| Message ID | 1441710324-11148-1-git-send-email-maxim.uvarov@linaro.org |
|---|---|
| State | New |
| Headers | show |
On 09/08/2015 01:05 PM, Maxim Uvarov wrote: > ODP has api to request available number of workers. Now > no need limit that inside application. > > Signed-off-by: Maxim Uvarov <maxim.uvarov@linaro.org> > --- > > v2: timer.c: do not substract -1 from number of workers. > > test/validation/common/odp_cunit_common.c | 7 ++-- > test/validation/common/odp_cunit_common.h | 2 -- > test/validation/scheduler/scheduler.c | 3 -- > test/validation/shmem/shmem.c | 6 ++-- > test/validation/synchronizers/synchronizers.c | 48 ++++++++++----------------- > test/validation/timer/timer.c | 10 ++---- > 6 files changed, 27 insertions(+), 49 deletions(-) > > diff --git a/test/validation/common/odp_cunit_common.c b/test/validation/common/odp_cunit_common.c > index d995ad3..01ea87b 100644 > --- a/test/validation/common/odp_cunit_common.c > +++ b/test/validation/common/odp_cunit_common.c > @@ -9,7 +9,7 @@ > #include <odp_cunit_common.h> > #include <odp/helper/linux.h> > /* Globals */ > -static odph_linux_pthread_t thread_tbl[MAX_WORKERS]; > +static odph_linux_pthread_t *thread_tbl; > > /* > * global init/term functions which may be registered > @@ -26,9 +26,11 @@ static struct { > int odp_cunit_thread_create(void *func_ptr(void *), pthrd_arg *arg) > { > odp_cpumask_t cpumask; > + int num; > > /* Create and init additional threads */ > - odp_cpumask_def_worker(&cpumask, arg->numthrds); > + num = odp_cpumask_def_worker(&cpumask, arg->numthrds); > + thread_tbl = calloc(sizeof(odph_linux_pthread_t), num); Maybe add a check of the return value here? Everything else looks good Reviewed-by: Nicolas Morey-Chaisemartin <nmorey@kalray.eu> > > return odph_linux_pthread_create(thread_tbl, &cpumask, func_ptr, > (void *)arg); > @@ -39,6 +41,7 @@ int odp_cunit_thread_exit(pthrd_arg *arg) > { > /* Wait for other threads to exit */ > odph_linux_pthread_join(thread_tbl, arg->numthrds); > + free(thread_tbl); > > return 0; > } > diff --git a/test/validation/common/odp_cunit_common.h b/test/validation/common/odp_cunit_common.h > index 6cafaaa..f94b44e 100644 > --- a/test/validation/common/odp_cunit_common.h > +++ b/test/validation/common/odp_cunit_common.h > @@ -16,8 +16,6 @@ > #include <stdint.h> > #include "CUnit/Basic.h" > > -#define MAX_WORKERS 32 /**< Maximum number of work threads */ > - > /* the function, called by module main(), to run the testsuites: */ > int odp_cunit_run(CU_SuiteInfo testsuites[]); > > diff --git a/test/validation/scheduler/scheduler.c b/test/validation/scheduler/scheduler.c > index 1874889..e12895d 100644 > --- a/test/validation/scheduler/scheduler.c > +++ b/test/validation/scheduler/scheduler.c > @@ -8,7 +8,6 @@ > #include "odp_cunit_common.h" > #include "scheduler.h" > > -#define MAX_WORKERS_THREADS 32 > #define MSG_POOL_SIZE (4 * 1024 * 1024) > #define QUEUES_PER_PRIO 16 > #define BUF_SIZE 64 > @@ -1018,8 +1017,6 @@ int scheduler_suite_init(void) > memset(globals, 0, sizeof(test_globals_t)); > > globals->num_workers = odp_cpumask_def_worker(&mask, 0); > - if (globals->num_workers > MAX_WORKERS) > - globals->num_workers = MAX_WORKERS; > > shm = odp_shm_reserve(SHM_THR_ARGS_NAME, sizeof(thread_args_t), > ODP_CACHE_LINE_SIZE, 0); > diff --git a/test/validation/shmem/shmem.c b/test/validation/shmem/shmem.c > index 6dc579a..dfa5310 100644 > --- a/test/validation/shmem/shmem.c > +++ b/test/validation/shmem/shmem.c > @@ -49,6 +49,7 @@ void shmem_test_odp_shm_sunnyday(void) > pthrd_arg thrdarg; > odp_shm_t shm; > test_shared_data_t *test_shared_data; > + odp_cpumask_t mask; > > shm = odp_shm_reserve(TESTNAME, > sizeof(test_shared_data_t), ALIGE_SIZE, 0); > @@ -67,10 +68,7 @@ void shmem_test_odp_shm_sunnyday(void) > test_shared_data->foo = TEST_SHARE_FOO; > test_shared_data->bar = TEST_SHARE_BAR; > > - thrdarg.numthrds = odp_cpu_count(); > - > - if (thrdarg.numthrds > MAX_WORKERS) > - thrdarg.numthrds = MAX_WORKERS; > + thrdarg.numthrds = odp_cpumask_def_worker(&mask, 0); > > odp_cunit_thread_create(run_shm_thread, &thrdarg); > odp_cunit_thread_exit(&thrdarg); > diff --git a/test/validation/synchronizers/synchronizers.c b/test/validation/synchronizers/synchronizers.c > index 0a31a40..914b37e 100644 > --- a/test/validation/synchronizers/synchronizers.c > +++ b/test/validation/synchronizers/synchronizers.c > @@ -45,7 +45,7 @@ typedef struct { > > typedef struct { > /* Global variables */ > - uint32_t g_num_threads; > + uint32_t g_num_workers; > uint32_t g_iterations; > uint32_t g_verbose; > uint32_t g_max_num_cores; > @@ -169,7 +169,7 @@ static uint32_t barrier_test(per_thread_mem_t *per_thread_mem, > > thread_num = odp_thread_id(); > global_mem = per_thread_mem->global_mem; > - num_threads = global_mem->g_num_threads; > + num_threads = global_mem->g_num_workers; > iterations = BARRIER_ITERATIONS; > > barrier_errs = 0; > @@ -710,7 +710,7 @@ static void barrier_test_init(void) > { > uint32_t num_threads, idx; > > - num_threads = global_mem->g_num_threads; > + num_threads = global_mem->g_num_workers; > > for (idx = 0; idx < NUM_TEST_BARRIERS; idx++) { > odp_barrier_init(&global_mem->test_barriers[idx], num_threads); > @@ -924,7 +924,7 @@ void synchronizers_test_no_barrier_functional(void) > { > pthrd_arg arg; > > - arg.numthrds = global_mem->g_num_threads; > + arg.numthrds = global_mem->g_num_workers; > barrier_test_init(); > odp_cunit_thread_create(no_barrier_functional_test, &arg); > odp_cunit_thread_exit(&arg); > @@ -934,7 +934,7 @@ void synchronizers_test_barrier_functional(void) > { > pthrd_arg arg; > > - arg.numthrds = global_mem->g_num_threads; > + arg.numthrds = global_mem->g_num_workers; > barrier_test_init(); > odp_cunit_thread_create(barrier_functional_test, &arg); > odp_cunit_thread_exit(&arg); > @@ -951,7 +951,7 @@ void synchronizers_test_no_lock_functional(void) > { > pthrd_arg arg; > > - arg.numthrds = global_mem->g_num_threads; > + arg.numthrds = global_mem->g_num_workers; > odp_cunit_thread_create(no_lock_functional_test, &arg); > odp_cunit_thread_exit(&arg); > } > @@ -966,7 +966,7 @@ void synchronizers_test_spinlock_api(void) > { > pthrd_arg arg; > > - arg.numthrds = global_mem->g_num_threads; > + arg.numthrds = global_mem->g_num_workers; > odp_cunit_thread_create(spinlock_api_tests, &arg); > odp_cunit_thread_exit(&arg); > } > @@ -975,7 +975,7 @@ void synchronizers_test_spinlock_functional(void) > { > pthrd_arg arg; > > - arg.numthrds = global_mem->g_num_threads; > + arg.numthrds = global_mem->g_num_workers; > odp_spinlock_init(&global_mem->global_spinlock); > odp_cunit_thread_create(spinlock_functional_test, &arg); > odp_cunit_thread_exit(&arg); > @@ -992,7 +992,7 @@ void synchronizers_test_ticketlock_api(void) > { > pthrd_arg arg; > > - arg.numthrds = global_mem->g_num_threads; > + arg.numthrds = global_mem->g_num_workers; > odp_cunit_thread_create(ticketlock_api_tests, &arg); > odp_cunit_thread_exit(&arg); > } > @@ -1001,7 +1001,7 @@ void synchronizers_test_ticketlock_functional(void) > { > pthrd_arg arg; > > - arg.numthrds = global_mem->g_num_threads; > + arg.numthrds = global_mem->g_num_workers; > odp_ticketlock_init(&global_mem->global_ticketlock); > > odp_cunit_thread_create(ticketlock_functional_test, &arg); > @@ -1019,7 +1019,7 @@ void synchronizers_test_rwlock_api(void) > { > pthrd_arg arg; > > - arg.numthrds = global_mem->g_num_threads; > + arg.numthrds = global_mem->g_num_workers; > odp_cunit_thread_create(rwlock_api_tests, &arg); > odp_cunit_thread_exit(&arg); > } > @@ -1028,7 +1028,7 @@ void synchronizers_test_rwlock_functional(void) > { > pthrd_arg arg; > > - arg.numthrds = global_mem->g_num_threads; > + arg.numthrds = global_mem->g_num_workers; > odp_rwlock_init(&global_mem->global_rwlock); > odp_cunit_thread_create(rwlock_functional_test, &arg); > odp_cunit_thread_exit(&arg); > @@ -1044,7 +1044,7 @@ int synchronizers_suite_init(void) > { > uint32_t num_threads, idx; > > - num_threads = global_mem->g_num_threads; > + num_threads = global_mem->g_num_workers; > odp_barrier_init(&global_mem->global_barrier, num_threads); > for (idx = 0; idx < NUM_RESYNC_BARRIERS; idx++) > odp_barrier_init(&global_mem->barrier_array[idx], num_threads); > @@ -1054,7 +1054,6 @@ int synchronizers_suite_init(void) > > int synchronizers_init(void) > { > - uint32_t workers_count, max_threads; > int ret = 0; > odp_cpumask_t mask; > > @@ -1078,25 +1077,12 @@ int synchronizers_init(void) > global_mem = odp_shm_addr(global_shm); > memset(global_mem, 0, sizeof(global_shared_mem_t)); > > - global_mem->g_num_threads = MAX_WORKERS; > + global_mem->g_num_workers = odp_cpumask_def_worker(&mask, 0); > global_mem->g_iterations = MAX_ITERATIONS; > global_mem->g_verbose = VERBOSE; > > - workers_count = odp_cpumask_def_worker(&mask, 0); > - > - max_threads = (workers_count >= MAX_WORKERS) ? > - MAX_WORKERS : workers_count; > - > - if (max_threads < global_mem->g_num_threads) { > - printf("Requested num of threads is too large\n"); > - printf("reducing from %" PRIu32 " to %" PRIu32 "\n", > - global_mem->g_num_threads, > - max_threads); > - global_mem->g_num_threads = max_threads; > - } > - > - printf("Num of threads used = %" PRIu32 "\n", > - global_mem->g_num_threads); > + printf("Num of workers used = %" PRIu32 "\n", > + global_mem->g_num_workers); > > return ret; > } > @@ -1158,7 +1144,7 @@ static void test_atomic_functional(void *func_ptr(void *)) > { > pthrd_arg arg; > > - arg.numthrds = global_mem->g_num_threads; > + arg.numthrds = global_mem->g_num_workers; > test_atomic_init(); > test_atomic_store(); > odp_cunit_thread_create(func_ptr, &arg); > diff --git a/test/validation/timer/timer.c b/test/validation/timer/timer.c > index 7a8b98a..bcba3d4 100644 > --- a/test/validation/timer/timer.c > +++ b/test/validation/timer/timer.c > @@ -34,12 +34,6 @@ static odp_timer_pool_t tp; > /** @private Count of timeouts delivered too late */ > static odp_atomic_u32_t ndelivtoolate; > > -/** @private min() function */ > -static int min(int a, int b) > -{ > - return a < b ? a : b; > -} > - > /* @private Timer helper structure */ > struct test_timer { > odp_timer_t tim; /* Timer handle */ > @@ -441,10 +435,12 @@ void timer_test_odp_timer_all(void) > int rc; > odp_pool_param_t params; > odp_timer_pool_param_t tparam; > + odp_cpumask_t mask; > + > /* Reserve at least one core for running other processes so the timer > * test hopefully can run undisturbed and thus get better timing > * results. */ > - int num_workers = min(odp_cpu_count() - 1, MAX_WORKERS); > + int num_workers = odp_cpumask_def_worker(&mask, 0) - 1; > /* On a single-CPU machine run at least one thread */ > if (num_workers < 1) > num_workers = 1;
On 09/09/15 10:24, Nicolas Morey-Chaisemartin wrote: > > On 09/08/2015 01:05 PM, Maxim Uvarov wrote: >> ODP has api to request available number of workers. Now >> no need limit that inside application. >> >> Signed-off-by: Maxim Uvarov <maxim.uvarov@linaro.org> >> --- >> >> v2: timer.c: do not substract -1 from number of workers. >> >> test/validation/common/odp_cunit_common.c | 7 ++-- >> test/validation/common/odp_cunit_common.h | 2 -- >> test/validation/scheduler/scheduler.c | 3 -- >> test/validation/shmem/shmem.c | 6 ++-- >> test/validation/synchronizers/synchronizers.c | 48 ++++++++++----------------- >> test/validation/timer/timer.c | 10 ++---- >> 6 files changed, 27 insertions(+), 49 deletions(-) >> >> diff --git a/test/validation/common/odp_cunit_common.c b/test/validation/common/odp_cunit_common.c >> index d995ad3..01ea87b 100644 >> --- a/test/validation/common/odp_cunit_common.c >> +++ b/test/validation/common/odp_cunit_common.c >> @@ -9,7 +9,7 @@ >> #include <odp_cunit_common.h> >> #include <odp/helper/linux.h> >> /* Globals */ >> -static odph_linux_pthread_t thread_tbl[MAX_WORKERS]; >> +static odph_linux_pthread_t *thread_tbl; >> >> /* >> * global init/term functions which may be registered >> @@ -26,9 +26,11 @@ static struct { >> int odp_cunit_thread_create(void *func_ptr(void *), pthrd_arg *arg) >> { >> odp_cpumask_t cpumask; >> + int num; >> >> /* Create and init additional threads */ >> - odp_cpumask_def_worker(&cpumask, arg->numthrds); >> + num = odp_cpumask_def_worker(&cpumask, arg->numthrds); >> + thread_tbl = calloc(sizeof(odph_linux_pthread_t), num); > Maybe add a check of the return value here? Everything else looks good > Reviewed-by: Nicolas Morey-Chaisemartin <nmorey@kalray.eu> I can add but usually over commit is turned on and all malloc() / calloc() pass and real allocation postponed to memory access. So you never know if you allocated memory or kernel will kill app on access. Maxim. >> >> return odph_linux_pthread_create(thread_tbl, &cpumask, func_ptr, >> (void *)arg); >> @@ -39,6 +41,7 @@ int odp_cunit_thread_exit(pthrd_arg *arg) >> { >> /* Wait for other threads to exit */ >> odph_linux_pthread_join(thread_tbl, arg->numthrds); >> + free(thread_tbl); >> >> return 0; >> } >> diff --git a/test/validation/common/odp_cunit_common.h b/test/validation/common/odp_cunit_common.h >> index 6cafaaa..f94b44e 100644 >> --- a/test/validation/common/odp_cunit_common.h >> +++ b/test/validation/common/odp_cunit_common.h >> @@ -16,8 +16,6 @@ >> #include <stdint.h> >> #include "CUnit/Basic.h" >> >> -#define MAX_WORKERS 32 /**< Maximum number of work threads */ >> - >> /* the function, called by module main(), to run the testsuites: */ >> int odp_cunit_run(CU_SuiteInfo testsuites[]); >> >> diff --git a/test/validation/scheduler/scheduler.c b/test/validation/scheduler/scheduler.c >> index 1874889..e12895d 100644 >> --- a/test/validation/scheduler/scheduler.c >> +++ b/test/validation/scheduler/scheduler.c >> @@ -8,7 +8,6 @@ >> #include "odp_cunit_common.h" >> #include "scheduler.h" >> >> -#define MAX_WORKERS_THREADS 32 >> #define MSG_POOL_SIZE (4 * 1024 * 1024) >> #define QUEUES_PER_PRIO 16 >> #define BUF_SIZE 64 >> @@ -1018,8 +1017,6 @@ int scheduler_suite_init(void) >> memset(globals, 0, sizeof(test_globals_t)); >> >> globals->num_workers = odp_cpumask_def_worker(&mask, 0); >> - if (globals->num_workers > MAX_WORKERS) >> - globals->num_workers = MAX_WORKERS; >> >> shm = odp_shm_reserve(SHM_THR_ARGS_NAME, sizeof(thread_args_t), >> ODP_CACHE_LINE_SIZE, 0); >> diff --git a/test/validation/shmem/shmem.c b/test/validation/shmem/shmem.c >> index 6dc579a..dfa5310 100644 >> --- a/test/validation/shmem/shmem.c >> +++ b/test/validation/shmem/shmem.c >> @@ -49,6 +49,7 @@ void shmem_test_odp_shm_sunnyday(void) >> pthrd_arg thrdarg; >> odp_shm_t shm; >> test_shared_data_t *test_shared_data; >> + odp_cpumask_t mask; >> >> shm = odp_shm_reserve(TESTNAME, >> sizeof(test_shared_data_t), ALIGE_SIZE, 0); >> @@ -67,10 +68,7 @@ void shmem_test_odp_shm_sunnyday(void) >> test_shared_data->foo = TEST_SHARE_FOO; >> test_shared_data->bar = TEST_SHARE_BAR; >> >> - thrdarg.numthrds = odp_cpu_count(); >> - >> - if (thrdarg.numthrds > MAX_WORKERS) >> - thrdarg.numthrds = MAX_WORKERS; >> + thrdarg.numthrds = odp_cpumask_def_worker(&mask, 0); >> >> odp_cunit_thread_create(run_shm_thread, &thrdarg); >> odp_cunit_thread_exit(&thrdarg); >> diff --git a/test/validation/synchronizers/synchronizers.c b/test/validation/synchronizers/synchronizers.c >> index 0a31a40..914b37e 100644 >> --- a/test/validation/synchronizers/synchronizers.c >> +++ b/test/validation/synchronizers/synchronizers.c >> @@ -45,7 +45,7 @@ typedef struct { >> >> typedef struct { >> /* Global variables */ >> - uint32_t g_num_threads; >> + uint32_t g_num_workers; >> uint32_t g_iterations; >> uint32_t g_verbose; >> uint32_t g_max_num_cores; >> @@ -169,7 +169,7 @@ static uint32_t barrier_test(per_thread_mem_t *per_thread_mem, >> >> thread_num = odp_thread_id(); >> global_mem = per_thread_mem->global_mem; >> - num_threads = global_mem->g_num_threads; >> + num_threads = global_mem->g_num_workers; >> iterations = BARRIER_ITERATIONS; >> >> barrier_errs = 0; >> @@ -710,7 +710,7 @@ static void barrier_test_init(void) >> { >> uint32_t num_threads, idx; >> >> - num_threads = global_mem->g_num_threads; >> + num_threads = global_mem->g_num_workers; >> >> for (idx = 0; idx < NUM_TEST_BARRIERS; idx++) { >> odp_barrier_init(&global_mem->test_barriers[idx], num_threads); >> @@ -924,7 +924,7 @@ void synchronizers_test_no_barrier_functional(void) >> { >> pthrd_arg arg; >> >> - arg.numthrds = global_mem->g_num_threads; >> + arg.numthrds = global_mem->g_num_workers; >> barrier_test_init(); >> odp_cunit_thread_create(no_barrier_functional_test, &arg); >> odp_cunit_thread_exit(&arg); >> @@ -934,7 +934,7 @@ void synchronizers_test_barrier_functional(void) >> { >> pthrd_arg arg; >> >> - arg.numthrds = global_mem->g_num_threads; >> + arg.numthrds = global_mem->g_num_workers; >> barrier_test_init(); >> odp_cunit_thread_create(barrier_functional_test, &arg); >> odp_cunit_thread_exit(&arg); >> @@ -951,7 +951,7 @@ void synchronizers_test_no_lock_functional(void) >> { >> pthrd_arg arg; >> >> - arg.numthrds = global_mem->g_num_threads; >> + arg.numthrds = global_mem->g_num_workers; >> odp_cunit_thread_create(no_lock_functional_test, &arg); >> odp_cunit_thread_exit(&arg); >> } >> @@ -966,7 +966,7 @@ void synchronizers_test_spinlock_api(void) >> { >> pthrd_arg arg; >> >> - arg.numthrds = global_mem->g_num_threads; >> + arg.numthrds = global_mem->g_num_workers; >> odp_cunit_thread_create(spinlock_api_tests, &arg); >> odp_cunit_thread_exit(&arg); >> } >> @@ -975,7 +975,7 @@ void synchronizers_test_spinlock_functional(void) >> { >> pthrd_arg arg; >> >> - arg.numthrds = global_mem->g_num_threads; >> + arg.numthrds = global_mem->g_num_workers; >> odp_spinlock_init(&global_mem->global_spinlock); >> odp_cunit_thread_create(spinlock_functional_test, &arg); >> odp_cunit_thread_exit(&arg); >> @@ -992,7 +992,7 @@ void synchronizers_test_ticketlock_api(void) >> { >> pthrd_arg arg; >> >> - arg.numthrds = global_mem->g_num_threads; >> + arg.numthrds = global_mem->g_num_workers; >> odp_cunit_thread_create(ticketlock_api_tests, &arg); >> odp_cunit_thread_exit(&arg); >> } >> @@ -1001,7 +1001,7 @@ void synchronizers_test_ticketlock_functional(void) >> { >> pthrd_arg arg; >> >> - arg.numthrds = global_mem->g_num_threads; >> + arg.numthrds = global_mem->g_num_workers; >> odp_ticketlock_init(&global_mem->global_ticketlock); >> >> odp_cunit_thread_create(ticketlock_functional_test, &arg); >> @@ -1019,7 +1019,7 @@ void synchronizers_test_rwlock_api(void) >> { >> pthrd_arg arg; >> >> - arg.numthrds = global_mem->g_num_threads; >> + arg.numthrds = global_mem->g_num_workers; >> odp_cunit_thread_create(rwlock_api_tests, &arg); >> odp_cunit_thread_exit(&arg); >> } >> @@ -1028,7 +1028,7 @@ void synchronizers_test_rwlock_functional(void) >> { >> pthrd_arg arg; >> >> - arg.numthrds = global_mem->g_num_threads; >> + arg.numthrds = global_mem->g_num_workers; >> odp_rwlock_init(&global_mem->global_rwlock); >> odp_cunit_thread_create(rwlock_functional_test, &arg); >> odp_cunit_thread_exit(&arg); >> @@ -1044,7 +1044,7 @@ int synchronizers_suite_init(void) >> { >> uint32_t num_threads, idx; >> >> - num_threads = global_mem->g_num_threads; >> + num_threads = global_mem->g_num_workers; >> odp_barrier_init(&global_mem->global_barrier, num_threads); >> for (idx = 0; idx < NUM_RESYNC_BARRIERS; idx++) >> odp_barrier_init(&global_mem->barrier_array[idx], num_threads); >> @@ -1054,7 +1054,6 @@ int synchronizers_suite_init(void) >> >> int synchronizers_init(void) >> { >> - uint32_t workers_count, max_threads; >> int ret = 0; >> odp_cpumask_t mask; >> >> @@ -1078,25 +1077,12 @@ int synchronizers_init(void) >> global_mem = odp_shm_addr(global_shm); >> memset(global_mem, 0, sizeof(global_shared_mem_t)); >> >> - global_mem->g_num_threads = MAX_WORKERS; >> + global_mem->g_num_workers = odp_cpumask_def_worker(&mask, 0); >> global_mem->g_iterations = MAX_ITERATIONS; >> global_mem->g_verbose = VERBOSE; >> >> - workers_count = odp_cpumask_def_worker(&mask, 0); >> - >> - max_threads = (workers_count >= MAX_WORKERS) ? >> - MAX_WORKERS : workers_count; >> - >> - if (max_threads < global_mem->g_num_threads) { >> - printf("Requested num of threads is too large\n"); >> - printf("reducing from %" PRIu32 " to %" PRIu32 "\n", >> - global_mem->g_num_threads, >> - max_threads); >> - global_mem->g_num_threads = max_threads; >> - } >> - >> - printf("Num of threads used = %" PRIu32 "\n", >> - global_mem->g_num_threads); >> + printf("Num of workers used = %" PRIu32 "\n", >> + global_mem->g_num_workers); >> >> return ret; >> } >> @@ -1158,7 +1144,7 @@ static void test_atomic_functional(void *func_ptr(void *)) >> { >> pthrd_arg arg; >> >> - arg.numthrds = global_mem->g_num_threads; >> + arg.numthrds = global_mem->g_num_workers; >> test_atomic_init(); >> test_atomic_store(); >> odp_cunit_thread_create(func_ptr, &arg); >> diff --git a/test/validation/timer/timer.c b/test/validation/timer/timer.c >> index 7a8b98a..bcba3d4 100644 >> --- a/test/validation/timer/timer.c >> +++ b/test/validation/timer/timer.c >> @@ -34,12 +34,6 @@ static odp_timer_pool_t tp; >> /** @private Count of timeouts delivered too late */ >> static odp_atomic_u32_t ndelivtoolate; >> >> -/** @private min() function */ >> -static int min(int a, int b) >> -{ >> - return a < b ? a : b; >> -} >> - >> /* @private Timer helper structure */ >> struct test_timer { >> odp_timer_t tim; /* Timer handle */ >> @@ -441,10 +435,12 @@ void timer_test_odp_timer_all(void) >> int rc; >> odp_pool_param_t params; >> odp_timer_pool_param_t tparam; >> + odp_cpumask_t mask; >> + >> /* Reserve at least one core for running other processes so the timer >> * test hopefully can run undisturbed and thus get better timing >> * results. */ >> - int num_workers = min(odp_cpu_count() - 1, MAX_WORKERS); >> + int num_workers = odp_cpumask_def_worker(&mask, 0) - 1; >> /* On a single-CPU machine run at least one thread */ >> if (num_workers < 1) >> num_workers = 1;
