[OpenMP] Initial testsuite change to purely llvm-lit based testing

[jlpeyton]

Summary

This change introduces a check-libomp target which is based upon llvm's lit test infrastructure. Each test (generated from the University of Houston's OpenMP testsuite) is compiled and then run. For each test, an exit status of 0 indicates success and non-zero indicates failure. This way, FileCheck is not needed. I've added a bit of logic to generate symlinks (libiomp5 and libgomp) in the build tree so that gcc can be tested as well.

[gribozavr]

runtime/test/api/has_openmp.c

2

Instead of duplicating the RUN line everywhere, I'd suggest to introduce a substitution, say, %omp-run-simple, that would be used in all "typical" tests. Then, the tests that need special handling will have a custom run line.

[jlpeyton]

Addressing Dmitri Gribenko's suggestion by creating %libomp-compile-and-run substitution.

[gribozavr]

runtime/test/lit.cfg

54

You'd need a space between the compiler and the flags.

Also, why not use the %clang and %cflags in the %libomp-compile-and-run substitution?

[jlpeyton]

You'd need a space between the compiler and the flags.

There was a prepended space in the config.test_cflags variable, but I have moved it to make it clearer.

Also, why not use the %clang and %cflags in the %libomp-compile-and-run substitution?

I could not get it to substitute the values. llvm-lit would try to execute '%clang' which of course failed, and %cflags would not substitute either. Fortunately, %t and %s did substitute like normal.

[gribozavr]

I could not get it to substitute the values. llvm-lit would try to execute '%clang' which of course failed, and %cflags would not substitute either. Fortunately, %t and %s did substitute like normal.

I think that's because lit applies substitutions first-to-last. Try putting %clang and %cflags last.

[jlpeyton]

I think that's because lit applies substitutions first-to-last. Try putting %clang and %cflags last.

That did the trick, thanks!

[hfinkel]

I see code in the lit/cmake files to handle windows, does this currently work? The tests seem to include unistd.h.

[jlpeyton]

I see code in the lit/cmake files to handle windows, does this currently work? The tests seem to include unistd.h.

Currently, it is Unices and Mac only. I only put the Windows logic in there in anticipation that it will be supported at some point. Right now, users can't easily use libomp.so on Windows because the clang front end doesn't include it when using -fopenmp (libomp.lib has to be explicitly linked).

[hfinkel]

In D11821#224562, @jlpeyton wrote:

I see code in the lit/cmake files to handle windows, does this currently work? The tests seem to include unistd.h.

Currently, it is Unices and Mac only. I only put the Windows logic in there in anticipation that it will be supported at some point. Right now, users can't easily use libomp.so on Windows because the clang front end doesn't include it when using -fopenmp (libomp.lib has to be explicitly linked).

Okay (although there is a patch out to fix the libomp.lib problem, the larger problem is that the tests themselves won't work because of the unistd.h dependency). Although we normally frown on committing dead code, in this particular case, the couple of lines seems straightforward, so I don't have a strong opinion. We should probably have it explicitly fail on Windows, however, with some informative 'Regression tests not supported yet' message instead of having the user run all of the tests and having them all appear to fail. We can remove this once we actually update the tests to compile on Windows.

(I've also added Chandler to the review; he told me he'd look at this soon).

[jlpeyton]

Okay (although there is a patch out to fix the libomp.lib problem, the larger problem is that the tests themselves won't work because of the unistd.h dependency)

I've rid the testsuite of unistd.h (which isn't needed) and added another my_sleep() function which is Windows compatible.

[hfinkel]

runtime/test/omp_my_sleep.h

13

Why do we need windows.h here? clock() is in time.h.

53

I don't understand why we're sleeping this way -- it is not really sleeping, but rather busy-waiting on the clock. Shouldn't we use something like nanosleep() instead?

[hfinkel]

runtime/test/atomic/omp_atomic.c

5

my_sleep is not called in this file.

runtime/test/critical/omp_critical.c

4

my_sleep is not called in this file.

runtime/test/tasking/omp_task_if.c

22

Please remove commented-out code.

26

Here too (or maybe this shouldn't be commented?)

runtime/test/worksharing/for/omp_for_schedule_dynamic.c

15

my_sleep is not called in this file.

[jlpeyton]

Why do we need windows.h here? clock() is in time.h.
I don't understand why we're sleeping this way -- it is not really sleeping, but rather busy-waiting on the clock. Shouldn't we use something like nanosleep() instead?

I've changed the my_sleep() function to actually sleep instead of busy wait. Windows uses Sleep() (which requires windows.h now) and Unices will use nanosleep().

my_sleep is not called in this file.

Removed.

my_sleep is not called in this file.

Removed.

Please remove commented-out code.
Here too (or maybe this shouldn't be commented?)

Changed to critical section

my_sleep is not called in this file.

Removed.

[jlpeyton]

Ping

[hfinkel]

LGTM, thanks!

[• jhowarth]

Can we get a back port of Commit rL248211 to the openmp 3.7 release branch for the 3.7.1 release? It applies cleanly here over openmp 3.7.0 and shows perfect results...

Testing Time: 27.55s

Expected Passes    : 66

[100%] Built target check-libomp

against clang 3.7.0.

[• jhowarth]

There is one issue with rL248211. On darwin, the check-libomp test suite run depends on DYLD_LIBRARY_PATH being passed along through make. This isn't true in El Capitan due to the System Integrity Protection pruning of DYLD_LIBRARY_PATH.

https://developer.apple.com/library/prerelease/ios/documentation/Security/Conceptual/System_Integrity_Protection_Guide/System_Integrity_Protection_Guide.pdf

The fix is to modify the cmake build such that "-Wl,-rpath,$libomp_path" is appended to LIBOMP_TEST_CFLAGS on darwin.

[• jhowarth]

One other issue. It appears that the tolerances used in runtime/test/api/omp_get_wtime.c are too tight for OS X 10.11. With or without SIP enabled, I get a return code between 2 to 6 on that test. Opening it up to...

return ((measured_time > 0.98 * wait_time) && (measured_time < 1.02 * wait_time)) ;

has the return code oscillate between returning 0 or 1.

return ((measured_time > 0.97 * wait_time) && (measured_time < 1.03 * wait_time)) ;

on my MacPro 3,1 dual quad-core is required under 10.11 to have the return code always report back as 0.

runtime/CMakeLists.txt

elseif("${LIBOMP_LIB_TYPE}" STREQUAL "stubs")
   set(STUBS_LIBRARY TRUE)
 endif()
 
 # Setting directory names
 set(LIBOMP_BASE_DIR ${CMAKE_CURRENT_SOURCE_DIR})
 set(LIBOMP_SRC_DIR ${LIBOMP_BASE_DIR}/src)
 set(LIBOMP_TOOLS_DIR ${LIBOMP_BASE_DIR}/tools)
 set(LIBOMP_INC_DIR ${LIBOMP_SRC_DIR}/include/${LIBOMP_OMP_VERSION})
+set(LIBOMP_BINARY_DIR ${CMAKE_CURRENT_BINARY_DIR})
 
 # Enabling Fortran if it is needed
 if(${LIBOMP_FORTRAN_MODULES})
   enable_language(Fortran)
 endif()
 # Enable MASM Compiler if it is needed (Windows only)
 if(WIN32)
   enable_language(ASM_MASM)
if(${LIBOMP_OMPT_SUPPORT})
     libomp_say("Use OMPT-blame       -- ${LIBOMP_OMPT_BLAME}")
     libomp_say("Use OMPT-trace       -- ${LIBOMP_OMPT_TRACE}")
   endif()
   libomp_say("Use Adaptive locks   -- ${LIBOMP_USE_ADAPTIVE_LOCKS}")
   libomp_say("Use quad precision   -- ${LIBOMP_USE_QUAD_PRECISION}")
 endif()
 
 add_subdirectory(src)
+add_subdirectory(test)
 

runtime/src/CMakeLists.txt

 add_library(omp SHARED ${LIBOMP_SOURCE_FILES})
 set_target_properties(omp PROPERTIES
   PREFIX "" SUFFIX "" OUTPUT_NAME "${LIBOMP_LIB_FILE}"
   LINK_FLAGS "${LIBOMP_CONFIGURED_LDFLAGS}"
   LINKER_LANGUAGE ${LIBOMP_LINKER_LANGUAGE}
   SKIP_BUILD_RPATH true # have Mac linker -install_name just be "-install_name libomp.dylib"
 )
 
+# Get the library's location within the build tree for the unit tester
+get_target_property(LIBOMP_LIBRARY_DIR omp LIBRARY_OUTPUT_DIRECTORY)
+if(NOT LIBOMP_LIBRARY_DIR)
+  set(LIBOMP_LIBRARY_DIR ${CMAKE_CURRENT_BINARY_DIR})
+  set(LIBOMP_LIBRARY_DIR ${CMAKE_CURRENT_BINARY_DIR} PARENT_SCOPE)
+else()
+  set(LIBOMP_LIBRARY_DIR ${LIBOMP_LIBRARY_DIR} PARENT_SCOPE)
+endif()
+
+# Add symbolic links to libomp
+if(NOT WIN32)
+  add_custom_command(TARGET omp POST_BUILD
+    COMMAND ${CMAKE_COMMAND} -E create_symlink ${LIBOMP_LIB_FILE}
+      libgomp${CMAKE_SHARED_LIBRARY_SUFFIX}
+    COMMAND ${CMAKE_COMMAND} -E create_symlink ${LIBOMP_LIB_FILE}
+      libiomp5${CMAKE_SHARED_LIBRARY_SUFFIX}
+    WORKING_DIRECTORY ${LIBOMP_LIBRARY_DIR}
+  )
+endif()
+
 # Linking command will include libraries in LIBOMP_CONFIGURED_LIBFLAGS
 libomp_get_libflags(LIBOMP_CONFIGURED_LIBFLAGS)
 target_link_libraries(omp ${LIBOMP_CONFIGURED_LIBFLAGS} ${CMAKE_DL_LIBS})
 
 # Create *.inc and omp.h before compiling any sources
 # objects depend on : .inc files and omp.h (and ompt.h if LIBOMP_OMPT_SUPPORT is on)
 # This way the *.inc and omp.h are generated before any compilations take place
 set(LIBOMP_NEEDED_HEADERS kmp_i18n_id.inc kmp_i18n_default.inc omp.h)

runtime/test/CMakeLists.txt

+# CMakeLists.txt file for unit testing OpenMP Library
+include(FindPythonInterp)
+include(CheckTypeSize)
+if(NOT PYTHONINTERP_FOUND)
+  libomp_warning_say("Could not find Python.")
+  libomp_warning_say("The check-libomp target will not be available!")
+  return()
+endif()
+
+set(LIBOMP_TEST_CFLAGS "" CACHE STRING
+  "Extra compiler flags to send to the test compiler")
+
+if(${LIBOMP_STANDALONE_BUILD})
+  # Make sure we can use the console pool for recent cmake and ninja > 1.5
+  if(CMAKE_VERSION VERSION_LESS 3.1.20141117)
+    set(cmake_3_2_USES_TERMINAL)
+  else()
+    set(cmake_3_2_USES_TERMINAL USES_TERMINAL)
+  endif()
+  set(LIBOMP_TEST_COMPILER ${CMAKE_C_COMPILER} CACHE STRING
+    "Compiler to use for testing OpenMP library")
+  set(LIBOMP_TEST_OPENMP_FLAG -fopenmp CACHE STRING
+    "OpenMP compiler flag to use for testing OpenMP library")
+  set(LIBOMP_LLVM_LIT_EXECUTABLE "" CACHE STRING
+    "Path to llvm-lit")
+  find_program(LIT_EXECUTABLE NAMES llvm-lit ${LIBOMP_LLVM_LIT_EXECUTABLE})
+  if(NOT LIT_EXECUTABLE)
+    libomp_warning_say("Cannot find llvm-lit.")
+    libomp_warning_say("Please put llvm-lit in your PATH or set LIBOMP_LLVM_LIT_EXECUTABLE to its full path")
+    libomp_warning_say("The check-libomp target will not be available!")
+    return()
+  endif()
+  # Set lit arguments
+  # The -j 1 lets the actual tests run with the entire machine.
+  # We have one test thread that spawns the tests serially.  This allows
+  # Each test to use the entire machine.
+  set(LIBOMP_LIT_ARGS_DEFAULT "-sv --show-unsupported --show-xfail -j 1")
+  if(MSVC OR XCODE)
+    set(LIBOMP_LIT_ARGS_DEFAULT "${LIBOMP_LIT_ARGS_DEFAULT} --no-progress-bar")
+  endif()
+  set(LIBOMP_LIT_ARGS "${LIBOMP_LIT_ARGS_DEFAULT}" CACHE STRING
+    "Default options for lit")
+  separate_arguments(LIBOMP_LIT_ARGS)
+  add_custom_target(check-libomp
+    COMMAND ${LIT_EXECUTABLE} ${LIBOMP_LIT_ARGS} ${CMAKE_CURRENT_BINARY_DIR}
+    DEPENDS omp
+    COMMENT "Running libomp tests"
+    ${cmake_3_2_USES_TERMINAL}
+  )
+else()
+  # LLVM source tree build, test just-built clang
+  if(NOT MSVC)
+    set(LIBOMP_TEST_COMPILER ${LLVM_RUNTIME_OUTPUT_INTDIR}/clang)
+  else()
+    set(LIBOMP_TEST_COMPILER ${LLVM_RUNTIME_OUTPUT_INTDIR}/clang.exe)
+  endif()
+  set(LIBOMP_TEST_OPENMP_FLAG -fopenmp=libomp)
+  # Use add_lit_testsuite() from LLVM CMake.
+  add_lit_testsuite(check-libomp 
+    "Running libomp tests"
+    ${CMAKE_CURRENT_BINARY_DIR}
+    ARGS "-j 1"
+    DEPENDS omp
+  )
+endif()
+
+# Configure the lit.site.cfg.in file
+set(AUTO_GEN_COMMENT "## Autogenerated by libomp configuration.\n# Do not edit!")
+configure_file(lit.site.cfg.in lit.site.cfg @ONLY)
+

runtime/test/api/has_openmp.c

+// RUN: %libomp-compile-and-run
+#include <stdio.h>
+#include <stdlib.h>
+#include "omp_testsuite.h"
+
+int test_has_openmp()
+{
+  int rvalue = 0;
+#ifdef _OPENMP
+  rvalue = 1;
+#endif
+  return (rvalue);
+}
+
+int main()
+{
+  int i;
+  int num_failed=0;
+  if(!test_has_openmp()) {
+    num_failed++;
+  }
+  return num_failed;
+}

runtime/test/api/omp_get_num_threads.c

+// RUN: %libomp-compile-and-run
+#include <stdio.h>
+#include "omp_testsuite.h"
+
+int test_omp_get_num_threads()
+{
+  /* checks that omp_get_num_threads is equal to the number of
+     threads */
+  int nthreads_lib;
+  int nthreads = 0;
+
+  nthreads_lib = -1;
+
+  #pragma omp parallel 
+  {
+    #pragma omp critical
+    {
+      nthreads++;
+    } /* end of critical */
+    #pragma omp single
+    { 
+      nthreads_lib = omp_get_num_threads ();
+    }  /* end of single */
+  } /* end of parallel */
+  return (nthreads == nthreads_lib);
+}
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for(i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_get_num_threads()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/api/omp_get_wtick.c

+// RUN: %libomp-compile-and-run
+#include <stdio.h>
+#include "omp_testsuite.h"
+
+int test_omp_get_wtick()
+{
+  double tick;
+  tick = -1.;
+  tick = omp_get_wtick ();
+  return ((tick > 0.0) && (tick < 0.01));
+}
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for(i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_get_wtick()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/api/omp_get_wtime.c

+// RUN: %libomp-compile-and-run
+#include <stdio.h>
+#include <stdlib.h>
+#include "omp_testsuite.h"
+#include "omp_my_sleep.h"
+
+int test_omp_get_wtime()
+{
+  double start;
+  double end;
+  double measured_time;
+  double wait_time = 0.25; 
+  start = 0;
+  end = 0;
+  start = omp_get_wtime();
+  my_sleep (wait_time); 
+  end = omp_get_wtime();
+  measured_time = end-start;
+  return ((measured_time > 0.99 * wait_time) && (measured_time < 1.01 * wait_time)) ;
+}
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for(i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_get_wtime()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/api/omp_in_parallel.c

+// RUN: %libomp-compile-and-run
+#include <stdio.h>
+#include "omp_testsuite.h"
+
+/*
+ * Checks that false is returned when called from serial region
+ * and true is returned when called within parallel region. 
+ */
+int test_omp_in_parallel()
+{
+  int serial;
+  int isparallel;
+
+  serial = 1;
+  isparallel = 0;
+  serial = omp_in_parallel();
+
+  #pragma omp parallel
+  {
+    #pragma omp single
+    {
+      isparallel = omp_in_parallel();
+    }
+  }
+  return (!(serial) && isparallel);
+}
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for(i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_in_parallel()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/atomic/omp_atomic.c

