1
1
openmpi/ompi/mca/coll/libnbc/nbc_igather.c
Torsten Hoefler 6b22641669 added LibNBC (http://www.unixer.de/NBC) as collv1 (blocking) component.
I know it does not make much sense but one can play around with the
performance. Numbers are available at http://www.unixer.de/research/nbcoll/perf/.
This is the first step towards collv2. Next step includes the addition
of non-blocking functions to the MPI-Layer and the collv1 interface.

It implements all MPI-1 collective algorithms in a non-blocking manner.
However, the collv1 interface does not allow non-blocking collectives so
that all collectives are used blocking by the ompi-glue layer.

I wanted to add LibNBC as a separate subdirectory, but I could not
convince the buildsystem (and had not the time). So the component looks
pretty messy. It would be great if somebody could explain me how to move
all nbc*{c,h}, and {hb,dict}*{c,h} to a seperate subdirectory.

It's .ompi_ignored because I did not test it exhaustively yet.

This commit was SVN r11401.
2006-08-24 16:47:18 +00:00

118 строки
4.1 KiB
C

#include "nbc.h"
#ifdef NBC_CACHE_SCHEDULE
/* tree comparison function for schedule cache */
int NBC_Gather_args_compare(NBC_Gather_args *a, NBC_Gather_args *b, void *param) {
if( (a->sendbuf == b->sendbuf) &&
(a->sendcount == b->sendcount) &&
(a->sendtype == b->sendtype) &&
(a->recvbuf == b->recvbuf) &&
(a->recvcount == b->recvcount) &&
(a->recvtype == b->recvtype) &&
(a->root == b->root) ) {
return 0;
}
if( a->sendbuf < b->sendbuf ) {
return -1;
}
return +1;
}
#endif
/* simple linear MPI_Igather */
int NBC_Igather(void* sendbuf, int sendcount, MPI_Datatype sendtype, void* recvbuf, int recvcount, MPI_Datatype recvtype, int root, MPI_Comm comm, NBC_Handle* handle) {
int rank, p, res, i;
MPI_Aint rcvext;
NBC_Schedule *schedule;
char *rbuf, inplace;
#ifdef NBC_CACHE_SCHEDULE
NBC_Gather_args *args, *found, search;
#endif
NBC_IN_PLACE(sendbuf, recvbuf, inplace);
res = NBC_Init_handle(handle, comm);
if(res != NBC_OK) { printf("Error in NBC_Init_handle(%i)\n", res); return res; }
res = MPI_Comm_rank(comm, &rank);
if (MPI_SUCCESS != res) { printf("MPI Error in MPI_Comm_rank() (%i)\n", res); return res; }
res = MPI_Comm_size(comm, &p);
if (MPI_SUCCESS != res) { printf("MPI Error in MPI_Comm_rank() (%i)\n", res); return res; }
res = MPI_Type_extent(recvtype, &rcvext);
if (MPI_SUCCESS != res) { printf("MPI Error in MPI_Type_extent() (%i)\n", res); return res; }
handle->tmpbuf = NULL;
if((rank == root) && (!inplace)) {
rbuf = ((char *)recvbuf) + (rank*recvcount*rcvext);
/* if I am the root - just copy the message (only without MPI_IN_PLACE) */
res = NBC_Copy(sendbuf, sendcount, sendtype, rbuf, recvcount, recvtype, comm);
if (NBC_OK != res) { printf("Error in NBC_Copy() (%i)\n", res); return res; }
}
#ifdef NBC_CACHE_SCHEDULE
/* search schedule in communicator specific tree */
search.sendbuf=sendbuf;
search.sendcount=sendcount;
search.sendtype=sendtype;
search.recvbuf=recvbuf;
search.recvcount=recvcount;
search.recvtype=recvtype;
search.root=root;
found = hb_tree_search(handle->comminfo->NBC_Dict[NBC_GATHER], &search);
if(found == NULL) {
#endif
schedule = malloc(sizeof(NBC_Schedule));
if (NULL == schedule) { printf("Error in malloc() (%i)\n", res); return res; }
res = NBC_Sched_create(schedule);
if(res != NBC_OK) { printf("Error in NBC_Sched_create (%i)\n", res); return res; }
/* send to root */
if(rank != root) {
/* send msg to root */
res = NBC_Sched_send(sendbuf, false, sendcount, sendtype, root, schedule);
if (NBC_OK != res) { printf("Error in NBC_Sched_send() (%i)\n", res); return res; }
} else {
for(i=0;i<p;i++) {
rbuf = ((char *)recvbuf) + (i*recvcount*rcvext);
if(i != root) {
/* root receives message to the right buffer */
res = NBC_Sched_recv(rbuf, false, recvcount, recvtype, i, schedule);
if (NBC_OK != res) { printf("Error in NBC_Sched_recv() (%i)\n", res); return res; }
}
}
}
res = NBC_Sched_commit(schedule);
if (NBC_OK != res) { printf("Error in NBC_Sched_commit() (%i)\n", res); return res; }
#ifdef NBC_CACHE_SCHEDULE
/* save schedule to tree */
args = malloc(sizeof(NBC_Gather_args));
args->sendbuf=sendbuf;
args->sendcount=sendcount;
args->sendtype=sendtype;
args->recvbuf=recvbuf;
args->recvcount=recvcount;
args->recvtype=recvtype;
args->root=root;
args->schedule=schedule;
res = hb_tree_insert (handle->comminfo->NBC_Dict[NBC_GATHER], args, args, 0);
if(res != 0) printf("error in dict_insert() (%i)\n", res);
/* increase number of elements for A2A */
if(++handle->comminfo->NBC_Dict_size[NBC_GATHER] > SCHED_DICT_UPPER) {
NBC_SchedCache_dictwipe(handle->comminfo->NBC_Dict[NBC_GATHER], &handle->comminfo->NBC_Dict_size[NBC_GATHER]);
}
} else {
/* found schedule */
schedule=found->schedule;
}
#endif
res = NBC_Start(handle, schedule);
if (NBC_OK != res) { printf("Error in NBC_Start() (%i)\n", res); return res; }
return NBC_OK;
}