On 09/09/2015 09:49 AM, Maxim Uvarov wrote: > On 09/09/15 10:24, Nicolas Morey-Chaisemartin wrote: >> >> On 09/08/2015 01:05 PM, Maxim Uvarov wrote: >>> ODP has api to request available number of workers. Now >>> no need limit that inside application. >>> >>> Signed-off-by: Maxim Uvarov <maxim.uvarov@linaro.org> >>> --- >>> >>> v2: timer.c: do not substract -1 from number of workers. >>> >>> test/validation/common/odp_cunit_common.c | 7 ++-- >>> test/validation/common/odp_cunit_common.h | 2 -- >>> test/validation/scheduler/scheduler.c | 3 -- >>> test/validation/shmem/shmem.c | 6 ++-- >>> test/validation/synchronizers/synchronizers.c | 48 ++++++++++----------------- >>> test/validation/timer/timer.c | 10 ++---- >>> 6 files changed, 27 insertions(+), 49 deletions(-) >>> >>> diff --git a/test/validation/common/odp_cunit_common.c b/test/validation/common/odp_cunit_common.c >>> index d995ad3..01ea87b 100644 >>> --- a/test/validation/common/odp_cunit_common.c >>> +++ b/test/validation/common/odp_cunit_common.c >>> @@ -9,7 +9,7 @@ >>> #include <odp_cunit_common.h> >>> #include <odp/helper/linux.h> >>> /* Globals */ >>> -static odph_linux_pthread_t thread_tbl[MAX_WORKERS]; >>> +static odph_linux_pthread_t *thread_tbl; >>> /* >>> * global init/term functions which may be registered >>> @@ -26,9 +26,11 @@ static struct { >>> int odp_cunit_thread_create(void *func_ptr(void *), pthrd_arg *arg) >>> { >>> odp_cpumask_t cpumask; >>> + int num; >>> /* Create and init additional threads */ >>> - odp_cpumask_def_worker(&cpumask, arg->numthrds); >>> + num = odp_cpumask_def_worker(&cpumask, arg->numthrds); >>> + thread_tbl = calloc(sizeof(odph_linux_pthread_t), num); >> Maybe add a check of the return value here? Everything else looks good >> Reviewed-by: Nicolas Morey-Chaisemartin <nmorey@kalray.eu> > > I can add but usually over commit is turned on and all malloc() / calloc() pass and real allocation > postponed to memory access. So you never know if you allocated memory or kernel will kill app > on access. > > Maxim. > On linux yes. On smaller HW platform (like ours) we don't use overcommit. As it's part of the common validation suite, it'd make sense to check it. Nicolas
On 09/09/15 11:02, Nicolas Morey-Chaisemartin wrote: > > On 09/09/2015 09:49 AM, Maxim Uvarov wrote: >> On 09/09/15 10:24, Nicolas Morey-Chaisemartin wrote: >>> On 09/08/2015 01:05 PM, Maxim Uvarov wrote: >>>> ODP has api to request available number of workers. Now >>>> no need limit that inside application. >>>> >>>> Signed-off-by: Maxim Uvarov <maxim.uvarov@linaro.org> >>>> --- >>>> >>>> v2: timer.c: do not substract -1 from number of workers. >>>> >>>> test/validation/common/odp_cunit_common.c | 7 ++-- >>>> test/validation/common/odp_cunit_common.h | 2 -- >>>> test/validation/scheduler/scheduler.c | 3 -- >>>> test/validation/shmem/shmem.c | 6 ++-- >>>> test/validation/synchronizers/synchronizers.c | 48 ++++++++++----------------- >>>> test/validation/timer/timer.c | 10 ++---- >>>> 6 files changed, 27 insertions(+), 49 deletions(-) >>>> >>>> diff --git a/test/validation/common/odp_cunit_common.c b/test/validation/common/odp_cunit_common.c >>>> index d995ad3..01ea87b 100644 >>>> --- a/test/validation/common/odp_cunit_common.c >>>> +++ b/test/validation/common/odp_cunit_common.c >>>> @@ -9,7 +9,7 @@ >>>> #include <odp_cunit_common.h> >>>> #include <odp/helper/linux.h> >>>> /* Globals */ >>>> -static odph_linux_pthread_t thread_tbl[MAX_WORKERS]; >>>> +static odph_linux_pthread_t *thread_tbl; >>>> /* >>>> * global init/term functions which may be registered >>>> @@ -26,9 +26,11 @@ static struct { >>>> int odp_cunit_thread_create(void *func_ptr(void *), pthrd_arg *arg) >>>> { >>>> odp_cpumask_t cpumask; >>>> + int num; >>>> /* Create and init additional threads */ >>>> - odp_cpumask_def_worker(&cpumask, arg->numthrds); >>>> + num = odp_cpumask_def_worker(&cpumask, arg->numthrds); >>>> + thread_tbl = calloc(sizeof(odph_linux_pthread_t), num); >>> Maybe add a check of the return value here? Everything else looks good >>> Reviewed-by: Nicolas Morey-Chaisemartin <nmorey@kalray.eu> >> I can add but usually over commit is turned on and all malloc() / calloc() pass and real allocation >> postponed to memory access. So you never know if you allocated memory or kernel will kill app >> on access. >> >> Maxim. >> > On linux yes. On smaller HW platform (like ours) we don't use overcommit. > As it's part of the common validation suite, it'd make sense to check it. > > Nicolas > Ok, in that case it's reasonable to account that. Will send v3. Maxim.
