• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 // RUN: %clang_cc1 -verify -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix CHECK --check-prefix CHECK-64
2 // RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
3 // RUN: %clang_cc1 -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix CHECK --check-prefix CHECK-64
4 // RUN: %clang_cc1 -verify -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix CHECK --check-prefix CHECK-32
5 // RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
6 // RUN: %clang_cc1 -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix CHECK --check-prefix CHECK-32
7 // expected-no-diagnostics
8 #ifndef HEADER
9 #define HEADER
10 
11 
12 // CHECK-DAG: [[GA:@.+]] = global double 1.000000e+00
13 // CHECK-DAG: [[GB:@.+]] = global double 2.000000e+00
14 // CHECK-DAG: [[GC:@.+]] = global double 3.000000e+00
15 // CHECK-DAG: [[GD:@.+]] = global double 4.000000e+00
16 // CHECK-DAG: [[FA:@.+]] = internal global float 5.000000e+00
17 // CHECK-DAG: [[FB:@.+]] = internal global float 6.000000e+00
18 // CHECK-DAG: [[FC:@.+]] = internal global float 7.000000e+00
19 // CHECK-DAG: [[FD:@.+]] = internal global float 8.000000e+00
20 // CHECK-DAG: [[BA:@.+]] = internal global float 9.000000e+00
21 // CHECK-DAG: [[BB:@.+]] = internal global float 1.000000e+01
22 // CHECK-DAG: [[BC:@.+]] = internal global float 1.100000e+01
23 // CHECK-DAG: [[BD:@.+]] = internal global float 1.200000e+01
24 // CHECK-DAG: [[TBA:@.+]] = {{.*}}global float 1.700000e+01
25 // CHECK-DAG: [[TBB:@.+]] = {{.*}}global float 1.800000e+01
26 // CHECK-DAG: [[TBC:@.+]] = {{.*}}global float 1.900000e+01
27 // CHECK-DAG: [[TBD:@.+]] = {{.*}}global float 2.000000e+01
28 
29 double Ga = 1.0;
30 double Gb = 2.0;
31 double Gc = 3.0;
32 double Gd = 4.0;
33 
34 // CHECK: define {{.*}} @{{.*}}foo{{.*}}(
35 // CHECK-SAME: i16 {{[^,]*}}[[A:%[^,]+]],
36 // CHECK-SAME: i16 {{[^,]*}}[[B:%[^,]+]],
37 // CHECK-SAME: i16 {{[^,]*}}[[C:%[^,]+]],
38 // CHECK-SAME: i16 {{[^,]*}}[[D:%[^,]+]])
39 // CHECK: [[LA:%.+]] = alloca i16
40 // CHECK: [[LB:%.+]] = alloca i16
41 // CHECK: [[LC:%.+]] = alloca i16
42 // CHECK: [[LD:%.+]] = alloca i16
foo(short a,short b,short c,short d)43 int foo(short a, short b, short c, short d){
44   static float Sa = 5.0;
45   static float Sb = 6.0;
46   static float Sc = 7.0;
47   static float Sd = 8.0;
48 
49   // CHECK-DAG:    [[VALLB:%.+]] = load i16, i16* [[LB]],
50   // CHECK-64-DAG: [[VALGB:%.+]] = load double, double* [[GB]],
51   // CHECK-DAG:    [[VALFB:%.+]] = load float, float* [[FB]],
52   // CHECK-64-DAG: [[VALGC:%.+]] = load double, double* [[GC]],
53   // CHECK-DAG:    [[VALLC:%.+]] = load i16, i16* [[LC]],
54   // CHECK-DAG:    [[VALFC:%.+]] = load float, float* [[FC]],
55   // CHECK-DAG:    [[VALLD:%.+]] = load i16, i16* [[LD]],
56   // CHECK-64-DAG: [[VALGD:%.+]] = load double, double* [[GD]],
57   // CHECK-DAG:    [[VALFD:%.+]] = load float, float* [[FD]],
58 
59   // 3 local vars being captured.
