ports/openmpi
ports/openmpi
1
1
Форкнуть 0
Код Релизы Активность

move temp buffer allocation out of the iteration loop - i.e. always use the

Просмотр исходного кода 
same temp loop.  The algorithm is rather synchronous already...

This commit was SVN r17831.
Этот коммит содержится в:
Rich Graham 2008-03-16 03:20:46 +00:00
родитель a1766b29f6
Коммит afcd1016fd
1 изменённых файлов: 74 добавлений и 39 удалений
Показать статистику Скачать Patch файл Скачать Diff файл Показать все файлы Свернуть все файлы

113
ompi/mca/coll/sm2/coll_sm2_allreduce.c
Просмотреть файл

@ -406,21 +406,17 @@ int mca_coll_sm2_allreduce_intra_recursive_doubling(void *sbuf, void *rbuf,
count_processed=0;
/* debug */
t0=opal_sys_timer_get_cycles();
/* end debug */
sm_buffer_desc=alloc_sm2_shared_buffer(sm_module);
/* debug */
t1=opal_sys_timer_get_cycles();
/* end debug */
/* get a pointer to the shared-memory working buffer */
/* NOTE: starting with a rather synchronous approach */
for( stripe_number=0 ; stripe_number < n_data_segments ; stripe_number++ ) {
/* debug */
t0=opal_sys_timer_get_cycles();
/* end debug */
sm_buffer_desc=alloc_sm2_shared_buffer(sm_module);
sm_buffer=sm_buffer_desc->base_segment_address;
if( NULL == sm_buffer) {
rc=OMPI_ERR_OUT_OF_RESOURCE;
goto Error;
}
/* debug */
t1=opal_sys_timer_get_cycles();
/* end debug */
/* get number of elements to process in this stripe */
count_this_stripe=n_dts_per_buffer;
if( count_processed + count_this_stripe > count )
@ -605,6 +601,9 @@ int mca_coll_sm2_allreduce_intra_recursive_doubling(void *sbuf, void *rbuf,
return OMPI_ERROR;
}
/* signal that I am done */
my_ctl_pointer->flag=tag;
} else {
tag=base_tag+my_exchange_node->n_tags-1;
@ -619,6 +618,18 @@ int mca_coll_sm2_allreduce_intra_recursive_doubling(void *sbuf, void *rbuf,
*/
my_ctl_pointer->flag=tag;
/* wait until child is done to move on - this buffer will
* be reused for the next stripe, so don't want to move
* on too quick.
*/
extra_rank=my_exchange_node->rank_extra_source;
extra_ctl_pointer=
sm_buffer_desc->proc_memory[extra_rank].control_region;
/* wait until remote data is read */
while( extra_ctl_pointer->flag < tag ) {
opal_progress();
}
}
}
@ -632,25 +643,30 @@ int mca_coll_sm2_allreduce_intra_recursive_doubling(void *sbuf, void *rbuf,
if( 0 != rc ) {
return OMPI_ERROR;
}
/* debug */
t9=opal_sys_timer_get_cycles();
timers[5]+=(t9-t8);
/* end debug */
/* "free" the shared-memory working buffer */
rc=free_sm2_shared_buffer(sm_module);
if( OMPI_SUCCESS != rc ) {
goto Error;
}
/* debug */
t10=opal_sys_timer_get_cycles();
timers[6]+=(t10-t9);
/* end debug */
/* update the count of elements processed */
count_processed+=count_this_stripe;
}
/* debug */
t9=opal_sys_timer_get_cycles();
timers[5]+=(t9-t8);
/* end debug */
/* "free" the shared-memory working buffer */
rc=free_sm2_shared_buffer(sm_module);
if( OMPI_SUCCESS != rc ) {
goto Error;
}
/* debug */
t10=opal_sys_timer_get_cycles();
timers[6]+=(t10-t9);
/* end debug */
/* return */
return rc;
@ -734,15 +750,25 @@ int mca_coll_sm2_allreduce_intra_recursive_doubling(void *sbuf, void *rbuf,
/* get a pointer to the shared-memory working buffer */
/* NOTE: starting with a rather synchronous approach */
/* debug */
t0=opal_sys_timer_get_cycles();
/* end debug */
/* use the same set of buffers for a single reduction */
sm_buffer_desc=alloc_sm2_shared_buffer(sm_module);
/* get pointers to my work buffers */
my_ctl_pointer=sm_buffer_desc->proc_memory[my_rank].control_region;
my_write_pointer=sm_buffer_desc->proc_memory[my_rank].data_segment;
my_read_pointer=my_write_pointer+len_data_buffer;
my_tmp_data_buffer[0]=my_write_pointer;
my_tmp_data_buffer[1]=my_read_pointer;
/* debug */
t1=opal_sys_timer_get_cycles();
/* end debug */
for( stripe_number=0 ; stripe_number < n_data_segments ; stripe_number++ ) {
/* debug */
t0=opal_sys_timer_get_cycles();
/* end debug */
/* debug */
t1=opal_sys_timer_get_cycles();
/* end debug */
/* get number of elements to process in this stripe */
count_this_stripe=n_dts_per_buffer;
if( count_processed + count_this_stripe > count )
@ -755,12 +781,6 @@ int mca_coll_sm2_allreduce_intra_recursive_doubling(void *sbuf, void *rbuf,
base_tag=sm_module->collective_tag;
sm_module->collective_tag+=my_exchange_node->n_tags;
/* get pointers to my work buffers */
my_ctl_pointer=sm_buffer_desc->proc_memory[my_rank].control_region;
my_write_pointer=sm_buffer_desc->proc_memory[my_rank].data_segment;
my_read_pointer=my_write_pointer+len_data_buffer;
my_tmp_data_buffer[0]=my_write_pointer;
my_tmp_data_buffer[1]=my_read_pointer;
/* debug */
t2=opal_sys_timer_get_cycles();
timers[0]+=(t2-t1);
@ -823,7 +843,6 @@ int mca_coll_sm2_allreduce_intra_recursive_doubling(void *sbuf, void *rbuf,
/* loop over data exchanges */
for(exchange=0 ; exchange < my_exchange_node->n_exchanges ; exchange++) {
/* debug */
t4=opal_sys_timer_get_cycles();
/* end debug */
@ -927,6 +946,10 @@ int mca_coll_sm2_allreduce_intra_recursive_doubling(void *sbuf, void *rbuf,
return OMPI_ERROR;
}
/* signal that I am done */
my_ctl_pointer->flag=tag;
} else {
tag=base_tag+my_exchange_node->n_tags-1;
@ -941,6 +964,18 @@ int mca_coll_sm2_allreduce_intra_recursive_doubling(void *sbuf, void *rbuf,
*/
my_ctl_pointer->flag=tag;
/* wait until child is done to move on - this buffer will
* be reused for the next stripe, so don't want to move
* on too quick.
*/
extra_rank=my_exchange_node->rank_extra_source;
extra_ctl_pointer=
sm_buffer_desc->proc_memory[extra_rank].control_region;
/* wait until remote data is read */
while(! ( extra_ctl_pointer->flag < tag ) ) {
opal_progress();
}
}
}