+// RUN: %libomp-compile-and-run
+#include <stdio.h>
+#include <math.h>
+#include "omp_testsuite.h"
+#include "omp_my_sleep.h"
+
+#define DOUBLE_DIGITS 20  /* dt^DOUBLE_DIGITS */
+#define MAX_FACTOR 10
+#define KNOWN_PRODUCT 3628800  /* 10! */
+
+int test_omp_atomic()
+{
+  int sum;
+  int diff;
+  double dsum = 0;
+  double dt = 0.5;  /* base of geometric row for + and - test*/
+  double ddiff;
+  int product;
+  int x;
+  int *logics;
+  int bit_and = 1;
+  int bit_or = 0;
+  int exclusiv_bit_or = 0;
+  int j;
+  int known_sum;
+  int known_diff;
+  int known_product;
+  int result = 0;
+  int logic_and = 1;
+  int logic_or = 0;
+  double dknown_sum;
+  double rounding_error = 1.E-9;
+  double dpt, div;
+  int logicsArray[LOOPCOUNT];
+  logics = logicsArray;
+  
+  sum = 0;
+  diff = 0;
+  product = 1;
+
+  // sum of integers test
+  #pragma omp parallel
+  {
+    int i;
+    #pragma omp for
+    for (i = 1; i <= LOOPCOUNT; i++) {
+      #pragma omp atomic
+      sum += i;
+    }
+  
+  }
+  known_sum = (LOOPCOUNT * (LOOPCOUNT + 1)) / 2;
+  if (known_sum != sum)
+  {
+    fprintf(stderr, 
+      "Error in sum with integers: Result was %d instead of %d.\n", 
+      sum, known_sum);
+    result++;
+  }
+  
+  // difference of integers test
+  #pragma omp parallel
+  {
+    int i;
+    #pragma omp for
+    for (i = 0; i < LOOPCOUNT; i++) {
+      #pragma omp atomic
+      diff -= i;
+    }
+  }
+  known_diff = ((LOOPCOUNT - 1) * LOOPCOUNT) / 2 * -1;
+  if (diff != known_diff)
+  {
+    fprintf (stderr,
+      "Error in difference with integers: Result was %d instead of 0.\n",
+      diff);
+    result++;
+  }
+
+  // sum of doubles test
+  dsum = 0;
+  dpt = 1;
+  for (j = 0; j < DOUBLE_DIGITS; ++j) {
+    dpt *= dt;
+  }
+  dknown_sum = (1 - dpt) / (1 -dt);
+  #pragma omp parallel
+  {
+    int i;
+    #pragma omp for
+    for (i = 0; i < DOUBLE_DIGITS; ++i) {
+      #pragma omp atomic
+      dsum += pow (dt, i);
+    }
+  }
+  if (dsum != dknown_sum && (fabs (dsum - dknown_sum) > rounding_error)) {
+    fprintf (stderr, "Error in sum with doubles: Result was %f"
+      " instead of: %f (Difference: %E)\n",
+      dsum, dknown_sum, dsum - dknown_sum);
+    result++;
+  }
+
+  // difference of doubles test
+  dpt = 1;
+  for (j = 0; j < DOUBLE_DIGITS; ++j) {
+    dpt *= dt;
+  }
+  ddiff = (1 - dpt) / (1 - dt);
+  #pragma omp parallel
+  {
+    int i;
+    #pragma omp for
+    for (i = 0; i < DOUBLE_DIGITS; ++i) {
+      #pragma omp atomic
+      ddiff -= pow (dt, i);
+    }
+  } 
+  if (fabs (ddiff) > rounding_error) {
+    fprintf (stderr,
+      "Error in difference with doubles: Result was %E instead of 0.0\n",
+      ddiff);
+    result++;
+  }
+
+  // product of integers test
+  #pragma omp parallel
+  {
+    int i;
+    #pragma omp for
+    for (i = 1; i <= MAX_FACTOR; i++) {
+      #pragma omp atomic
+      product *= i;
+    }
+  }
+  known_product = KNOWN_PRODUCT;
+  if (known_product != product) {
+    fprintf (stderr,
+      "Error in product with integers: Result was %d instead of %d\n",
+      product, known_product);
+    result++;
+  }
+
+  // division of integers test
+  product = KNOWN_PRODUCT;
+  #pragma omp parallel
+  {
+     int i;
+    #pragma omp for
+    for (i = 1; i <= MAX_FACTOR; ++i) {
+      #pragma omp atomic
+      product /= i;
+    }
+  }
+  if (product != 1) {
+    fprintf (stderr,
+      "Error in product division with integers: Result was %d"
+      " instead of 1\n",
+      product);
+    result++;
+  }
+  
+  // division of doubles test
+  div = 5.0E+5;
+  #pragma omp parallel
+  {
+    int i;
+    #pragma omp for
+    for (i = 1; i <= MAX_FACTOR; i++) {
+      #pragma omp atomic
+      div /= i;
+    }
+  }
+  if (fabs(div-0.137787) >= 1.0E-4 ) {
+    result++;
+    fprintf (stderr, "Error in division with double: Result was %f"
+      " instead of 0.137787\n", div);
+  }
+
+  // ++ test
+  x = 0;
+  #pragma omp parallel
+  {
+    int i;
+    #pragma omp for
+    for (i = 0; i < LOOPCOUNT; ++i) {
+      #pragma omp atomic
+      x++;
+    }
+  }
+  if (x != LOOPCOUNT) {
+    result++;
+    fprintf (stderr, "Error in ++\n");
+  }
+
+  // -- test
+  #pragma omp parallel
+  {
+    int i;
+    #pragma omp for
+    for (i = 0; i < LOOPCOUNT; ++i) {
+      #pragma omp atomic
+      x--;
+    }
+  }
+  if (x != 0) {
+    result++;
+    fprintf (stderr, "Error in --\n");
+  }
+
+  // bit-and test part 1
+  for (j = 0; j < LOOPCOUNT; ++j) {
+    logics[j] = 1;
+  }
+  bit_and = 1;
+  #pragma omp parallel
+  {
+     int i;
+    #pragma omp for
+    for (i = 0; i < LOOPCOUNT; ++i) {
+      #pragma omp atomic
+      bit_and &= logics[i];
+    }
+  }
+  if (!bit_and) {
+    result++;
+    fprintf (stderr, "Error in BIT AND part 1\n");
+  }
+
+  // bit-and test part 2
+  bit_and = 1;
+  logics[LOOPCOUNT / 2] = 0;
+  #pragma omp parallel
+  {
+    int i;
+    #pragma omp for
+    for (i = 0; i < LOOPCOUNT; ++i) {
+      #pragma omp atomic
+      bit_and &= logics[i];
+    }
+  }
+  if (bit_and) {
+    result++;
+    fprintf (stderr, "Error in BIT AND part 2\n");
+  }
+
+  // bit-or test part 1
+  for (j = 0; j < LOOPCOUNT; j++) {
+    logics[j] = 0;
+  }
+  bit_or = 0;
+  #pragma omp parallel
+  {
+    int i;
+    #pragma omp for
+    for (i = 0; i < LOOPCOUNT; ++i) {
+      #pragma omp atomic
+      bit_or |= logics[i];
+    }
+  }
+  if (bit_or) {
+    result++;
+    fprintf (stderr, "Error in BIT OR part 1\n");
+  }
+
+  // bit-or test part 2
+  bit_or = 0;
+  logics[LOOPCOUNT / 2] = 1;
+  #pragma omp parallel
+  {
+    
+    int i;
+    #pragma omp for
+    for (i = 0; i < LOOPCOUNT; ++i) {
+      #pragma omp atomic
+      bit_or |= logics[i];
+    }
+  }
+  if (!bit_or) {
+    result++;
+    fprintf (stderr, "Error in BIT OR part 2\n");
+  }
+
+  // bit-xor test part 1
+  for (j = 0; j < LOOPCOUNT; j++) {
+    logics[j] = 0;
+  }
+  exclusiv_bit_or = 0;
+  #pragma omp parallel
+  {
+    int i;
+    #pragma omp for
+    for (i = 0; i < LOOPCOUNT; ++i) {
+      #pragma omp atomic
+      exclusiv_bit_or ^= logics[i];
+    }
+  }
+  if (exclusiv_bit_or) {
+    result++;
+    fprintf (stderr, "Error in EXCLUSIV BIT OR part 1\n");
+  }
+
+  // bit-xor test part 2
+  exclusiv_bit_or = 0;
+  logics[LOOPCOUNT / 2] = 1;
+  #pragma omp parallel
+  {
+    int i;
+    #pragma omp for
+    for (i = 0; i < LOOPCOUNT; ++i) {
+      #pragma omp atomic
+      exclusiv_bit_or ^= logics[i];
+    }
+    
+  }
+  if (!exclusiv_bit_or) {
+    result++;
+    fprintf (stderr, "Error in EXCLUSIV BIT OR part 2\n");
+  }
+
+  // left shift test
+  x = 1;
+  #pragma omp parallel
+  {
+    int i;
+    #pragma omp for
+    for (i = 0; i < 10; ++i) {
+      #pragma omp atomic
+      x <<= 1;
+    }
+    
+  }
+  if ( x != 1024) {
+    result++;
+    fprintf (stderr, "Error in <<\n");
+    x = 1024;
+  }
+
+  // right shift test
+  #pragma omp parallel
+  {
+    int i;
+    #pragma omp for
+    for (i = 0; i < 10; ++i) {
+      #pragma omp atomic
+      x >>= 1;
+    }
+  }
+  if (x != 1) {
+    result++;
+    fprintf (stderr, "Error in >>\n");
+  }
+
+  return (result == 0);
+} // test_omp_atomic()
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for(i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_atomic()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/barrier/omp_barrier.c

+// RUN: %libomp-compile-and-run
+#include <stdio.h>
+#include "omp_testsuite.h"
+#include "omp_my_sleep.h"
+
+int test_omp_barrier()
+{
+  int result1;
+  int result2;
+  result1 = 0;
+  result2 = 0;
+
+  #pragma omp parallel
+  {
+    int rank;
+    rank = omp_get_thread_num ();
+    if (rank ==1) {
+      my_sleep(SLEEPTIME);
+      result2 = 3;
+    }
+    #pragma omp barrier
+    if (rank == 2) {
+      result1 = result2;
+    }
+  }
+  return (result1 == 3);
+}
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for(i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_barrier()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/critical/omp_critical.c

+// RUN: %libomp-compile-and-run
+#include <stdio.h>
+#include "omp_testsuite.h"
+#include "omp_my_sleep.h"
+
+int test_omp_critical()
+{
+  int sum;
+  int known_sum;
+
+  sum=0;
+  #pragma omp parallel
+  {
+    int mysum=0;
+    int i;
+    #pragma omp for
+    for (i = 0; i < 1000; i++)
+      mysum = mysum + i;
+
+    #pragma omp critical
+    sum = mysum +sum;
+  }
+  known_sum = 999 * 1000 / 2;
+  return (known_sum == sum);
+}
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for(i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_critical()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/flush/omp_flush.c

+// RUN: %libomp-compile-and-run
+#include <stdio.h>
+#include "omp_testsuite.h"
+#include "omp_my_sleep.h"
+
+int test_omp_flush()
+{
+  int result1;
+  int result2;
+  int dummy;
+
+  result1 = 0;
+  result2 = 0;
+
+  #pragma omp parallel
+  {
+    int rank;
+    rank = omp_get_thread_num ();
+    #pragma omp barrier
+    if (rank == 1) {
+      result2 = 3;
+      #pragma omp flush (result2)
+      dummy = result2;
+    }
+    if (rank == 0) {
+      my_sleep(SLEEPTIME);
+      #pragma omp flush (result2)
+      result1 = result2;
+    }
+  }  /* end of parallel */
+  return ((result1 == result2) && (result2 == dummy) && (result2 == 3));
+}
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for (i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_flush()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/lit.cfg

+# -*- Python -*- vim: set ft=python ts=4 sw=4 expandtab tw=79:
+# Configuration file for the 'lit' test runner.
+
+import os
+import lit.formats
+
+# Tell pylint that we know config and lit_config exist somewhere.
+if 'PYLINT_IMPORT' in os.environ:
+    config = object()
+    lit_config = object()
+
+def append_dynamic_library_path(name, value, sep):
+    if name in config.environment:
+        config.environment[name] = value + sep + config.environment[name]
+    else:
+        config.environment[name] = value
+
+# name: The name of this test suite.
+config.name = 'libomp'
+
+# suffixes: A list of file extensions to treat as test files.
+config.suffixes = ['.c']
+
+# test_source_root: The root path where tests are located.
+config.test_source_root = os.path.dirname(__file__)
+
+# test_exec_root: The root object directory where output is placed
+config.test_exec_root = config.libomp_obj_root
+
+# test format
+config.test_format = lit.formats.ShTest()
+
+# compiler flags
+config.test_cflags = config.test_openmp_flag + \
+    " -I " + config.test_source_root + \
+    " -I " + config.omp_header_directory + \
+    " -L " + config.library_dir + \
+    " " + config.test_extra_cflags
+
+# Setup environment to find dynamic library at runtime
+if config.operating_system == 'Windows':
+    append_dynamic_library_path('PATH', config.library_dir, ";")
+elif config.operating_system == 'Darwin':
+    append_dynamic_library_path('DYLD_LIBRARY_PATH', config.library_dir, ":")
+else: # Unices
+    append_dynamic_library_path('LD_LIBRARY_PATH', config.library_dir, ":")
+
+# substitutions
+config.substitutions.append(("%libomp-compile-and-run", \
+    "%clang %cflags %s -o %t -lm && %t"))
+config.substitutions.append(("%clang", config.test_compiler))
+config.substitutions.append(("%openmp_flag", config.test_openmp_flag))
+config.substitutions.append(("%cflags", config.test_cflags))
+

runtime/test/lit.site.cfg.in

+@AUTO_GEN_COMMENT@
+
+config.test_compiler = "@LIBOMP_TEST_COMPILER@"
+config.test_openmp_flag = "@LIBOMP_TEST_OPENMP_FLAG@"
+config.test_extra_cflags = "@LIBOMP_TEST_CFLAGS@"
+config.libomp_obj_root = "@CMAKE_CURRENT_BINARY_DIR@"
+config.library_dir = "@LIBOMP_LIBRARY_DIR@"
+config.omp_header_directory = "@LIBOMP_BINARY_DIR@/src"
+config.operating_system = "@CMAKE_SYSTEM_NAME@"
+
+# Let the main config do the real work.
+lit_config.load_config(config, "@LIBOMP_BASE_DIR@/test/lit.cfg")
+

runtime/test/lock/omp_lock.c

+// RUN: %libomp-compile-and-run
+#include <stdio.h>
+#include "omp_testsuite.h"
+
+omp_lock_t lck;
+
+int test_omp_lock()
+{
+  int nr_threads_in_single = 0;
+  int result = 0;
+  int nr_iterations = 0;
+  int i;
+
+  omp_init_lock(&lck);
+  #pragma omp parallel shared(lck)
+  {
+    #pragma omp for
+    for(i = 0; i < LOOPCOUNT; i++) {
+      omp_set_lock(&lck);
+      #pragma omp flush
+      nr_threads_in_single++;
+      #pragma omp flush       
+      nr_iterations++;
+      nr_threads_in_single--;
+      result = result + nr_threads_in_single;
+      omp_unset_lock(&lck);
+    }
+  }
+  omp_destroy_lock(&lck);
+
+  return ((result == 0) && (nr_iterations == LOOPCOUNT));
+}
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for(i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_lock()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/lock/omp_nest_lock.c

+// RUN: %libomp-compile-and-run
+#include <stdio.h>
+#include "omp_testsuite.h"
+
+omp_nest_lock_t lck;
+
+int test_omp_nest_lock()
+{
+  int nr_threads_in_single = 0;
+  int result = 0;
+  int nr_iterations = 0;
+  int i;
+
+  omp_init_nest_lock(&lck);
+  #pragma omp parallel shared(lck)  
+  {
+    #pragma omp for
+    for(i = 0; i < LOOPCOUNT; i++) {
+      omp_set_nest_lock(&lck);
+      #pragma omp flush
+      nr_threads_in_single++;
+      #pragma omp flush       
+      nr_iterations++;
+      nr_threads_in_single--;
+      result = result + nr_threads_in_single;
+      omp_unset_nest_lock(&lck);
+    }
+  }
+  omp_destroy_nest_lock(&lck);
+
+  return ((result == 0) && (nr_iterations == LOOPCOUNT));
+}
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for(i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_nest_lock()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/lock/omp_test_lock.c

+// RUN: %libomp-compile-and-run
+#include <stdio.h>
+#include "omp_testsuite.h"
+
+omp_lock_t lck;
+
+int test_omp_test_lock()
+{
+  int nr_threads_in_single = 0;
+  int result = 0;
+  int nr_iterations = 0;
+  int i;
+
+  omp_init_lock (&lck);
+  #pragma omp parallel shared(lck)  
+  {
+    #pragma omp for
+    for (i = 0; i < LOOPCOUNT; i++) {
+      while (!omp_test_lock (&lck))
+      {};
+      #pragma omp flush
+      nr_threads_in_single++;
+      #pragma omp flush       
+      nr_iterations++;
+      nr_threads_in_single--;
+      result = result + nr_threads_in_single;
+      omp_unset_lock (&lck);
+    }
+  }
+  omp_destroy_lock(&lck);
+  return ((result == 0) && (nr_iterations == LOOPCOUNT));
+}
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for(i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_test_lock()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/lock/omp_test_nest_lock.c

+// RUN: %libomp-compile-and-run
+#include <stdio.h>
+#include "omp_testsuite.h"
+
+static omp_nest_lock_t lck;
+
+int test_omp_test_nest_lock()
+{
+  int nr_threads_in_single = 0;
+  int result = 0;
+  int nr_iterations = 0;
+  int i;
+
+  omp_init_nest_lock (&lck);
+  #pragma omp parallel shared(lck) 
+  {
+    #pragma omp for
+    for (i = 0; i < LOOPCOUNT; i++)
+    {
+      /*omp_set_lock(&lck);*/
+      while(!omp_test_nest_lock (&lck))
+      {};
+      #pragma omp flush
+      nr_threads_in_single++;
+      #pragma omp flush       
+      nr_iterations++;
+      nr_threads_in_single--;
+      result = result + nr_threads_in_single;
+      omp_unset_nest_lock (&lck);
+    }
+  }
+  omp_destroy_nest_lock (&lck);
+  return ((result == 0) && (nr_iterations == LOOPCOUNT));
+}
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for(i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_test_nest_lock()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/master/omp_master.c

+// RUN: %libomp-compile-and-run
+#include <stdio.h>
+#include "omp_testsuite.h"
+
+int test_omp_master()
+{
+  int nthreads;
+  int executing_thread;
+
+  nthreads = 0;
+  executing_thread = -1;
+
+  #pragma omp parallel
+  {
+    #pragma omp master 
+    {
+      #pragma omp critical
+      {
+        nthreads++;
+      }
+      executing_thread = omp_get_thread_num();
+    } /* end of master*/
+  } /* end of parallel*/
+  return ((nthreads == 1) && (executing_thread == 0));
+}
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for(i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_master()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/master/omp_master_3.c

+// RUN: %libomp-compile-and-run
+#include <stdio.h>
+#include "omp_testsuite.h"
+
+int test_omp_master_3()
+{
+  int nthreads;
+  int executing_thread;
+  int tid_result = 0; /* counts up the number of wrong thread no. for
+               the master thread. (Must be 0) */
+  nthreads = 0;
+  executing_thread = -1;
+
+  #pragma omp parallel
+  {
+    #pragma omp master 
+    {
+      int tid = omp_get_thread_num();
+      if (tid != 0) {
+        #pragma omp critical
+        { tid_result++; }
+      }
+      #pragma omp critical
+      {
+        nthreads++;
+      }
+      executing_thread = omp_get_thread_num ();
+    } /* end of master*/
+  } /* end of parallel*/
+  return ((nthreads == 1) && (executing_thread == 0) && (tid_result == 0));
+}
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for(i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_master_3()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/omp_my_sleep.h

+#ifndef MY_SLEEP_H
+#define MY_SLEEP_H
+
+/*! Utility function to have a sleep function with better resolution and
+ *  which only stops one thread. */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <errno.h>
+#include <time.h>
+
+#if defined(_WIN32)
+# include <windows.h>
+
+// Windows version of sleeping
+// According to Windows documentation: clock() "Calculates the *wall-clock*
+//   time used by the calling process" 
+// So it can be used as a millisecond resolution timer.
+static void my_sleep(double sleeptime) {
+  clock_t start;
+  clock_t real;
+  start = clock();
+  if(start == -1) {
+    fprintf("error: my_sleep(): elapsed time is unavailable\r\n");
+    exit(-1);
+  }
+  real = start;
+  while(((double)(real-start)/(double)CLOCKS_PER_SEC) < sleeptime) {
+    real = clock();
+    if(real == -1) {
+      fprintf("error: my_sleep(): elapsed time is unavailable\r\n");
+      exit(-1);
+    }
+  }
+}
+
+#else // _WIN32
+
+// Unices use this my_sleep() function
+# include <sys/time.h>
+
+static void my_sleep(double sleeptime) {
+  struct timeval tv;
+  struct timezone tzp;
+  double start;
+  double real;
+  if(gettimeofday(&tv,&tzp)!=0) {
+    perror("get_time: ");
+    exit(-1);
+  }
+  start = (double)tv.tv_sec + ((double)tv.tv_usec/1000000.0);
+  real = start;
+  while((real-start) < sleeptime) {
+    if(gettimeofday(&tv,&tzp)!=0) {
+      perror("get_time: ");
+      exit(-1);
+    }
+    real = (double)tv.tv_sec + ((double)tv.tv_usec/1000000.0);
+  }
+}
+
+#endif // _WIN32
+
+#endif // MY_SLEEP_H

runtime/test/omp_testsuite.h

+/* Global headerfile of the OpenMP Testsuite */
+
+#ifndef OMP_TESTSUITE_H
+#define OMP_TESTSUITE_H
+
+#include <stdio.h>
+#include <omp.h>
+
+/* General                                                */
+/**********************************************************/
+#define LOOPCOUNT 1000 /* Number of iterations to slit amongst threads */
+#define REPETITIONS 10 /* Number of times to run each test */
+
+/* following times are in seconds */
+#define SLEEPTIME 0.1
+
+/* Definitions for tasks                                  */
+/**********************************************************/
+#define NUM_TASKS 25
+#define MAX_TASKS_PER_THREAD 5
+
+#endif

runtime/test/parallel/omp_nested.c

+// RUN: %libomp-compile-and-run
+#include <stdio.h>
+#include "omp_testsuite.h"
+
+/*
+ * Test if the compiler supports nested parallelism
+ * By Chunhua Liao, University of Houston
+ * Oct. 2005
+ */
+int test_omp_nested()
+{
+  int counter = 0;
+#ifdef _OPENMP
+  omp_set_nested(1);
+#endif
+
+  #pragma omp parallel shared(counter)
+  {
+    #pragma omp critical
+    counter++;
+    #pragma omp parallel
+    {
+      #pragma omp critical
+      counter--;
+    }
+  }
+  return (counter != 0);
+}
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for(i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_nested()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/parallel/omp_parallel_copyin.c

+// RUN: %libomp-compile-and-run
+#include <stdio.h>
+#include <stdlib.h>
+#include "omp_testsuite.h"
+
+static int sum1 = 789;
+#pragma omp threadprivate(sum1)
+
+int test_omp_parallel_copyin()
+{
+  int sum, num_threads;
+  int known_sum;
+
+  sum = 0;
+  sum1 = 7;
+  num_threads = 0;
+
+  #pragma omp parallel copyin(sum1)
+  {
+    /*printf("sum1=%d\n",sum1);*/
+    int i;
+    #pragma omp for 
+    for (i = 1; i < 1000; i++) {
+      sum1 = sum1 + i;
+    } /*end of for*/
+    #pragma omp critical
+    {
+      sum = sum + sum1;
+      num_threads++;
+    } /*end of critical*/
+  } /* end of parallel*/  
+  known_sum = (999 * 1000) / 2 + 7 * num_threads;
+  return (known_sum == sum);
+}
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for(i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_parallel_copyin()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/parallel/omp_parallel_default.c

