Squashed 'thirdParty/cuda_memtest/' changes from 3091fa6..0e27280

0e27280 Merge pull request #6 from Flamefire/dev
8c8de37 Add space
17550d1 Merge pull request #2 from ComputationalRadiationPhysics/topic-codeCleanup
278e29c Remove Unused Parameter arg
162fa90 Fix (un)singed compares and return types
e843493 CMake: Add Correct Warning and Fix missing -pthread
06419c2 Merge pull request #3 from ComputationalRadiationPhysics/topic-travis
2997e4f Merge pull request #4 from ComputationalRadiationPhysics/fix-headerGuard
e524a78 Add .travis.yml file for CI
c821814 Fix Broken __MEMTEST_H__ Header Guard

git-subtree-dir: thirdParty/cuda_memtest
git-subtree-split: 0e27280cacfd77aec1fce6a7f82236239c02e5d1

.travis.yml

@@ -0,0 +1,40 @@
+language: cpp
+
+compiler:
+  - gcc
+  - clang
+
+env:
+  global:
+    - INSTALL_DIR=~/mylibs
+    - NVML_FILE=cuda_346.46_gdk_linux.run
+    - NVML_LINK=http://developer.download.nvidia.com/compute/cuda/7_0/Prod/local_installers/
+  matrix:
+    - USE_NVML=1
+    - USE_NVML=0
+
+script:
+  - mkdir build_tmp && cd build_tmp
+# CUDA 4.0 on travis... work-around missing atomicAdd
+  - cmake -DCMAKE_INSTALL_PREFIX=$INSTALL_DIR -DCUDA_ARCH=sm_10 -DSAME_NVCC_FLAGS_IN_SUBPROJECTS=ON $TRAVIS_BUILD_DIR
+  - make
+  - make install
+
+before_script:
+  - sudo apt-get update -qq
+  - sudo apt-get install -qq build-essential
+  - sudo apt-get install -qq gcc-4.4 g++-4.4
+  - sudo update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-4.4 60 --slave /usr/bin/g++ g++ /usr/bin/g++-4.4
+  - gcc --version && g++ --version
+  - sudo apt-get install -qq nvidia-common
+  - sudo apt-get install -qq nvidia-current
+  - sudo apt-get install -qq nvidia-cuda-toolkit nvidia-cuda-dev
+  - if [ $USE_NVML -eq 1 ]; then wget $NVML_LINK$NVML_FILE && chmod u+x $NVML_FILE && sudo ./$NVML_FILE --silent --installdir=/ ; fi
+  - if [ $USE_NVML -eq 1 ]; then export CMAKE_PREFIX_PATH=/usr/src/gdk/nvml/lib/ ; fi
+  - sudo find /usr/ -name libcuda*.so
+  - sudo find /usr/ -name nvml.h
+  - sudo find /usr/ -name libnvidia*.so
+  - sudo find /usr/ -name libnvml*.so
+
+after_script:
+  - ls -halR $INSTALL_DIR

CMakeLists.txt

@@ -31,6 +31,9 @@ find_package(CUDA REQUIRED)
 if(SAME_NVCC_FLAGS_IN_SUBPROJECTS)
   set(CUDA_ARCH sm_13 CACHE STRING "set GPU architecture" )
   set(CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS} ${nvcc_flags} -arch=${CUDA_ARCH})
+  if(CUDA_ARCH STREQUAL "sm_10")
+    add_definitions(-DSM_10=1)
+  endif()
 
   if(CUDA_SHOW_CODELINES)
     set(CUDA_NVCC_FLAGS "${CUDA_NVCC_FLAGS}" --source-in-ptx -Xcompiler -rdynamic -lineinfo)
@@ -69,34 +72,34 @@ endif()
 ################################################################################
 
 find_package(Threads REQUIRED)
+list(APPEND LIBS ${CMAKE_THREAD_LIBS_INIT})
+
+# work around for Hypnos (Ubuntu 14.04) with gcc 4.8.2
+if(NOT CMAKE_THREAD_LIBS_INIT)
+    list(APPEND LIBS "-pthread")
+endif()
 
