \
6 // RUN:   -faltivec-src-compat=xl -triple powerpc-unknown-unknown -S -emit-llvm %s -o - | FileCheck %s
7 // RUN: %clang -mcpu=pwr8 -faltivec-src-compat=xl --target=powerpc-unknown-unknown -S -emit-llvm %s -o - | FileCheck %s
8 // RUN: %clang -mcpu=pwr9 -faltivec-src-compat=xl --target=powerpc-unknown-unknown -S -emit-llvm %s -o - | FileCheck %s
9 
10 // CHECK-LABEL: @ui8(
11 // CHECK:         [[A_ADDR:%.*]] = alloca <16 x i8>, align 16
12 // CHECK-NEXT:    [[B_ADDR:%.*]] = alloca <16 x i8>, align 16
13 // CHECK-NEXT:    store <16 x i8> [[A:%.*]], <16 x i8>* [[A_ADDR]], align 16
14 // CHECK-NEXT:    store <16 x i8> [[B:%.*]], <16 x i8>* [[B_ADDR]], align 16
15 // CHECK-NEXT:    [[TMP0:%.*]] = load <16 x i8>, <16 x i8>* [[A_ADDR]], align 16
16 // CHECK-NEXT:    [[TMP1:%.*]] = load <16 x i8>, <16 x i8>* [[B_ADDR]], align 16
17 // CHECK-NEXT:    [[TMP2:%.*]] = call i32 @llvm.ppc.altivec.vcmpequb.p(i32 2, <16 x i8> [[TMP0]], <16 x i8> [[TMP1]])
18 // CHECK-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0
19 // CHECK-NEXT:    [[TMP3:%.*]] = zext i1 [[TOBOOL]] to i64
20 // CHECK-NEXT:    [[COND:%.*]] = select i1 [[TOBOOL]], i32 3, i32 7
21 // CHECK-NEXT:    ret i32 [[COND]]
22 //
23 // ERROR: error: used type '__attribute__((__vector_size__(16 * sizeof(char)))) char' (vector of 16 'char' values) where arithmetic or pointer type is required
24 int ui8(vector unsigned char a, vector unsigned char b) {
25   return a == b ? 3 : 7;
26 }
27 
28 // CHECK-LABEL: @si8(
29 // CHECK:         [[A_ADDR:%.*]] = alloca <16 x i8>, align 16
30 // CHECK-NEXT:    [[B_ADDR:%.*]] = alloca <16 x i8>, align 16
31 // CHECK-NEXT:    store <16 x i8> [[A:%.*]], <16 x i8>* [[A_ADDR]], align 16
32 // CHECK-NEXT:    store <16 x i8> [[B:%.*]], <16 x i8>* [[B_ADDR]], align 16
33 // CHECK-NEXT:    [[TMP0:%.*]] = load <16 x i8>, <16 x i8>* [[A_ADDR]], align 16
34 // CHECK-NEXT:    [[TMP1:%.*]] = load <16 x i8>, <16 x i8>* [[B_ADDR]], align 16
35 // CHECK-NEXT:    [[TMP2:%.*]] = call i32 @llvm.ppc.altivec.vcmpequb.p(i32 2, <16 x i8> [[TMP0]], <16 x i8> [[TMP1]])
36 // CHECK-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0
37 // CHECK-NEXT:    [[TMP3:%.*]] = zext i1 [[TOBOOL]] to i64
38 // CHECK-NEXT:    [[COND:%.*]] = select i1 [[TOBOOL]], i32 3, i32 7
39 // CHECK-NEXT:    ret i32 [[COND]]
40 //
41 // ERROR: error: used type '__attribute__((__vector_size__(16 * sizeof(char)))) char' (vector of 16 'char' values) where arithmetic or pointer type is required
42 int si8(vector signed char a, vector signed char b) {
43   return a == b ? 3 : 7;
44 }
45 
46 // CHECK-LABEL: @ui16(
47 // CHECK:         [[A_ADDR:%.*]] = alloca <8 x i16>, align 16
48 // CHECK-NEXT:    [[B_ADDR:%.*]] = alloca <8 x i16>, align 16
49 // CHECK-NEXT:    store <8 x i16> [[A:%.*]], <8 x i16>* [[A_ADDR]], align 16
50 // CHECK-NEXT:    store <8 x i16> [[B:%.*]], <8 x i16>* [[B_ADDR]], align 16
51 // CHECK-NEXT:    [[TMP0:%.*]] = load <8 x i16>, <8 x i16>* [[A_ADDR]], align 16
52 // CHECK-NEXT:    [[TMP1:%.*]] = load <8 x i16>, <8 x i16>* [[B_ADDR]], align 16
53 // CHECK-NEXT:    [[TMP2:%.*]] = call i32 @llvm.ppc.altivec.vcmpequh.p(i32 2, <8 x i16> [[TMP0]], <8 x i16> [[TMP1]])
54 // CHECK-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0
55 // CHECK-NEXT:    [[TMP3:%.*]] = zext i1 [[TOBOOL]] to i64
56 // CHECK-NEXT:    [[COND:%.*]] = select i1 [[TOBOOL]], i32 3, i32 7
57 // CHECK-NEXT:    ret i32 [[COND]]
58 //