+// RUN: %libomp-compile-and-run
+#include <stdio.h>
+#include "omp_testsuite.h"
+
+int test_omp_parallel_default()
+{
+  int i;
+  int sum;
+  int mysum;
+  int known_sum;
+  sum =0;
+  known_sum = (LOOPCOUNT * (LOOPCOUNT + 1)) / 2 ;
+
+  #pragma omp parallel default(shared) private(i) private(mysum)
+  {
+    mysum = 0;
+    #pragma omp for
+    for (i = 1; i <= LOOPCOUNT; i++) {
+      mysum = mysum + i;
+    } 
+    #pragma omp critical
+    {
+      sum = sum + mysum;
+    }   /* end of critical */
+  }   /* end of parallel */
+  if (known_sum != sum) {
+    fprintf(stderr, "KNOWN_SUM = %d; SUM = %d\n", known_sum, sum);
+  }
+  return (known_sum == sum);
+}
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for(i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_parallel_default()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/parallel/omp_parallel_firstprivate.c

+// RUN: %libomp-compile-and-run
+#include <stdio.h>
+#include <stdlib.h>
+#include "omp_testsuite.h"
+
+//static int sum1 = 789;
+
+int test_omp_parallel_firstprivate()
+{
+  int sum, num_threads,sum1;
+  int known_sum;
+
+  sum = 0;
+  sum1=7;
+  num_threads = 0;
+
+  #pragma omp parallel firstprivate(sum1)
+  {
+    /*printf("sum1=%d\n",sum1);*/
+    int i;
+    #pragma omp for 
+    for (i = 1; i < 1000; i++) {
+      sum1 = sum1 + i;
+    } /*end of for*/
+    #pragma omp critical
+    {
+      sum = sum + sum1;
+      num_threads++;
+    } /*end of critical*/
+  } /* end of parallel*/  
+  known_sum = (999 * 1000) / 2 + 7 * num_threads;
+  return (known_sum == sum);
+}
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for(i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_parallel_firstprivate()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/parallel/omp_parallel_if.c

+// RUN: %libomp-compile-and-run
+#include <stdio.h>
+#include "omp_testsuite.h"
+
+int test_omp_parallel_if()
+{
+  int i;
+  int sum;
+  int known_sum;
+  int mysum;
+  int control=1;
+
+  sum =0;
+  known_sum = (LOOPCOUNT * (LOOPCOUNT + 1)) / 2 ;
+  #pragma omp parallel private(i) if(control==0)
+  {
+    mysum = 0;
+    for (i = 1; i <= LOOPCOUNT; i++) {
+      mysum = mysum + i;
+    } 
+    #pragma omp critical
+    {
+      sum = sum + mysum;
+    }
+  }
+  return (known_sum == sum);
+}
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for(i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_parallel_if()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/parallel/omp_parallel_num_threads.c

+// RUN: %libomp-compile-and-run
+#include <stdio.h>
+#include "omp_testsuite.h"
+
+int test_omp_parallel_num_threads()
+{
+  int num_failed;
+  int threads;
+  int nthreads;
+  int max_threads = 0;
+
+  num_failed = 0;
+
+  /* first we check how many threads are available */
+  #pragma omp parallel
+  {
+    #pragma omp master
+    max_threads = omp_get_num_threads ();
+  }
+
+  /* we increase the number of threads from one to maximum:*/
+  for(threads = 1; threads <= max_threads; threads++) {
+    nthreads = 0;
+    #pragma omp parallel reduction(+:num_failed) num_threads(threads)
+    {
+      num_failed = num_failed + !(threads == omp_get_num_threads());
+      #pragma omp atomic
+      nthreads += 1;
+    }
+    num_failed = num_failed + !(nthreads == threads);
+  }
+  return (!num_failed);
+}
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for(i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_parallel_num_threads()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/parallel/omp_parallel_private.c

+// RUN: %libomp-compile-and-run
+#include <stdio.h>
+#include <stdlib.h>
+#include "omp_testsuite.h"
+
+//static int sum1 = 789;
+
+int test_omp_parallel_private()
+{
+  int sum, num_threads,sum1;
+  int known_sum;
+
+  sum = 0;
+  num_threads = 0;
+
+  #pragma omp parallel private(sum1)
+  {
+    int i;
+    sum1 = 7;
+    /*printf("sum1=%d\n",sum1);*/
+    #pragma omp for 
+    for (i = 1; i < 1000; i++) {
+      sum1 = sum1 + i;
+    }
+    #pragma omp critical
+    {
+      sum = sum + sum1;
+      num_threads++;
+    }
+  }
+  known_sum = (999 * 1000) / 2 + 7 * num_threads;
+  return (known_sum == sum);
+}
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for(i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_parallel_private()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/parallel/omp_parallel_reduction.c

+// RUN: %libomp-compile-and-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! */
+
+int test_omp_parallel_reduction()
+{
+  int sum;
+  int known_sum;
+  double dsum;
+  double dknown_sum;
+  double dt=0.5; /* base of geometric row for + and - test*/
+  double rounding_error= 1.E-9;
+  int diff;
+  double ddiff;
+  int product;
+  int known_product;
+  int logic_and;
+  int logic_or;
+  int bit_and;
+  int bit_or;
+  int exclusiv_bit_or;
+  int logics[LOOPCOUNT];
+  int i;
+  double dpt;
+  int result;
+
+  sum =0;
+  dsum=0;
+  product=1;
+  logic_and=1;
+  logic_or=0;
+  bit_and=1;
+  bit_or=0;
+  exclusiv_bit_or=0;
+  result=0;
+  dt = 1./3.;
+  known_sum = (LOOPCOUNT*(LOOPCOUNT+1))/2;
+
+  /* Tests for integers */
+  #pragma omp parallel for schedule(dynamic,1) private(i) reduction(+:sum)
+  for (i=1;i<=LOOPCOUNT;i++) {
+    sum=sum+i;
+  }
+
+  if(known_sum!=sum) {
+    result++;
+    fprintf(stderr,"Error in sum with integers: Result was %d instead of %d\n",sum,known_sum); 
+  }
+
+  diff = (LOOPCOUNT*(LOOPCOUNT+1))/2;
+  #pragma omp parallel for schedule(dynamic,1) private(i) reduction(-:diff)
+  for (i=1;i<=LOOPCOUNT;++i) {
+    diff=diff-i;
+  }
+
+  if(diff != 0) {
+    result++;
+    fprintf(stderr,"Error in difference with integers: Result was %d instead of 0.\n",diff);
+  }
+
+  /* Tests for doubles */
+  dsum=0;
+  dpt=1;
+  for (i=0;i<DOUBLE_DIGITS;++i) {
+    dpt*=dt;
+  }
+  dknown_sum = (1-dpt)/(1-dt);
+  #pragma omp parallel for schedule(dynamic,1) private(i) reduction(+:dsum)
+  for (i=0;i<DOUBLE_DIGITS;++i) {
+    dsum += pow(dt,i);
+  }
+
+  if( fabs(dsum-dknown_sum) > rounding_error ) {
+    result++; 
+    fprintf(stderr,"Error in sum with doubles: Result was %f instead of %f (Difference: %E)\n",dsum,dknown_sum, dsum-dknown_sum);
+  }
+
+  dpt=1;
+
+  for (i=0;i<DOUBLE_DIGITS;++i) {
+    dpt*=dt;
+  }
+  fprintf(stderr,"\n");
+  ddiff = (1-dpt)/(1-dt);
+  #pragma omp parallel for schedule(dynamic,1) private(i) reduction(-:ddiff)
+  for (i=0;i<DOUBLE_DIGITS;++i) {
+    ddiff -= pow(dt,i);
+  }
+  if( fabs(ddiff) > rounding_error) {
+    result++;
+    fprintf(stderr,"Error in Difference with doubles: Result was %E instead of 0.0\n",ddiff);
+  }
+
+  /* Tests for product of integers */
+  #pragma omp parallel for schedule(dynamic,1) private(i) reduction(*:product)
+  for(i=1;i<=MAX_FACTOR;i++) {
+    product *= i;
+  }
+
+  known_product = KNOWN_PRODUCT;
+  if(known_product != product) {
+    result++;
+    fprintf(stderr,"Error in Product with integers: Result was %d instead of %d\n\n",product,known_product);
+  }
+
+  /* Tests for logical and */
+  for(i=0;i<LOOPCOUNT;i++) {
+    logics[i]=1;
+  }
+
+  #pragma omp parallel for schedule(dynamic,1) private(i) reduction(&&:logic_and)
+  for(i=0;i<LOOPCOUNT;++i) {
+    logic_and = (logic_and && logics[i]);
+  }
+  if(!logic_and) {
+    result++;
+    fprintf(stderr,"Error in logic AND part 1.\n");
+  }
+
+  logic_and = 1;
+  logics[LOOPCOUNT/2]=0;
+
+  #pragma omp parallel for schedule(dynamic,1) private(i) reduction(&&:logic_and)
+  for(i=0;i<LOOPCOUNT;++i) {
+    logic_and = logic_and && logics[i];
+  }
+  if(logic_and) {
+    result++;
+    fprintf(stderr,"Error in logic AND part 2.\n");
+  }
+
+  /* Tests for logical or */
+  for(i=0;i<LOOPCOUNT;i++) {
+    logics[i]=0;
+  }
+
+  #pragma omp parallel for schedule(dynamic,1) private(i) reduction(||:logic_or)
+  for(i=0;i<LOOPCOUNT;++i) {
+    logic_or = logic_or || logics[i];
+  }
+  if(logic_or) {
+    result++;
+    fprintf(stderr,"Error in logic OR part 1.\n");
+  }
+  logic_or = 0;
+  logics[LOOPCOUNT/2]=1;
+
+  #pragma omp parallel for schedule(dynamic,1) private(i) reduction(||:logic_or)
+  for(i=0;i<LOOPCOUNT;++i) {
+    logic_or = logic_or || logics[i];
+  }
+  if(!logic_or) {
+    result++;
+    fprintf(stderr,"Error in logic OR part 2.\n");
+  }
+
+  /* Tests for bitwise and */
+  for(i=0;i<LOOPCOUNT;++i) {
+    logics[i]=1;
+  }
+
+  #pragma omp parallel for schedule(dynamic,1) private(i) reduction(&:bit_and)
+  for(i=0;i<LOOPCOUNT;++i) {
+    bit_and = (bit_and & logics[i]);
+  }
+  if(!bit_and) {
+    result++;
+    fprintf(stderr,"Error in BIT AND part 1.\n");
+  }
+
+  bit_and = 1;
+  logics[LOOPCOUNT/2]=0;
+
+  #pragma omp parallel for schedule(dynamic,1) private(i) reduction(&:bit_and)
+  for(i=0;i<LOOPCOUNT;++i) {
+    bit_and = bit_and & logics[i];
+  }
+  if(bit_and) {
+    result++;
+    fprintf(stderr,"Error in BIT AND part 2.\n");
+  }
+
+  for(i=0;i<LOOPCOUNT;i++) {
+    logics[i]=0;
+  }
+
+  /* Tests for bitwise or */
+  #pragma omp parallel for schedule(dynamic,1) private(i) reduction(|:bit_or)
+  for(i=0;i<LOOPCOUNT;++i) {
+    bit_or = bit_or | logics[i];
+  }
+  if(bit_or) {
+    result++;
+    fprintf(stderr,"Error in BIT OR part 1\n");
+  }
+  bit_or = 0;
+  logics[LOOPCOUNT/2]=1;
+
+  #pragma omp parallel for schedule(dynamic,1) private(i) reduction(|:bit_or)
+  for(i=0;i<LOOPCOUNT;++i) {
+    bit_or = bit_or | logics[i];
+  }
+  if(!bit_or) {
+    result++;
+    fprintf(stderr,"Error in BIT OR part 2\n");
+  }
+
+  for(i=0;i<LOOPCOUNT;i++) {
+    logics[i]=0;
+  }
+
+  /* Tests for bitwise xor */
+  #pragma omp parallel for schedule(dynamic,1) private(i) reduction(^:exclusiv_bit_or)
+  for(i=0;i<LOOPCOUNT;++i) {
+    exclusiv_bit_or = exclusiv_bit_or ^ logics[i];
+  }
+  if(exclusiv_bit_or) {
+    result++;
+    fprintf(stderr,"Error in EXCLUSIV BIT OR part 1\n");
+  }
+
+  exclusiv_bit_or = 0;
+  logics[LOOPCOUNT/2]=1;
+
+  #pragma omp parallel for schedule(dynamic,1) private(i) reduction(^:exclusiv_bit_or)
+  for(i=0;i<LOOPCOUNT;++i) {
+    exclusiv_bit_or = exclusiv_bit_or ^ logics[i];
+  }
+  if(!exclusiv_bit_or) {
+    result++;
+    fprintf(stderr,"Error in EXCLUSIV BIT OR part 2\n");
+  }
+
+  /*printf("\nResult:%d\n",result);*/
+  return (result==0);
+}
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for(i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_parallel_reduction()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/parallel/omp_parallel_shared.c

+// RUN: %libomp-compile-and-run
+#include <stdio.h>
+#include "omp_testsuite.h"
+
+int test_omp_parallel_shared()
+{
+  int i;
+  int sum;
+  int known_sum;
+
+  sum = 0;
+  known_sum = (LOOPCOUNT * (LOOPCOUNT + 1)) / 2 ;
+
+  #pragma omp parallel private(i) shared(sum) 
+  {
+
+    int mysum = 0;
+    #pragma omp for
+    for (i = 1; i <= LOOPCOUNT; i++) {
+      mysum = mysum + i;
+    } 
+    #pragma omp critical
+    {
+      sum = sum + mysum;
+    }
+
+
+  }
+  if (known_sum != sum) {
+    fprintf(stderr, "KNOWN_SUM = %d; SUM = %d\n", known_sum, sum);
+  }
+  return (known_sum == sum);
+}
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for(i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_parallel_shared()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/tasking/omp_task.c

+// RUN: %libomp-compile-and-run
+#include <stdio.h>
+#include <math.h>
+#include "omp_testsuite.h"
+#include "omp_my_sleep.h"
+
+int test_omp_task()
+{
+  int tids[NUM_TASKS];
+  int i;
+
+  #pragma omp parallel 
+  {
+    #pragma omp single
+    {
+      for (i = 0; i < NUM_TASKS; i++) {
+        /* First we have to store the value of the loop index in a new variable
+         * which will be private for each task because otherwise it will be overwritten
+         * if the execution of the task takes longer than the time which is needed to 
+         * enter the next step of the loop!
+         */
+        int myi;
+        myi = i;
+        #pragma omp task
+        {
+          my_sleep (SLEEPTIME);
+          tids[myi] = omp_get_thread_num();
+        } /* end of omp task */
+      } /* end of for */
+    } /* end of single */
+  } /*end of parallel */
+
+  /* Now we ckeck if more than one thread executed the tasks. */
+  for (i = 1; i < NUM_TASKS; i++) {
+    if (tids[0] != tids[i])
+      return 1;
+  }
+  return 0;
+} /* end of check_parallel_for_private */
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for(i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_task()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/tasking/omp_task_final.c

+// RUN: %libomp-compile-and-run
+#include <stdio.h>
+#include <math.h>
+#include "omp_testsuite.h"
+#include "omp_my_sleep.h"
+
+int test_omp_task_final()
+{
+  int tids[NUM_TASKS];
+  int in_final[NUM_TASKS];
+  int error;
+  #pragma omp parallel 
+  {
+    #pragma omp single
+    {
+      int i;
+      for (i = 0; i < NUM_TASKS; i++) {
+        /* First we have to store the value of the loop index in a new variable
+         * which will be private for each task because otherwise it will be overwritten
+         * if the execution of the task takes longer than the time which is needed to 
+         * enter the next step of the loop!
+         */
+        int myi;
+        myi = i;
+        #pragma omp task final(i>=10) firstprivate(myi) shared(tids) shared(in_final)
+        {
+          my_sleep (SLEEPTIME);
+          tids[myi] = omp_get_thread_num();
+          in_final[myi] = omp_in_final();
+        }
+      }
+    }
+  }
+
+  /* Now we check if more than one thread executed the tasks. */
+  int i;
+  for (i = 10; i < NUM_TASKS; i++) {
+    fprintf(stderr, "tids[10] = %d , tids[%d] = %d\n", tids[10], i, tids[i]);
+    fprintf(stderr, "in_final[%d] = %d\n", i, in_final[i]);
+    if (tids[10] != tids[i])
+      error++;
+  }
+  fprintf(stderr, "\n");
+  return (error==0);
+} /* end of check_parallel_for_private */
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for(i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_task_final()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/tasking/omp_task_firstprivate.c

+// RUN: %libomp-compile-and-run
+#include <stdio.h>
+#include <math.h>
+#include "omp_testsuite.h"
+
+int test_omp_task_firstprivate()
+{
+  int i;
+  int sum = 1234;
+  int known_sum;
+  int result = 0; /* counts the wrong sums from tasks */
+
+  known_sum = 1234 + (LOOPCOUNT * (LOOPCOUNT + 1)) / 2;
+
+  #pragma omp parallel
+  {
+    #pragma omp single
+    {
+      for (i = 0; i < NUM_TASKS; i++) {
+        #pragma omp task firstprivate(sum)
+        {
+          int j;
+          for (j = 0; j <= LOOPCOUNT; j++) {
+            #pragma omp flush
+            sum += j;
+          }
+
+          /* check if calculated sum was right */
+          if (sum != known_sum) {
+            #pragma omp critical 
+            { result++; }
+          }
+        } /* omp task */
+      } /* for loop */
+    } /* omp single */
+  } /* omp parallel */
+  return (result == 0);
+}
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for(i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_task_firstprivate()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/tasking/omp_task_if.c

+// RUN: %libomp-compile-and-run
+#include <stdio.h>
+#include <math.h>
+#include "omp_testsuite.h"
+#include "omp_my_sleep.h"
+
+int test_omp_task_if()
+{
+  int condition_false;
+  int count;
+  int result;
+
+  count=0;
+  condition_false = (count == 1);
+  #pragma omp parallel 
+  {
+    #pragma omp single
+    {
+      #pragma omp task if (condition_false) shared(count, result)
+      {
+        my_sleep (SLEEPTIME);
+        //#pragma omp flush (count)
+        result = (0 == count);
+      } /* end of omp task */
+      count = 1;
+      //#pragma omp flush (count)
+    } /* end of single */
+  } /*end of parallel */
+  return result;
+}
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for(i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_task_if()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/tasking/omp_task_imp_firstprivate.c

+// RUN: %libomp-compile-and-run
+#include <stdio.h>
+#include <math.h>
+#include "omp_testsuite.h"
+
+/* Utility function do spend some time in a loop */
+int test_omp_task_imp_firstprivate()
+{
+  int i=5;
+  int k = 0;
+  int result = 0;
+  int task_result = 1;
+  #pragma omp parallel firstprivate(i)
+  {
+    #pragma omp single
+    {
+      for (k = 0; k < NUM_TASKS; k++) {
+        #pragma omp task shared(result , task_result)
+        {
+          int j;
+          //check if i is private
+          if(i != 5)
+            task_result = 0;
+          for(j = 0; j < NUM_TASKS; j++)
+            i++;
+          //this should be firstprivate implicitly
+        }
+      }
+      #pragma omp taskwait
+      result = (task_result && i==5);
+    }
+  }
+  return result;
+}
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for(i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_task_imp_firstprivate()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/tasking/omp_task_private.c

