[OpenMP][NFC] lit: Run barrier/reduction tests for each barrier implementation

[t-msn]

Summary

There are several barrier algorithms in libomp and each barrier also impliments
its reduction scheme. Therefore basic barrier/reduction tests should be run
for each barrier.

Update following tests' lit to run the tests for each barrier currently supported
in libomp (linear, tree, hyper, hierachical and distribution):

barrier/omp_barrier.c
ompt/synchronization/reduction/empty_reduce.c
ompt/synchronization/reduction/tree_reduce.c
parallel/omp_parallel_reduction.c

Depends on D123193

[t-msn]

There are following reduction tests:

$ LIT_OPTS="--show-pass --filter=reduction" ninja check-openmp
********************
Passed Tests (15):
  libarcher :: reduction/parallel-reduction-nowait.c
  libarcher :: reduction/parallel-reduction.c
  libomp :: misc_bugs/teams-reduction.c
  libomp :: ompt/synchronization/reduction/empty_reduce.c
  libomp :: ompt/synchronization/reduction/tree_reduce.c
  libomp :: parallel/omp_parallel_reduction.c
  libomp :: tasking/kmp_task_reduction_nest.cpp
  libomp :: tasking/task_reduction1.c
  libomp :: tasking/task_reduction2.c
  libomp :: tasking/task_reduction3.c
  libomp :: tasking/task_reduction4.c
  libomp :: worksharing/for/omp_for_reduction.c
  libomp :: worksharing/for/omp_parallel_for_reduction.c
  libomp :: worksharing/sections/omp_parallel_sections_reduction.c
  libomp :: worksharing/sections/omp_sections_reduction.c

I thought these threes are enough to check each barrier's basic functionality:

libomp :: ompt/synchronization/reduction/empty_reduce.c
libomp :: ompt/synchronization/reduction/tree_reduce.c
libomp :: parallel/omp_parallel_reduction.c

Let me know if this is wrong. Dependency of D123193 is to pass the tests.

openmp/runtime/test/barrier/omp_barrier.c

+// default is hyper barrier
 // RUN: %libomp-compile-and-run
 // RUN: %libomp-compile && env KMP_BLOCKTIME=infinite %libomp-run
-// RUN: %libomp-compile && env KMP_PLAIN_BARRIER_PATTERN='hierarchical,hierarchical' KMP_FORKJOIN_BARRIER_PATTERN='hierarchical,hierarchical' %libomp-run
-// RUN: %libomp-compile && env KMP_BLOCKTIME=infinite KMP_PLAIN_BARRIER_PATTERN='hierarchical,hierarchical' KMP_FORKJOIN_BARRIER_PATTERN='hierarchical,hierarchical' %libomp-run
-// RUN: %libomp-compile && env KMP_PLAIN_BARRIER_PATTERN='dist,dist' KMP_FORKJOIN_BARRIER_PATTERN='dist,dist' KMP_REDUCTION_BARRIER_PATTERN='dist,dist' %libomp-run
-// RUN: %libomp-compile && env KMP_BLOCKTIME=infinite KMP_PLAIN_BARRIER_PATTERN='dist,dist' KMP_FORKJOIN_BARRIER_PATTERN='dist,dist' KMP_REDUCTION_BARRIER_PATTERN='dist,dist' %libomp-run
+//
+// check other barriers
+// RUN: %libomp-compile && env %linear-barrier %libomp-run
+// RUN: %libomp-compile && env %linear-barrier KMP_BLOCKTIME=infinite %libomp-run
+// RUN: %libomp-compile && env %tree-barrier %libomp-run
+// RUN: %libomp-compile && env %tree-barrier KMP_BLOCKTIME=infinite %libomp-run
+// RUN: %libomp-compile && env %hier-barrier %libomp-run
+// RUN: %libomp-compile && env %hier-barrier KMP_BLOCKTIME=infinite %libomp-run
+// RUN: %libomp-compile && env %dist-barrier %libomp-run
+// RUN: %libomp-compile && env %dist-barrier KMP_BLOCKTIME=infinite %libomp-run
 #include <stdio.h>
 #include "omp_testsuite.h"
 #include "omp_my_sleep.h"
 
 int test_omp_barrier()
 {
   int result1;
   int result2;

openmp/runtime/test/lit.cfg

else:
     # If testing the runtime within an LLVM tree, then always include omp.h
     # directory associated with the new clang compiler.
     config.substitutions.append(("%flags-use-compiler-omp-h",
         config.test_flags))
 config.substitutions.append(("%flags", config.test_flags))
 config.substitutions.append(("%python", '"%s"' % (sys.executable)))
 config.substitutions.append(("%not", config.test_not))
 