60 
61   // CHECK-DAG: store i16 [[VALLB]], i16* [[CONVLB:%.+]],
62   // CHECK-DAG: [[CONVLB]] = bitcast i[[sz:64|32]]* [[CADDRLB:%.+]] to i16*
63   // CHECK-DAG: [[CVALLB:%.+]] = load i[[sz]], i[[sz]]* [[CADDRLB]],
64   // CHECK-DAG: [[CPTRLB:%.+]] = inttoptr i[[sz]] [[CVALLB]] to i8*
65   // CHECK-DAG: store i8* [[CPTRLB]], i8** [[GEPLB:%.+]],
66   // CHECK-DAG: [[GEPLB]] = getelementptr inbounds [9 x i8*], [9 x i8*]* %{{.+}}, i32 0, i32 {{[0-8]}}
67 
68   // CHECK-DAG: store i16 [[VALLC]], i16* [[CONVLC:%.+]],
69   // CHECK-DAG: [[CONVLC]] = bitcast i[[sz]]* [[CADDRLC:%.+]] to i16*
70   // CHECK-DAG: [[CVALLC:%.+]] = load i[[sz]], i[[sz]]* [[CADDRLC]],
71   // CHECK-DAG: [[CPTRLC:%.+]] = inttoptr i[[sz]] [[CVALLC]] to i8*
72   // CHECK-DAG: store i8* [[CPTRLC]], i8** [[GEPLC:%.+]],
73   // CHECK-DAG: [[GEPLC]] = getelementptr inbounds [9 x i8*], [9 x i8*]* %{{.+}}, i32 0, i32 {{[0-8]}}
74 
75   // CHECK-DAG: store i16 [[VALLD]], i16* [[CONVLD:%.+]],
76   // CHECK-DAG: [[CONVLD]] = bitcast i[[sz]]* [[CADDRLD:%.+]] to i16*
77   // CHECK-DAG: [[CVALLD:%.+]] = load i[[sz]], i[[sz]]* [[CADDRLD]],
78   // CHECK-DAG: [[CPTRLD:%.+]] = inttoptr i[[sz]] [[CVALLD]] to i8*
79   // CHECK-DAG: store i8* [[CPTRLD]], i8** [[GEPLD:%.+]],
80   // CHECK-DAG: [[GEPLD]] = getelementptr inbounds [9 x i8*], [9 x i8*]* %{{.+}}, i32 0, i32 {{[0-8]}}
81 
82   // 3 static vars being captured.
83 
84   // CHECK-DAG: store float [[VALFB]], float* [[CONVFB:%.+]],
85   // CHECK-DAG: [[CONVFB]] = bitcast i[[sz]]* [[CADDRFB:%.+]] to float*
86   // CHECK-DAG: [[CVALFB:%.+]] = load i[[sz]], i[[sz]]* [[CADDRFB]],
87   // CHECK-DAG: [[CPTRFB:%.+]] = inttoptr i[[sz]] [[CVALFB]] to i8*
88   // CHECK-DAG: store i8* [[CPTRFB]], i8** [[GEPFB:%.+]],
89   // CHECK-DAG: [[GEPFB]] = getelementptr inbounds [9 x i8*], [9 x i8*]* %{{.+}}, i32 0, i32 {{[0-8]}}
90 
91   // CHECK-DAG: store float [[VALFC]], float* [[CONVFC:%.+]],
92   // CHECK-DAG: [[CONVFC]] = bitcast i[[sz]]* [[CADDRFC:%.+]] to float*
93   // CHECK-DAG: [[CVALFC:%.+]] = load i[[sz]], i[[sz]]* [[CADDRFC]],
94   // CHECK-DAG: [[CPTRFC:%.+]] = inttoptr i[[sz]] [[CVALFC]] to i8*
95   // CHECK-DAG: store i8* [[CPTRFC]], i8** [[GEPFC:%.+]],
96   // CHECK-DAG: [[GEPFC]] = getelementptr inbounds [9 x i8*], [9 x i8*]* %{{.+}}, i32 0, i32 {{[0-8]}}
97 
98   // CHECK-DAG: store float [[VALFD]], float* [[CONVFD:%.+]],
99   // CHECK-DAG: [[CONVFD]] = bitcast i[[sz]]* [[CADDRFD:%.+]] to float*
100   // CHECK-DAG: [[CVALFD:%.+]] = load i[[sz]], i[[sz]]* [[CADDRFD]],
101   // CHECK-DAG: [[CPTRFD:%.+]] = inttoptr i[[sz]] [[CVALFD]] to i8*
102   // CHECK-DAG: store i8* [[CPTRFD]], i8** [[GEPFD:%.+]],
103   // CHECK-DAG: [[GEPFD]] = getelementptr inbounds [9 x i8*], [9 x i8*]* %{{.+}}, i32 0, i32 {{[0-8]}}
104 
105   // 3 static global vars being captured.