59 // ERROR: error: used type '__attribute__((__vector_size__(8 * sizeof(short)))) short' (vector of 8 'short' values) where arithmetic or pointer type is required
60 int ui16(vector unsigned short a, vector unsigned short b) {
61   return a == b ? 3 : 7;
62 }
63 
64 // CHECK-LABEL: @si16(
65 // CHECK:         [[A_ADDR:%.*]] = alloca <8 x i16>, align 16
66 // CHECK-NEXT:    [[B_ADDR:%.*]] = alloca <8 x i16>, align 16
67 // CHECK-NEXT:    store <8 x i16> [[A:%.*]], <8 x i16>* [[A_ADDR]], align 16
68 // CHECK-NEXT:    store <8 x i16> [[B:%.*]], <8 x i16>* [[B_ADDR]], align 16
69 // CHECK-NEXT:    [[TMP0:%.*]] = load <8 x i16>, <8 x i16>* [[A_ADDR]], align 16
70 // CHECK-NEXT:    [[TMP1:%.*]] = load <8 x i16>, <8 x i16>* [[B_ADDR]], align 16
71 // CHECK-NEXT:    [[TMP2:%.*]] = call i32 @llvm.ppc.altivec.vcmpequh.p(i32 2, <8 x i16> [[TMP0]], <8 x i16> [[TMP1]])
72 // CHECK-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0
73 // CHECK-NEXT:    [[TMP3:%.*]] = zext i1 [[TOBOOL]] to i64
74 // CHECK-NEXT:    [[COND:%.*]] = select i1 [[TOBOOL]], i32 3, i32 7
75 // CHECK-NEXT:    ret i32 [[COND]]
76 //
77 // ERROR: error: used type '__attribute__((__vector_size__(8 * sizeof(short)))) short' (vector of 8 'short' values) where arithmetic or pointer type is required
78 int si16(vector signed short a, vector signed short b) {
79   return a == b ? 3 : 7;
80 }
81 
82 // CHECK-LABEL: @ui32(
83 // CHECK:         [[A_ADDR:%.*]] = alloca <4 x i32>, align 16
84 // CHECK-NEXT:    [[B_ADDR:%.*]] = alloca <4 x i32>, align 16
85 // CHECK-NEXT:    store <4 x i32> [[A:%.*]], <4 x i32>* [[A_ADDR]], align 16
86 // CHECK-NEXT:    store <4 x i32> [[B:%.*]], <4 x i32>* [[B_ADDR]], align 16
87 // CHECK-NEXT:    [[TMP0:%.*]] = load <4 x i32>, <4 x i32>* [[A_ADDR]], align 16
88 // CHECK-NEXT:    [[TMP1:%.*]] = load <4 x i32>, <4 x i32>* [[B_ADDR]], align 16
89 // CHECK-NEXT:    [[TMP2:%.*]] = call i32 @llvm.ppc.altivec.vcmpequw.p(i32 2, <4 x i32> [[TMP0]], <4 x i32> [[TMP1]])
90 // CHECK-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0
91 // CHECK-NEXT:    [[TMP3:%.*]] = zext i1 [[TOBOOL]] to i64
92 // CHECK-NEXT:    [[COND:%.*]] = select i1 [[TOBOOL]], i32 3, i32 7
93 // CHECK-NEXT:    ret i32 [[COND]]
94 //
95 // ERROR: error: used type '__attribute__((__vector_size__(4 * sizeof(long)))) long' (vector of 4 'long' values) where arithmetic or pointer type is required
96 int ui32(vector unsigned int a, vector unsigned int b) {
97   return a == b ? 3 : 7;
98 }
99 
100 // CHECK-LABEL: @si32(
101 // CHECK:         [[A_ADDR:%.*]] = alloca <4 x i32>, align 16
102 // CHECK-NEXT:    [[B_ADDR:%.*]] = alloca <4 x i32>, align 16
103 // CHECK-NEXT:    store <4 x i32> [[A:%.*]], <4 x i32>* [[A_ADDR]], align 16
104 // CHECK-NEXT:    store <4 x i32> [[B:%.*]], <4 x i32>* [[B_ADDR]], align 16
105 // CHECK-NEXT:    [[TMP0:%.*]] = load <4 x i32>, <4 x i32>* [[A_ADDR]], align 16
106 // CHECK-NEXT:    [[TMP1:%.*]] = load <4 x i32>, <4 x i32>* [[B_ADDR]], align 16
107 // CHECK-NEXT:    [[TMP2:%.*]] = call i32 @llvm.ppc.altivec.vcmpequw.p(i32 2, <4 x i32> [[TMP0]], <4 x i32> [[TMP1]])
108 // CHECK-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0
109 // CHECK-NEXT:    [[TMP3:%.*]] = zext i1 [[TOBOOL]] to i64
110 // CHECK-NEXT:    [[COND:%.*]] = select i1 [[TOBOOL]], i32 3, i32 7
111 // CHECK-NEXT:    ret i32 [[COND]]
112 //
113 // ERROR: error: used type '__attribute__((__vector_size__(4 * sizeof(long)))) long' (vector of 4 'long' values) where arithmetic or pointer type is required
114 int si32(vector signed int a, vector signed int b) {