+# shortcut for barrier enviroments
+config.substitutions.append(("%linear-barrier",
+        "KMP_PLAIN_BARRIER_PATTERN=linear,linear \
+        KMP_FORKJOIN_BARRIER_PATTERN=linear,linear \
+        KMP_REDUCTION_BARRIER_PATTERN=linear,linear"))
+config.substitutions.append(("%tree-barrier",
+    "KMP_PLAIN_BARRIER_PATTERN=tre,tree \
+     KMP_FORKJOIN_BARRIER_PATTERN=tree,tree \
+     KMP_REDUCTION_BARRIER_PATTERN=tree,tree"))
+config.substitutions.append(("%hier-barrier",
+    "KMP_PLAIN_BARRIER_PATTERN=hier,hier \
+    KMP_FORKJOIN_BARRIER_PATTERN=hier,hier \
+    KMP_REDUCTION_BARRIER_PATTERN=hier,hier"))
+config.substitutions.append(("%dist-barrier",
+    "KMP_PLAIN_BARRIER_PATTERN=dist,dist \
+    KMP_FORKJOIN_BARRIER_PATTERN=dist,dist \
+    KMP_REDUCTION_BARRIER_PATTERN=dist,dist"))
+
 if config.has_ompt:
     config.substitutions.append(("FileCheck", "tee %%t.out | %s" % config.test_filecheck))
     config.substitutions.append(("%sort-threads", "sort -n -s"))
     if config.operating_system == 'Windows':
         # No such environment variable on Windows.
         config.substitutions.append(("%preload-tool", "true ||"))
         config.substitutions.append(("%no-as-needed-flag", "-Wl,--no-as-needed"))
     elif config.operating_system == 'Darwin':
         config.substitutions.append(("%preload-tool", "env DYLD_INSERT_LIBRARIES=%T/tool.so"))
         # No such linker flag on Darwin.
         config.substitutions.append(("%no-as-needed-flag", ""))
     else:
         config.substitutions.append(("%preload-tool", "env LD_PRELOAD=%T/tool.so"))
         config.substitutions.append(("%no-as-needed-flag", "-Wl,--no-as-needed"))
 else:
     config.substitutions.append(("FileCheck", config.test_filecheck))

openmp/runtime/test/ompt/synchronization/reduction/empty_reduce.c

 // RUN: %libomp-compile-and-run | FileCheck %s
 // RUN: %libomp-compile -DNOWAIT && %libomp-run | FileCheck %s
+//
+// RUN: %libomp-compile && env %linear-barrier %libomp-run | FileCheck %s
+// RUN: %libomp-compile -DNOWAIT && env %linear-barrier %libomp-run | FileCheck %s
+// RUN: %libomp-compile && env %tree-barrier %libomp-run | FileCheck %s
+// RUN: %libomp-compile -DNOWAIT && env %tree-barrier %libomp-run | FileCheck %s
+// RUN: %libomp-compile && env %hier-barrier %libomp-run | FileCheck %s
+// RUN: %libomp-compile -DNOWAIT && env %hier-barrier %libomp-run | FileCheck %s
+// RUN: %libomp-compile && env %dist-barrier %libomp-run | FileCheck %s
+// RUN: %libomp-compile -DNOWAIT && env %dist-barrier %libomp-run | FileCheck %s
+//
 // REQUIRES: ompt
 // UNSUPPORTED: gcc
 #include "callback.h"
 #include <omp.h>
 
 #ifdef NOWAIT
 #define FOR_CLAUSE nowait
 #else

openmp/runtime/test/ompt/synchronization/reduction/tree_reduce.c

-// RUN: %libomp-compile-and-run | %sort-threads | FileCheck %s
-// RUN: %libomp-compile -DNOWAIT && %libomp-run | %sort-threads | FileCheck %s
+// RUN: %libomp-compile-and-run | %sort-threads | FileCheck --check-prefixes=CHECK,MULTIPLE %s
+// RUN: %libomp-compile -DNOWAIT && env %libomp-run | %sort-threads | FileCheck --check-prefixes=CHECK,MULTIPLE %s
+//
+// RUN: %libomp-compile && env %linear-barrier %libomp-run | %sort-threads | FileCheck --check-prefixes=CHECK,MULTIPLE %s
+// RUN: %libomp-compile -DNOWAIT && env %linear-barrier %libomp-run | %sort-threads | FileCheck --check-prefixes=CHECK,MULTIPLE  %s
+// RUN: %libomp-compile && env %tree-barrier %libomp-run | %sort-threads | FileCheck --check-prefixes=CHECK,MULTIPLE %s
+// RUN: %libomp-compile -DNOWAIT && env %tree-barrier %libomp-run | %sort-threads | FileCheck --check-prefixes=CHECK,MULTIPLE %s
+// RUN: %libomp-compile && env %hier-barrier %libomp-run | %sort-threads | FileCheck --check-prefixes=CHECK,MULTIPLE %s
+// RUN: %libomp-compile -DNOWAIT && env %hier-barrier %libomp-run | %sort-threads | FileCheck --check-prefixes=CHECK,MULTIPLE %s
+// As dist barrier calls 1 pair of callbacks for several reductions. only match to CHECK
+// RUN: %libomp-compile && env %dist-barrier %libomp-run | %sort-threads | FileCheck --check-prefixes=CHECK %s
+// RUN: %libomp-compile -DNOWAIT && env %dist-barrier %libomp-run | %sort-threads | FileCheck --check-prefixes=CHECK %s
+//
 // REQUIRES: ompt
 // UNSUPPORTED: gcc
 #include "callback.h"
 #include <omp.h>
 