106 
107   // CHECK-64-DAG: store double [[VALGB]], double* [[CONVGB:%.+]],
108   // CHECK-64-DAG: [[CONVGB]] = bitcast i[[sz]]* [[CADDRGB:%.+]] to double*
109   // CHECK-64-DAG: [[CVALGB:%.+]] = load i[[sz]], i[[sz]]* [[CADDRGB]],
110   // CHECK-64-DAG: [[CPTRGB:%.+]] = inttoptr i[[sz]] [[CVALGB]] to i8*
111   // CHECK-64-DAG: store i8* [[CPTRGB]], i8** [[GEPGB:%.+]],
112   // CHECK-32-DAG: store i8* bitcast (double* @Gb to i8*), i8** [[GEPGB:%.+]],
113   // CHECK-DAG: [[GEPGB]] = getelementptr inbounds [9 x i8*], [9 x i8*]* %{{.+}}, i32 0, i32 {{[0-8]}}
114 
115   // CHECK-64-DAG: store double [[VALGC]], double* [[CONVGC:%.+]],
116   // CHECK-64-DAG: [[CONVGC]] = bitcast i[[sz]]* [[CADDRGC:%.+]] to double*
117   // CHECK-64-DAG: [[CVALGC:%.+]] = load i[[sz]], i[[sz]]* [[CADDRGC]],
118   // CHECK-64-DAG: [[CPTRGC:%.+]] = inttoptr i[[sz]] [[CVALGC]] to i8*
119   // CHECK-64-DAG: store i8* [[CPTRGC]], i8** [[GEPGC:%.+]],
120   // CHECK-32-DAG: store i8* bitcast (double* @Gc to i8*), i8** [[GEPGC:%.+]],
121   // CHECK-DAG: [[GEPGC]] = getelementptr inbounds [9 x i8*], [9 x i8*]* %{{.+}}, i32 0, i32 {{[0-8]}}
122 
123   // CHECK-64-DAG: store double [[VALGD]], double* [[CONVGD:%.+]],
124   // CHECK-64-DAG: [[CONVGD]] = bitcast i[[sz]]* [[CADDRGD:%.+]] to double*
125   // CHECK-64-DAG: [[CVALGD:%.+]] = load i[[sz]], i[[sz]]* [[CADDRGD]],
126   // CHECK-64-DAG: [[CPTRGD:%.+]] = inttoptr i[[sz]] [[CVALGD]] to i8*
127   // CHECK-64-DAG: store i8* [[CPTRGD]], i8** [[GEPGD:%.+]],
128   // CHECK-32-DAG: store i8* bitcast (double* @Gd to i8*), i8** [[GEPGD:%.+]],
129   // CHECK-DAG: [[GEPGD]] = getelementptr inbounds [9 x i8*], [9 x i8*]* %{{.+}}, i32 0, i32 {{[0-8]}}
130 
131   // CHECK: call i32 @__tgt_target
132   // CHECK: call void [[OFFLOADF:@.+]](
133   // Capture b, Gb, Sb, Gc, c, Sc, d, Gd, Sd
134   #pragma omp target if(Ga>0.0 && a>0 && Sa>0.0)
135   {
136     b += 1;
137     Gb += 1.0;
138     Sb += 1.0;
139 
140     // CHECK: define internal void [[OFFLOADF]]({{.+}} {{.*}}%{{.+}}, {{.+}} {{.*}}%{{.+}}, {{.+}} {{.*}}%{{.+}}, {{.+}} {{.*}}%{{.+}}, {{.+}} {{.*}}%{{.+}}, {{.+}} {{.*}}%{{.+}}, {{.+}} {{.*}}%{{.+}}, {{.+}} {{.*}}%{{.+}}, {{.+}} {{.*}}%{{.+}})
141     // The parallel region only uses 3 captures.
142     // CHECK:     call {{.*}}@__kmpc_fork_call(%ident_t* {{.+}}, i32 {{.+}}, void (i32*, i32*, ...)* bitcast ({{.*}}[[PARF:@.+]] to {{.*}}), {{.+}}* %{{.+}}, {{.+}}* %{{.+}}, {{.+}}* %{{.+}})
143     // CHECK:     call void @.omp_outlined.(i32* %{{.+}}, i32* %{{.+}}, {{.+}}* %{{.+}}, {{.+}}* %{{.+}}, {{.+}}* %{{.+}})
144     // Capture d, Gd, Sd,
145 
146     // CHECK: define internal void [[PARF]](i32* noalias %{{.*}}, i32* noalias %{{.*}},
147     #pragma omp parallel if(Gc>0.0 && c>0 && Sc>0.0)
148     {
149       d += 1;
150       Gd += 1.0;
151       Sd += 1.0;
152     }
153   }
154   return a + b + c + d + (int)Sa + (int)Sb + (int)Sc + (int)Sd;
155 }
156 
157 // CHECK: define {{.*}} @{{.*}}bar{{.*}}(
158 // CHECK-SAME: i16 {{[^,]*}}[[A:%[^,]+]],
159 // CHECK-SAME: i16 {{[^,]*}}[[B:%[^,]+]],
160 // CHECK-SAME: i16 {{[^,]*}}[[C:%[^,]+]],
161 // CHECK-SAME: i16 {{[^,]*}}[[D:%[^,]+]])
162 // CHECK: [[LA:%.+]] = alloca i16
163 // CHECK: [[LB:%.+]] = alloca i16
164 // CHECK: [[LC:%.+]] = alloca i16
