// Function is called once by every concurrent thread.
static void exec( QthreadExec & exec , const void * arg )
{
const ParallelFor & self = * ((const ParallelFor *) arg );
const WorkRange range( self.m_policy, exec.worker_rank(), exec.worker_size() );
ParallelFor::template exec_range< WorkTag > ( self.m_functor , range.begin() , range.end() );
// All threads wait for completion.
exec.exec_all_barrier();
}
static void exec( QthreadExec & exec , const void * arg )
{
const ParallelReduce & self = * ((const ParallelReduce *) arg );
const WorkRange range( self.m_policy, exec.worker_rank(), exec.worker_size() );
ParallelReduce::template exec_range< WorkTag >(
self.m_functor, range.begin(), range.end(),
ValueInit::init( ReducerConditional::select(self.m_functor , self.m_reducer)
, exec.exec_all_reduce_value() ) );
exec.template exec_all_reduce< FunctorType, ReducerType, WorkTag >( self.m_functor, self.m_reducer );
}
// Function is called once by every concurrent thread.
static void execute( QthreadExec & exec , const void * arg )
{
const ParallelFor & self = * ((const ParallelFor *) arg );
const Policy range( self.m_policy , exec.worker_rank() , exec.worker_size() );
const typename Policy::member_type work_end = range.end();
for ( typename Policy::member_type iwork = range.begin() ; iwork < work_end ; ++iwork ) {
self.m_func( iwork );
}
// All threads wait for completion.
exec.exec_all_barrier();
}
static void execute( QthreadExec & exec , const void * arg )
{
const ParallelReduce & self = * ((const ParallelReduce *) arg );
const Policy range( self.m_policy , exec.worker_rank() , exec.worker_size() );
// Initialize thread-local value
typename Reduce::reference_type update = Reduce::init( self.m_func , exec.exec_all_reduce_value() );
const typename Policy::member_type work_end = range.end();
for ( typename Policy::member_type iwork = range.begin() ; iwork < work_end ; ++iwork ) {
self.m_func( iwork , update );
}
exec.exec_all_reduce( self.m_func );
}