 #ifdef NOWAIT
 #define FOR_CLAUSE nowait
 #else
#pragma omp for reduction(+ : sum, a, b) FOR_CLAUSE
   // CHECK: {{^}}[[MASTER_ID]]: ompt_event_implicit_task_begin:
   // CHECK-SAME: parallel_id=[[PARALLEL_ID]], task_id=[[TASK_ID:[0-9]+]]
 
   // order and distribution to threads not determined
   // CHECK: {{^}}{{[0-f]+}}: ompt_event_reduction_begin:
   // CHECK-SAME: parallel_id=[[PARALLEL_ID]], task_id={{[0-9]+}}
   // CHECK: {{^}}{{[0-f]+}}: ompt_event_reduction_end:
   // CHECK-SAME: parallel_id=[[PARALLEL_ID]], task_id={{[0-9]+}}
-  // CHECK: {{^}}{{[0-f]+}}: ompt_event_reduction_begin:
-  // CHECK-SAME: parallel_id=[[PARALLEL_ID]], task_id={{[0-9]+}}
-  // CHECK: {{^}}{{[0-f]+}}: ompt_event_reduction_end:
-  // CHECK-SAME: parallel_id=[[PARALLEL_ID]], task_id={{[0-9]+}}
-  // CHECK: {{^}}{{[0-f]+}}: ompt_event_reduction_begin:
-  // CHECK-SAME: parallel_id=[[PARALLEL_ID]], task_id={{[0-9]+}}
-  // CHECK: {{^}}{{[0-f]+}}: ompt_event_reduction_end:
-  // CHECK-SAME: parallel_id=[[PARALLEL_ID]], task_id={{[0-9]+}}
-  // CHECK: {{^}}{{[0-f]+}}: ompt_event_reduction_begin:
-  // CHECK-SAME: parallel_id=[[PARALLEL_ID]], task_id={{[0-9]+}}
-  // CHECK: {{^}}{{[0-f]+}}: ompt_event_reduction_end:
-  // CHECK-SAME: parallel_id=[[PARALLEL_ID]], task_id={{[0-9]+}}
+  //
+  // spec says there needs at least 1 pair of callbacks.
+  // therefore, following matches are optional
+  // MULTIPLE: {{^}}{{[0-f]+}}: ompt_event_reduction_begin:
+  // MULTIPLE-SAME: parallel_id=[[PARALLEL_ID]], task_id={{[0-9]+}}
+  // MULTIPLE: {{^}}{{[0-f]+}}: ompt_event_reduction_end:
+  // MULTIPLE-SAME: parallel_id=[[PARALLEL_ID]], task_id={{[0-9]+}}
+  // MULTIPLE: {{^}}{{[0-f]+}}: ompt_event_reduction_begin:
+  // MULTIPLE-SAME: parallel_id=[[PARALLEL_ID]], task_id={{[0-9]+}}
+  // MULTIPLE: {{^}}{{[0-f]+}}: ompt_event_reduction_end:
+  // MULTIPLE-SAME: parallel_id=[[PARALLEL_ID]], task_id={{[0-9]+}}
+  // MULTIPLE: {{^}}{{[0-f]+}}: ompt_event_reduction_begin:
+  // MULTIPLE-SAME: parallel_id=[[PARALLEL_ID]], task_id={{[0-9]+}}
+  // MULTIPLE: {{^}}{{[0-f]+}}: ompt_event_reduction_end:
+  // MULTIPLE-SAME: parallel_id=[[PARALLEL_ID]], task_id={{[0-9]+}}
 
   return 0;
 }

openmp/runtime/test/parallel/omp_parallel_reduction.c

 // RUN: %libomp-compile-and-run
+// RUN: %libomp-compile && env %linear-barrier %libomp-run
+// RUN: %libomp-compile && env %tree-barrier %libomp-run
+// RUN: %libomp-compile && env %hier-barrier %libomp-run
+// RUN: %libomp-compile && env %dist-barrier %libomp-run
 #include <stdio.h>
 #include <math.h>
 #include "omp_testsuite.h"
 
 #define DOUBLE_DIGITS 20    /* dt^DOUBLE_DIGITS */
 #define MAX_FACTOR 10
 #define KNOWN_PRODUCT 3628800  /* 10! */