165 // CHECK: [[LD:%.+]] = alloca i16
bar(short a,short b,short c,short d)166 int bar(short a, short b, short c, short d){
167   static float Sa = 9.0;
168   static float Sb = 10.0;
169   static float Sc = 11.0;
170   static float Sd = 12.0;
171 
172   // CHECK: call void {{.*}}@__kmpc_fork_call(%ident_t* {{.+}}, i32 {{.+}}, void (i32*, i32*, ...)* bitcast ({{.*}}[[PARF:@.+]] to {{.*}}), i16* %{{.+}}, i16* %{{.+}}, i16* %{{.+}}, i16* %{{.+}})
173   // CHECK: define internal void [[PARF]](i32* noalias %{{.*}}, i32* noalias %{{.*}}, i16* dereferenceable(2) [[A:%.+]], i16* dereferenceable(2) [[B:%.+]], i16* dereferenceable(2) [[C:%.+]], i16* dereferenceable(2) [[D:%.+]])
174   // Capture a, b, c, d
175   // CHECK: [[ALLOCLA:%.+]] = alloca i16
176   // CHECK: [[ALLOCLB:%.+]] = alloca i16
177   // CHECK: [[ALLOCLC:%.+]] = alloca i16
178   // CHECK: [[ALLOCLD:%.+]] = alloca i16
179   // CHECK: [[LLA:%.+]] = load i16*, i16** [[ALLOCLA]],
180   // CHECK: [[LLB:%.+]] = load i16*, i16** [[ALLOCLB]],
181   // CHECK: [[LLC:%.+]] = load i16*, i16** [[ALLOCLC]],
182   // CHECK: [[LLD:%.+]] = load i16*, i16** [[ALLOCLD]],
183   #pragma omp parallel
184   {
185     // CHECK-DAG:    [[VALLB:%.+]] = load i16, i16* [[LLB]],
186     // CHECK-64-DAG: [[VALGB:%.+]] = load double, double* [[GB]],
187     // CHECK-DAG:    [[VALFB:%.+]] = load float, float* [[BB]],
188     // CHECK-64-DAG: [[VALGC:%.+]] = load double, double* [[GC]],
189     // CHECK-DAG:    [[VALLC:%.+]] = load i16, i16* [[LLC]],
190     // CHECK-DAG:    [[VALFC:%.+]] = load float, float* [[BC]],
191     // CHECK-DAG:    [[VALLD:%.+]] = load i16, i16* [[LLD]],
192     // CHECK-64-DAG: [[VALGD:%.+]] = load double, double* [[GD]],
193     // CHECK-DAG:    [[VALFD:%.+]] = load float, float* [[BD]],
194 
195     // 3 local vars being captured.
196 
197     // CHECK-DAG: store i16 [[VALLB]], i16* [[CONVLB:%.+]],
198     // CHECK-DAG: [[CONVLB]] = bitcast i[[sz:64|32]]* [[CADDRLB:%.+]] to i16*
199     // CHECK-DAG: [[CVALLB:%.+]] = load i[[sz]], i[[sz]]* [[CADDRLB]],
200     // CHECK-DAG: [[CPTRLB:%.+]] = inttoptr i[[sz]] [[CVALLB]] to i8*
201     // CHECK-DAG: store i8* [[CPTRLB]], i8** [[GEPLB:%.+]],
202     // CHECK-DAG: [[GEPLB]] = getelementptr inbounds [9 x i8*], [9 x i8*]* %{{.+}}, i32 0, i32 {{[0-8]}}
203 
204     // CHECK-DAG: store i16 [[VALLC]], i16* [[CONVLC:%.+]],
205     // CHECK-DAG: [[CONVLC]] = bitcast i[[sz]]* [[CADDRLC:%.+]] to i16*
206     // CHECK-DAG: [[CVALLC:%.+]] = load i[[sz]], i[[sz]]* [[CADDRLC]],
207     // CHECK-DAG: [[CPTRLC:%.+]] = inttoptr i[[sz]] [[CVALLC]] to i8*
208     // CHECK-DAG: store i8* [[CPTRLC]], i8** [[GEPLC:%.+]],
209     // CHECK-DAG: [[GEPLC]] = getelementptr inbounds [9 x i8*], [9 x i8*]* %{{.+}}, i32 0, i32 {{[0-8]}}
210 
211     // CHECK-DAG: store i16 [[VALLD]], i16* [[CONVLD:%.+]],
212     // CHECK-DAG: [[CONVLD]] = bitcast i[[sz]]* [[CADDRLD:%.+]] to i16*
213     // CHECK-DAG: [[CVALLD:%.+]] = load i[[sz]], i[[sz]]* [[CADDRLD]],
214     // CHECK-DAG: [[CPTRLD:%.+]] = inttoptr i[[sz]] [[CVALLD]] to i8*
215     // CHECK-DAG: store i8* [[CPTRLD]], i8** [[GEPLD:%.+]],
216     // CHECK-DAG: [[GEPLD]] = getelementptr inbounds [9 x i8*], [9 x i8*]* %{{.+}}, i32 0, i32 {{[0-8]}}
217 
218     // 3 static vars being captured.
