1
1
openmpi/ompi/mca/coll/libnbc/nbc_iscan.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

121 строка
4.4 KiB
C

#include "nbc.h"
#ifdef NBC_CACHE_SCHEDULE
/* tree comparison function for schedule cache */
int NBC_Scan_args_compare(NBC_Scan_args *a, NBC_Scan_args *b, void *param) {
if( (a->sendbuf == b->sendbuf) &&
(a->recvbuf == b->recvbuf) &&
(a->count == b->count) &&
(a->datatype == b->datatype) &&
(a->op == b->op) ) {
return 0;
}
if( a->sendbuf < b->sendbuf ) {
return -1;
}
return +1;
}
#endif
/* linear iscan
* working principle:
* 1. each node (but node 0) receives from left neigbor
* 2. performs op
* 3. all but rank p-1 do sends to it's right neigbor and exits
*
*/
int NBC_Iscan(void* sendbuf, void* recvbuf, int count, MPI_Datatype datatype, MPI_Op op, MPI_Comm comm, NBC_Handle* handle) {
int rank, p, res;
MPI_Aint ext;
NBC_Schedule *schedule;
#ifdef NBC_CACHE_SCHEDULE
NBC_Scan_args *args, *found, search;
#endif
char inplace;
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_size() (%i)\n", res); return res; }
res = MPI_Type_extent(datatype, &ext);
if (MPI_SUCCESS != res) { printf("MPI Error in MPI_Type_extent() (%i)\n", res); return res; }
handle->tmpbuf = malloc(ext*count);
if(handle->tmpbuf == NULL) { printf("Error in malloc()\n"); return NBC_OOR; }
if((rank == 0) && !inplace) {
/* copy data to receivebuf */
res = NBC_Copy(sendbuf, count, datatype, recvbuf, count, datatype, 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.recvbuf=recvbuf;
search.count=count;
search.datatype=datatype;
search.op=op;
found = hb_tree_search(handle->comminfo->NBC_Dict[NBC_SCAN], &search);
if(found == NULL) {
#endif
schedule = malloc(sizeof(NBC_Schedule));
if (NULL == schedule) { printf("Error in malloc()\n"); return res; }
res = NBC_Sched_create(schedule);
if(res != NBC_OK) { printf("Error in NBC_Sched_create (%i)\n", res); return res; }
if(rank != 0) {
res = NBC_Sched_recv(0, true, count, datatype, rank-1, schedule);
if (NBC_OK != res) { free(handle->tmpbuf); printf("Error in NBC_Sched_recv() (%i)\n", res); return res; }
/* we have to wait until we have the data */
res = NBC_Sched_barrier(schedule);
if (NBC_OK != res) { free(handle->tmpbuf); printf("Error in NBC_Sched_barrier() (%i)\n", res); return res; }
/* perform the reduce in my local buffer */
res = NBC_Sched_op(recvbuf, false, sendbuf, false, 0, true, count, datatype, op, schedule);
if (NBC_OK != res) { free(handle->tmpbuf); printf("Error in NBC_Sched_op() (%i)\n", res); return res; }
/* this cannot be done until handle->tmpbuf is unused :-( */
res = NBC_Sched_barrier(schedule);
if (NBC_OK != res) { free(handle->tmpbuf); printf("Error in NBC_Sched_barrier() (%i)\n", res); return res; }
}
if(rank != p-1) {
res = NBC_Sched_send(recvbuf, false, count, datatype, rank+1, schedule);
if (NBC_OK != res) { free(handle->tmpbuf); printf("Error in NBC_Sched_send() (%i)\n", res); return res; }
}
res = NBC_Sched_commit(schedule);
if (NBC_OK != res) { free(handle->tmpbuf); printf("Error in NBC_Sched_commit() (%i)\n", res); return res; }
#ifdef NBC_CACHE_SCHEDULE
/* save schedule to tree */
args = malloc(sizeof(NBC_Alltoall_args));
args->sendbuf=sendbuf;
args->recvbuf=recvbuf;
args->count=count;
args->datatype=datatype;
args->op=op;
args->schedule=schedule;
res = hb_tree_insert (handle->comminfo->NBC_Dict[NBC_SCAN], 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_SCAN] > SCHED_DICT_UPPER) {
NBC_SchedCache_dictwipe(handle->comminfo->NBC_Dict[NBC_SCAN], &handle->comminfo->NBC_Dict_size[NBC_SCAN]);
}
} else {
/* found schedule */
schedule=found->schedule;
}
#endif
res = NBC_Start(handle, schedule);
if (NBC_OK != res) { free(handle->tmpbuf); printf("Error in NBC_Start() (%i)\n", res); return res; }
/* tmpbuf is freed with the handle */
return NBC_OK;
}