+// RUN: %libomp-compile-and-run
+#include <stdio.h>
+#include <math.h>
+#include "omp_testsuite.h"
+
+/* Utility function do spend some time in a loop */
+int test_omp_task_private()
+{
+  int i;
+  int known_sum;
+  int sum = 0;
+  int result = 0; /* counts the wrong sums from tasks */
+
+  known_sum = (LOOPCOUNT * (LOOPCOUNT + 1)) / 2;
+
+  #pragma omp parallel
+  {
+    #pragma omp single
+    {
+      for (i = 0; i < NUM_TASKS; i++) {
+        #pragma omp task private(sum) shared(result, known_sum)
+        {
+          int j;
+          //if sum is private, initialize to 0
+          sum = 0;
+          for (j = 0; j <= LOOPCOUNT; j++) {
+            #pragma omp flush
+            sum += j;
+          }
+          /* check if calculated sum was right */
+          if (sum != known_sum) {
+            #pragma omp critical 
+            result++;
+          }
+        } /* end of omp task */
+      } /* end of for */
+    } /* end of single */
+  } /* end of parallel*/
+  return (result == 0);
+}
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for(i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_task_private()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/tasking/omp_task_shared.c

+// RUN: %libomp-compile-and-run
+#include <stdio.h>
+#include <math.h>
+#include "omp_testsuite.h"
+
+/* Utility function do spend some time in a loop */
+int test_omp_task_imp_shared()
+{
+  int i;
+  int k = 0;
+  int result = 0;
+  i=0;
+
+  #pragma omp parallel
+  {
+    #pragma omp single
+    for (k = 0; k < NUM_TASKS; k++) {
+      #pragma omp task shared(i)
+      {
+        #pragma omp atomic
+        i++;
+        //this should be shared implicitly
+      }
+    }
+  }
+  result = i;
+  return ((result == NUM_TASKS));
+}
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for(i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_task_imp_shared()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/tasking/omp_task_untied.c

+// RUN: %libomp-compile-and-run
+#include <stdio.h>
+#include <math.h>
+#include "omp_testsuite.h"
+#include "omp_my_sleep.h"
+
+int test_omp_task_untied()
+{
+  int i;
+  int count;
+  int start_tid[NUM_TASKS];
+  int current_tid[NUM_TASKS];
+
+  count = 0;
+
+  /*initialization*/
+  for (i=0; i< NUM_TASKS; i++){
+    start_tid[i]=0;
+    current_tid[i]=0;
+  }
+
+  #pragma omp parallel firstprivate(i)
+  {
+    #pragma omp single
+    {
+      for (i = 0; i < NUM_TASKS; i++) {
+        int myi = i;
+        #pragma omp task untied
+        {
+          my_sleep(SLEEPTIME);
+          start_tid[myi] = omp_get_thread_num();
+          current_tid[myi] = omp_get_thread_num();
+
+          #pragma omp taskwait
+
+          if((start_tid[myi] % 2) != 0) {
+            my_sleep(SLEEPTIME);
+            current_tid[myi] = omp_get_thread_num();
+          } else {
+            current_tid[myi] = omp_get_thread_num();
+          }
+        } /*end of omp task */
+      } /* end of for */
+    } /* end of single */
+  } /* end of parallel */
+
+  for (i=0;i<NUM_TASKS; i++) {
+    //fprintf(stderr, "start_tid[%d]=%d, current_tid[%d]=%d\n",
+     // i, start_tid[i], i , current_tid[i]);
+    if (current_tid[i] == start_tid[i])
+      count++;
+  }
+  return (count<NUM_TASKS);
+}
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for(i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_task_untied()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/tasking/omp_taskwait.c

+// RUN: %libomp-compile-and-run
+#include <stdio.h>
+#include <math.h>
+#include "omp_testsuite.h"
+#include "omp_my_sleep.h"
+
+int test_omp_taskwait()
+{
+  int result1 = 0;   /* Stores number of not finished tasks after the taskwait */
+  int result2 = 0;   /* Stores number of wrong array elements at the end */
+  int array[NUM_TASKS];
+  int i;
+
+  /* fill array */
+  for (i = 0; i < NUM_TASKS; i++) 
+    array[i] = 0;
+
+  #pragma omp parallel 
+  {
+    #pragma omp single
+    {
+      for (i = 0; i < NUM_TASKS; i++) {
+        /* First we have to store the value of the loop index in a new variable
+         * which will be private for each task because otherwise it will be overwritten
+         * if the execution of the task takes longer than the time which is needed to 
+         * enter the next step of the loop!
+         */
+        int myi;
+        myi = i;
+        #pragma omp task
+        {
+          my_sleep (SLEEPTIME);
+          array[myi] = 1;
+        } /* end of omp task */
+      } /* end of for */
+      #pragma omp taskwait
+      /* check if all tasks were finished */
+      for (i = 0; i < NUM_TASKS; i++) 
+        if (array[i] != 1)
+          result1++;
+
+      /* generate some more tasks which now shall overwrite 
+       * the values in the tids array */
+      for (i = 0; i < NUM_TASKS; i++) {
+        int myi;
+        myi = i;
+        #pragma omp task
+        {
+          array[myi] = 2;
+        } /* end of omp task */
+      } /* end of for */
+    } /* end of single */
+  } /*end of parallel */
+
+  /* final check, if all array elements contain the right values: */
+  for (i = 0; i < NUM_TASKS; i++) {
+    if (array[i] != 2)
+      result2++;
+  }
+  return ((result1 == 0) && (result2 == 0));
+}
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for(i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_taskwait()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/tasking/omp_taskyield.c

+// RUN: %libomp-compile-and-run
+#include <stdio.h>
+#include <math.h>
+#include "omp_testsuite.h"
+#include "omp_my_sleep.h"
+
+int test_omp_taskyield()
+{
+  int i;
+  int count = 0;
+  int start_tid[NUM_TASKS];
+  int current_tid[NUM_TASKS];
+
+  for (i=0; i< NUM_TASKS; i++) {
+    start_tid[i]=0;
+    current_tid[i]=0;
+  }
+
+  #pragma omp parallel
+  {
+    #pragma omp single
+    {
+      for (i = 0; i < NUM_TASKS; i++) {
+        int myi = i;
+        #pragma omp task untied
+        {
+          my_sleep(SLEEPTIME);
+          start_tid[myi] = omp_get_thread_num();
+          #pragma omp taskyield
+          if((start_tid[myi] %2) ==0){
+            my_sleep(SLEEPTIME);
+            current_tid[myi] = omp_get_thread_num();
+          } /*end of if*/
+        } /* end of omp task */
+      } /* end of for */
+    } /* end of single */
+  } /* end of parallel */
+  for (i=0;i<NUM_TASKS; i++) {
+    //printf("start_tid[%d]=%d, current_tid[%d]=%d\n",
+      //i, start_tid[i], i , current_tid[i]);
+    if (current_tid[i] == start_tid[i])
+      count++;
+  }
+  return (count<NUM_TASKS);
+}
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for(i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_taskyield()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/threadprivate/omp_threadprivate.c

+// RUN: %libomp-compile-and-run
+/*
+ * Threadprivate is tested in 2 ways:
+ * 1. The global variable declared as threadprivate should have
+ *  local copy for each thread. Otherwise race condition and 
+ *  wrong result.
+ * 2. If the value of local copy is retained for the two adjacent
+ *  parallel regions
+ */
+#include "omp_testsuite.h"
+#include <stdlib.h>
+#include <stdio.h>
+
+static int sum0=0;
+static int myvalue = 0;
+
+#pragma omp threadprivate(sum0)
+#pragma omp threadprivate(myvalue)
+
+int test_omp_threadprivate()
+{
+  int sum = 0;
+  int known_sum;
+  int i; 
+  int iter;
+  int *data;
+  int size;
+  int num_failed = 0;
+  int my_random;
+  omp_set_dynamic(0);
+
+  #pragma omp parallel private(i) 
+  {
+    sum0 = 0;
+    #pragma omp for 
+    for (i = 1; i <= LOOPCOUNT; i++) {
+      sum0 = sum0 + i;
+    } /*end of for*/
+    #pragma omp critical
+    {
+      sum = sum + sum0;
+    } /*end of critical */
+  } /* end of parallel */  
+  known_sum = (LOOPCOUNT * (LOOPCOUNT + 1)) / 2;
+  if (known_sum != sum ) {
+    fprintf (stderr, " known_sum = %d, sum = %d\n", known_sum, sum);
+  }
+
+  /* the next parallel region is just used to get the number of threads*/
+  omp_set_dynamic(0);
+  #pragma omp parallel
+  {
+    #pragma omp master
+    {
+      size=omp_get_num_threads();
+      data=(int*) malloc(size*sizeof(int));
+    }
+  }/* end parallel*/
+
+  srand(45);
+  for (iter = 0; iter < 100; iter++) {
+    my_random = rand(); /* random number generator is 
+                 called inside serial region*/
+
+    /* the first parallel region is used to initialiye myvalue
+       and the array with my_random+rank */
+    #pragma omp parallel
+    {
+      int rank;
+      rank = omp_get_thread_num ();
+      myvalue = data[rank] = my_random + rank;
+    }
+
+    /* the second parallel region verifies that the 
+       value of "myvalue" is retained */
+    #pragma omp parallel reduction(+:num_failed)
+    {
+      int rank;
+      rank = omp_get_thread_num ();
+      num_failed = num_failed + (myvalue != data[rank]);
+      if(myvalue != data[rank]) {
+        fprintf (stderr, " myvalue = %d, data[rank]= %d\n",
+          myvalue, data[rank]);
+      }
+    }
+  }
+  free (data);
+  return (known_sum == sum) && !num_failed;
+} /* end of check_threadprivate*/
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for(i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_threadprivate()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/threadprivate/omp_threadprivate_for.c

+// RUN: %libomp-compile-and-run
+#include "omp_testsuite.h"
+#include <stdlib.h>
+#include <stdio.h>
+
+static int i;
+#pragma omp threadprivate(i)
+
+int test_omp_threadprivate_for()
+{
+  int known_sum;
+  int sum;
+
+  known_sum = (LOOPCOUNT * (LOOPCOUNT + 1)) / 2;
+  sum = 0;
+
+  #pragma omp parallel
+  {
+    int sum0 = 0;
+    #pragma omp for
+    for (i = 1; i <= LOOPCOUNT; i++) {
+      sum0 = sum0 + i;
+    }
+    #pragma omp critical
+    {
+      sum = sum + sum0;
+    }
+  } /* end of parallel */  
+
+  if (known_sum != sum ) {
+    fprintf(stderr, " known_sum = %d, sum = %d\n", known_sum, sum);
+  }
+  return (known_sum == sum);
+} /* end of check_threadprivate*/
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for(i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_threadprivate_for()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/worksharing/for/omp_for_collapse.c

+// RUN: %libomp-compile-and-run
+#include <stdio.h>
+#include <math.h>
+#include "omp_testsuite.h"
+
+/* Utility function to check that i is increasing monotonically 
+   with each call */
+static int check_i_islarger (int i)
+{
+  static int last_i;
+  int islarger;
+  if (i==1)
+    last_i=0;
+  islarger = ((i >= last_i)&&(i - last_i<=1));
+  last_i = i;
+  return (islarger);
+}
+
+int test_omp_for_collapse()
+{
+  int is_larger = 1;
+
+  #pragma omp parallel
+  {
+    int i,j;
+    int my_islarger = 1;
+    #pragma omp for private(i,j) schedule(static,1) collapse(2) ordered
+    for (i = 1; i < 100; i++) {
+      for (j =1; j <100; j++) {
+        #pragma omp ordered
+        my_islarger = check_i_islarger(i)&&my_islarger;
+      }
+    }
+    #pragma omp critical
+    is_larger = is_larger && my_islarger;
+  }
+  return (is_larger);
+}
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for(i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_for_collapse()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/worksharing/for/omp_for_firstprivate.c

+// RUN: %libomp-compile-and-run
+#include <stdio.h>
+#include <math.h>
+#include "omp_testsuite.h"
+
+int sum1;
+#pragma omp threadprivate(sum1)
+
+int test_omp_for_firstprivate()
+{
+  int sum;
+  int sum0;
+  int known_sum;
+  int threadsnum;
+
+  sum = 0;
+  sum0 = 12345;
+  sum1 = 0;
+
+  #pragma omp parallel
+  {
+    #pragma omp single
+    {
+      threadsnum=omp_get_num_threads();
+    }
+    /* sum0 = 0; */
+  
+    int i;
+    #pragma omp for firstprivate(sum0)
+    for (i = 1; i <= LOOPCOUNT; i++) {
+      sum0 = sum0 + i;
+      sum1 = sum0;
+    }  /* end of for */
+  
+    #pragma omp critical
+    {
+      sum = sum + sum1;
+    }  /* end of critical */
+  }  /* end of parallel */  
+  known_sum = 12345* threadsnum+ (LOOPCOUNT * (LOOPCOUNT + 1)) / 2;
+  return (known_sum == sum);
+}
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for(i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_for_firstprivate()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/worksharing/for/omp_for_lastprivate.c

+// RUN: %libomp-compile-and-run
+#include <stdio.h>
+#include <math.h>
+#include "omp_testsuite.h"
+
+int sum0;
+#pragma omp threadprivate(sum0)
+
+int test_omp_for_lastprivate()
+{
+  int sum = 0;
+  int known_sum;
+  int i0;
+  
+  i0 = -1;
+
+  #pragma omp parallel
+  {
+    sum0 = 0;
+    {  /* Begin of orphaned block */
+      int i;
+      #pragma omp for schedule(static,7) lastprivate(i0)
+      for (i = 1; i <= LOOPCOUNT; i++) {
+        sum0 = sum0 + i;
+        i0 = i;
+      }  /* end of for */
+    }  /* end of orphaned block */
+
+    #pragma omp critical
+    {
+      sum = sum + sum0;
+    }  /* end of critical */
+  }  /* end of parallel */  
+
+  known_sum = (LOOPCOUNT * (LOOPCOUNT + 1)) / 2;
+  fprintf(stderr, "known_sum = %d , sum = %d\n",known_sum,sum);
+  fprintf(stderr, "LOOPCOUNT = %d , i0 = %d\n",LOOPCOUNT,i0);
+  return ((known_sum == sum) && (i0 == LOOPCOUNT));
+}
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for (i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_for_lastprivate()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/worksharing/for/omp_for_nowait.c

+// RUN: %libomp-compile-and-run
+#include <stdio.h>
+#include "omp_testsuite.h"
+#include "omp_my_sleep.h"
+
+int test_omp_for_nowait()
+{
+  int result;
+  int count;
+  int j;
+  int myarray[LOOPCOUNT];
+
+  result = 0;
+  count = 0;
+
+  #pragma omp parallel 
+  {
+    int rank;
+    int i;
+
+    rank = omp_get_thread_num();
+
+    #pragma omp for nowait 
+    for (i = 0; i < LOOPCOUNT; i++) {
+      if (i == 0) {
+        my_sleep(SLEEPTIME);
+        count = 1;
+        #pragma omp flush(count)
+      }
+    }
+    
+    #pragma omp for
+    for (i = 0; i < LOOPCOUNT; i++) {
+      #pragma omp flush(count)
+      if (count == 0)
+        result = 1;
+    }
+  }
+  return result;
+}
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for(i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_for_nowait()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/worksharing/for/omp_for_ordered.c

+// RUN: %libomp-compile-and-run
+#include <stdio.h>
+#include <math.h>
+#include "omp_testsuite.h"
+
+static int last_i = 0;
+
+/* Utility function to check that i is increasing monotonically 
+   with each call */
+static int check_i_islarger (int i)
+{
+  int islarger;
+  islarger = (i > last_i);
+  last_i = i;
+  return (islarger);
+}
+
+int test_omp_for_ordered()
+{
+  int sum;
+  int is_larger = 1;
+  int known_sum;
+
+  last_i = 0;
+  sum = 0;
+
+  #pragma omp parallel
+  {
+    int i;
+    int my_islarger = 1;
+    #pragma omp for schedule(static,1) ordered
+    for (i = 1; i < 100; i++) {
+      #pragma omp ordered
+      {
+        my_islarger = check_i_islarger(i) && my_islarger;
+        sum = sum + i;
+      }
+    }
+    #pragma omp critical
+    {
+      is_larger = is_larger && my_islarger;
+    }
+  }
+
+  known_sum=(99 * 100) / 2;
+  return ((known_sum == sum) && is_larger);
+}
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for(i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_for_ordered()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/worksharing/for/omp_for_private.c

+// RUN: %libomp-compile-and-run
+#include <stdio.h>
+#include <math.h>
+#include "omp_testsuite.h"
+
+/* Utility function do spend some time in a loop */
+static void do_some_work()
+{
+  int i;
+  double sum = 0;
+  for(i = 0; i < 1000; i++){
+  sum += sqrt ((double) i);
+  }
+}
+
+int sum1;
+#pragma omp threadprivate(sum1)
+
+int test_omp_for_private()
+{
+  int sum = 0;
+  int sum0;
+  int known_sum;
+
+  sum0 = 0;  /* setting (global) sum0 = 0 */
+
+  #pragma omp parallel
+  {
+    sum1 = 0;  /* setting sum1 in each thread to 0 */
+    {  /* begin of orphaned block */
+      int i;
+      #pragma omp for private(sum0) schedule(static,1)
+      for (i = 1; i <= LOOPCOUNT; i++) {
+        sum0 = sum1;
+        #pragma omp flush
+        sum0 = sum0 + i;
+        do_some_work ();
+        #pragma omp flush
+        sum1 = sum0;
+      }
+    }  /* end of orphaned block */
+
+    #pragma omp critical
+    {
+      sum = sum + sum1;
+    }  /*end of critical*/
+  }  /* end of parallel*/  
+  known_sum = (LOOPCOUNT * (LOOPCOUNT + 1)) / 2;
+  return (known_sum == sum);
+}                
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for(i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_for_private()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/worksharing/for/omp_for_reduction.c

