1a021b8f2d
allgather. The new collectives provide a signifigant performance increase over tuned for small and medium messages. We are initially setting the priority lower than tuned until this has had some time to soak in the trunk. Please set coll_ml_priority to 90 for MTT runs. Credit for this work goes to Manjunath Gorentla Venkata (ORNL), Pavel Shamis (ORNL), and Nathan Hjelm (LANL). Commit details (for reference): Import ORNL's collectives for MPI_Allreduce, MPI_Reduce, and MPI_Allgather. We need to take the basesmuma header into account when calculating the ptpcoll small message thresholds. Add a define to bcol.h indicating the maximum header size so we can take the header into account while not making ptpcoll dependent on information from basesmuma. This resolves an issue with allreduce where ptpcoll overwrites the header of the next buffer in the basesmuma bank. Fix reduce and make a sequential collective launcher in coll_ml_inlines.h The root calculation for reduce was wrong for any root != 0. There are four possibilities for the root: - The root is not the current process but is in the current hierarchy. In this case the root is the index of the global root as specified in the root vector. - The root is not the current process and is not in the next level of the hierarchy. In this case 0 must be the local root since this process will never communicate with the real root. - The root is not the current process but will be in next level of the hierarchy. In this case the current process must be the root. - I am the root. The root is my index. Tested with IMB which rotates the root on every call to MPI_Reduce. Consider IMB the reproducer for the issue this commit solves. Make the bcast algorithm decision an enumerated variable Resolve various asset failures when destructing coll ml requests. Two issues: - Always reset the request to be invalid before returning it to the free list. This will avoid an asset in ompi_request_t's destructor. OMPI_REQUEST_FINI does this (and also releases the fortran handle index). - Never explicitly construct or destruct the superclass of an opal object. This screws up the class function tables and will cause either an assert failure or a segmentation fault when destructing coll ml requests. Cleanup allgather. I removed the duplicate non-blocking and blocking functions and modeled the cleanup after what I found in allreduce. Also cleaned up the code somewhat. Don't bother copying from the send to the recieve buffer in bcol_basesmuma_allreduce_intra_fanin_fanout if the pointers are the same. The eliminates a warning about memcpy and aliasing and avoids an unnecessary call to memcpy. Alwasy call CHECK_AND_RELEASE on memsync collectives. There was a call to OBJ_RELEASE on the collective communicator but because CHECK_AND_RECYLCE was never called there was not matching call to OBJ_RELEASE. This caused coll ml to leak communicators. Make allreduce use the sequential collective launcher in coll_ml_inlines.h Just launch the next collective in the component progress. I am a little unsure about this patch. There appears to be some sort of race between collectives that causes buffer exhaustion in some cases (IMB Allreduce is a reproducer). Changing progress to only launch the next bcol seems to resolve the issue but might not be the best fix. Note that I see little-no performance penalty for this change. Fix allreduce when there are extra sources. There was an issue with the buffer offset calculation when there are extra sources. In the case of extra sources == 1 the offset was set to buffer_size (just past the header of the next buffer). I adjusted the buffer size to take into accoun the maximum header size (see the earlier commit that added this) and simplified the offset calculation. Make reduce/allreduce non-blocking. This is required for MPI_Comm_idup to work correctly. This has been tested with various layouts using the ibm testsuite and imb and appears to have the same performance as the old blocking version. Fix allgather for non-contiguous layouts and simplify parsing the topology. Some things in this patch: - There were several comments to the effect that level 0 of the hierarchy MUST contain all of the ranks. At least one function made this assumption but it was not true. I changed the sbgp components and the coll ml initization code to enforce this requirement. - Ensure that hierarchy level 0 has the ranks in the correct scatter gather order. This removes the need for a separate sort list and fixes the offset calculation for allgather. - There were several passes over the hierarchy to determine properties of the hierarchy. I eliminated these extra passes and the memory allocation associated with them and calculate the tree properties on the fly. The same DFS recursion also handles the re-order of level 0. All these changes have been verified with MPI_Allreduce, MPI_Reduce, and MPI_Allgather. All functions now pass all IBM/Open MPI, and IMB tests. coll/ml: correct pointer usage for MPI_BOTTOM Since contiguous datatypes are copied via memcpy (bypassing the convertor) we need to adjust for the lb of the datatype. This corrects problems found testing code that uses MPI_BOTTOM (NULL) as the send pointer. Add fallback collectives for allreduce and reduce. cmr=v1.7.5:reviewer=pasha This commit was SVN r30363.
146 строки
4.7 KiB
C
146 строки
4.7 KiB
C
/*
|
|
* Copyright (c) 2009-2012 Oak Ridge National Laboratory. All rights reserved.
|
|
* Copyright (c) 2009-2012 Mellanox Technologies. All rights reserved.
|
|
* $COPYRIGHT$
|
|
*
|
|
* Additional copyrights may follow
|
|
*
|
|
* $HEADER$
|
|
*/
|
|
/** @file */
|
|
|
|
#include "ompi_config.h"
|
|
|
|
#include "ompi/constants.h"
|
|
#include "opal/threads/mutex.h"
|
|
#include "ompi/communicator/communicator.h"
|
|
#include "ompi/mca/coll/coll.h"
|
|
#include "ompi/mca/bcol/bcol.h"
|
|
#include "opal/sys/atomic.h"
|
|
#include "ompi/mca/coll/ml/coll_ml.h"
|
|
#include "ompi/mca/coll/ml/coll_ml_hier_algorithms.h"
|
|
|
|
coll_ml_collective_description_t *collective_op;
|
|
bcol_fn_args_t fn_arguments;
|
|
mca_coll_ml_descriptor_t *msg_desc;
|
|
|
|
static int coll_ml_setup_ibarrier_instance_recursive_doubling(
|
|
mca_coll_ml_descriptor_t *msg_desc,
|
|
coll_ml_collective_description_t *collective_op)
|
|
{
|
|
/* local variables */
|
|
ret=OMPI_SUCCESS;
|
|
int i_fn,cnt;
|
|
|
|
/* initialize function arguments */
|
|
|
|
/* mark all routines as not yet started - need this, so that
|
|
* when we try to progress the barrier, we know where to pickup
|
|
* when a function is called - MOVE this into the setup function.