219 
220     // CHECK-DAG: store float [[VALFB]], float* [[CONVFB:%.+]],
221     // CHECK-DAG: [[CONVFB]] = bitcast i[[sz]]* [[CADDRFB:%.+]] to float*
222     // CHECK-DAG: [[CVALFB:%.+]] = load i[[sz]], i[[sz]]* [[CADDRFB]],
223     // CHECK-DAG: [[CPTRFB:%.+]] = inttoptr i[[sz]] [[CVALFB]] to i8*
224     // CHECK-DAG: store i8* [[CPTRFB]], i8** [[GEPFB:%.+]],
225     // CHECK-DAG: [[GEPFB]] = getelementptr inbounds [9 x i8*], [9 x i8*]* %{{.+}}, i32 0, i32 {{[0-8]}}
226 
227     // CHECK-DAG: store float [[VALFC]], float* [[CONVFC:%.+]],
228     // CHECK-DAG: [[CONVFC]] = bitcast i[[sz]]* [[CADDRFC:%.+]] to float*
229     // CHECK-DAG: [[CVALFC:%.+]] = load i[[sz]], i[[sz]]* [[CADDRFC]],
230     // CHECK-DAG: [[CPTRFC:%.+]] = inttoptr i[[sz]] [[CVALFC]] to i8*
231     // CHECK-DAG: store i8* [[CPTRFC]], i8** [[GEPFC:%.+]],
232     // CHECK-DAG: [[GEPFC]] = getelementptr inbounds [9 x i8*], [9 x i8*]* %{{.+}}, i32 0, i32 {{[0-8]}}
233 
234     // CHECK-DAG: store float [[VALFD]], float* [[CONVFD:%.+]],
235     // CHECK-DAG: [[CONVFD]] = bitcast i[[sz]]* [[CADDRFD:%.+]] to float*
236     // CHECK-DAG: [[CVALFD:%.+]] = load i[[sz]], i[[sz]]* [[CADDRFD]],
237     // CHECK-DAG: [[CPTRFD:%.+]] = inttoptr i[[sz]] [[CVALFD]] to i8*
238     // CHECK-DAG: store i8* [[CPTRFD]], i8** [[GEPFD:%.+]],
239     // CHECK-DAG: [[GEPFD]] = getelementptr inbounds [9 x i8*], [9 x i8*]* %{{.+}}, i32 0, i32 {{[0-8]}}
240 
241     // 3 static global vars being captured.
242 
243     // CHECK-64-DAG: store double [[VALGB]], double* [[CONVGB:%.+]],
244     // CHECK-64-DAG: [[CONVGB]] = bitcast i[[sz]]* [[CADDRGB:%.+]] to double*
245     // CHECK-64-DAG: [[CVALGB:%.+]] = load i[[sz]], i[[sz]]* [[CADDRGB]],
246     // CHECK-64-DAG: [[CPTRGB:%.+]] = inttoptr i[[sz]] [[CVALGB]] to i8*
247     // CHECK-64-DAG: store i8* [[CPTRGB]], i8** [[GEPGB:%.+]],
248     // CHECK-32-DAG: store i8* bitcast (double* @Gb to i8*), i8** [[GEPGB:%.+]],
249     // CHECK-DAG: [[GEPGB]] = getelementptr inbounds [9 x i8*], [9 x i8*]* %{{.+}}, i32 0, i32 {{[0-8]}}
250 
251     // CHECK-64-DAG: store double [[VALGC]], double* [[CONVGC:%.+]],
252     // CHECK-64-DAG: [[CONVGC]] = bitcast i[[sz]]* [[CADDRGC:%.+]] to double*
253     // CHECK-64-DAG: [[CVALGC:%.+]] = load i[[sz]], i[[sz]]* [[CADDRGC]],
254     // CHECK-64-DAG: [[CPTRGC:%.+]] = inttoptr i[[sz]] [[CVALGC]] to i8*
255     // CHECK-64-DAG: store i8* [[CPTRGC]], i8** [[GEPGC:%.+]],
256     // CHECK-32-DAG: store i8* bitcast (double* @Gc to i8*), i8** [[GEPGC:%.+]],
257     // CHECK-DAG: [[GEPGC]] = getelementptr inbounds [9 x i8*], [9 x i8*]* %{{.+}}, i32 0, i32 {{[0-8]}}
258 
259     // CHECK-64-DAG: store double [[VALGD]], double* [[CONVGD:%.+]],
260     // CHECK-64-DAG: [[CONVGD]] = bitcast i[[sz]]* [[CADDRGD:%.+]] to double*
261     // CHECK-64-DAG: [[CVALGD:%.+]] = load i[[sz]], i[[sz]]* [[CADDRGD]],
262     // CHECK-64-DAG: [[CPTRGD:%.+]] = inttoptr i[[sz]] [[CVALGD]] to i8*
263     // CHECK-64-DAG: store i8* [[CPTRGD]], i8** [[GEPGD:%.+]],
264     // CHECK-32-DAG: store i8* bitcast (double* @Gd to i8*), i8** [[GEPGD:%.+]],
265     // CHECK-DAG: [[GEPGD]] = getelementptr inbounds [9 x i8*], [9 x i8*]* %{{.+}}, i32 0, i32 {{[0-8]}}
266 
267     // CHECK: call i32 @__tgt_target
268     // CHECK: call void [[OFFLOADF:@.+]](
269     // Capture b, Gb, Sb, Gc, c, Sc, d, Gd, Sd