+// RUN: %libomp-compile-and-run
+#include <stdio.h>
+#include <stdlib.h>
+#include <math.h>
+#include "omp_testsuite.h"
+
+#define DOUBLE_DIGITS 20  /* dt^DOUBLE_DIGITS */
+#define MAX_FACTOR 10
+#define KNOWN_PRODUCT 3628800  /* 10! */
+
+int test_omp_for_reduction ()
+{
+  double dt;
+  int sum;
+  int diff;
+  int product = 1;
+  double dsum;
+  double dknown_sum;
+  double ddiff;
+  int logic_and;
+  int logic_or;
+  int bit_and;
+  int bit_or;
+  int exclusiv_bit_or;
+  int *logics;
+  int i;
+  int known_sum;
+  int known_product;
+  double rounding_error = 1.E-9; /* over all rounding error to be
+                    ignored in the double tests */
+  double dpt;
+  int result = 0;
+  int logicsArray[LOOPCOUNT];
+
+  /* Variables for integer tests */
+  sum = 0;
+  product = 1;
+  known_sum = (LOOPCOUNT * (LOOPCOUNT + 1)) / 2;
+  /* variabels for double tests */
+  dt = 1. / 3.;  /* base of geometric row for + and - test*/
+  dsum = 0.;
+  /* Variabeles for logic  tests */
+  logics = logicsArray;
+  logic_and = 1;
+  logic_or = 0;
+  /* Variabeles for bit operators tests */
+  bit_and = 1;
+  bit_or = 0;
+  /* Variables for exclusiv bit or */
+  exclusiv_bit_or = 0;
+
+  /************************************************************************/
+  /** Tests for integers                         **/
+  /************************************************************************/
+
+  /**** Testing integer addition ****/
+  #pragma omp parallel 
+  {
+    int j;
+    #pragma omp for schedule(dynamic,1) reduction(+:sum)
+    for (j = 1; j <= LOOPCOUNT; j++) {
+      sum = sum + j;
+    }
+  }
+  if (known_sum != sum) {
+    result++;
+    fprintf (stderr, "Error in sum with integers: Result was %d"
+      " instead of %d.\n", sum, known_sum); 
+  }
+
+  /**** Testing integer subtracton ****/
+  diff = (LOOPCOUNT * (LOOPCOUNT + 1)) / 2;
+  #pragma omp parallel 
+  {
+    int j;
+    #pragma omp for schedule(dynamic,1) reduction(-:diff)
+    for (j = 1; j <= LOOPCOUNT; j++) {
+      diff = diff - j;
+    }
+  }
+  if (diff != 0) {
+    result++;
+    fprintf (stderr, "Error in difference with integers: Result was %d"
+      " instead of 0.\n", diff);
+  }
+
+  /**** Testing integer multiplication ****/
+  #pragma omp parallel 
+  {
+    int j;
+    #pragma omp for schedule(dynamic,1) reduction(*:product)
+    for (j = 1; j <= MAX_FACTOR; j++) {
+      product *= j;
+    }
+  }
+  known_product = KNOWN_PRODUCT;
+  if(known_product != product) {
+    result++;
+    fprintf (stderr,"Error in Product with integers: Result was %d"
+      " instead of %d\n",product,known_product);
+  }
+
+  /************************************************************************/
+  /** Tests for doubles                          **/
+  /************************************************************************/
+
+  /**** Testing double addition ****/
+  dsum = 0.;
+  dpt = 1.;
+  for (i = 0; i < DOUBLE_DIGITS; ++i) {
+    dpt *= dt;
+  }
+  dknown_sum = (1 - dpt) / (1 - dt);
+  #pragma omp parallel 
+  {
+    int j;
+    #pragma omp for schedule(dynamic,1) reduction(+:dsum)
+    for (j = 0; j < DOUBLE_DIGITS; j++) {  
+      dsum += pow (dt, j);
+    }
+  }
+  if (fabs (dsum - dknown_sum) > rounding_error) {
+    result++; 
+    fprintf (stderr, "\nError in sum with doubles: Result was %f"
+      " instead of: %f (Difference: %E)\n",
+      dsum, dknown_sum, dsum-dknown_sum);
+  }
+
+  /**** Testing double subtraction ****/
+  ddiff = (1 - dpt) / (1 - dt);
+  #pragma omp parallel 
+  {
+    int j;
+    #pragma omp for schedule(dynamic,1) reduction(-:ddiff)
+    for (j = 0; j < DOUBLE_DIGITS; ++j) {
+      ddiff -= pow (dt, j);
+    }
+  }
+  if (fabs (ddiff) > rounding_error) {
+    result++;
+    fprintf (stderr, "Error in Difference with doubles: Result was %E"
+      " instead of 0.0\n", ddiff);
+  }
+
+
+  /************************************************************************/
+  /** Tests for logical values                       **/
+  /************************************************************************/
+
+  /**** Testing logic and ****/
+  for (i = 0; i < LOOPCOUNT; i++) {
+    logics[i] = 1;
+  }
+
+  #pragma omp parallel 
+  {
+    int j;
+    #pragma omp for schedule(dynamic,1) reduction(&&:logic_and)
+    for (j = 0; j < LOOPCOUNT; ++j) {
+      logic_and = (logic_and && logics[j]);
+    }
+  }
+  if(!logic_and) {
+    result++;
+    fprintf (stderr, "Error in logic AND part 1\n");
+  }
+
+  logic_and = 1;
+  logics[LOOPCOUNT / 2] = 0;
+
+  #pragma omp parallel 
+  {
+    int j;
+    #pragma omp for schedule(dynamic,1) reduction(&&:logic_and)
+    for (j = 0; j < LOOPCOUNT; ++j) {
+      logic_and = logic_and && logics[j];
+    }
+  }
+  if(logic_and) {
+    result++;
+    fprintf (stderr, "Error in logic AND part 2\n");
+  }
+
+  /**** Testing logic or ****/
+  for (i = 0; i < LOOPCOUNT; i++) {
+    logics[i] = 0;
+  }
+
+  #pragma omp parallel 
+  {
+    int j;
+    #pragma omp for schedule(dynamic,1) reduction(||:logic_or)  
+    for (j = 0; j < LOOPCOUNT; ++j) {
+      logic_or = logic_or || logics[j];
+    }
+  }
+  if (logic_or) {
+    result++;
+    fprintf (stderr, "Error in logic OR part 1\n");
+  }
+
+  logic_or = 0;
+  logics[LOOPCOUNT / 2] = 1;
+
+  #pragma omp parallel 
+  {
+    int j;
+    #pragma omp for schedule(dynamic,1) reduction(||:logic_or)
+    for (j = 0; j < LOOPCOUNT; ++j) {
+      logic_or = logic_or || logics[j];
+    }
+  }
+  if(!logic_or) {
+    result++;
+    fprintf (stderr, "Error in logic OR part 2\n");
+  }
+
+  /************************************************************************/
+  /** Tests for bit values                         **/
+  /************************************************************************/
+
+  /**** Testing bit and ****/
+  for (i = 0; i < LOOPCOUNT; ++i) {
+    logics[i] = 1;
+  }
+
+  #pragma omp parallel 
+  {
+    int j;
+    #pragma omp for schedule(dynamic,1) reduction(&:bit_and)  
+    for (j = 0; j < LOOPCOUNT; ++j) {
+      bit_and = (bit_and & logics[j]);
+    }
+  }
+  if (!bit_and) {
+    result++;
+    fprintf (stderr, "Error in BIT AND part 1\n");
+  }
+
+  bit_and = 1;
+  logics[LOOPCOUNT / 2] = 0;
+
+  #pragma omp parallel 
+  {
+    int j;
+    #pragma omp for schedule(dynamic,1) reduction(&:bit_and)
+    for (j = 0; j < LOOPCOUNT; ++j) {
+      bit_and = bit_and & logics[j];
+    }
+  }
+  if (bit_and) {
+    result++;
+    fprintf (stderr, "Error in BIT AND part 2\n");
+  }
+
+  /**** Testing bit or ****/
+  for (i = 0; i < LOOPCOUNT; i++) {
+    logics[i] = 0;
+  }
+
+  #pragma omp parallel 
+  {
+    int j;
+    #pragma omp for schedule(dynamic,1) reduction(|:bit_or)
+    for (j = 0; j < LOOPCOUNT; ++j) {
+      bit_or = bit_or | logics[j];
+    }
+  }
+  if (bit_or) {
+    result++;
+    fprintf (stderr, "Error in BIT OR part 1\n");
+  }
+
+  bit_or = 0;
+  logics[LOOPCOUNT / 2] = 1;
+
+  #pragma omp parallel 
+  {
+    int j;
+    #pragma omp for schedule(dynamic,1) reduction(|:bit_or)
+    for (j = 0; j < LOOPCOUNT; ++j) {
+      bit_or = bit_or | logics[j];
+    }
+  }
+  if (!bit_or) {
+    result++;
+    fprintf (stderr, "Error in BIT OR part 2\n");
+  }
+
+  /**** Testing exclusive bit or ****/
+  for (i = 0; i < LOOPCOUNT; i++) {
+    logics[i] = 0;
+  }
+
+  #pragma omp parallel 
+  {
+    int j;
+    #pragma omp for schedule(dynamic,1) reduction(^:exclusiv_bit_or)
+    for (j = 0; j < LOOPCOUNT; ++j) {
+      exclusiv_bit_or = exclusiv_bit_or ^ logics[j];
+    }
+  }
+  if (exclusiv_bit_or) {
+    result++;
+    fprintf (stderr, "Error in EXCLUSIV BIT OR part 1\n");
+  }
+
+  exclusiv_bit_or = 0;
+  logics[LOOPCOUNT / 2] = 1;
+
+  #pragma omp parallel 
+  {
+    int j;
+    #pragma omp for schedule(dynamic,1) reduction(^:exclusiv_bit_or)
+    for (j = 0; j < LOOPCOUNT; ++j) {
+      exclusiv_bit_or = exclusiv_bit_or ^ logics[j];
+    }
+  }
+  if (!exclusiv_bit_or) {
+    result++;
+    fprintf (stderr, "Error in EXCLUSIV BIT OR part 2\n");
+  }
+
+  return (result == 0);
+  free (logics);
+}
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for(i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_for_reduction()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/worksharing/for/omp_for_schedule_auto.c

+// RUN: %libomp-compile-and-run
+#include <stdio.h>
+#include <math.h>
+#include "omp_testsuite.h"
+
+int sum1;
+#pragma omp threadprivate(sum1)
+
+int test_omp_for_auto()
+{
+  int sum;
+  int sum0;
+  int known_sum;
+  int threadsnum;
+
+  sum = 0;
+  sum0 = 12345;
+  sum1 = 0;
+
+  #pragma omp parallel
+  {
+    #pragma omp single
+    {
+      threadsnum=omp_get_num_threads();
+    }
+    /* sum0 = 0; */
+  
+    int i;
+    #pragma omp for firstprivate(sum0) schedule(auto)
+    for (i = 1; i <= LOOPCOUNT; i++) {
+      sum0 = sum0 + i;
+      sum1 = sum0;
+    }
+  
+    #pragma omp critical
+    {
+      sum = sum + sum1;
+    }
+  }
+
+  known_sum = 12345 * threadsnum + (LOOPCOUNT * (LOOPCOUNT + 1)) / 2;
+  return (known_sum == sum);
+}
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for(i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_for_auto()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/worksharing/for/omp_for_schedule_dynamic.c

+// RUN: %libomp-compile-and-run
+/*
+ * Test for dynamic scheduling with chunk size
+ * Method: caculate how many times the iteration space is dispatched
+ *     and judge if each dispatch has the requested chunk size
+ *     unless it is the last one.
+ * It is possible for two adjacent chunks are assigned to the same thread
+ * Modified by Chunhua Liao
+ */
+#include <stdio.h>
+#include <omp.h>
+#include <unistd.h>
+#include <stdlib.h>
+#include "omp_testsuite.h"
+#include "omp_my_sleep.h"
+
+#define CFDMAX_SIZE 100
+const int chunk_size = 7;
+
+int test_omp_for_schedule_dynamic()
+{
+  int tid;
+  int *tids;
+  int i;
+  int tidsArray[CFDMAX_SIZE];
+  int count = 0;
+  int tmp_count = 0; /*dispatch times*/
+  int *tmp;  /*store chunk size for each dispatch*/
+  int result = 0;
+  
+  tids = tidsArray;
+
+  #pragma omp parallel private(tid) shared(tids)
+  {        /* begin of parallel */
+    int tid;
+    tid = omp_get_thread_num ();
+    #pragma omp for schedule(dynamic,chunk_size)
+    for (i = 0; i < CFDMAX_SIZE; i++) {
+      tids[i] = tid;
+    }
+  }
+
+  for (i = 0; i < CFDMAX_SIZE - 1; ++i) {
+    if (tids[i] != tids[i + 1]) {
+      count++;
+    }
+  }
+
+  tmp = (int *) malloc (sizeof (int) * (count + 1));
+  tmp[0] = 1;
+
+  for (i = 0; i < CFDMAX_SIZE - 1; ++i) {
+    if (tmp_count > count) {
+      printf ("--------------------\nTestinternal Error: List too small!!!\n--------------------\n");  /* Error handling */
+      break;
+    }
+    if (tids[i] != tids[i + 1]) {
+      tmp_count++;
+      tmp[tmp_count] = 1;
+    } else {
+      tmp[tmp_count]++;
+    }
+  }
+  /* is dynamic statement working? */
+  for (i = 0; i < count; i++) {
+    if ((tmp[i]%chunk_size)!=0) {
+      /* it is possible for 2 adjacent chunks assigned to a same thread */
+      result++;
+      fprintf(stderr,"The intermediate dispatch has wrong chunksize.\n");
+      /* result += ((tmp[i] / chunk_size) - 1); */
+    }
+  }
+  if ((tmp[count]%chunk_size)!=(CFDMAX_SIZE%chunk_size)) { 
+    result++;
+    fprintf(stderr,"the last dispatch has wrong chunksize.\n");
+  }
+  /* for (int i=0;i<count+1;++i) printf("%d\t:=\t%d\n",i+1,tmp[i]); */
+  return (result==0);
+}
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for(i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_for_schedule_dynamic()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/worksharing/for/omp_for_schedule_guided.c

+// RUN: %libomp-compile-and-run
+
+/* Test for guided scheduling
+ * Ensure threads get chunks interleavely first
+ * Then judge the chunk sizes are decreasing to a stable value
+ * Modified by Chunhua Liao
+ * For example, 100 iteration on 2 threads, chunksize 7
+ * one line for each dispatch, 0/1 means thread id
+ * 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0  24
+ * 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1        18
+ * 0 0 0 0 0 0 0 0 0 0 0 0 0 0            14
+ * 1 1 1 1 1 1 1 1 1 1                10
+ * 0 0 0 0 0 0 0 0                   8
+ * 1 1 1 1 1 1 1                   7
+ * 0 0 0 0 0 0 0                   7
+ * 1 1 1 1 1 1 1                   7
+ * 0 0 0 0 0                     5
+*/
+#include <stdio.h>
+#include <unistd.h>
+#include <stdlib.h>
+#include "omp_testsuite.h"
+#include "omp_my_sleep.h"
+
+#define CFSMAX_SIZE 1000
+#define MAX_TIME  0.005
+
+#ifdef SLEEPTIME
+#undef SLEEPTIME
+#define SLEEPTIME 0.0001
+#endif
+
+int test_omp_for_schedule_guided()
+{
+  int * tids;
+  int * chunksizes;
+  int notout;
+  int maxiter;
+  int threads;
+  int i;
+  int result;
+
+  tids = (int *) malloc (sizeof (int) * (CFSMAX_SIZE + 1));
+  maxiter = 0;
+  result = 1;
+  notout = 1;
+
+  /* Testing if enough threads are available for this check. */
+  #pragma omp parallel
+  {
+    #pragma omp single
+    {
+      threads = omp_get_num_threads();
+    }
+  }
+
+  /* ensure there are at least two threads */
+  if (threads < 2) {
+    omp_set_num_threads(2);
+    threads = 2;
+  }
+
+  /* Now the real parallel work:  
+   * Each thread will start immediately with the first chunk.
+   */
+  #pragma omp parallel shared(tids,maxiter)
+  {  /* begin of parallel */
+    double count;
+    int tid;
+    int j;
+
+    tid = omp_get_thread_num ();
+
+    #pragma omp for nowait schedule(guided)
+    for(j = 0; j < CFSMAX_SIZE; ++j) {
+      count = 0.;
+      #pragma omp flush(maxiter)
+      if (j > maxiter) {
+        #pragma omp critical
+        {
+          maxiter = j;
+        }
+      }
+      /*printf ("thread %d sleeping\n", tid);*/
+      #pragma omp flush(maxiter,notout)  
+      while (notout && (count < MAX_TIME) && (maxiter == j)) {
+        #pragma omp flush(maxiter,notout)
+        my_sleep (SLEEPTIME);
+        count += SLEEPTIME;
+#ifdef VERBOSE
+        printf(".");
+#endif
+      }
+#ifdef VERBOSE
+      if (count > 0.) printf(" waited %lf s\n", count);
+#endif
+      /*printf ("thread %d awake\n", tid);*/
+      tids[j] = tid;
+#ifdef VERBOSE
+      printf("%d finished by %d\n",j,tid);
+#endif
+    } /* end of for */
+    notout = 0;
+    #pragma omp flush(maxiter,notout)
+  } /* end of parallel */
+
+  /*******************************************************
+   * evaluation of the values              *
+   *******************************************************/
+  {
+    int determined_chunksize = 1;
+    int last_threadnr = tids[0];
+    int global_chunknr = 0;
+    int openwork = CFSMAX_SIZE;
+    int expected_chunk_size;
+    int* local_chunknr = (int*)malloc(threads * sizeof(int));
+    double c = 1;
+
+    for (i = 0; i < threads; i++)
+      local_chunknr[i] = 0;
+
+    tids[CFSMAX_SIZE] = -1;
+
+    /*
+     * determine the number of global chunks
+     */
+    // fprintf(stderr,"# global_chunknr thread local_chunknr chunksize\n");
+    for(i = 1; i <= CFSMAX_SIZE; ++i) {
+      if (last_threadnr==tids[i]) { 
+        determined_chunksize++; 
+      } else {
+        /* fprintf(stderr, "%d\t%d\t%d\t%d\n", global_chunknr,
+           last_threadnr, local_chunknr[last_threadnr], m); */
+        global_chunknr++;
+        local_chunknr[last_threadnr]++;
+        last_threadnr = tids[i];
+        determined_chunksize = 1;
+      }
+    }
+    /* now allocate the memory for saving the sizes of the global chunks */
+    chunksizes = (int*)malloc(global_chunknr * sizeof(int));
+
+    /*
+    * Evaluate the sizes of the global chunks
+    */
+    global_chunknr = 0;
+    determined_chunksize = 1;
+    last_threadnr = tids[0];    
+    for (i = 1; i <= CFSMAX_SIZE; ++i) {
+      /* If the threadnumber was the same as before increase the
+       * detected chunksize for this chunk otherwise set the detected 
+       * chunksize again to one and save the number of the next 
+       * thread in last_threadnr. 
+       */
+      if (last_threadnr == tids[i]) { 
+        determined_chunksize++; 
+      } else {
+        chunksizes[global_chunknr] = determined_chunksize;
+        global_chunknr++;
+        local_chunknr[last_threadnr]++;
+        last_threadnr = tids[i];
+        determined_chunksize = 1;
+      }
+    }
+
+#ifdef VERBOSE
+    fprintf(stderr, "found\texpected\tconstant\n");
+#endif
+
+    /* identify the constant c for the exponential 
+       decrease of the chunksize */
+    expected_chunk_size = openwork / threads;
+    c = (double) chunksizes[0] / expected_chunk_size;
+    
+    for (i = 0; i < global_chunknr; i++) {
+      /* calculate the new expected chunksize */
+      if (expected_chunk_size > 1) 
+        expected_chunk_size = c * openwork / threads;
+#ifdef VERBOSE
+      fprintf(stderr, "%8d\t%8d\t%lf\n", chunksizes[i],
+        expected_chunk_size, c * chunksizes[i]/expected_chunk_size);
+#endif
+      /* check if chunksize is inside the rounding errors */
+      if (abs (chunksizes[i] - expected_chunk_size) >= 2) {
+        result = 0;
+#ifndef VERBOSE
+        fprintf(stderr, "Chunksize differed from expected "
+          "value: %d instead of %d\n", chunksizes[i],
+          expected_chunk_size);
+        return 0;
+#endif
+      } /* end if */
+
+#ifndef VERBOSE
+      if (expected_chunk_size - chunksizes[i] < 0)
+        fprintf(stderr, "Chunksize did not decrease: %d"
+          " instead of %d\n", chunksizes[i],expected_chunk_size);
+#endif
+
+      /* calculating the remaining amount of work */
+      openwork -= chunksizes[i];
+    }  
+  }
+  return result;
+}
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for(i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_for_schedule_guided()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/worksharing/for/omp_for_schedule_static.c

+// RUN: %libomp-compile-and-run
+#include <stdio.h>
+#include <unistd.h>
+#include <stdlib.h>
+#include "omp_testsuite.h"
+#include "omp_my_sleep.h"
+
+#define CFSMAX_SIZE 1000
+#define MAX_TIME 0.01
+
+#ifdef SLEEPTIME
+#undef SLEEPTIME
+#define SLEEPTIME 0.0005
+#endif
+
+int test_omp_for_schedule_static()
+{
+  int threads;
+  int i,lasttid;
+  int * tids;
+  int notout;
+  int maxiter;
+  int chunk_size;
+  int counter = 0;
+  int tmp_count=1;
+  int lastthreadsstarttid = -1;
+  int result = 1;
+
+  chunk_size = 7;
+  tids = (int *) malloc (sizeof (int) * (CFSMAX_SIZE + 1));
+  notout = 1;
+  maxiter = 0;
+
+  #pragma omp parallel shared(tids,counter)
+  {  /* begin of parallel*/
+    #pragma omp single
+    {
+      threads = omp_get_num_threads ();
+    }  /* end of single */
+  }  /* end of parallel */
+
+  if (threads < 2) {
+    omp_set_num_threads(2);
+    threads = 2;
+  }
+  fprintf (stderr,"Using an internal count of %d\nUsing a specified"
+    " chunksize of %d\n", CFSMAX_SIZE, chunk_size);
+  tids[CFSMAX_SIZE] = -1;  /* setting endflag */
+  #pragma omp parallel shared(tids)
+  { /* begin of parallel */
+    double count;
+    int tid;
+    int j;
+
+    tid = omp_get_thread_num ();
+
+    #pragma omp for nowait schedule(static,chunk_size)
+    for(j = 0; j < CFSMAX_SIZE; ++j) {
+      count = 0.;
+      #pragma omp flush(maxiter)
+      if (j > maxiter) {
+        #pragma omp critical
+        {
+          maxiter = j;
+        }
+      }
+      /*printf ("thread %d sleeping\n", tid);*/
+      while (notout && (count < MAX_TIME) && (maxiter == j)) {
+        #pragma omp flush(maxiter,notout)
+        my_sleep (SLEEPTIME);
+        count += SLEEPTIME;
+        printf(".");
+      }
+#ifdef VERBOSE
+      if (count > 0.) printf(" waited %lf s\n", count);
+#endif
+      /*printf ("thread %d awake\n", tid);*/
+      tids[j] = tid;
+#ifdef VERBOSE
+      printf("%d finished by %d\n",j,tid);
+#endif
+    } /* end of for */
+    notout = 0;
+    #pragma omp flush(maxiter,notout)
+  } /* end of parallel */
+
+  /**** analysing the data in array tids ****/
+
+  lasttid = tids[0];
+  tmp_count = 0; 
+
+  for (i = 0; i < CFSMAX_SIZE + 1; ++i) {
+    /* If the work  was done by the same thread increase tmp_count by one. */
+    if (tids[i] == lasttid) {
+      tmp_count++;
+#ifdef VERBOSE
+      fprintf (stderr, "%d: %d \n", i, tids[i]);
+#endif
+      continue;
+    }
+
+    /* Check if the next thread had has the right thread number. When finding 
+     * threadnumber -1 the end should be reached. 
+     */    
+    if (tids[i] == (lasttid + 1) % threads || tids[i] == -1) {
+      /* checking for the right chunk size */
+      if (tmp_count == chunk_size) {
+        tmp_count = 1;
+        lasttid = tids[i];
+#ifdef VERBOSE
+        fprintf (stderr, "OK\n");
+#endif
+      } else {
+        /* If the chunk size was wrong, check if the end was reached */
+        if (tids[i] == -1) {
+          if (i == CFSMAX_SIZE) {
+            fprintf (stderr, "Last thread had chunk size %d\n",
+              tmp_count);
+            break;
+          } else {
+            fprintf (stderr, "ERROR: Last thread (thread with"
+              " number -1) was found before the end.\n");
+            result = 0;
+          }
+        } else {
+          fprintf (stderr, "ERROR: chunk size was %d. (assigned"
+            " was %d)\n", tmp_count, chunk_size);
+          result = 0;
+        }
+      }
+    } else {
+      fprintf(stderr, "ERROR: Found thread with number %d (should be"
+        " inbetween 0 and %d).", tids[i], threads - 1);
+      result = 0;
+    }
+#ifdef VERBOSE
+    fprintf (stderr, "%d: %d \n", i, tids[i]);
+#endif
+  }
+
+  return result;
+}
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for(i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_for_schedule_static()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/worksharing/for/omp_for_schedule_static_3.c

