Homec4science
Diffusion lammps (efficient_neuronet) Browse

lammps/src/KSPACE/remap.cpp
7875009218a7efficient_neuronet

Last Change
Raw File

remap.cpp
View Options

/* ----------------------------------------------------------------------
LAMMPS - Large-scale Atomic/Molecular Massively Parallel Simulator
http://lammps.sandia.gov, Sandia National Laboratories
Steve Plimpton, sjplimp@sandia.gov
Copyright (2003) Sandia Corporation. Under the terms of Contract
DE-AC04-94AL85000 with Sandia Corporation, the U.S. Government retains
certain rights in this software. This software is distributed under
the GNU General Public License.
See the README file in the top-level LAMMPS directory.
------------------------------------------------------------------------- */
#include <stdio.h>
#include <stdlib.h>
#include "remap.h"
#define PACK_DATA FFT_SCALAR
#include "pack.h"
#define MIN(A,B) ((A) < (B) ? (A) : (B))
#define MAX(A,B) ((A) > (B) ? (A) : (B))
/* ----------------------------------------------------------------------
Data layout for 3d remaps:
data set of Nfast x Nmid x Nslow elements is owned by P procs
each element = nqty contiguous datums
on input, each proc owns a subsection of the elements
on output, each proc will own a (presumably different) subsection
my subsection must not overlap with any other proc's subsection,
i.e. the union of all proc's input (or output) subsections must
exactly tile the global Nfast x Nmid x Nslow data set
when called from C, all subsection indices are
C-style from 0 to N-1 where N = Nfast or Nmid or Nslow
when called from F77, all subsection indices are
F77-style from 1 to N where N = Nfast or Nmid or Nslow
a proc can own 0 elements on input or output
by specifying hi index < lo index
on both input and output, data is stored contiguously on a processor
with a fast-varying, mid-varying, and slow-varying index
------------------------------------------------------------------------- */
/* ----------------------------------------------------------------------
Perform 3d remap
Arguments:
in starting address of input data on this proc
out starting address of where output data for this proc
will be placed (can be same as in)
buf extra memory required for remap
if memory=0 was used in call to remap_3d_create_plan
then buf must be big enough to hold output result
i.e. nqty * (out_ihi-out_ilo+1) * (out_jhi-out_jlo+1) *
(out_khi-out_klo+1)
if memory=1 was used in call to remap_3d_create_plan
then buf is not used, can just be a dummy pointer
plan plan returned by previous call to remap_3d_create_plan
------------------------------------------------------------------------- */
void remap_3d(FFT_SCALAR *in, FFT_SCALAR *out, FFT_SCALAR *buf,
struct remap_plan_3d *plan)
{
// use point-to-point communication
if (!plan->usecollective) {
int i,isend,irecv;
FFT_SCALAR *scratch;
if (plan->memory == 0)
scratch = buf;
else
scratch = plan->scratch;
// post all recvs into scratch space
for (irecv = 0; irecv < plan->nrecv; irecv++)
MPI_Irecv(&scratch[plan->recv_bufloc[irecv]],plan->recv_size[irecv],
MPI_FFT_SCALAR,plan->recv_proc[irecv],0,
plan->comm,&plan->request[irecv]);
// send all messages to other procs
for (isend = 0; isend < plan->nsend; isend++) {
plan->pack(&in[plan->send_offset[isend]],
plan->sendbuf,&plan->packplan[isend]);