270     #pragma omp target if(Ga>0.0 && a>0 && Sa>0.0)
271     {
272       b += 1;
273       Gb += 1.0;
274       Sb += 1.0;
275 
276       // CHECK: define internal void [[OFFLOADF]]({{.+}} {{.*}}%{{.+}}, {{.+}} {{.*}}%{{.+}}, {{.+}} {{.*}}%{{.+}}, {{.+}} {{.*}}%{{.+}}, {{.+}} {{.*}}%{{.+}}, {{.+}} {{.*}}%{{.+}}, {{.+}} {{.*}}%{{.+}}, {{.+}} {{.*}}%{{.+}}, {{.+}} {{.*}}%{{.+}})
277       // CHECK: call void {{.*}}@__kmpc_fork_call(%ident_t* {{.+}}, i32 {{.+}}, void (i32*, i32*, ...)* bitcast ({{.*}}[[PARF:@.+]] to {{.*}})
278 
279       // CHECK: define internal void [[PARF]](i32* noalias %{{.*}}, i32* noalias %{{.*}}, {{.+}}* dereferenceable({{.+}}) %{{.+}}, {{.+}}* dereferenceable({{.+}}) %{{.+}}, {{.+}}* dereferenceable({{.+}}) %{{.+}})
280       // Capture d, Gd, Sd
281       #pragma omp parallel if(Gc>0.0 && c>0 && Sc>0.0)
282       {
283         d += 1;
284         Gd += 1.0;
285         Sd += 1.0;
286       }
287     }
288   }
289   return a + b + c + d + (int)Sa + (int)Sb + (int)Sc + (int)Sd;
290 }
291 
292 ///
293 /// Tests with template functions.
294 ///
295 
296 // CHECK: define {{.*}} @{{.*}}tbar2{{.*}}(
297 
298 // CHECK: define {{.*}} @{{.*}}tbar{{.*}}(
299 // CHECK-SAME: i16 {{[^,]*}}[[A:%[^,]+]],
300 // CHECK-SAME: i16 {{[^,]*}}[[B:%[^,]+]],
301 // CHECK-SAME: i16 {{[^,]*}}[[C:%[^,]+]],
302 // CHECK-SAME: i16 {{[^,]*}}[[D:%[^,]+]])
303 // CHECK: [[LA:%.+]] = alloca i16
304 // CHECK: [[LB:%.+]] = alloca i16
305 // CHECK: [[LC:%.+]] = alloca i16
306 // CHECK: [[LD:%.+]] = alloca i16
307 template<typename T>
tbar(T a,T b,T c,T d)308 int tbar(T a, T b, T c, T d){
309   static float Sa = 17.0;
310   static float Sb = 18.0;
311   static float Sc = 19.0;
312   static float Sd = 20.0;
313 
314   // CHECK: call void {{.*}}@__kmpc_fork_call(%ident_t* {{.+}}, i32 {{.+}}, void (i32*, i32*, ...)* bitcast ({{.*}}[[PARF:@.+]] to {{.*}}), i16* %{{.+}}, i16* %{{.+}}, i16* %{{.+}}, i16* %{{.+}})
315   // CHECK: define internal void [[PARF]](i32* noalias %{{.*}}, i32* noalias %{{.*}}, i16* dereferenceable(2) [[A:%.+]], i16* dereferenceable(2) [[B:%.+]], i16* dereferenceable(2) [[C:%.+]], i16* dereferenceable(2) [[D:%.+]])
316   // Capture a, b, c, d
317   // CHECK: [[ALLOCLA:%.+]] = alloca i16
318   // CHECK: [[ALLOCLB:%.+]] = alloca i16
319   // CHECK: [[ALLOCLC:%.+]] = alloca i16
320   // CHECK: [[ALLOCLD:%.+]] = alloca i16
321   // CHECK: [[LLA:%.+]] = load i16*, i16** [[ALLOCLA]],
322   // CHECK: [[LLB:%.+]] = load i16*, i16** [[ALLOCLB]],
323   // CHECK: [[LLC:%.+]] = load i16*, i16** [[ALLOCLC]],
324   // CHECK: [[LLD:%.+]] = load i16*, i16** [[ALLOCLD]],
325   #pragma omp parallel
326   {
327     // CHECK-DAG:    [[VALLB:%.+]] = load i16, i16* [[LLB]],
328     // CHECK-64-DAG: [[VALGB:%.+]] = load double, double* [[GB]],
329     // CHECK-DAG:    [[VALFB:%.+]] = load float, float* [[TBB]],
330     // CHECK-64-DAG: [[VALGC:%.+]] = load double, double* [[GC]],
331     // CHECK-DAG:    [[VALLC:%.+]] = load i16, i16* [[LLC]],
332     // CHECK-DAG:    [[VALFC:%.+]] = load float, float* [[TBC]],
333     // CHECK-DAG:    [[VALLD:%.+]] = load i16, i16* [[LLD]],
334     // CHECK-64-DAG: [[VALGD:%.+]] = load double, double* [[GD]],
335     // CHECK-DAG:    [[VALFD:%.+]] = load float, float* [[TBD]],
336 
337     // 3 local vars being captured.