+// RUN: %libomp-compile-and-run
+#include <stdio.h>
+#include <unistd.h>
+#include <stdlib.h>
+#include "omp_testsuite.h"
+#include "omp_my_sleep.h"
+
+#define CFSMAX_SIZE 1000
+#define MAX_TIME 0.01
+
+#ifdef SLEEPTIME
+#undef SLEEPTIME
+#define SLEEPTIME 0.0005
+#endif
+
+#define VERBOSE 0
+
+int test_omp_for_schedule_static_3()
+{
+  int threads;
+  int i,lasttid;
+
+  int * tids;
+  int * tids2;
+  int notout;
+  int maxiter;
+  int chunk_size;
+
+  int counter = 0;
+  int tmp_count=1;
+  int lastthreadsstarttid = -1;
+  int result = 1;
+  chunk_size = 7;
+
+  tids = (int *) malloc (sizeof (int) * (CFSMAX_SIZE + 1));
+  notout = 1;
+  maxiter = 0;
+
+  #pragma omp parallel shared(tids,counter)
+  {  /* begin of parallel*/
+    #pragma omp single
+    {
+      threads = omp_get_num_threads ();
+    }  /* end of single */
+  }  /* end of parallel */
+
+  /* Ensure that at least two threads are created */
+  if (threads < 2) {
+    omp_set_num_threads(2);
+    threads = 2;
+  } 
+  fprintf (stderr,"Using an internal count of %d\nUsing a"
+    " specified chunksize of %d\n", CFSMAX_SIZE, chunk_size);
+  tids[CFSMAX_SIZE] = -1;  /* setting endflag */
+
+  #pragma omp parallel shared(tids)
+  {  /* begin of parallel */
+    double count;
+    int tid;
+    int j;
+
+    tid = omp_get_thread_num ();
+
+    #pragma omp for nowait schedule(static,chunk_size)
+    for(j = 0; j < CFSMAX_SIZE; ++j) {
+      count = 0.;
+      #pragma omp flush(maxiter)
+      if (j > maxiter) {
+        #pragma omp critical
+        {
+          maxiter = j;
+        }
+      }
+      /*printf ("thread %d sleeping\n", tid);*/
+      while (notout && (count < MAX_TIME) && (maxiter == j)) {
+        #pragma omp flush(maxiter,notout)
+        my_sleep (SLEEPTIME);
+        count += SLEEPTIME;
+        printf(".");
+      }
+#ifdef VERBOSE
+      if (count > 0.) printf(" waited %lf s\n", count);
+#endif
+      /*printf ("thread %d awake\n", tid);*/
+      tids[j] = tid;
+#ifdef VERBOSE
+      printf("%d finished by %d\n",j,tid);
+#endif
+    } /* end of omp parallel for */
+
+    notout = 0;
+    #pragma omp flush(maxiter,notout)
+  } /* end of parallel */
+
+  /**** analysing the data in array tids ****/
+
+  lasttid = tids[0];
+  tmp_count = 0; 
+
+  for (i = 0; i < CFSMAX_SIZE + 1; ++i) {
+    /* If the work  was done by the same thread 
+       increase tmp_count by one. */
+    if (tids[i] == lasttid) {
+      tmp_count++;
+#ifdef VERBOSE
+      fprintf (stderr, "%d: %d \n", i, tids[i]);
+#endif
+      continue;
+    }
+
+    /* Check if the next thread had has the right thread number. 
+     * When finding threadnumber -1 the end should be reached. 
+     */    
+    if (tids[i] == (lasttid + 1) % threads || tids[i] == -1) {
+      /* checking for the right chunk size */
+      if (tmp_count == chunk_size) {
+        tmp_count = 1;
+        lasttid = tids[i];
+#ifdef VERBOSE
+        fprintf (stderr, "OK\n");
+#endif
+      } else {
+        /* If the chunk size was wrong, check if the end was reached */
+        if (tids[i] == -1) {
+          if (i == CFSMAX_SIZE) {
+            fprintf (stderr, "Last thread had chunk size %d\n",
+              tmp_count);
+            break;
+          } else {
+            fprintf (stderr, "ERROR: Last thread (thread with"
+              " number -1) was found before the end.\n");
+            result = 0;
+          }
+        } else {
+          fprintf (stderr, "ERROR: chunk size was %d. (assigned"
+            " was %d)\n", tmp_count, chunk_size);
+          result = 0;
+        }
+      }
+    } else {
+      fprintf(stderr, "ERROR: Found thread with number %d (should be"
+        " inbetween 0 and %d).", tids[i], threads - 1);
+      result = 0;
+    }
+#ifdef VERBOSE
+    fprintf (stderr, "%d: %d \n", i, tids[i]);
+#endif
+  }
+
+  /* Now we check if several loop regions in one parallel region have the
+   * same logical assignement of chunks to threads. We use the nowait 
+   * clause to increase the probability to get an error. */
+
+  /* First we allocate some more memmory */
+  free (tids);
+  tids = (int *) malloc (sizeof (int) * LOOPCOUNT);
+  tids2 = (int *) malloc (sizeof (int) * LOOPCOUNT);
+
+  #pragma omp parallel 
+  {
+    {
+      int n;
+      #pragma omp for schedule(static) nowait
+      for (n = 0; n < LOOPCOUNT; n++) {
+        if (LOOPCOUNT == n + 1 )
+          my_sleep(SLEEPTIME);
+
+        tids[n] = omp_get_thread_num();
+      }
+    }
+    {
+      int m;
+      #pragma omp for schedule(static) nowait
+      for (m = 1; m <= LOOPCOUNT; m++) {
+        tids2[m-1] = omp_get_thread_num();
+      }
+    }
+  }
+
+  for (i = 0; i < LOOPCOUNT; i++)
+  if (tids[i] != tids2[i]) {
+    fprintf (stderr, "Chunk no. %d was assigned once to thread %d and"
+      " later to thread %d.\n", i, tids[i],tids2[i]);
+    result = 0;
+  }
+
+  free (tids);
+  free (tids2);
+  return result;
+}
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for (i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_for_schedule_static_3()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/worksharing/for/omp_parallel_for_firstprivate.c

+// RUN: %libomp-compile-and-run
+#include <stdio.h>
+#include "omp_testsuite.h"
+
+int test_omp_parallel_for_firstprivate()
+{
+  int sum ;
+  int i2;
+  int i;
+  int known_sum;
+
+  sum=0;
+  i2=3;
+
+  #pragma omp parallel for reduction(+:sum) private(i) firstprivate(i2)
+  for (i = 1; i <= LOOPCOUNT; i++) {
+    sum = sum + (i + i2);
+  }
+
+  known_sum = (LOOPCOUNT * (LOOPCOUNT + 1)) / 2 + i2 * LOOPCOUNT;
+  return (known_sum == sum);
+} /* end of check_parallel_for_fistprivate */
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for(i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_parallel_for_firstprivate()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/worksharing/for/omp_parallel_for_if.c

+// RUN: %libomp-compile-and-run
+#include <stdio.h>
+#include <math.h>
+#include "omp_testsuite.h"
+
+int test_omp_parallel_for_if()
+{
+  int known_sum;
+  int num_threads;
+  int sum, sum2;
+  int i;
+  int control;
+
+  control = 0;
+  num_threads=0;
+  sum = 0;
+  sum2 = 0;
+
+  #pragma omp parallel for private(i) if (control==1)
+  for (i=0; i <= LOOPCOUNT; i++) {
+    num_threads = omp_get_num_threads();
+    sum = sum + i;
+  }
+
+  known_sum = (LOOPCOUNT * (LOOPCOUNT + 1)) / 2;
+  fprintf(stderr, "Number of threads determined by"
+    " omp_get_num_threads: %d\n", num_threads);
+  return (known_sum == sum && num_threads == 1);
+} /* end of check_parallel_for_private */
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for(i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_parallel_for_if()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/worksharing/for/omp_parallel_for_lastprivate.c

+// RUN: %libomp-compile-and-run
+#include <stdio.h>
+#include "omp_testsuite.h"
+
+int test_omp_parallel_for_lastprivate()
+{
+  int sum;
+  int i;
+  int i0;
+  int known_sum;
+
+  sum =0;
+  i0 = -1;
+
+  #pragma omp parallel for reduction(+:sum) \
+    schedule(static,7) private(i) lastprivate(i0)
+  for (i = 1; i <= LOOPCOUNT; i++) {
+    sum = sum + i;
+    i0 = i;
+  } /* end of parallel for */
+
+  known_sum = (LOOPCOUNT * (LOOPCOUNT + 1)) / 2;
+  return ((known_sum == sum) && (i0 == LOOPCOUNT));
+} /* end of check_parallel_for_lastprivate */
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for(i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_parallel_for_lastprivate()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/worksharing/for/omp_parallel_for_ordered.c

+// RUN: %libomp-compile-and-run
+#include <stdio.h>
+#include "omp_testsuite.h"
+
+static int last_i = 0;
+
+int i;
+#pragma omp threadprivate(i)
+
+/* Variable ii is used to avoid problems with a threadprivate variable used as a loop
+ * index. See test omp_threadprivate_for.
+ */
+static int ii;
+#pragma omp threadprivate(ii)
+
+/*! 
+  Utility function: returns true if the passed argument is larger than 
+  the argument of the last call of this function.
+ */
+static int check_i_islarger2(int i)
+{
+  int islarger;
+  islarger = (i > last_i);
+  last_i = i;
+  return (islarger);
+}
+
+int test_omp_parallel_for_ordered()
+{
+  int sum;
+  int is_larger;
+  int known_sum;
+  int i;
+
+  sum = 0;
+  is_larger = 1;
+  last_i = 0;
+  #pragma omp parallel for schedule(static,1) private(i) ordered
+  for (i = 1; i < 100; i++) {
+    ii = i;
+    #pragma omp ordered
+    {
+      is_larger = check_i_islarger2 (ii) && is_larger;
+      sum  = sum + ii;
+    }
+  }
+  known_sum = (99 * 100) / 2;
+  fprintf (stderr," known_sum = %d , sum = %d \n", known_sum, sum);
+  fprintf (stderr," is_larger = %d\n", is_larger);
+  return (known_sum == sum) && is_larger;
+}
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for(i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_parallel_for_ordered()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/worksharing/for/omp_parallel_for_private.c

+// RUN: %libomp-compile-and-run
+#include <stdio.h>
+#include <math.h>
+#include "omp_testsuite.h"
+
+/*! Utility function to spend some time in a loop */
+static void do_some_work (void)
+{
+  int i;
+  double sum = 0;
+  for(i = 0; i < 1000; i++){
+    sum += sqrt (i);
+  }
+}
+
+int test_omp_parallel_for_private()
+{
+  int sum;
+  int i;
+  int i2;
+  int known_sum;
+
+  sum =0;
+  i2=0;
+
+  #pragma omp parallel for reduction(+:sum) schedule(static,1) private(i) private(i2)
+  for (i=1;i<=LOOPCOUNT;i++)
+  {
+    i2 = i;
+    #pragma omp flush
+    do_some_work ();
+    #pragma omp flush
+    sum = sum + i2;
+  } /*end of for*/
+  known_sum = (LOOPCOUNT * (LOOPCOUNT + 1)) / 2;
+  return (known_sum == sum);
+} /* end of check_parallel_for_private */
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for(i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_parallel_for_private()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/worksharing/for/omp_parallel_for_reduction.c

+// RUN: %libomp-compile-and-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! */
+
+int test_omp_parallel_for_reduction()
+{
+  int sum;
+  int known_sum;
+  double dsum;
+  double dknown_sum;
+  double dt=0.5;  /* base of geometric row for + and - test*/
+  double rounding_error= 1.E-9;
+  int diff;
+  double ddiff;
+  int product;
+  int known_product;
+  int logic_and;
+  int logic_or;
+  int bit_and;
+  int bit_or;
+  int exclusiv_bit_or;
+  int logics[LOOPCOUNT];
+  int i;
+  double dpt;
+  int result;
+
+  sum =0;
+  dsum=0;
+  dt = 1./3.;
+  result = 0;
+  product = 1;
+  logic_and=1;
+  logic_or=0;
+  bit_and=1;
+  bit_or=0;
+  exclusiv_bit_or=0;
+
+  /* Tests for integers */
+  known_sum = (LOOPCOUNT*(LOOPCOUNT+1))/2;
+  #pragma omp parallel for schedule(dynamic,1) private(i) reduction(+:sum)
+  for (i=1;i<=LOOPCOUNT;i++) {
+    sum=sum+i;
+  }
+  if(known_sum!=sum) {
+    result++;
+    fprintf(stderr,"Error in sum with integers: Result was %d"
+      " instead of %d\n",sum,known_sum); 
+  }
+
+  diff = (LOOPCOUNT*(LOOPCOUNT+1))/2;
+  #pragma omp parallel for schedule(dynamic,1) private(i) reduction(-:diff)
+  for (i=1;i<=LOOPCOUNT;++i) {
+    diff=diff-i;
+  }
+  if(diff != 0) {
+    result++;
+    fprintf(stderr,"Error in difference with integers: Result was %d"
+      " instead of 0.\n",diff);
+  }
+
+  /* Tests for doubles */
+  dsum=0;
+  dpt=1;
+  for (i=0;i<DOUBLE_DIGITS;++i) {
+    dpt*=dt;
+  }
+  dknown_sum = (1-dpt)/(1-dt);
+  #pragma omp parallel for schedule(dynamic,1) private(i) reduction(+:dsum)
+  for (i=0;i<DOUBLE_DIGITS;++i) {
+    dsum += pow(dt,i);
+  }
+  if( fabs(dsum-dknown_sum) > rounding_error ) {
+    result++; 
+    fprintf(stderr,"Error in sum with doubles: Result was %f"
+      " instead of %f (Difference: %E)\n",
+      dsum, dknown_sum, dsum-dknown_sum);
+  }
+
+  dpt=1;
+
+  for (i=0;i<DOUBLE_DIGITS;++i) {
+    dpt*=dt;
+  }
+  fprintf(stderr,"\n");
+  ddiff = (1-dpt)/(1-dt);
+  #pragma omp parallel for schedule(dynamic,1) private(i) reduction(-:ddiff)
+  for (i=0;i<DOUBLE_DIGITS;++i) {
+    ddiff -= pow(dt,i);
+  }
+  if( fabs(ddiff) > rounding_error) {
+    result++;
+    fprintf(stderr,"Error in Difference with doubles: Result was %E"
+      " instead of 0.0\n",ddiff);
+  }
+
+  /* Tests for integers */
+  #pragma omp parallel for schedule(dynamic,1) private(i) reduction(*:product)
+  for(i=1;i<=MAX_FACTOR;i++) {
+    product *= i;
+  }
+  known_product = KNOWN_PRODUCT;
+  if(known_product != product) {
+    result++;
+    fprintf(stderr,"Error in Product with integers: Result was %d"
+      " instead of %d\n\n",product,known_product);
+  }
+
+  /* Tests for logic AND */
+  for(i=0;i<LOOPCOUNT;i++) {
+    logics[i]=1;
+  }
+
+  #pragma omp parallel for schedule(dynamic,1) private(i) \
+    reduction(&&:logic_and)
+  for(i=0;i<LOOPCOUNT;++i) {
+    logic_and = (logic_and && logics[i]);
+  }
+  if(!logic_and) {
+    result++;
+    fprintf(stderr,"Error in logic AND part 1.\n");
+  }
+
+  logic_and = 1;
+  logics[LOOPCOUNT/2]=0;
+
+  #pragma omp parallel for schedule(dynamic,1) private(i) \
+    reduction(&&:logic_and)
+  for(i=0;i<LOOPCOUNT;++i) {
+    logic_and = logic_and && logics[i];
+  }
+  if(logic_and) {
+    result++;
+    fprintf(stderr,"Error in logic AND part 2.\n");
+  }
+
+  /* Tests for logic OR */
+  for(i=0;i<LOOPCOUNT;i++) {
+    logics[i]=0;
+  }
+
+  #pragma omp parallel for schedule(dynamic,1) private(i) \
+    reduction(||:logic_or)
+  for(i=0;i<LOOPCOUNT;++i) {
+    logic_or = logic_or || logics[i];
+  }
+  if(logic_or) {
+    result++;
+    fprintf(stderr,"Error in logic OR part 1.\n");
+  }
+  logic_or = 0;
+  logics[LOOPCOUNT/2]=1;
+
+  #pragma omp parallel for schedule(dynamic,1) private(i) \
+    reduction(||:logic_or)
+  for(i=0;i<LOOPCOUNT;++i) {
+    logic_or = logic_or || logics[i];
+  }
+  if(!logic_or) {
+    result++;
+    fprintf(stderr,"Error in logic OR part 2.\n");
+  }
+
+  /* Tests for bitwise AND */
+  for(i=0;i<LOOPCOUNT;++i) {
+    logics[i]=1;
+  }
+
+  #pragma omp parallel for schedule(dynamic,1) private(i) \
+    reduction(&:bit_and)
+  for(i=0;i<LOOPCOUNT;++i) {
+    bit_and = (bit_and & logics[i]);
+  }
+  if(!bit_and) {
+    result++;
+    fprintf(stderr,"Error in BIT AND part 1.\n");
+  }
+
+  bit_and = 1;
+  logics[LOOPCOUNT/2]=0;
+
+  #pragma omp parallel for schedule(dynamic,1) private(i) \
+    reduction(&:bit_and)
+  for(i=0;i<LOOPCOUNT;++i) {
+    bit_and = bit_and & logics[i];
+  }
+  if(bit_and) {
+    result++;
+    fprintf(stderr,"Error in BIT AND part 2.\n");
+  }
+
+  /* Tests for bitwise OR */
+  for(i=0;i<LOOPCOUNT;i++) {
+    logics[i]=0;
+  }
+
+  #pragma omp parallel for schedule(dynamic,1) private(i) \
+    reduction(|:bit_or)
+  for(i=0;i<LOOPCOUNT;++i) {
+    bit_or = bit_or | logics[i];
+  }
+  if(bit_or) {
+    result++;
+    fprintf(stderr,"Error in BIT OR part 1\n");
+  }
+  bit_or = 0;
+  logics[LOOPCOUNT/2]=1;
+
+  #pragma omp parallel for schedule(dynamic,1) private(i) \
+    reduction(|:bit_or)
+  for(i=0;i<LOOPCOUNT;++i) {
+    bit_or = bit_or | logics[i];
+  }
+  if(!bit_or) {
+    result++;
+    fprintf(stderr,"Error in BIT OR part 2\n");
+  }
+
+  /* Tests for bitwise XOR */
+  for(i=0;i<LOOPCOUNT;i++) {
+    logics[i]=0;
+  }
+
+  #pragma omp parallel for schedule(dynamic,1) private(i) \
+    reduction(^:exclusiv_bit_or)
+  for(i=0;i<LOOPCOUNT;++i) {
+    exclusiv_bit_or = exclusiv_bit_or ^ logics[i];
+  }
+  if(exclusiv_bit_or) {
+    result++;
+    fprintf(stderr,"Error in EXCLUSIV BIT OR part 1\n");
+  }
+
+  exclusiv_bit_or = 0;
+  logics[LOOPCOUNT/2]=1;
+
+  #pragma omp parallel for schedule(dynamic,1) private(i) \
+    reduction(^:exclusiv_bit_or)
+  for(i=0;i<LOOPCOUNT;++i) {
+    exclusiv_bit_or = exclusiv_bit_or ^ logics[i];
+  }
+  if(!exclusiv_bit_or) {
+    result++;
+    fprintf(stderr,"Error in EXCLUSIV BIT OR part 2\n");
+  }
+
+  /*printf("\nResult:%d\n",result);*/
+  return (result==0);
+}
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for(i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_parallel_for_reduction()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/worksharing/sections/omp_parallel_sections_firstprivate.c