 
 ################################################################################
 # Build type (debug, release)
 ################################################################################
 
 option(CUDA_MEMTEST_RELEASE "disable all runtime asserts" ON)
-if(PIC_RELEASE)
+set(CUDA_NVCC_FLAGS "${CUDA_NVCC_FLAGS} -Xcompiler=-pthread")
+if(CUDA_MEMTEST_RELEASE)
     add_definitions(-DNDEBUG)
-    set(CUDA_NVCC_FLAGS "${CUDA_NVCC_FLAGS}" "-Xcompiler=-pthread")
-else(PIC_RELEASE)
+else(CUDA_MEMTEST_RELEASE)
     set(CMAKE_BUILD_TYPE Debug)
-    set(CUDA_NVCC_FLAGS "${CUDA_NVCC_FLAGS}" -g "-Xcompiler=-g,-pthread")
-endif(PIC_RELEASE)
+    set(CUDA_NVCC_FLAGS "${CUDA_NVCC_FLAGS} -Xcompiler=-g")
+    set(CUDA_NVCC_FLAGS "${CUDA_NVCC_FLAGS} -g")
+endif(CUDA_MEMTEST_RELEASE)
 
 
 ################################################################################
 # Warnings
 ################################################################################
 
-set(CMAKE_CXX_FLAGS_DEFAULT "-Wall")
-
-
-################################################################################
-# Configure include directories
-################################################################################
-
-set(CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS} ${INCLUDE_DIRECTORIES})
+set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wall")
 
 
 ################################################################################
@@ -109,12 +112,12 @@ cuda_add_executable(cuda_memtest
     cuda_memtest.cu
 )
 
-target_link_libraries(cuda_memtest ${LIBS} ${CUDA_CUDART_LIBRARY} ${CMAKE_THREAD_LIBS_INIT})
+target_link_libraries(cuda_memtest ${LIBS} ${CUDA_CUDART_LIBRARY})
 
 
 ################################################################################
 # Install cuda_memtest
 ################################################################################
 
 install(TARGETS cuda_memtest
-         RUNTIME DESTINATION bin)
+        RUNTIME DESTINATION bin)

cuda_memtest.cu

@@ -76,6 +76,7 @@ pthread_mutex_t atomic_mutex = PTHREAD_MUTEX_INITIALIZER;
 void run_tests(char*, unsigned int);
 extern void update_temperature(void);
 extern void allocate_small_mem(void);
+volatile int active_update_temperature;
 
 typedef struct arg_s{
     unsigned int device;
@@ -186,19 +187,19 @@ thread_func(void* _arg)
 
 
 void*
-temp_monitor_thread_func(void* arg)
+temp_monitor_thread_func(void*)
 {
     do{
-	update_temperature();
-	sleep(monitor_interval);
-    }while(1);
-
+        update_temperature();
+        sleep(monitor_interval);
+    }while(active_update_temperature);
+    return NULL;
 }
 
 
 void list_tests_info(void)
 {
-    int i;
+    size_t i;
     for (i = 0;i < DIM(cuda_memtests); i++){
 	printf("%s %s\n", cuda_memtests[i].desc, cuda_memtests[i].enabled?"":" ==disabled by default==");
     }
@@ -319,7 +320,7 @@ main(int argc, char** argv)
 	    if (i+1 >= argc){
 		usage(argv);
 	    }
-	    int idx = atoi(argv[i+1]);
+	    size_t idx = atoi(argv[i+1]);
 	    if (idx >= DIM(cuda_memtests)){
 		fprintf(stderr, "Error: invalid test id\n");
 		usage(argv);
@@ -334,7 +335,7 @@ main(int argc, char** argv)
 	    if (i+1 >= argc){
 		usage(argv);
 	    }
-	    int idx = atoi(argv[i+1]);
+	    size_t idx = atoi(argv[i+1]);
 	    if (idx >= DIM(cuda_memtests)){
 		fprintf(stderr, "Error: invalid test id\n");
 		usage(argv);
@@ -345,7 +346,7 @@ main(int argc, char** argv)
 	    continue;
 	}
 	if (strcmp(argv[i], "--disable_all") == 0){
-	    int k;
+	    size_t k;
 	    for (k=0;k < DIM(cuda_memtests);k++){
 		cuda_memtests[k].enabled = 0;
 	    }
@@ -475,7 +476,7 @@ main(int argc, char** argv)
 