115   return a == b ? 3 : 7;
116 }
117 
118 // CHECK-LABEL: @si64(
119 // CHECK:         [[A_ADDR:%.*]] = alloca <2 x i64>, align 16
120 // CHECK-NEXT:    [[B_ADDR:%.*]] = alloca <2 x i64>, align 16
121 // CHECK-NEXT:    store <2 x i64> [[A:%.*]], <2 x i64>* [[A_ADDR]], align 16
122 // CHECK-NEXT:    store <2 x i64> [[B:%.*]], <2 x i64>* [[B_ADDR]], align 16
123 // CHECK-NEXT:    [[TMP0:%.*]] = load <2 x i64>, <2 x i64>* [[A_ADDR]], align 16
124 // CHECK-NEXT:    [[TMP1:%.*]] = load <2 x i64>, <2 x i64>* [[B_ADDR]], align 16
125 // CHECK-NEXT:    [[TMP2:%.*]] = call i32 @llvm.ppc.altivec.vcmpequd.p(i32 2, <2 x i64> [[TMP0]], <2 x i64> [[TMP1]])
126 // CHECK-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0
127 // CHECK-NEXT:    [[TMP3:%.*]] = zext i1 [[TOBOOL]] to i64
128 // CHECK-NEXT:    [[COND:%.*]] = select i1 [[TOBOOL]], i32 3, i32 7
129 // CHECK-NEXT:    ret i32 [[COND]]
130 //
131 // ERROR: error: used type '__attribute__((__vector_size__(2 * sizeof(long long)))) long long' (vector of 2 'long long' values) where arithmetic or pointer type is required
132 int si64(vector long long a, vector long long b) {
133   return a == b ? 3 : 7;
134 }
135 
136 // CHECK-LABEL: @f32(
137 // CHECK:         [[A_ADDR:%.*]] = alloca <4 x float>, align 16
138 // CHECK-NEXT:    [[B_ADDR:%.*]] = alloca <4 x float>, align 16
139 // CHECK-NEXT:    store <4 x float> [[A:%.*]], <4 x float>* [[A_ADDR]], align 16
140 // CHECK-NEXT:    store <4 x float> [[B:%.*]], <4 x float>* [[B_ADDR]], align 16
141 // CHECK-NEXT:    [[TMP0:%.*]] = load <4 x float>, <4 x float>* [[A_ADDR]], align 16
142 // CHECK-NEXT:    [[TMP1:%.*]] = load <4 x float>, <4 x float>* [[B_ADDR]], align 16
143 // CHECK-NEXT:    [[TMP2:%.*]] = call i32 @llvm.ppc.altivec.vcmpeqfp.p(i32 2, <4 x float> [[TMP0]], <4 x float> [[TMP1]])
144 // CHECK-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0
145 // CHECK-NEXT:    [[TMP3:%.*]] = zext i1 [[TOBOOL]] to i64
146 // CHECK-NEXT:    [[COND:%.*]] = select i1 [[TOBOOL]], i32 3, i32 7
147 // CHECK-NEXT:    ret i32 [[COND]]
148 //
149 // ERROR: error: used type '__attribute__((__vector_size__(4 * sizeof(long)))) long' (vector of 4 'long' values) where arithmetic or pointer type is required
150 int f32(vector float a, vector float b) {
151   return a == b ? 3 : 7;
152 }
153 
154 // CHECK-LABEL: @f64(
155 // CHECK:         [[A_ADDR:%.*]] = alloca <2 x double>, align 16
156 // CHECK-NEXT:    [[B_ADDR:%.*]] = alloca <2 x double>, align 16
157 // CHECK-NEXT:    store <2 x double> [[A:%.*]], <2 x double>* [[A_ADDR]], align 16
158 // CHECK-NEXT:    store <2 x double> [[B:%.*]], <2 x double>* [[B_ADDR]], align 16
159 // CHECK-NEXT:    [[TMP0:%.*]] = load <2 x double>, <2 x double>* [[A_ADDR]], align 16
160 // CHECK-NEXT:    [[TMP1:%.*]] = load <2 x double>, <2 x double>* [[B_ADDR]], align 16
161 // CHECK-NEXT:    [[TMP2:%.*]] = call i32 @llvm.ppc.vsx.xvcmpeqdp.p(i32 2, <2 x double> [[TMP0]], <2 x double> [[TMP1]])
162 // CHECK-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0
163 // CHECK-NEXT:    [[TMP3:%.*]] = zext i1 [[TOBOOL]] to i64
164 // CHECK-NEXT:    [[COND:%.*]] = select i1 [[TOBOOL]], i32 3, i32 7
165 // CHECK-NEXT:    ret i32 [[COND]]
166 //
167 // ERROR: error: used type '__attribute__((__vector_size__(2 * sizeof(long long)))) long long' (vector of 2 'long long' values) where arithmetic or pointer type is required
168 int f64(vector double a, vector double b) {
169   return a == b ? 3 : 7;
170 }
171