+// RUN: %libomp-compile-and-run
+#include <stdio.h>
+#include "omp_testsuite.h"
+
+int test_omp_parallel_sections_firstprivate()
+{
+  int sum;
+  int sum0;
+  int known_sum;
+
+  sum =7;
+  sum0=11;
+
+  #pragma omp parallel sections firstprivate(sum0)
+  {
+    #pragma omp section 
+    {
+      #pragma omp critical
+      {
+        sum= sum+sum0;
+      }
+    }  
+    #pragma omp section
+    {
+      #pragma omp critical
+      {
+        sum= sum+sum0;
+      }
+    }
+    #pragma omp section
+    {
+      #pragma omp critical
+      {
+        sum= sum+sum0;
+      }
+    }         
+  }
+
+  known_sum=11*3+7;
+  return (known_sum==sum); 
+} /* end of check_section_firstprivate*/
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for(i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_parallel_sections_firstprivate()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/worksharing/sections/omp_parallel_sections_lastprivate.c

+// RUN: %libomp-compile-and-run
+#include <stdio.h>
+#include "omp_testsuite.h"
+
+int test_omp_parallel_sections_lastprivate()
+{
+  int sum;
+  int sum0;
+  int i;
+  int i0;
+  int known_sum;
+  sum =0;
+  sum0 = 0;
+  i0 = -1;
+
+  #pragma omp parallel sections private(i,sum0) lastprivate(i0)
+  {
+    #pragma omp section  
+    {
+      sum0=0;
+      for (i=1;i<400;i++) {
+        sum0=sum0+i;
+        i0=i;
+      }
+      #pragma omp critical
+      {
+        sum= sum+sum0;
+      }
+    }
+    #pragma omp section 
+    {
+      sum0=0;
+      for(i=400;i<700;i++) {
+        sum0=sum0+i;
+        i0=i;
+      }
+      #pragma omp critical
+      {
+        sum= sum+sum0;
+      }
+    }
+    #pragma omp section 
+    {
+      sum0=0;
+      for(i=700;i<1000;i++) {
+        sum0=sum0+i;
+        i0=i;
+      }
+      #pragma omp critical
+      {
+        sum= sum+sum0;
+      }
+    }
+  }
+
+  known_sum=(999*1000)/2;
+  return ((known_sum==sum) && (i0==999) );
+}
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for(i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_parallel_sections_lastprivate()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/worksharing/sections/omp_parallel_sections_private.c

+// RUN: %libomp-compile-and-run
+#include <stdio.h>
+#include "omp_testsuite.h"
+
+int test_omp_parallel_sections_private()
+{
+  int sum;
+  int sum0;
+  int i;
+  int known_sum;
+
+  sum = 7;
+  sum0=0;
+
+  #pragma omp parallel sections private(sum0, i)
+  {
+    #pragma omp section 
+    {
+      sum0=0;
+      for (i=1;i<400;i++)
+        sum0=sum0+i;
+      #pragma omp critical
+      {
+        sum= sum+sum0;
+      }
+    }
+    #pragma omp section
+    {
+      sum0=0;
+      for(i=400;i<700;i++)
+        sum0=sum0+i;
+      #pragma omp critical
+      {
+        sum= sum+sum0;
+      }
+    }
+    #pragma omp section
+    {
+      sum0=0;
+      for(i=700;i<1000;i++)
+        sum0=sum0+i;
+      #pragma omp critical
+      {
+        sum= sum+sum0;
+      }
+    }
+  }
+
+  known_sum=(999*1000)/2+7;
+  return (known_sum==sum); 
+} /* end of check_section_private*/
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for(i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_parallel_sections_private()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/worksharing/sections/omp_parallel_sections_reduction.c

+// RUN: %libomp-compile-and-run
+#include <stdio.h>
+#include <math.h>
+#include "omp_testsuite.h"
+
+int test_omp_parallel_sections_reduction()
+{
+  int sum;
+  int known_sum;  
+  double dpt;
+  double dsum;
+  double dknown_sum;
+  double dt=0.5; /* base of geometric row for + and - test*/
+  double rounding_error= 1.E-5;
+  int diff;
+  double ddiff;
+  int product;
+  int known_product;
+  int logic_and;
+  int bit_and;
+  int logic_or;
+  int bit_or;
+  int exclusiv_bit_or;
+  int logics[1000];
+  int i;
+  int result;
+
+  sum = 7;
+  dsum=0;
+  product =1;
+  dpt = 1;
+  logic_and=1;
+  bit_and=1;
+  logic_or=0;
+  bit_or=0;
+  exclusiv_bit_or=0;
+  result =0;
+  /*  int my_islarger;*/
+  /*int is_larger=1;*/
+
+  // Test summation of integers
+  known_sum = (999*1000)/2+7;
+  #pragma omp parallel sections private(i) reduction(+:sum)
+  {
+    #pragma omp section
+    {
+      for (i=1;i<300;i++) {
+        sum=sum+i;
+      }
+    }
+    #pragma omp section
+    {
+      for (i=300;i<700;i++) {
+        sum=sum+i;
+      }
+    }
+    #pragma omp section
+    {
+      for (i=700;i<1000;i++) {
+        sum=sum+i;
+      }
+    }
+  }
+  if(known_sum!=sum) {
+    result++;
+    fprintf(stderr,"Error in sum with integers: Result was %d"
+      " instead of %d.\n",sum, known_sum);
+  }
+
+  // Test differences of integers
+  diff = (999*1000)/2;
+  #pragma omp parallel sections private(i) reduction(-:diff)
+  {
+    #pragma omp section
+    {
+      for (i=1;i<300;i++) {
+        diff=diff-i;
+      }
+    }
+    #pragma omp section
+    {
+      for (i=300;i<700;i++) {
+        diff=diff-i;
+      }
+    }
+    #pragma omp section
+    {
+      for (i=700;i<1000;i++) {
+        diff=diff-i;
+      }
+    }
+  }
+  if(diff != 0) {
+    result++;
+    fprintf(stderr,"Error in Difference with integers: Result was %d"
+      " instead of 0.\n",diff);
+  }
+
+  // Test summation of doubles
+  for (i=0;i<20;++i) {
+    dpt*=dt;
+  }
+  dknown_sum = (1-dpt)/(1-dt);
+  #pragma omp parallel sections private(i) reduction(+:dsum)
+  {
+    #pragma omp section
+    {
+      for (i=0;i<6;++i) {
+        dsum += pow(dt,i);
+      }
+    }
+    #pragma omp section
+    {
+      for (i=6;i<12;++i) {
+        dsum += pow(dt,i);
+      }
+    }
+    #pragma omp section
+    {
+      for (i=12;i<20;++i) {
+        dsum += pow(dt,i);
+      }
+    }
+  }
+  if( fabs(dsum-dknown_sum) > rounding_error ) {
+    result++; 
+    fprintf(stderr,"Error in sum with doubles: Result was %f"
+      " instead of %f (Difference: %E)\n",
+      dsum, dknown_sum, dsum-dknown_sum);
+  }
+
+  // Test differences of doubles
+  dpt=1;
+  for (i=0;i<20;++i) {
+    dpt*=dt;
+  }
+  fprintf(stderr,"\n");
+  ddiff = (1-dpt)/(1-dt);
+  #pragma omp parallel sections private(i) reduction(-:ddiff)
+  {
+    #pragma omp section
+    {
+      for (i=0;i<6;++i) {
+        ddiff -= pow(dt,i);
+      }
+    }
+    #pragma omp section
+    {
+      for (i=6;i<12;++i) {
+        ddiff -= pow(dt,i);
+      }
+    }
+    #pragma omp section
+    {
+      for (i=12;i<20;++i) {
+        ddiff -= pow(dt,i);
+      }
+    }
+  }
+  if( fabs(ddiff) > rounding_error) {
+    result++;
+    fprintf(stderr,"Error in Difference with doubles: Result was %E"
+      " instead of 0.0\n",ddiff);
+  }
+
+  // Test product of integers
+  known_product = 3628800;
+  #pragma omp parallel sections private(i) reduction(*:product)
+  {
+    #pragma omp section
+    {  
+      for(i=1;i<3;i++) {
+        product *= i;
+      }
+    }
+    #pragma omp section
+    {
+      for(i=3;i<7;i++) {
+        product *= i;
+      }
+    }
+    #pragma omp section
+    {
+      for(i=7;i<11;i++) {
+        product *= i;
+      }
+    }
+  }
+  if(known_product != product) {
+    result++;
+    fprintf(stderr,"Error in Product with integers: Result was %d"
+      " instead of %d\n",product,known_product);
+  }
+
+  // Test logical AND
+  for(i=0;i<1000;i++) {
+    logics[i]=1;
+  }
+
+  #pragma omp parallel sections private(i) reduction(&&:logic_and)
+  {
+    #pragma omp section
+    {
+      for (i=1;i<300;i++) {
+        logic_and = (logic_and && logics[i]);
+      }
+    }
+    #pragma omp section
+    {
+      for (i=300;i<700;i++) {
+        logic_and = (logic_and && logics[i]);
+      }
+    }
+    #pragma omp section
+    {
+      for (i=700;i<1000;i++) {
+        logic_and = (logic_and && logics[i]);
+      }
+    }
+  }
+  if(!logic_and) {
+    result++;
+    fprintf(stderr,"Error in logic AND part 1\n");
+  }
+  logic_and = 1;
+  logics[501] = 0;
+
+  #pragma omp parallel sections private(i) reduction(&&:logic_and)
+  {
+    #pragma omp section
+    {
+      for (i=1;i<300;i++) {
+        logic_and = (logic_and && logics[i]);
+      }
+    }
+    #pragma omp section
+    {
+      for (i=300;i<700;i++) {
+        logic_and = (logic_and && logics[i]);
+      }
+    }
+    #pragma omp section
+    {
+      for (i=700;i<1000;i++) {
+        logic_and = (logic_and && logics[i]);
+      }
+    }
+  }
+  if(logic_and) {
+    result++;
+    fprintf(stderr,"Error in logic AND part 2");
+  }
+
+  // Test logical OR
+  for(i=0;i<1000;i++) {
+    logics[i]=0;
+  }
+
+  #pragma omp parallel sections private(i) reduction(||:logic_or)
+  {
+    #pragma omp section
+    {
+      for (i=1;i<300;i++) {
+        logic_or = (logic_or || logics[i]);
+      }
+    }
+    #pragma omp section
+    {
+      for (i=300;i<700;i++) {
+        logic_or = (logic_or || logics[i]);
+      }
+    }
+    #pragma omp section
+    {
+      for (i=700;i<1000;i++) {
+        logic_or = (logic_or || logics[i]);
+      }
+    }
+  }
+  if(logic_or) {
+    result++;
+    fprintf(stderr,"Error in logic OR part 1\n");
+  }
+
+  logic_or = 0;
+  logics[501]=1;
+
+  #pragma omp parallel sections private(i) reduction(||:logic_or)
+  {
+    #pragma omp section
+    {
+      for (i=1;i<300;i++) {
+        logic_or = (logic_or || logics[i]);
+      }
+    }
+    #pragma omp section
+    {
+      for (i=300;i<700;i++) {
+        logic_or = (logic_or || logics[i]);
+      }
+    }
+    #pragma omp section
+    {
+      for (i=700;i<1000;i++) {
+        logic_or = (logic_or || logics[i]);
+      }
+    }
+  }
+  if(!logic_or) {
+    result++;
+    fprintf(stderr,"Error in logic OR part 2\n");
+  }
+
+  // Test bitwise AND
+  for(i=0;i<1000;++i) {
+    logics[i]=1;
+  }
+
+  #pragma omp parallel sections private(i) reduction(&:bit_and)
+  {
+    #pragma omp section
+    {  
+      for(i=0;i<300;++i) {
+        bit_and = (bit_and & logics[i]);
+      }
+    }
+    #pragma omp section
+    {  
+      for(i=300;i<700;++i) {
+        bit_and = (bit_and & logics[i]);
+      }
+    }
+    #pragma omp section
+    {  
+      for(i=700;i<1000;++i) {
+        bit_and = (bit_and & logics[i]);
+      }
+    }
+  }
+  if(!bit_and) {
+    result++;
+    fprintf(stderr,"Error in BIT AND part 1\n");
+  }
+
+  bit_and = 1;
+  logics[501]=0;
+
+  #pragma omp parallel sections private(i) reduction(&:bit_and)
+  {
+    #pragma omp section
+    {
+      for(i=0;i<300;++i) {
+        bit_and = bit_and & logics[i];
+      }
+    }
+    #pragma omp section
+    {
+      for(i=300;i<700;++i) {
+        bit_and = bit_and & logics[i];
+      }
+    }
+    #pragma omp section
+    {
+      for(i=700;i<1000;++i) {
+        bit_and = bit_and & logics[i];
+      }
+    }
+  }
+  if(bit_and) {
+    result++;
+    fprintf(stderr,"Error in BIT AND part 2");
+  }
+
+  // Test bitwise OR
+  for(i=0;i<1000;i++) {
+    logics[i]=0;
+  }
+
+  #pragma omp parallel sections private(i) reduction(|:bit_or)
+  {
+    #pragma omp section
+    {
+      for(i=0;i<300;++i) {
+        bit_or = bit_or | logics[i];
+      }
+    }
+    #pragma omp section
+    {
+      for(i=300;i<700;++i) {
+        bit_or = bit_or | logics[i];
+      }
+    }
+    #pragma omp section
+    {
+      for(i=700;i<1000;++i) {
+        bit_or = bit_or | logics[i];
+      }
+    }
+  }
+  if(bit_or) {
+    result++;
+    fprintf(stderr,"Error in BIT OR part 1\n");
+  }
+  bit_or = 0;
+  logics[501]=1;
+
+  #pragma omp parallel sections private(i) reduction(|:bit_or)
+  {
+    #pragma omp section
+    {
+      for(i=0;i<300;++i) {
+        bit_or = bit_or | logics[i];
+      }
+    }
+    #pragma omp section
+    {
+      for(i=300;i<700;++i) {
+        bit_or = bit_or | logics[i];
+      }
+    }
+    #pragma omp section
+    {
+      for(i=700;i<1000;++i) {
+        bit_or = bit_or | logics[i];
+      }
+    }
+  }
+  if(!bit_or) {
+    result++;
+    fprintf(stderr,"Error in BIT OR part 2\n");
+  }
+
+  // Test bitwise XOR
+  for(i=0;i<1000;i++) {
+    logics[i]=0;
+  }
+
+  #pragma omp parallel sections private(i) reduction(^:exclusiv_bit_or)
+  {
+    #pragma omp section
+    {  
+      for(i=0;i<300;++i) {
+        exclusiv_bit_or = exclusiv_bit_or ^ logics[i];
+      }
+    }
+    #pragma omp section
+    {  
+      for(i=300;i<700;++i) {
+        exclusiv_bit_or = exclusiv_bit_or ^ logics[i];
+      }
+    }
+    #pragma omp section
+    {  
+      for(i=700;i<1000;++i) {
+        exclusiv_bit_or = exclusiv_bit_or ^ logics[i];
+      }
+    }
+  }
+  if(exclusiv_bit_or) {
+    result++;
+    fprintf(stderr,"Error in EXCLUSIV BIT OR part 1\n");
+  }
+
+  exclusiv_bit_or = 0;
+  logics[501]=1;
+
+  #pragma omp parallel sections private(i) reduction(^:exclusiv_bit_or)
+  {
+    #pragma omp section
+    {
+      for(i=0;i<300;++i) {
+        exclusiv_bit_or = exclusiv_bit_or ^ logics[i];
+      }
+    }
+    #pragma omp section
+    {
+      for(i=300;i<700;++i) {
+        exclusiv_bit_or = exclusiv_bit_or ^ logics[i];
+      }
+    }
+    #pragma omp section
+    {
+      for(i=700;i<1000;++i) {
+        exclusiv_bit_or = exclusiv_bit_or ^ logics[i];
+      }
+    }
+  }
+  if(!exclusiv_bit_or) {
+    result++;
+    fprintf(stderr,"Error in EXCLUSIV BIT OR part 2\n");
+  }
+
+  /*printf("\nResult:%d\n",result);*/
+  return (result==0);
+}
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for(i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_parallel_sections_reduction()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/worksharing/sections/omp_section_firstprivate.c

+// RUN: %libomp-compile-and-run
+#include <stdio.h>
+#include "omp_testsuite.h"
+
+int test_omp_section_firstprivate()
+{
+  int sum;
+  int sum0;
+  int known_sum;
+
+  sum0 = 11;
+  sum = 7;
+  #pragma omp parallel
+  {
+    #pragma omp  sections firstprivate(sum0)
+    {
+      #pragma omp section 
+      {
+        #pragma omp critical
+        {
+          sum = sum + sum0;
+        }
+      }  
+      #pragma omp section
+      {
+        #pragma omp critical
+        {
+          sum = sum + sum0;
+        }
+      }
+      #pragma omp section
+      {
+        #pragma omp critical
+        {
+          sum = sum + sum0;
+        }
+      }         
+    }
+  }
+  known_sum = 11 * 3 + 7;
+  return (known_sum == sum); 
+} /* end of check_section_firstprivate*/
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for(i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_section_firstprivate()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/worksharing/sections/omp_section_lastprivate.c

+// RUN: %libomp-compile-and-run
+#include <stdio.h>
+#include "omp_testsuite.h"
+
+int test_omp_section_lastprivate()
+{
+  int i0 = -1;
+  int sum = 0;
+  int i;
+  int sum0 = 0;
+  int known_sum;
+
+  i0 = -1;
+  sum = 0;
+
+  #pragma omp parallel
+  {
+    #pragma omp sections lastprivate(i0) private(i,sum0)
+    {
+      #pragma omp section  
+      {
+        sum0 = 0;
+        for (i = 1; i < 400; i++)
+        {
+          sum0 = sum0 + i;
+          i0 = i;
+        }
+        #pragma omp critical
+        {
+          sum = sum + sum0;
+        } /*end of critical*/
+      } /* end of section */
+      #pragma omp section 
+      {
+        sum0 = 0;
+        for(i = 400; i < 700; i++)
+        {
+          sum0 = sum0 + i;
+          i0 = i;
+        }
+        #pragma omp critical
+        {
+          sum = sum + sum0;
+        } /*end of critical*/
+      }
+      #pragma omp section 
+      {
+        sum0 = 0;
+        for(i = 700; i < 1000; i++)
+        {
+          sum0 = sum0 + i;
+          i0 = i;
+        }
+        #pragma omp critical
+        {
+          sum = sum + sum0;
+        } /*end of critical*/
+      } /* end of section */
+    } /* end of sections*/
+  } /* end of parallel*/  
+  known_sum = (999 * 1000) / 2;
+  return ((known_sum == sum) && (i0 == 999) );
+}
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for(i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_section_lastprivate()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/worksharing/sections/omp_section_private.c

