Merge pull request #20 from ajarmusch/master

Fixes

Author: ajarmusch

README.md

@@ -1,6 +1,8 @@
 # OpenACCV-V
 This repository is updated with OpenACC test codes validating and verifying implementations of features and its conformance to the OpenACC specification. We are continuously adding and editing tests to conform to the latest version of the [OpenACC Specification](https://www.openacc.org/specification). 
 
+Consult our website for more details on results and our project [OpenACC V&V Website](https://crpl.cis.udel.edu/oaccvv/).
+
 ## Executing program
 
 In order to run the suite, we have provided a Python script, infrastructure.py that can be run. It is recommended to use Python 3.3 or later. Once Python has been loaded into your environment, the script can be invoked with the command  

Tests/parallel_loop_auto.c

@@ -1,68 +1,113 @@
 #include "acc_testsuite.h"
 #ifndef T1
 //T1:parallel,loop,combined-constructs,syntactic,V:2.0-2.7
+//data independent, treated as a independent clause
 int test1(){
     int err = 0;
     srand(SEED);
-    real_t * a = (real_t *)malloc(n * sizeof(real_t));
-    real_t * a_copy = (real_t *)malloc(n * sizeof(real_t));
-    real_t * b = (real_t *)malloc(n * sizeof(real_t));
+    real_t * value = (real_t *)malloc(n * sizeof(real_t));
+    real_t * empty = (real_t *)malloc(n * sizeof(real_t));
 
     for (int x = 0; x < n; ++x){
-        a[x] = rand() / (real_t)(RAND_MAX / 10);
-        b[x] = 0;
+        value[x] = rand() / (real_t)(RAND_MAX / 10);
+        empty[x] = 0;
     }
 
-    #pragma acc data copyin(a[0:n]) copyout(b[0:n])
+    #pragma acc data copyin(value[0:n]) copy(empty[0:n])
     {
       #pragma acc parallel loop auto
       for (int x = 0; x < n; ++x){
-        b[x] = a[x];
+        empty[x] = value[x];
       }
     }
 
     for (int x = 0; x < n; ++x){
-      if (fabs(b[x] - a[x]) > PRECISION){
+      if (fabs(empty[x] - value[x]) > PRECISION){
         err = 1;
       }
     }
 
+    free(value);
+    free(empty);
+
     return err;
 }
 #endif
 
 #ifndef T2
-//T2:parallel,loop,combined-constructs,V:2.0-2.7
-int test2(){
+//T2:parallel,loop,combined-constructs,syntactic,V:2.0-2.7
+//data dependent, treated with as a seq clause. Added the num_gangs clause with 1
+int test3(){
     int err = 0;
     srand(SEED);
-    real_t * a = (real_t *)malloc(n * sizeof(real_t));
-    real_t * a_copy = (real_t *)malloc(n * sizeof(real_t));
-    real_t * b = (real_t *)malloc(n * sizeof(real_t));
+    real_t * device = (real_t *)malloc(n * sizeof(real_t));
+    real_t * host = (real_t *)malloc(n * sizeof(real_t));
 
     for (int x = 0; x < n; ++x){
-        a[x] = rand() / (real_t)(RAND_MAX / 10);
-        a_copy[x] = a[x];
+        device[x] = rand() / (real_t)(RAND_MAX / 10);
+        host[x] = device[x];
     }
 
-    #pragma acc data copy(a[0:n])
+    #pragma acc data copy(device[0:n])
     {
-      #pragma acc parallel loop auto
+      #pragma acc parallel loop num_gangs(1) auto
+      for (int x = 1; x < n; ++x){
+        device[x] = device[x - 1] + device[x];
+      }
+    }
+
+    real_t rolling_total = 0.0;
+    for (int x = 0; x < n; ++x){
+      rolling_total += host[x];
+      if (fabs(rolling_total - device[x]) > PRECISION){
+        err = 1;
+      }
+    }
+
+    free(device);
+    free(host);
+
+    return err;
+}
+#endif
+
+#ifndef T3
+//T3:parallel,loop,combined-constructs,V:2.0-2.7
+//data dependent, treated with as a seq clause. 
+int test3(){
+    int err = 0;
+    srand(SEED);
+    real_t * device = (real_t *)malloc(n * sizeof(real_t));
+    real_t * host = (real_t *)malloc(n * sizeof(real_t));
+
+    for (int x = 0; x < n; ++x){
+        device[x] = rand() / (real_t)(RAND_MAX / 10);
+        host[x] = device[x];
+    }
+
+    #pragma acc data copy(device[0:n])
+    {
+      #pragma acc parallel loop num_gangs(1) vector worker auto
       for (int x = 1; x < n; ++x){
-        a[x] = a[x - 1] + a[x];
+        device[x] = device[x - 1] + device[x];
       }
     }
 
     real_t rolling_total = 0.0;
     for (int x = 0; x < n; ++x){
-      rolling_total += a_copy[x];
-      if (fabs(rolling_total - a[x]) > PRECISION){
+      rolling_total += host[x];
+      if (fabs(rolling_total - device[x]) > PRECISION){
         err = 1;
       }
     }
 
+    free(device);
+    free(host);
+
     return err;
 }
+
+
 #endif
 
 int main(){
@@ -71,19 +116,28 @@ int main(){
 #ifndef T1
     failed = 0;
     for (int x = 0; x < NUM_TEST_CALLS; ++x){
-        failed = failed + test1();
+        failed +=  test1();
     }
-    if (failed != 0){
-        failcode = failcode + (1 << 0);
+    if (failed){
+        failcode +=  (1 << 0);
     }
 #endif
 #ifndef T2
     failed = 0;
     for (int x = 0; x < NUM_TEST_CALLS; ++x){
-        failed = failed + test2();
+        failed += test2();
+    }
+    if (failed){
+        failcode += (1 << 1);
+    }
+#endif
+#ifndef T3
+    failed = 0;
+    for (int x = 0; x < NUM_TEST_CALLS; ++x){
+        failed += test3();
     }
-    if (failed != 0){
-        failcode = failcode + (1 << 1);
+    if (failed){
+        failcode += (1 << 2);
     }
 #endif
     return failcode;