MPI_Send(plan->sendbuf,plan->send_size[isend],MPI_FFT_SCALAR,
plan->send_proc[isend],0,plan->comm);
}
// copy in -> scratch -> out for self data
if (plan->self) {
isend = plan->nsend;
irecv = plan->nrecv;
plan->pack(&in[plan->send_offset[isend]],
&scratch[plan->recv_bufloc[irecv]],
&plan->packplan[isend]);
plan->unpack(&scratch[plan->recv_bufloc[irecv]],
&out[plan->recv_offset[irecv]],&plan->unpackplan[irecv]);
}
// unpack all messages from scratch -> out
for (i = 0; i < plan->nrecv; i++) {
MPI_Waitany(plan->nrecv,plan->request,&irecv,MPI_STATUS_IGNORE);
plan->unpack(&scratch[plan->recv_bufloc[irecv]],
&out[plan->recv_offset[irecv]],&plan->unpackplan[irecv]);
}
// use All2Allv collective for remap communication
} else {
if (plan->commringlen > 0) {
int isend,irecv;
// create send and recv buffers for alltoallv collective
int sendBufferSize = 0;
int recvBufferSize = 0;
for (int i=0;i<plan->nsend;i++)
sendBufferSize += plan->send_size[i];
for (int i=0;i<plan->nrecv;i++)
recvBufferSize += plan->recv_size[i];
FFT_SCALAR *packedSendBuffer
= (FFT_SCALAR *) malloc(sizeof(FFT_SCALAR) * sendBufferSize);
FFT_SCALAR *packedRecvBuffer
= (FFT_SCALAR *) malloc(sizeof(FFT_SCALAR) * recvBufferSize);
int *sendcnts = (int *) malloc(sizeof(int) * plan->commringlen);
int *rcvcnts = (int *) malloc(sizeof(int) * plan->commringlen);
int *sdispls = (int *) malloc(sizeof(int) * plan->commringlen);
int *rdispls = (int *) malloc(sizeof(int) * plan->commringlen);
int *nrecvmap = (int *) malloc(sizeof(int) * plan->commringlen);
// create and populate send data, count and displacement buffers
int currentSendBufferOffset = 0;
for (isend = 0; isend < plan->commringlen; isend++) {
sendcnts[isend] = 0;
sdispls[isend] = 0;
int foundentry = 0;
for (int i=0;(i<plan->nsend && !foundentry); i++) {
if (plan->send_proc[i] == plan->commringlist[isend]) {
foundentry = 1;
sendcnts[isend] = plan->send_size[i];
sdispls[isend] = currentSendBufferOffset;
plan->pack(&in[plan->send_offset[i]],
&packedSendBuffer[currentSendBufferOffset],
&plan->packplan[i]);
currentSendBufferOffset += plan->send_size[i];
}
}
}
// create and populate recv count and displacement buffers
int currentRecvBufferOffset = 0;
for (irecv = 0; irecv < plan->commringlen; irecv++) {
rcvcnts[irecv] = 0;
rdispls[irecv] = 0;
nrecvmap[irecv] = -1;
int foundentry = 0;
for (int i=0;(i<plan->nrecv && !foundentry); i++) {
if (plan->recv_proc[i] == plan->commringlist[irecv]) {
foundentry = 1;
rcvcnts[irecv] = plan->recv_size[i];
rdispls[irecv] = currentRecvBufferOffset;
currentRecvBufferOffset += plan->recv_size[i];
nrecvmap[irecv] = i;
}
}
}
MPI_Alltoallv(packedSendBuffer, sendcnts, sdispls,
MPI_FFT_SCALAR, packedRecvBuffer, rcvcnts,
rdispls, MPI_FFT_SCALAR, plan->comm);
// unpack the data from the recv buffer into out
currentRecvBufferOffset = 0;
for (irecv = 0; irecv < plan->commringlen; irecv++) {
if (nrecvmap[irecv] > -1) {
plan->unpack(&packedRecvBuffer[currentRecvBufferOffset],
&out[plan->recv_offset[nrecvmap[irecv]]],
&plan->unpackplan[nrecvmap[irecv]]);
currentRecvBufferOffset += plan->recv_size[nrecvmap[irecv]];
}
}
// free temporary data structures
free(sendcnts);
free(rcvcnts);
free(sdispls);
free(rdispls);
free(nrecvmap);
free(packedSendBuffer);
free(packedRecvBuffer);
}
}
}
/* ----------------------------------------------------------------------