|
|
*/
|
|
for(i_fn=0 ; i_fn < collective_op->n_functions ; i_fn++ ) {
|
|
msg_desc->fragment.fn_args[i_fn].function_status=FUNCTION_NOT_STARTED;
|
|
}
|
|
|
|
/* setup the fanin root */
|
|
for(i_fn=0 ; i_fn < collective_op->
|
|
alg_params.coll_fn.ibarrier_recursive_doubling.n_fanin_steps ;
|
|
i_fn++ ) {
|
|
mca_bcol_base_module_t *bcol_mod=
|
|
msg_desc->local_comm_description->functions[i_fn].bcol_module;
|
|
/* the lowest rank in the group will be the root */
|
|
msg_desc->fragment.fn_args[i_fn].root=0;
|
|
}
|
|
|
|
/* setup the fanout root */
|
|
cnt=alg_params.coll_fn.ibarrier_recursive_doubling.n_fanin_steps+
|
|
alg_params.coll_fn.ibarrier_recursive_doubling.n_recursive_doubling_steps;
|
|
for(i_fn=cnt ; i_fn < cnt + collective_op->
|
|
alg_params.coll_fn.ibarrier_recursive_doubling.n_fanin_steps ;
|
|
i_fn++ ) {
|
|
mca_bcol_base_module_t *bcol_mod=
|
|
msg_desc->local_comm_description->functions[i_fn].bcol_module;
|
|
/* the lowest rank in the group will be the root */
|
|
msg_desc->fragment.fn_args[i_fn].root=0;
|
|
}
|
|
|
|
/* successful completion */
|
|
return ret;
|
|
}
|
|
|
|
/**
|
|
* Hierarchical blocking barrier
|
|
*/
|
|
int mca_coll_ml_nb_barrier_intra( struct ompi_communicator_t *comm,
|
|
ompi_request_t ** request, mca_coll_base_module_t *module)
|
|
{
|
|
/* local variables */
|
|
int ret=OMPI_SUCCESS;
|
|
mca_coll_ml_module_t *ml_module;
|
|
uint64_t sequence_number;
|
|
int i_fn;
|
|
coll_ml_collective_description_t *collective_op;
|
|
bcol_fn_args_t fn_arguments;
|
|
mca_coll_ml_descriptor_t *msg_desc;
|
|
|
|
ml_module=(mca_coll_ml_module_t *) module;
|
|
/* debug */
|
|
fprintf(stderr," mca_coll_ml_nb_barrier_intra called \n");
|
|
fflush(stderr);
|
|
/* end debug */
|
|
|
|
/* grab full message descriptor - RLG: this is really not optimal,
|
|
* as we may be doing too much initialization if the collective
|
|
* routine completes on the first call to progress which is called
|
|
* within this routine. Need to see if we can be more efficient
|
|
* here. The current issue is that the only way that I can think
|
|
* to do this now is with two separate code paths, which I want to
|
|
* avoid at this stage.
|
|
*/
|
|
OMPI_FREE_LIST_GET(&(ml_module->message_descriptors),
|
|
msg_desc,ret);
|
|
if( OMPI_SUCCESS != ret) {
|
|
goto Error;
|
|
}
|
|
|
|
/* get message sequence number */
|
|
sequence_number=OPAL_THREAD_ADD64(
|
|
&(ml_module->no_data_collective_sequence_num),1);
|
|
fn_arguments.sequence_num=sequence_number;
|
|
|
|
|
|
/* get pointer to schedule - only one algorithm at this stage */
|
|
collective_op=&(ml_module->hierarchical_algorithms[BCOL_NB_BARRIER][0]);
|
|
|
|
/* call setup function - RLG: right now this is totally extra,
|
|
* but if we are going to have more than one algorithm,
|
|
* this is a better way to do this. */
|
|
coll_ml_setup_ibarrier_instance_recursive_doubling(
|
|
msg_desc,collective_op);
|
|
|
|
/* call the progress function to actually start the barrier */
|
|
|
|
/* recycle resources - RLG: need to think about this one */
|
|
|
|
#if 0
|
|
/* run barrier */
|
|
/* need to add bcol context for the call */
|
|
for( i_fn =0 ; i_fn < collective_op->n_functions ; i_fn++ ) {
|
|
mca_bcol_base_module_t *bcol_module=
|
|
collective_op->functions[i_fn].bcol_module;
|
|
/* for barrier, all we need is the group information that is
|
|
* captured in the bcol module
|
|
*/
|
|
ret=collective_op->functions[i_fn].fn(&fn_arguments,
|
|
NULL,NULL,bcol_module);
|
|
if( OMPI_SUCCESS != ret) {
|
|
} goto Error;
|
|
}
|
|
#endif
|
|
|
|
return OMPI_SUCCESS;
|
|
|
|
Error:
|
|
return ret;
|
|
}
|