diff --git a/test/validation/common/odp_cunit_common.c b/test/validation/common/odp_cunit_common.c index d995ad3..01ea87b 100644 --- a/test/validation/common/odp_cunit_common.c +++ b/test/validation/common/odp_cunit_common.c @@ -9,7 +9,7 @@ #include <odp_cunit_common.h> #include <odp/helper/linux.h> /* Globals */ -static odph_linux_pthread_t thread_tbl[MAX_WORKERS]; +static odph_linux_pthread_t *thread_tbl; /* * global init/term functions which may be registered @@ -26,9 +26,11 @@ static struct { int odp_cunit_thread_create(void *func_ptr(void *), pthrd_arg *arg) { odp_cpumask_t cpumask; + int num; /* Create and init additional threads */ - odp_cpumask_def_worker(&cpumask, arg->numthrds); + num = odp_cpumask_def_worker(&cpumask, arg->numthrds); + thread_tbl = calloc(sizeof(odph_linux_pthread_t), num); return odph_linux_pthread_create(thread_tbl, &cpumask, func_ptr, (void *)arg); @@ -39,6 +41,7 @@ int odp_cunit_thread_exit(pthrd_arg *arg) { /* Wait for other threads to exit */ odph_linux_pthread_join(thread_tbl, arg->numthrds); + free(thread_tbl); return 0; } diff --git a/test/validation/common/odp_cunit_common.h b/test/validation/common/odp_cunit_common.h index 6cafaaa..f94b44e 100644 --- a/test/validation/common/odp_cunit_common.h +++ b/test/validation/common/odp_cunit_common.h @@ -16,8 +16,6 @@ #include <stdint.h> #include "CUnit/Basic.h" -#define MAX_WORKERS 32 /**< Maximum number of work threads */ - /* the function, called by module main(), to run the testsuites: */ int odp_cunit_run(CU_SuiteInfo testsuites[]); diff --git a/test/validation/scheduler/scheduler.c b/test/validation/scheduler/scheduler.c index 1874889..e12895d 100644 --- a/test/validation/scheduler/scheduler.c +++ b/test/validation/scheduler/scheduler.c @@ -8,7 +8,6 @@ #include "odp_cunit_common.h" #include "scheduler.h" -#define MAX_WORKERS_THREADS 32 #define MSG_POOL_SIZE (4 * 1024 * 1024) #define QUEUES_PER_PRIO 16 #define BUF_SIZE 64 @@ -1018,8 +1017,6 @@ int scheduler_suite_init(void) memset(globals, 0, sizeof(test_globals_t)); globals->num_workers = odp_cpumask_def_worker(&mask, 0); - if (globals->num_workers > MAX_WORKERS) - globals->num_workers = MAX_WORKERS; shm = odp_shm_reserve(SHM_THR_ARGS_NAME, sizeof(thread_args_t), ODP_CACHE_LINE_SIZE, 0); diff --git a/test/validation/shmem/shmem.c b/test/validation/shmem/shmem.c index 6dc579a..dfa5310 100644 --- a/test/validation/shmem/shmem.c +++ b/test/validation/shmem/shmem.c @@ -49,6 +49,7 @@ void shmem_test_odp_shm_sunnyday(void) pthrd_arg thrdarg; odp_shm_t shm; test_shared_data_t *test_shared_data; + odp_cpumask_t mask; shm = odp_shm_reserve(TESTNAME, sizeof(test_shared_data_t), ALIGE_SIZE, 0); @@ -67,10 +68,7 @@ void shmem_test_odp_shm_sunnyday(void) test_shared_data->foo = TEST_SHARE_FOO; test_shared_data->bar = TEST_SHARE_BAR; - thrdarg.numthrds = odp_cpu_count(); - - if (thrdarg.numthrds > MAX_WORKERS) - thrdarg.numthrds = MAX_WORKERS; + thrdarg.numthrds = odp_cpumask_def_worker(&mask, 0); odp_cunit_thread_create(run_shm_thread, &thrdarg); odp_cunit_thread_exit(&thrdarg); diff --git a/test/validation/synchronizers/synchronizers.c b/test/validation/synchronizers/synchronizers.c index 0a31a40..914b37e 100644 --- a/test/validation/synchronizers/synchronizers.c +++ b/test/validation/synchronizers/synchronizers.c @@ -45,7 +45,7 @@ typedef struct { typedef struct { /* Global variables */ - uint32_t g_num_threads; + uint32_t g_num_workers; uint32_t g_iterations; uint32_t g_verbose; uint32_t g_max_num_cores; @@ -169,7 +169,7 @@ static uint32_t barrier_test(per_thread_mem_t *per_thread_mem, thread_num = odp_thread_id(); global_mem = per_thread_mem->global_mem; - num_threads = global_mem->g_num_threads; + num_threads = global_mem->g_num_workers; iterations = BARRIER_ITERATIONS; barrier_errs = 0; @@ -710,7 +710,7 @@ static void barrier_test_init(void) { uint32_t num_threads, idx; - num_threads = global_mem->g_num_threads; + num_threads = global_mem->g_num_workers; for (idx = 0; idx < NUM_TEST_BARRIERS; idx++) { odp_barrier_init(&global_mem->test_barriers[idx], num_threads); @@ -924,7 +924,7 @@ void synchronizers_test_no_barrier_functional(void) { pthrd_arg arg; - arg.numthrds = global_mem->g_num_threads; + arg.numthrds = global_mem->g_num_workers; barrier_test_init(); odp_cunit_thread_create(no_barrier_functional_test, &arg); odp_cunit_thread_exit(&arg); @@ -934,7 +934,7 @@ void synchronizers_test_barrier_functional(void) { pthrd_arg arg; - arg.numthrds = global_mem->g_num_threads; + arg.numthrds = global_mem->g_num_workers; barrier_test_init(); odp_cunit_thread_create(barrier_functional_test, &arg); odp_cunit_thread_exit(&arg); @@ -951,7 +951,7 @@ void synchronizers_test_no_lock_functional(void) { pthrd_arg arg; - arg.numthrds = global_mem->g_num_threads; + arg.numthrds = global_mem->g_num_workers; odp_cunit_thread_create(no_lock_functional_test, &arg); odp_cunit_thread_exit(&arg); } @@ -966,7 +966,7 @@ void synchronizers_test_spinlock_api(void) { pthrd_arg arg; - arg.numthrds = global_mem->g_num_threads; + arg.numthrds = global_mem->g_num_workers; odp_cunit_thread_create(spinlock_api_tests, &arg); odp_cunit_thread_exit(&arg); } @@ -975,7 +975,7 @@ void synchronizers_test_spinlock_functional(void) { pthrd_arg arg; - arg.numthrds = global_mem->g_num_threads; + arg.numthrds = global_mem->g_num_workers; odp_spinlock_init(&global_mem->global_spinlock); odp_cunit_thread_create(spinlock_functional_test, &arg); odp_cunit_thread_exit(&arg); @@ -992,7 +992,7 @@ void synchronizers_test_ticketlock_api(void) { pthrd_arg arg; - arg.numthrds = global_mem->g_num_threads; + arg.numthrds = global_mem->g_num_workers; odp_cunit_thread_create(ticketlock_api_tests, &arg); odp_cunit_thread_exit(&arg); } @@ -1001,7 +1001,7 @@ void synchronizers_test_ticketlock_functional(void) { pthrd_arg arg; - arg.numthrds = global_mem->g_num_threads; + arg.numthrds = global_mem->g_num_workers; odp_ticketlock_init(&global_mem->global_ticketlock); odp_cunit_thread_create(ticketlock_functional_test, &arg); @@ -1019,7 +1019,7 @@ void synchronizers_test_rwlock_api(void) { pthrd_arg arg; - arg.numthrds = global_mem->g_num_threads; + arg.numthrds = global_mem->g_num_workers; odp_cunit_thread_create(rwlock_api_tests, &arg); odp_cunit_thread_exit(&arg); } @@ -1028,7 +1028,7 @@ void synchronizers_test_rwlock_functional(void) { pthrd_arg arg; - arg.numthrds = global_mem->g_num_threads; + arg.numthrds = global_mem->g_num_workers; odp_rwlock_init(&global_mem->global_rwlock); odp_cunit_thread_create(rwlock_functional_test, &arg); odp_cunit_thread_exit(&arg); @@ -1044,7 +1044,7 @@ int synchronizers_suite_init(void) { uint32_t num_threads, idx; - num_threads = global_mem->g_num_threads; + num_threads = global_mem->g_num_workers; odp_barrier_init(&global_mem->global_barrier, num_threads); for (idx = 0; idx < NUM_RESYNC_BARRIERS; idx++) odp_barrier_init(&global_mem->barrier_array[idx], num_threads); @@ -1054,7 +1054,6 @@ int synchronizers_suite_init(void) int synchronizers_init(void) { - uint32_t workers_count, max_threads; int ret = 0; odp_cpumask_t mask; @@ -1078,25 +1077,12 @@ int synchronizers_init(void) global_mem = odp_shm_addr(global_shm); memset(global_mem, 0, sizeof(global_shared_mem_t)); - global_mem->g_num_threads = MAX_WORKERS; + global_mem->g_num_workers = odp_cpumask_def_worker(&mask, 0); global_mem->g_iterations = MAX_ITERATIONS; global_mem->g_verbose = VERBOSE; - workers_count = odp_cpumask_def_worker(&mask, 0); - - max_threads = (workers_count >= MAX_WORKERS) ? - MAX_WORKERS : workers_count; - - if (max_threads < global_mem->g_num_threads) { - printf("Requested num of threads is too large\n"); - printf("reducing from %" PRIu32 " to %" PRIu32 "\n", - global_mem->g_num_threads, - max_threads); - global_mem->g_num_threads = max_threads; - } - - printf("Num of threads used = %" PRIu32 "\n", - global_mem->g_num_threads); + printf("Num of workers used = %" PRIu32 "\n", + global_mem->g_num_workers); return ret; } @@ -1158,7 +1144,7 @@ static void test_atomic_functional(void *func_ptr(void *)) { pthrd_arg arg; - arg.numthrds = global_mem->g_num_threads; + arg.numthrds = global_mem->g_num_workers; test_atomic_init(); test_atomic_store(); odp_cunit_thread_create(func_ptr, &arg); diff --git a/test/validation/timer/timer.c b/test/validation/timer/timer.c index 7a8b98a..bcba3d4 100644 --- a/test/validation/timer/timer.c +++ b/test/validation/timer/timer.c @@ -34,12 +34,6 @@ static odp_timer_pool_t tp; /** @private Count of timeouts delivered too late */ static odp_atomic_u32_t ndelivtoolate; -/** @private min() function */ -static int min(int a, int b) -{ - return a < b ? a : b; -} - /* @private Timer helper structure */ struct test_timer { odp_timer_t tim; /* Timer handle */ @@ -441,10 +435,12 @@ void timer_test_odp_timer_all(void) int rc; odp_pool_param_t params; odp_timer_pool_param_t tparam; + odp_cpumask_t mask; + /* Reserve at least one core for running other processes so the timer * test hopefully can run undisturbed and thus get better timing * results. */ - int num_workers = min(odp_cpu_count() - 1, MAX_WORKERS); + int num_workers = odp_cpumask_def_worker(&mask, 0) - 1; /* On a single-CPU machine run at least one thread */ if (num_workers < 1) num_workers = 1;
ODP has api to request available number of workers. Now no need limit that inside application. Signed-off-by: Maxim Uvarov <maxim.uvarov@linaro.org> --- v2: timer.c: do not substract -1 from number of workers. test/validation/common/odp_cunit_common.c | 7 ++-- test/validation/common/odp_cunit_common.h | 2 -- test/validation/scheduler/scheduler.c | 3 -- test/validation/shmem/shmem.c | 6 ++-- test/validation/synchronizers/synchronizers.c | 48 ++++++++++----------------- test/validation/timer/timer.c | 10 ++---- 6 files changed, 27 insertions(+), 49 deletions(-)