Create plan for performing a 3d remap
Arguments:
comm MPI communicator for the P procs which own the data
in_ilo,in_ihi input bounds of data I own in fast index
in_jlo,in_jhi input bounds of data I own in mid index
in_klo,in_khi input bounds of data I own in slow index
out_ilo,out_ihi output bounds of data I own in fast index
out_jlo,out_jhi output bounds of data I own in mid index
out_klo,out_khi output bounds of data I own in slow index
nqty # of datums per element
permute permutation in storage order of indices on output
0 = no permutation
1 = permute once = mid->fast, slow->mid, fast->slow
2 = permute twice = slow->fast, fast->mid, mid->slow
memory user provides buffer memory for remap or system does
0 = user provides memory
1 = system provides memory
precision precision of data
1 = single precision (4 bytes per datum)
2 = double precision (8 bytes per datum)
usecollective whether to use collective MPI or point-to-point
------------------------------------------------------------------------- */
struct remap_plan_3d *remap_3d_create_plan(
MPI_Comm comm,
int in_ilo, int in_ihi, int in_jlo, int in_jhi,
int in_klo, int in_khi,
int out_ilo, int out_ihi, int out_jlo, int out_jhi,
int out_klo, int out_khi,
int nqty, int permute, int memory, int precision, int usecollective)
{
struct remap_plan_3d *plan;
struct extent_3d *inarray, *outarray;
struct extent_3d in,out,overlap;
int i,iproc,nsend,nrecv,ibuf,size,me,nprocs;
// query MPI info
MPI_Comm_rank(comm,&me);
MPI_Comm_size(comm,&nprocs);
// allocate memory for plan data struct
plan = (struct remap_plan_3d *) malloc(sizeof(struct remap_plan_3d));
if (plan == NULL) return NULL;
plan->usecollective = usecollective;
// store parameters in local data structs
in.ilo = in_ilo;
in.ihi = in_ihi;
in.isize = in.ihi - in.ilo + 1;
in.jlo = in_jlo;
in.jhi = in_jhi;
in.jsize = in.jhi - in.jlo + 1;
in.klo = in_klo;
in.khi = in_khi;
in.ksize = in.khi - in.klo + 1;
out.ilo = out_ilo;
out.ihi = out_ihi;
out.isize = out.ihi - out.ilo + 1;
out.jlo = out_jlo;
out.jhi = out_jhi;
out.jsize = out.jhi - out.jlo + 1;
out.klo = out_klo;
out.khi = out_khi;
out.ksize = out.khi - out.klo + 1;
// combine output extents across all procs
inarray = (struct extent_3d *) malloc(nprocs*sizeof(struct extent_3d));
if (inarray == NULL) return NULL;
outarray = (struct extent_3d *) malloc(nprocs*sizeof(struct extent_3d));
if (outarray == NULL) return NULL;
MPI_Allgather(&out,sizeof(struct extent_3d),MPI_BYTE,
outarray,sizeof(struct extent_3d),MPI_BYTE,comm);
// count send collides, including self
nsend = 0;
iproc = me;
for (i = 0; i < nprocs; i++) {
iproc++;
if (iproc == nprocs) iproc = 0;
nsend += remap_3d_collide(&in,&outarray[iproc],&overlap);
}
// malloc space for send info
if (nsend) {
plan->pack = pack_3d;
plan->send_offset = (int *) malloc(nsend*sizeof(int));
plan->send_size = (int *) malloc(nsend*sizeof(int));
plan->send_proc = (int *) malloc(nsend*sizeof(int));
plan->packplan = (struct pack_plan_3d *)
malloc(nsend*sizeof(struct pack_plan_3d));
if (plan->send_offset == NULL || plan->send_size == NULL ||
plan->send_proc == NULL || plan->packplan == NULL) return NULL;
}
// store send info, with self as last entry
nsend = 0;
iproc = me;
for (i = 0; i < nprocs; i++) {
iproc++;
if (iproc == nprocs) iproc = 0;
if (remap_3d_collide(&in,&outarray[iproc],&overlap)) {