338 
339     // CHECK-DAG: store i16 [[VALLB]], i16* [[CONVLB:%.+]],
340     // CHECK-DAG: [[CONVLB]] = bitcast i[[sz:64|32]]* [[CADDRLB:%.+]] to i16*
341     // CHECK-DAG: [[CVALLB:%.+]] = load i[[sz]], i[[sz]]* [[CADDRLB]],
342     // CHECK-DAG: [[CPTRLB:%.+]] = inttoptr i[[sz]] [[CVALLB]] to i8*
343     // CHECK-DAG: store i8* [[CPTRLB]], i8** [[GEPLB:%.+]],
344     // CHECK-DAG: [[GEPLB]] = getelementptr inbounds [9 x i8*], [9 x i8*]* %{{.+}}, i32 0, i32 {{[0-8]}}
345 
346     // CHECK-DAG: store i16 [[VALLC]], i16* [[CONVLC:%.+]],
347     // CHECK-DAG: [[CONVLC]] = bitcast i[[sz]]* [[CADDRLC:%.+]] to i16*
348     // CHECK-DAG: [[CVALLC:%.+]] = load i[[sz]], i[[sz]]* [[CADDRLC]],
349     // CHECK-DAG: [[CPTRLC:%.+]] = inttoptr i[[sz]] [[CVALLC]] to i8*
350     // CHECK-DAG: store i8* [[CPTRLC]], i8** [[GEPLC:%.+]],
351     // CHECK-DAG: [[GEPLC]] = getelementptr inbounds [9 x i8*], [9 x i8*]* %{{.+}}, i32 0, i32 {{[0-8]}}
352 
353     // CHECK-DAG: store i16 [[VALLD]], i16* [[CONVLD:%.+]],
354     // CHECK-DAG: [[CONVLD]] = bitcast i[[sz]]* [[CADDRLD:%.+]] to i16*
355     // CHECK-DAG: [[CVALLD:%.+]] = load i[[sz]], i[[sz]]* [[CADDRLD]],
356     // CHECK-DAG: [[CPTRLD:%.+]] = inttoptr i[[sz]] [[CVALLD]] to i8*
357     // CHECK-DAG: store i8* [[CPTRLD]], i8** [[GEPLD:%.+]],
358     // CHECK-DAG: [[GEPLD]] = getelementptr inbounds [9 x i8*], [9 x i8*]* %{{.+}}, i32 0, i32 {{[0-8]}}
359 
360     // 3 static vars being captured.
361 
362     // CHECK-DAG: store float [[VALFB]], float* [[CONVFB:%.+]],
363     // CHECK-DAG: [[CONVFB]] = bitcast i[[sz]]* [[CADDRFB:%.+]] to float*
364     // CHECK-DAG: [[CVALFB:%.+]] = load i[[sz]], i[[sz]]* [[CADDRFB]],
365     // CHECK-DAG: [[CPTRFB:%.+]] = inttoptr i[[sz]] [[CVALFB]] to i8*
366     // CHECK-DAG: store i8* [[CPTRFB]], i8** [[GEPFB:%.+]],
367     // CHECK-DAG: [[GEPFB]] = getelementptr inbounds [9 x i8*], [9 x i8*]* %{{.+}}, i32 0, i32 {{[0-8]}}
368 
369     // CHECK-DAG: store float [[VALFC]], float* [[CONVFC:%.+]],
370     // CHECK-DAG: [[CONVFC]] = bitcast i[[sz]]* [[CADDRFC:%.+]] to float*
371     // CHECK-DAG: [[CVALFC:%.+]] = load i[[sz]], i[[sz]]* [[CADDRFC]],
372     // CHECK-DAG: [[CPTRFC:%.+]] = inttoptr i[[sz]] [[CVALFC]] to i8*
373     // CHECK-DAG: store i8* [[CPTRFC]], i8** [[GEPFC:%.+]],
374     // CHECK-DAG: [[GEPFC]] = getelementptr inbounds [9 x i8*], [9 x i8*]* %{{.+}}, i32 0, i32 {{[0-8]}}
375 
376     // CHECK-DAG: store float [[VALFD]], float* [[CONVFD:%.+]],
377     // CHECK-DAG: [[CONVFD]] = bitcast i[[sz]]* [[CADDRFD:%.+]] to float*
378     // CHECK-DAG: [[CVALFD:%.+]] = load i[[sz]], i[[sz]]* [[CADDRFD]],
379     // CHECK-DAG: [[CPTRFD:%.+]] = inttoptr i[[sz]] [[CVALFD]] to i8*
380     // CHECK-DAG: store i8* [[CPTRFD]], i8** [[GEPFD:%.+]],
381     // CHECK-DAG: [[GEPFD]] = getelementptr inbounds [9 x i8*], [9 x i8*]* %{{.+}}, i32 0, i32 {{[0-8]}}
382 
383     // 3 static global vars being captured.
