void tdaxpby( unsigned n ,
double a , const double * x ,
double b , double * y ,
int block )
{
int p_size ;
TPI_Size( & p_size );
{
unsigned *tmp = malloc( p_size );
struct TaskXY data = { a , b , x , y , n , tmp };
int i ;
for ( i = 0 ; i < p_size ; ++i ) { tmp[i] = i ; }
if ( 0 < block ) {
TPI_Run( & task_axpby_work_block , & data , 0 );
}
else if ( block < 0 ) {
TPI_Set_lock_size( p_size );
TPI_Run( & task_axpby_work_steal , & data , 0 );
}
else {
TPI_Run( & task_axpby_work , & data , 0 );
}
free(tmp);
}
}
int TPI_Finalize()
{
ThreadPool * const pool = local_thread_pool();
int result = ! pool ? TPI_ERROR_ACTIVE : 0 ;
if ( ! result ) {
--( pool->m_number_threads );
if ( 0 < pool->m_number_threads ) {
pthread_mutex_lock( & pool->m_pool_lock );
pool->m_work_end_count = -1 ; /* Trigger to terminate */
pthread_mutex_lock( & pool->m_pool_lock_run );
pthread_cond_broadcast( & pool->m_pool_cond_run );
pthread_mutex_unlock( & pool->m_pool_lock_run );
pthread_cond_wait( & pool->m_pool_cond , & pool->m_pool_lock );
pthread_mutex_unlock( & pool->m_pool_lock );
}
pool->m_work_size = 0 ;
pool->m_work_begin_count = 0 ;
pool->m_work_end_count = 0 ;
pool->m_number_threads = 0 ;
TPI_Set_lock_size( 0 );
}
return result ;
}
void txddot1( double * s , unsigned n , const double * x )
{
struct TaskX data ;
data.x_sum = s ;
data.x_beg = x ;
data.number = n ;
TPI_Set_lock_size( 1 );
TPI_Run( & task_xddot_x_work , & data , 0 );
}
inline
int Set_lock_size( int n ) { return TPI_Set_lock_size( n ); }