plan->send_proc[nsend] = iproc;
plan->send_offset[nsend] = nqty *
((overlap.klo-in.klo)*in.jsize*in.isize +
((overlap.jlo-in.jlo)*in.isize + overlap.ilo-in.ilo));
plan->packplan[nsend].nfast = nqty*overlap.isize;
plan->packplan[nsend].nmid = overlap.jsize;
plan->packplan[nsend].nslow = overlap.ksize;
plan->packplan[nsend].nstride_line = nqty*in.isize;
plan->packplan[nsend].nstride_plane = nqty*in.jsize*in.isize;
plan->packplan[nsend].nqty = nqty;
plan->send_size[nsend] = nqty*overlap.isize*overlap.jsize*overlap.ksize;
nsend++;
}
}
// plan->nsend = # of sends not including self
if (nsend && plan->send_proc[nsend-1] == me) {
if (plan->usecollective) // for collectives include self in nsend list
plan->nsend = nsend;
else
plan->nsend = nsend - 1;
} else
plan->nsend = nsend;
// combine input extents across all procs
MPI_Allgather(&in,sizeof(struct extent_3d),MPI_BYTE,
inarray,sizeof(struct extent_3d),MPI_BYTE,comm);
// count recv collides, including self
nrecv = 0;
iproc = me;
for (i = 0; i < nprocs; i++) {
iproc++;
if (iproc == nprocs) iproc = 0;
nrecv += remap_3d_collide(&out,&inarray[iproc],&overlap);
}
// malloc space for recv info
if (nrecv) {
if (permute == 0)
plan->unpack = unpack_3d;
else if (permute == 1) {
if (nqty == 1)
plan->unpack = unpack_3d_permute1_1;
else if (nqty == 2)
plan->unpack = unpack_3d_permute1_2;
else
plan->unpack = unpack_3d_permute1_n;
}
else if (permute == 2) {
if (nqty == 1)
plan->unpack = unpack_3d_permute2_1;
else if (nqty == 2)
plan->unpack = unpack_3d_permute2_2;
else
plan->unpack = unpack_3d_permute2_n;
}
plan->recv_offset = (int *) malloc(nrecv*sizeof(int));
plan->recv_size = (int *) malloc(nrecv*sizeof(int));
plan->recv_proc = (int *) malloc(nrecv*sizeof(int));
plan->recv_bufloc = (int *) malloc(nrecv*sizeof(int));
plan->request = (MPI_Request *) malloc(nrecv*sizeof(MPI_Request));
plan->unpackplan = (struct pack_plan_3d *)
malloc(nrecv*sizeof(struct pack_plan_3d));
if (plan->recv_offset == NULL || plan->recv_size == NULL ||
plan->recv_proc == NULL || plan->recv_bufloc == NULL ||
plan->request == NULL || plan->unpackplan == NULL) return NULL;
}
// store recv info, with self as last entry
ibuf = 0;
nrecv = 0;
iproc = me;
for (i = 0; i < nprocs; i++) {
iproc++;
if (iproc == nprocs) iproc = 0;
if (remap_3d_collide(&out,&inarray[iproc],&overlap)) {
plan->recv_proc[nrecv] = iproc;
plan->recv_bufloc[nrecv] = ibuf;
if (permute == 0) {
plan->recv_offset[nrecv] = nqty *
((overlap.klo-out.klo)*out.jsize*out.isize +
(overlap.jlo-out.jlo)*out.isize + (overlap.ilo-out.ilo));
plan->unpackplan[nrecv].nfast = nqty*overlap.isize;
plan->unpackplan[nrecv].nmid = overlap.jsize;
plan->unpackplan[nrecv].nslow = overlap.ksize;
plan->unpackplan[nrecv].nstride_line = nqty*out.isize;
plan->unpackplan[nrecv].nstride_plane = nqty*out.jsize*out.isize;
plan->unpackplan[nrecv].nqty = nqty;
}
else if (permute == 1) {
plan->recv_offset[nrecv] = nqty *
((overlap.ilo-out.ilo)*out.ksize*out.jsize +
(overlap.klo-out.klo)*out.jsize + (overlap.jlo-out.jlo));
plan->unpackplan[nrecv].nfast = overlap.isize;
plan->unpackplan[nrecv].nmid = overlap.jsize;
plan->unpackplan[nrecv].nslow = overlap.ksize;
plan->unpackplan[nrecv].nstride_line = nqty*out.jsize;
plan->unpackplan[nrecv].nstride_plane = nqty*out.ksize*out.jsize;
plan->unpackplan[nrecv].nqty = nqty;
}
else {