384 
385     // CHECK-64-DAG: store double [[VALGB]], double* [[CONVGB:%.+]],
386     // CHECK-64-DAG: [[CONVGB]] = bitcast i[[sz]]* [[CADDRGB:%.+]] to double*
387     // CHECK-64-DAG: [[CVALGB:%.+]] = load i[[sz]], i[[sz]]* [[CADDRGB]],
388     // CHECK-64-DAG: [[CPTRGB:%.+]] = inttoptr i[[sz]] [[CVALGB]] to i8*
389     // CHECK-64-DAG: store i8* [[CPTRGB]], i8** [[GEPGB:%.+]],
390     // CHECK-32-DAG: store i8* bitcast (double* @Gb to i8*), i8** [[GEPGB:%.+]],
391     // CHECK-DAG: [[GEPGB]] = getelementptr inbounds [9 x i8*], [9 x i8*]* %{{.+}}, i32 0, i32 {{[0-8]}}
392 
393     // CHECK-64-DAG: store double [[VALGC]], double* [[CONVGC:%.+]],
394     // CHECK-64-DAG: [[CONVGC]] = bitcast i[[sz]]* [[CADDRGC:%.+]] to double*
395     // CHECK-64-DAG: [[CVALGC:%.+]] = load i[[sz]], i[[sz]]* [[CADDRGC]],
396     // CHECK-64-DAG: [[CPTRGC:%.+]] = inttoptr i[[sz]] [[CVALGC]] to i8*
397     // CHECK-64-DAG: store i8* [[CPTRGC]], i8** [[GEPGC:%.+]],
398     // CHECK-32-DAG: store i8* bitcast (double* @Gc to i8*), i8** [[GEPGC:%.+]],
399     // CHECK-DAG: [[GEPGC]] = getelementptr inbounds [9 x i8*], [9 x i8*]* %{{.+}}, i32 0, i32 {{[0-8]}}
400 
401     // CHECK-64-DAG: store double [[VALGD]], double* [[CONVGD:%.+]],
402     // CHECK-64-DAG: [[CONVGD]] = bitcast i[[sz]]* [[CADDRGD:%.+]] to double*
403     // CHECK-64-DAG: [[CVALGD:%.+]] = load i[[sz]], i[[sz]]* [[CADDRGD]],
404     // CHECK-64-DAG: [[CPTRGD:%.+]] = inttoptr i[[sz]] [[CVALGD]] to i8*
405     // CHECK-64-DAG: store i8* [[CPTRGD]], i8** [[GEPGD:%.+]],
406     // CHECK-32-DAG: store i8* bitcast (double* @Gd to i8*), i8** [[GEPGD:%.+]],
407     // CHECK-DAG: [[GEPGD]] = getelementptr inbounds [9 x i8*], [9 x i8*]* %{{.+}}, i32 0, i32 {{[0-8]}}
408 
409     // CHECK: call i32 @__tgt_target
410     // CHECK: call void [[OFFLOADF:@.+]](
411     // Capture b, Gb, Sb, Gc, c, Sc, d, Gd, Sd
412     #pragma omp target if(Ga>0.0 && a>0 && Sa>0.0)
413     {
414       b += 1;
415       Gb += 1.0;
416       Sb += 1.0;
417 
418       // CHECK: define internal void [[OFFLOADF]]({{.+}} {{.*}}%{{.+}}, {{.+}} {{.*}}%{{.+}}, {{.+}} {{.*}}%{{.+}}, {{.+}} {{.*}}%{{.+}}, {{.+}} {{.*}}%{{.+}}, {{.+}} {{.*}}%{{.+}}, {{.+}} {{.*}}%{{.+}}, {{.+}} {{.*}}%{{.+}}, {{.+}} {{.*}}%{{.+}})
419       // CHECK: call void {{.*}}@__kmpc_fork_call(%ident_t* {{.+}}, i32 {{.+}}, void (i32*, i32*, ...)* bitcast ({{.*}}[[PARF:@.+]] to {{.*}})
420 
421       // CHECK: define internal void [[PARF]](i32* noalias %{{.*}}, i32* noalias %{{.*}}, {{.+}}* dereferenceable({{.+}}) %{{.+}}, {{.+}}* dereferenceable({{.+}}) %{{.+}}, {{.+}}* dereferenceable({{.+}}) %{{.+}})
422       // Capture d, Gd, Sd
423       #pragma omp parallel if(Gc>0.0 && c>0 && Sc>0.0)
424       {
425         d += 1;
426         Gd += 1.0;
427         Sd += 1.0;
428       }
429     }
430   }
431   return a + b + c + d + (int)Sa + (int)Sb + (int)Sc + (int)Sd;
432 }
433 
tbar2(short a,short b,short c,short d)434 int tbar2(short a, short b, short c, short d){
435   return tbar(a, b, c, d);
436 }
437 
438 #endif
439