 	if (strcmp(argv[i], "--stress") == 0){
 	    //equal to "--disable_all --enable_test 10 --exit_on_error"
-	    int k;
+	    size_t k;
 	    for (k=0;k < DIM(cuda_memtests);k++){
 		cuda_memtests[k].enabled = 0;
 	    }
@@ -500,6 +501,7 @@ main(int argc, char** argv)
     }
     pthread_t temp_pid;
     if (monitor_temp){
+	active_update_temperature = 1;
 	if (pthread_create(&temp_pid, NULL, temp_monitor_thread_func, NULL)  != 0){
 	    printf("ERROR: creating thread for temperature monitoring failed\n");
 	    exit(ERR_GENERAL);
@@ -552,6 +554,7 @@ main(int argc, char** argv)
 	exit(ERR_BAD_STATE);
     }
 
+    active_update_temperature = 0;
 
     for(i=0;i < num_gpus;i++){
 	pthread_join(pid[i], NULL);

cuda_memtest.h

@@ -39,7 +39,7 @@
  */
 
 #ifndef __MEMTEST_H__
-#define __MEMTEST_H_
+#define __MEMTEST_H__
 
 #include <stdio.h>
 #include <stdint.h>
@@ -80,12 +80,12 @@ extern void get_driver_info(char* info, unsigned int len);
 #define PRINTF(fmt,...) do{						\
 	if (monitor_temp){						\
 	    pthread_mutex_lock(&mutex);					\
-	    printf("[%s][%s][%d][%d C]:"fmt, time_string(), hostname, gpu_idx, gpu_temp[gpu_idx],##__VA_ARGS__); \
+	    printf("[%s][%s][%d][%d C]:" fmt, time_string(), hostname, gpu_idx, gpu_temp[gpu_idx],##__VA_ARGS__); \
 	    pthread_mutex_unlock(&mutex);				\
 	}								\
 	else{								\
 	    pthread_mutex_lock(&mutex);					\
-	    printf("[%s][%s][%d]:"fmt, time_string(), hostname, gpu_idx, ##__VA_ARGS__); \
+	    printf("[%s][%s][%d]:" fmt, time_string(), hostname, gpu_idx, ##__VA_ARGS__); \
 	    pthread_mutex_unlock(&mutex);					\
 	}								\
 	fflush(stdout);							\
@@ -94,10 +94,10 @@ extern void get_driver_info(char* info, unsigned int len);
 
 #define FPRINTF(fmt,...) do{						\
 	if (monitor_temp){					\
-	    fprintf(stderr, "[%s][%s][%d][%d C]:"fmt, time_string(), hostname, gpu_idx, gpu_temp[gpu_idx],##__VA_ARGS__); \
+	    fprintf(stderr, "[%s][%s][%d][%d C]:" fmt, time_string(), hostname, gpu_idx, gpu_temp[gpu_idx],##__VA_ARGS__); \
 	}								\
 	else{								\
-	    fprintf(stderr, "[%s][%s][%d]:"fmt, time_string(), hostname, gpu_idx, ##__VA_ARGS__); \
+	    fprintf(stderr, "[%s][%s][%d]:" fmt, time_string(), hostname, gpu_idx, ##__VA_ARGS__); \
 	}								\
 	fflush(stderr);							\
     } while(0)

tests.cu

@@ -588,7 +588,7 @@ test0(char* ptr, unsigned int tot_num_blocks)
     kernel_test0_global_read<<<1, 1>>>(ptr, end_ptr, err_count, err_addr, err_expect, err_current, err_second_read); SYNC_CUERR;
     error_checking("test0 on global address",  0);
 
-    for(int ite = 0;ite < num_iterations; ite++){
+    for(unsigned int ite = 0;ite < num_iterations; ite++){
 	for (i=0;i < tot_num_blocks; i+= GRIDSIZE){
 	    dim3 grid;
 	    grid.x= GRIDSIZE;