plan->recv_offset[nrecv] = nqty *
((overlap.jlo-out.jlo)*out.isize*out.ksize +
(overlap.ilo-out.ilo)*out.ksize + (overlap.klo-out.klo));
plan->unpackplan[nrecv].nfast = overlap.isize;
plan->unpackplan[nrecv].nmid = overlap.jsize;
plan->unpackplan[nrecv].nslow = overlap.ksize;
plan->unpackplan[nrecv].nstride_line = nqty*out.ksize;
plan->unpackplan[nrecv].nstride_plane = nqty*out.isize*out.ksize;
plan->unpackplan[nrecv].nqty = nqty;
}
plan->recv_size[nrecv] = nqty*overlap.isize*overlap.jsize*overlap.ksize;
ibuf += plan->recv_size[nrecv];
nrecv++;
}
}
// create sub-comm rank list
if (plan->usecollective) {
plan->commringlist = NULL;
// merge recv and send rank lists
// ask Steve Plimpton about method to more accurately determine
// maximum number of procs contributing to pencil
int maxcommsize = nprocs;
int *commringlist = (int *) malloc(maxcommsize*sizeof(int));
int commringlen = 0;
for (int i = 0; i < nrecv; i++) {
commringlist[i] = plan->recv_proc[i];
commringlen++;
}
for (int i = 0; i < nsend; i++) {
int foundentry = 0;
for (int j=0;j<commringlen;j++)
if (commringlist[j] == plan->send_proc[i]) foundentry = 1;
if (!foundentry) {
commringlist[commringlen] = plan->send_proc[i];
commringlen++;
}
}
// sort initial commringlist
int swap = 0;
for (int c = 0 ; c < (commringlen - 1); c++) {
for (int d = 0 ; d < commringlen - c - 1; d++) {
if (commringlist[d] > commringlist[d+1]) {
swap = commringlist[d];
commringlist[d] = commringlist[d+1];
commringlist[d+1] = swap;
}
}
}
// collide all inarray extents for the comm ring with all output
// extents and all outarray extents for the comm ring with all input
// extents - if there is a collison add the rank to the comm ring,
// keep iterating until nothing is added to commring
int commringappend = 1;
while (commringappend) {
int newcommringlen = commringlen;
commringappend = 0;
for (int i=0;i<commringlen;i++) {
for (int j=0;j<nprocs;j++) {
if (remap_3d_collide(&inarray[commringlist[i]],
&outarray[j],&overlap)) {
int alreadyinlist = 0;
for (int k=0;k<newcommringlen;k++) {
if (commringlist[k] == j) {
alreadyinlist = 1;
}
}
if (!alreadyinlist) {
commringlist[newcommringlen++] = j;
commringappend = 1;
}
}
if (remap_3d_collide(&outarray[commringlist[i]],
&inarray[j],&overlap)) {
int alreadyinlist = 0;
for (int k=0;k<newcommringlen;k++) {
if (commringlist[k] == j) alreadyinlist = 1;
}
if (!alreadyinlist) {
commringlist[newcommringlen++] = j;
commringappend = 1;
}
}
}
}
commringlen = newcommringlen;
}
// sort the final commringlist
for (int c = 0 ; c < ( commringlen - 1 ); c++) {
for (int d = 0 ; d < commringlen - c - 1; d++) {
if (commringlist[d] > commringlist[d+1]) {
swap = commringlist[d];
commringlist[d] = commringlist[d+1];
commringlist[d+1] = swap;
}
}
}
// resize commringlist to final size
commringlist = (int *) realloc(commringlist, commringlen*sizeof(int));
// set the plan->commringlist
plan->commringlen = commringlen;
plan->commringlist = commringlist;
}
// plan->nrecv = # of recvs not including self
// for collectives include self in the nsend list
if (nrecv && plan->recv_proc[nrecv-1] == me) {
if (plan->usecollective) plan->nrecv = nrecv;
else plan->nrecv = nrecv - 1;
} else plan->nrecv = nrecv;
// init remaining fields in remap plan
plan->memory = memory;
if (nrecv == plan->nrecv) plan->self = 0;
else plan->self = 1;