+// RUN: %libomp-compile-and-run
+#include <stdio.h>
+#include "omp_testsuite.h"
+
+int test_omp_section_private()
+{
+  int sum;
+  int sum0;
+  int i;
+  int known_sum;
+
+  sum = 7;
+  sum0 = 0;
+
+  #pragma omp parallel
+  {
+    #pragma omp sections private(sum0,i)
+    {
+      #pragma omp section 
+      {
+        sum0 = 0;
+        for (i = 1; i < 400; i++)
+          sum0 = sum0 + i;
+        #pragma omp critical
+        {
+          sum = sum + sum0;
+        }
+      }  
+      #pragma omp section
+      {
+        sum0 = 0;
+        for (i = 400; i < 700; i++)
+          sum0 = sum0 + i;
+        #pragma omp critical
+        {
+          sum = sum + sum0;
+        }
+      }
+      #pragma omp section
+      {
+        sum0 = 0;
+        for (i = 700; i < 1000; i++)
+          sum0 = sum0 + i;
+        #pragma omp critical
+        {
+          sum = sum + sum0;
+        }
+      }         
+    } /*end of sections*/
+  } /* end of parallel */
+  known_sum = (999 * 1000) / 2 + 7;
+  return (known_sum == sum); 
+} /* end of check_section_private*/
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for(i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_section_private()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/worksharing/sections/omp_sections_nowait.c

+// RUN: %libomp-compile-and-run
+#include <stdio.h>
+#include "omp_testsuite.h"
+#include "omp_my_sleep.h"
+
+int test_omp_sections_nowait()
+{
+  int result;
+  int count;
+  int j;
+
+  result = 0;
+  count = 0;
+
+  #pragma omp parallel 
+  {
+    int rank;
+    rank = omp_get_thread_num ();
+    #pragma omp sections nowait
+    {
+      #pragma omp section
+      {
+        fprintf(stderr, "Thread nr %d enters first section"
+          " and gets sleeping.\n", rank);
+        my_sleep(SLEEPTIME);
+        count = 1;
+        fprintf(stderr, "Thread nr %d woke up an set"
+          " count to 1.\n", rank);
+        #pragma omp flush(count)
+      }
+      #pragma omp section
+      {
+        fprintf(stderr, "Thread nr %d executed work in the"
+          " first section.\n", rank);
+      }
+    }
+    /* Begin of second sections environment */
+    #pragma omp sections
+    {
+      #pragma omp section
+      {
+        fprintf(stderr, "Thread nr %d executed work in the"
+          " second section.\n", rank);
+      }
+      #pragma omp section
+      {
+        fprintf(stderr, "Thread nr %d executed work in the"
+          " second section and controls the value of count\n", rank);
+        if (count == 0)
+          result = 1;
+        fprintf(stderr, "count was %d\n", count);
+      }
+    }
+  }
+  return result;
+}
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for(i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_sections_nowait()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/worksharing/sections/omp_sections_reduction.c

+// RUN: %libomp-compile-and-run
+#include <stdio.h>
+#include <math.h>
+#include "omp_testsuite.h"
+
+int test_omp_sections_reduction()
+{
+  int sum;
+  int known_sum;
+  double dpt,dsum;
+  double dknown_sum;
+  double dt=0.5; /* base of geometric row for + and - test*/
+  double rounding_error= 1.E-9;
+  int diff;
+  double ddiff;
+  int product;
+  int known_product;
+  int logic_and;
+  int bit_and;
+  int logic_or;
+  int bit_or;
+  int exclusiv_bit_or;
+  int logics[1000];
+  int i;
+  int result;
+  /* int my_islarger; */
+  /*int is_larger=1;*/
+  sum =7;
+  dpt =1;
+  dsum=0;
+  product =1;
+  logic_and=1;
+  bit_and=1;
+  logic_or=0;
+  bit_or=0;
+  exclusiv_bit_or=0;
+  result = 0;
+  dt = 1./3.;
+
+  known_sum = (999*1000)/2+7;
+  #pragma omp parallel
+  {
+    #pragma omp sections private(i) reduction(+:sum)
+    {
+      #pragma omp section
+      {
+        for (i=1;i<300;i++) {
+          sum=sum+i;
+        }
+      }
+      #pragma omp section
+      {
+        for (i=300;i<700;i++) {
+          sum=sum+i;
+        }
+      }
+      #pragma omp section
+      {
+        for (i=700;i<1000;i++) {
+          sum=sum+i;
+        }
+      }
+    }
+  }
+  if(known_sum!=sum) {
+    ++result;
+    fprintf(stderr,"Error in sum with integers: Result was %d"
+      " instead of %d\n", sum,known_sum);
+  }
+
+  diff = (999*1000)/2;
+  #pragma omp parallel
+  {
+    #pragma omp sections private(i) reduction(-:diff)
+    {
+      #pragma omp section
+      {
+        for (i=1;i<300;i++) {
+          diff=diff-i;
+        }
+      }
+      #pragma omp section
+      {
+        for (i=300;i<700;i++) {
+          diff=diff-i;
+        }
+      }
+      #pragma omp section
+      {
+        for (i=700;i<1000;i++) {
+          diff=diff-i;
+        }
+      }
+    }
+  }
+  if(diff != 0) {
+    result++;
+    fprintf(stderr,"Error in Difference with integers: Result was %d"
+      " instead of 0.\n",diff);
+  }
+
+  for (i=0;i<20;++i) {
+    dpt*=dt;
+  }
+  dknown_sum = (1-dpt)/(1-dt);
+  #pragma omp parallel
+  {
+    #pragma omp sections private(i) reduction(+:dsum)
+    {
+      #pragma omp section
+      {
+        for (i=0;i<6;++i) {
+          dsum += pow(dt,i);
+        }
+      }
+      #pragma omp section
+      {
+        for (i=6;i<12;++i) {
+          dsum += pow(dt,i);
+        }
+      }
+      #pragma omp section
+      {
+        for (i=12;i<20;++i) {
+          dsum += pow(dt,i);
+        }
+      }
+    }
+  }
+  if( fabs(dsum-dknown_sum) > rounding_error ) {
+    result++; 
+    fprintf(stderr,"Error in sum with doubles: Result was %f"
+      " instead of %f (Difference: %E)\n",
+      dsum, dknown_sum, dsum-dknown_sum);
+  }
+
+  dpt=1;
+  for (i=0;i<20;++i) {
+    dpt*=dt;
+  }
+  fprintf(stderr,"\n");
+  ddiff = (1-dpt)/(1-dt);
+  #pragma omp parallel
+  {
+    #pragma omp sections private(i) reduction(-:ddiff)
+    {
+      #pragma omp section
+      {
+        for (i=0;i<6;++i) {
+          ddiff -= pow(dt,i);
+        }
+      }
+      #pragma omp section
+      {
+        for (i=6;i<12;++i) {
+          ddiff -= pow(dt,i);
+        }
+      }
+      #pragma omp section
+      {
+        for (i=12;i<20;++i) {
+          ddiff -= pow(dt,i);
+        }
+      }
+    }
+  }
+
+  if(fabs(ddiff) > rounding_error) {
+    result++;
+    fprintf(stderr,"Error in Difference with doubles: Result was %E"
+      " instead of 0.0\n",ddiff);
+  }
+
+  known_product = 3628800;
+  #pragma omp parallel
+  {
+    #pragma omp sections private(i) reduction(*:product)
+    {
+      #pragma omp section
+      {  
+        for(i=1;i<3;i++) {
+          product *= i;
+        }
+      }
+      #pragma omp section
+      {
+        for(i=3;i<7;i++) {
+          product *= i;
+        }
+      }
+      #pragma omp section
+      {
+        for(i=7;i<11;i++) {
+          product *= i;
+        }
+      }
+    }
+  }
+  if(known_product != product) {
+    result++;
+    fprintf(stderr,"Error in Product with integers: Result was %d"
+      " instead of %d\n",product,known_product);
+  }
+
+  for(i=0;i<1000;i++) {
+    logics[i]=1;
+  }
+
+  #pragma omp parallel
+  {
+    #pragma omp sections private(i) reduction(&&:logic_and)
+    {
+      #pragma omp section
+      {
+        for (i=1;i<300;i++) {
+          logic_and = (logic_and && logics[i]);
+        }
+      }
+      #pragma omp section
+      {
+        for (i=300;i<700;i++) {
+          logic_and = (logic_and && logics[i]);
+        }
+      }
+      #pragma omp section
+      {
+        for (i=700;i<1000;i++) {
+          logic_and = (logic_and && logics[i]);
+        }
+      }
+    }
+  }
+  if(!logic_and) {
+    result++;
+    fprintf(stderr,"Error in logic AND part 1\n");
+  }
+
+  logic_and = 1;
+  logics[501] = 0;
+
+  #pragma omp parallel
+  {
+    #pragma omp sections private(i) reduction(&&:logic_and)
+    {
+      #pragma omp section
+      {
+        for (i=1;i<300;i++) {
+          logic_and = (logic_and && logics[i]);
+        }
+      }
+      #pragma omp section
+      {
+        for (i=300;i<700;i++) {
+          logic_and = (logic_and && logics[i]);
+        }
+      }
+      #pragma omp section
+      {
+        for (i=700;i<1000;i++) {
+          logic_and = (logic_and && logics[i]);
+        }
+      }
+    }
+  }
+  if(logic_and) {
+    result++;
+    fprintf(stderr,"Error in logic AND part 2\n");
+  }
+
+  for(i=0;i<1000;i++) {
+    logics[i]=0;
+  }
+
+  #pragma omp parallel 
+  {
+    #pragma omp sections private(i) reduction(||:logic_or)
+    {
+      #pragma omp section
+      {
+        for (i=1;i<300;i++) {
+          logic_or = (logic_or || logics[i]);
+        }
+      }
+      #pragma omp section
+      {
+        for (i=300;i<700;i++) {
+          logic_or = (logic_or || logics[i]);
+        }
+      }
+      #pragma omp section
+      {
+        for (i=700;i<1000;i++) {
+          logic_or = (logic_or || logics[i]);
+        }
+      }
+    }
+  }
+  if(logic_or) {
+    result++;
+    fprintf(stderr,"\nError in logic OR part 1\n");
+  }
+
+  logic_or = 0;
+  logics[501]=1;
+
+  #pragma omp parallel 
+  {
+    #pragma omp sections private(i) reduction(||:logic_or)
+    {
+      #pragma omp section
+      {
+        for (i=1;i<300;i++) {
+          logic_or = (logic_or || logics[i]);
+        }
+      }
+      #pragma omp section
+      {
+        for (i=300;i<700;i++) {
+          logic_or = (logic_or || logics[i]);
+        }
+      }
+      #pragma omp section
+      {
+        for (i=700;i<1000;i++) {
+          logic_or = (logic_or || logics[i]);
+        }
+      }
+    }
+  }
+  if(!logic_or) {
+    result++;
+    fprintf(stderr,"Error in logic OR part 2\n");
+  }
+
+  for(i=0;i<1000;++i) {
+    logics[i]=1;
+  }
+
+  #pragma omp parallel 
+  {
+    #pragma omp sections private(i) reduction(&:bit_and)
+    {
+      #pragma omp section
+      {  
+        for(i=0;i<300;++i) {
+          bit_and = (bit_and & logics[i]);
+        }
+      }
+      #pragma omp section
+      {  
+        for(i=300;i<700;++i) {
+          bit_and = (bit_and & logics[i]);
+        }
+      }
+      #pragma omp section
+      {  
+        for(i=700;i<1000;++i) {
+          bit_and = (bit_and & logics[i]);
+        }
+      }
+    }
+  }
+  if(!bit_and) {
+    result++;
+    fprintf(stderr,"Error in BIT AND part 1\n");
+  }
+
+  bit_and = 1;
+  logics[501]=0;
+
+  #pragma omp parallel 
+  {
+    #pragma omp sections private(i) reduction(&:bit_and)
+    {
+      #pragma omp section
+      {
+        for(i=0;i<300;++i) {
+          bit_and = bit_and & logics[i];
+        }
+      }
+      #pragma omp section
+      {
+        for(i=300;i<700;++i) {
+          bit_and = bit_and & logics[i];
+        }
+      }
+      #pragma omp section
+      {
+        for(i=700;i<1000;++i) {
+          bit_and = bit_and & logics[i];
+        }
+      }
+    }
+  }
+  if(bit_and) {
+    result++;
+    fprintf(stderr,"Error in BIT AND part 2\n");
+  }
+
+  for(i=0;i<1000;i++) {
+    logics[i]=0;
+  }
+
+  #pragma omp parallel 
+  {
+    #pragma omp sections private(i) reduction(|:bit_or)
+    {
+      #pragma omp section
+      {
+        for(i=0;i<300;++i) {
+          bit_or = bit_or | logics[i];
+        }
+      }
+      #pragma omp section
+      {
+        for(i=300;i<700;++i) {
+          bit_or = bit_or | logics[i];
+        }
+      }
+      #pragma omp section
+      {
+        for(i=700;i<1000;++i) {
+          bit_or = bit_or | logics[i];
+        }
+      }
+    }
+  }
+  if(bit_or) {
+    result++;
+    fprintf(stderr,"Error in BIT OR part 1\n");
+  }
+  bit_or = 0;
+  logics[501]=1;
+
+  #pragma omp parallel 
+  {
+    #pragma omp sections private(i) reduction(|:bit_or)
+    {
+      #pragma omp section
+      {
+        for(i=0;i<300;++i) {
+          bit_or = bit_or | logics[i];
+        }
+      }
+      #pragma omp section
+      {
+        for(i=300;i<700;++i) {
+          bit_or = bit_or | logics[i];
+        }
+      }
+      #pragma omp section
+      {
+        for(i=700;i<1000;++i) {
+          bit_or = bit_or | logics[i];
+        }
+      }
+    }
+  }
+  if(!bit_or) {
+    result++;
+    fprintf(stderr,"Error in BIT OR part 2\n");
+  }
+
+  for(i=0;i<1000;i++) {
+    logics[i]=0;
+  }
+
+  #pragma omp parallel 
+  {
+    #pragma omp sections private(i) reduction(^:exclusiv_bit_or)
+    {
+      #pragma omp section
+      {  
+        for(i=0;i<300;++i) {
+          exclusiv_bit_or = exclusiv_bit_or ^ logics[i];
+        }
+      }
+      #pragma omp section
+      {  
+        for(i=300;i<700;++i) {
+          exclusiv_bit_or = exclusiv_bit_or ^ logics[i];
+        }
+      }
+      #pragma omp section
+      {  
+        for(i=700;i<1000;++i) {
+          exclusiv_bit_or = exclusiv_bit_or ^ logics[i];
+        }
+      }
+    }
+  }
+  if(exclusiv_bit_or) {
+    result++;
+    fprintf(stderr,"Error in EXCLUSIV BIT OR part 1\n");
+  }
+
+  exclusiv_bit_or = 0;
+  logics[501]=1;
+
+  #pragma omp parallel 
+  {
+    #pragma omp sections private(i) reduction(^:exclusiv_bit_or)
+    {
+      #pragma omp section
+      {
+        for(i=0;i<300;++i) {
+          exclusiv_bit_or = exclusiv_bit_or ^ logics[i];
+        }
+      }
+      #pragma omp section
+      {
+        for(i=300;i<700;++i) {
+          exclusiv_bit_or = exclusiv_bit_or ^ logics[i];
+        }
+      }
+      #pragma omp section
+      {
+        for(i=700;i<1000;++i) {
+          exclusiv_bit_or = exclusiv_bit_or ^ logics[i];
+        }
+      }
+    }
+  }
+  if(!exclusiv_bit_or) {
+    result++;
+    fprintf(stderr,"Error in EXCLUSIV BIT OR part 2\n");
+  }
+
+  /*printf("\nResult:%d\n",result);*/
+  return (result==0);
+}
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for(i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_sections_reduction()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/worksharing/single/omp_single.c

+// RUN: %libomp-compile-and-run
+#include <stdio.h>
+#include "omp_testsuite.h"
+
+int test_omp_single()
+{
+  int nr_threads_in_single;
+  int result;
+  int nr_iterations;
+  int i;
+
+  nr_threads_in_single = 0;
+  result = 0;
+  nr_iterations = 0;
+
+  #pragma omp parallel private(i)
+  {
+    for (i = 0; i < LOOPCOUNT; i++) {
+      #pragma omp single 
+      {  
+        #pragma omp flush
+        nr_threads_in_single++;
+        #pragma omp flush             
+        nr_iterations++;
+        nr_threads_in_single--;
+        result = result + nr_threads_in_single;
+      }
+    }
+  }
+  return ((result == 0) && (nr_iterations == LOOPCOUNT));
+} /* end of check_single*/
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for(i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_single()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/worksharing/single/omp_single_copyprivate.c

+// RUN: %libomp-compile-and-run
+#include "omp_testsuite.h"
+
+#define DEBUG_TEST 0
+
+int j;
+#pragma omp threadprivate(j)
+
+int test_omp_single_copyprivate()
+{
+  int result;
+  int nr_iterations;
+
+  result = 0;
+  nr_iterations = 0;
+  #pragma omp parallel
+  {
+    int i;
+    for (i = 0; i < LOOPCOUNT; i++)
+    {
+#if DEBUG_TEST
+         int thread;
+         thread = omp_get_thread_num ();
+#endif
+      #pragma omp single copyprivate(j)
+      {
+        nr_iterations++;
+        j = i;
+#if DEBUG_TEST
+        printf ("thread %d assigns, j = %d, i = %d\n", thread, j, i);
+#endif
+      }
+#if DEBUG_TEST
+      #pragma omp barrier
+#endif
+      #pragma omp critical
+      {
+#if DEBUG_TEST
+        printf ("thread = %d, j = %d, i = %d\n", thread, j, i);
+#endif
+        result = result + j - i;
+      }
+      #pragma omp barrier
+    } /* end of for */
+  } /* end of parallel */
+  return ((result == 0) && (nr_iterations == LOOPCOUNT));
+}
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for(i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_single_copyprivate()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/worksharing/single/omp_single_nowait.c

+// RUN: %libomp-compile-and-run
+#include <stdio.h>
+#include "omp_testsuite.h"
+
+int my_iterations;
+#pragma omp threadprivate(my_iterations)
+
+int test_omp_single_nowait()
+{
+  int nr_iterations;
+  int total_iterations = 0;
+  int i;
+
+  nr_iterations = 0;
+  my_iterations = 0;
+
+  #pragma omp parallel private(i)
+  {
+    for (i = 0; i < LOOPCOUNT; i++) {
+      #pragma omp single nowait
+      {
+        #pragma omp atomic  
+        nr_iterations++;
+      }
+    }
+  }
+
+  #pragma omp parallel private(i) 
+  {
+    my_iterations = 0;
+    for (i = 0; i < LOOPCOUNT; i++) {
+      #pragma omp single nowait
+      {
+        my_iterations++;
+      }
+    }
+    #pragma omp critical
+    {
+      total_iterations += my_iterations;
+    }
+
+  }
+  return ((nr_iterations == LOOPCOUNT) && (total_iterations == LOOPCOUNT));
+} /* end of check_single_nowait*/
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for(i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_single_nowait()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}

runtime/test/worksharing/single/omp_single_private.c

+// RUN: %libomp-compile-and-run
+#include <stdio.h>
+#include "omp_testsuite.h"
+
+int myit = 0;
+#pragma omp threadprivate(myit)
+int myresult = 0;
+#pragma omp threadprivate(myresult)
+
+int test_omp_single_private()
+{
+  int nr_threads_in_single;
+  int result;
+  int nr_iterations;
+  int i;
+
+  myit = 0;
+  nr_threads_in_single = 0;
+  nr_iterations = 0;
+  result = 0;
+
+  #pragma omp parallel private(i)
+  {
+    myresult = 0;
+    myit = 0;
+    for (i = 0; i < LOOPCOUNT; i++) {
+      #pragma omp single private(nr_threads_in_single) nowait
+      {  
+        nr_threads_in_single = 0;
+        #pragma omp flush
+        nr_threads_in_single++;
+        #pragma omp flush             
+        myit++;
+        myresult = myresult + nr_threads_in_single;
+      }
+    }
+    #pragma omp critical
+    {
+      result += nr_threads_in_single;
+      nr_iterations += myit;
+    }
+  }
+  return ((result == 0) && (nr_iterations == LOOPCOUNT));
+} /* end of check_single private */ 
+
+int main()
+{
+  int i;
+  int num_failed=0;
+
+  for(i = 0; i < REPETITIONS; i++) {
+    if(!test_omp_single_private()) {
+      num_failed++;
+    }
+  }
+  return num_failed;
+}