// free locally malloced space
free(inarray);
free(outarray);
// find biggest send message (not including self) and malloc space for it
plan->sendbuf = NULL;
size = 0;
for (nsend = 0; nsend < plan->nsend; nsend++)
size = MAX(size,plan->send_size[nsend]);
if (size) {
plan->sendbuf = (FFT_SCALAR *) malloc(size*sizeof(FFT_SCALAR));
if (plan->sendbuf == NULL) return NULL;
}
// if requested, allocate internal scratch space for recvs,
// only need it if I will receive any data (including self)
plan->scratch = NULL;
if (memory == 1) {
if (nrecv > 0) {
plan->scratch =
(FFT_SCALAR *) malloc(nqty*out.isize*out.jsize*out.ksize *
sizeof(FFT_SCALAR));
if (plan->scratch == NULL) return NULL;
}
}
// if using collective and the commringlist is NOT empty create a
// communicator for the plan based off an MPI_Group created with
// ranks from the commringlist
if ((plan->usecollective && (plan->commringlen > 0))) {
MPI_Group orig_group, new_group;
MPI_Comm_group(comm, &orig_group);
MPI_Group_incl(orig_group, plan->commringlen,
plan->commringlist, &new_group);
MPI_Comm_create(comm, new_group, &plan->comm);
}
// if using collective and the comm ring list is empty create
// a communicator for the plan with an empty group
else if ((plan->usecollective) && (plan->commringlen == 0)) {
MPI_Comm_create(comm, MPI_GROUP_EMPTY, &plan->comm);
}
// not using collective - dup comm
else MPI_Comm_dup(comm,&plan->comm);
// return pointer to plan
return plan;
}
/* ----------------------------------------------------------------------
Destroy a 3d remap plan
------------------------------------------------------------------------- */
void remap_3d_destroy_plan(struct remap_plan_3d *plan)
{
// free MPI communicator
if (!((plan->usecollective) && (plan->commringlen == 0)))
MPI_Comm_free(&plan->comm);
if (plan->usecollective) {
if (plan->commringlist != NULL)
free(plan->commringlist);
}
// free internal arrays
if (plan->nsend || plan->self) {
free(plan->send_offset);
free(plan->send_size);
free(plan->send_proc);
free(plan->packplan);
if (plan->sendbuf) free(plan->sendbuf);
}
if (plan->nrecv || plan->self) {
free(plan->recv_offset);
free(plan->recv_size);
free(plan->recv_proc);
free(plan->recv_bufloc);
free(plan->request);
free(plan->unpackplan);
if (plan->scratch) free(plan->scratch);
}
// free plan itself
free(plan);
}
/* ----------------------------------------------------------------------
collide 2 sets of indices to determine overlap
compare bounds of block1 with block2 to see if they overlap
return 1 if they do and put bounds of overlapping section in overlap
return 0 if they do not overlap
------------------------------------------------------------------------- */
int remap_3d_collide(struct extent_3d *block1, struct extent_3d *block2,
struct extent_3d *overlap)
{
overlap->ilo = MAX(block1->ilo,block2->ilo);
overlap->ihi = MIN(block1->ihi,block2->ihi);
overlap->jlo = MAX(block1->jlo,block2->jlo);
overlap->jhi = MIN(block1->jhi,block2->jhi);
overlap->klo = MAX(block1->klo,block2->klo);
overlap->khi = MIN(block1->khi,block2->khi);
if (overlap->ilo > overlap->ihi ||
overlap->jlo > overlap->jhi ||
overlap->klo > overlap->khi) return 0;
overlap->isize = overlap->ihi - overlap->ilo + 1;
overlap->jsize = overlap->jhi - overlap->jlo + 1;
overlap->ksize = overlap->khi - overlap->klo + 1;
return 1;
}
c4science · Help