ICODE-ADC/lammps
4
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
Files
1254d579f76074ac54efb9cd0ddb4372fc3f8be3
lammps/src/grid2d.cpp
Axel Kohlmeyer b459d0c9b9 apply clang-format
2024-07-22 23:58:10 -04:00

1957 lines
64 KiB
C++
Raw Blame History

/* ----------------------------------------------------------------------
LAMMPS - Large-scale Atomic/Molecular Massively Parallel Simulator
https://www.lammps.org/, Sandia National Laboratories
LAMMPS development team: developers@lammps.org
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 "grid2d.h"
#include "comm.h"
#include "domain.h"
#include "error.h"
#include "fix.h"
#include "irregular.h"
#include "kspace.h"
#include "math_extra.h"
#include "memory.h"
#include "pair.h"
#include <cstring>
using namespace LAMMPS_NS;
static constexpr int DELTA = 16;
static constexpr int OFFSET = 16384;
/* ----------------------------------------------------------------------
NOTES:
if o indices for ghosts are < 0 or hi indices are >= N,
then grid is treated as periodic in that dimension,
comm is done across the periodic boundaries
tiled implementations only work for RCB, not general tilings
b/c RCB tree is used to find neighboring tiles
------------------------------------------------------------------------- */
/* ----------------------------------------------------------------------
constructor to create a 2d distributed grid
Grid3d assigns owned/ghost cells to each proc via setup_grid()
it MUST be called after constructor
gcomm = caller's communicator
gnx,gny,gnz = global grid size
------------------------------------------------------------------------- */
Grid2d::Grid2d(LAMMPS *lmp, MPI_Comm gcomm, int gnx, int gny) :
Pointers(lmp), swap(nullptr), requests(nullptr), srequest(nullptr), rrequest(nullptr),
sresponse(nullptr), rresponse(nullptr), send(nullptr), recv(nullptr), copy(nullptr),
send_remap(nullptr), recv_remap(nullptr), overlap_procs(nullptr), xsplit(nullptr),
ysplit(nullptr), zsplit(nullptr), grid2proc(nullptr), rcbinfo(nullptr), overlap_list(nullptr)
{
gridcomm = gcomm;
MPI_Comm_rank(gridcomm, &me);
MPI_Comm_size(gridcomm, &nprocs);
nx = gnx;
ny = gny;
// default settings, can be overridden by set() methods
// these affect assignment of owned and ghost cells
maxdist = 0.0;
stencil_grid_lo = stencil_grid_hi = 0;
stencil_atom_lo = stencil_atom_hi = 0;
shift_grid = 0.5;
shift_atom_lo = shift_atom_hi = 0.0;
yextra = 0;
yfactor = 1.0;
}
/* ----------------------------------------------------------------------
alternate constructor to create a 2d distributed grid
caller assigns owned/ghost cells to each proc
setup_grid() must NOT be called
gcomm = caller's communicator
gnx,gny = global grid size
i xy lo/hi = extent of owned grid cells on this proc
o xy lo/hi = extent of owned+ghost grid cells on this proc
owned and ghost indices are inclusive
owned indices range from 0 to N-1
ghost indices can extend < 0 or >= N
------------------------------------------------------------------------- */
Grid2d::Grid2d(LAMMPS *lmp, MPI_Comm gcomm, int gnx, int gny, int ixlo, int ixhi, int iylo,
int iyhi, int oxlo, int oxhi, int oylo, int oyhi) :
Pointers(lmp), swap(nullptr), requests(nullptr), srequest(nullptr), rrequest(nullptr),
sresponse(nullptr), rresponse(nullptr), send(nullptr), recv(nullptr), copy(nullptr),
send_remap(nullptr), recv_remap(nullptr), overlap_procs(nullptr), xsplit(nullptr),
ysplit(nullptr), zsplit(nullptr), grid2proc(nullptr), rcbinfo(nullptr), overlap_list(nullptr)
{
gridcomm = gcomm;
MPI_Comm_rank(gridcomm, &me);
MPI_Comm_size(gridcomm, &nprocs);
nx = gnx;
ny = gny;
// store owned/ghost indices provided by caller
inxlo = ixlo;
inxhi = ixhi;
inylo = iylo;
inyhi = iyhi;
outxlo = oxlo;
outxhi = oxhi;
outylo = oylo;
outyhi = oyhi;
// additional intialization
// other constructor invokes this from setup_grid()
Grid2d::initialize();
}
/* ---------------------------------------------------------------------- */
Grid2d::~Grid2d()
{
// brick comm data structs
for (int i = 0; i < nswap; i++) {
memory->destroy(swap[i].packlist);
memory->destroy(swap[i].unpacklist);
}
memory->sfree(swap);
delete[] xsplit;
delete[] ysplit;
memory->destroy(grid2proc);
// tiled comm data structs
for (int i = 0; i < nsend; i++) memory->destroy(send[i].packlist);
memory->sfree(send);
for (int i = 0; i < nrecv; i++) memory->destroy(recv[i].unpacklist);
memory->sfree(recv);
for (int i = 0; i < ncopy; i++) {
memory->destroy(copy[i].packlist);
memory->destroy(copy[i].unpacklist);
}
memory->sfree(copy);
delete[] requests;
delete[] requests_remap;
memory->sfree(rcbinfo);
// remap data structs
deallocate_remap();
}
// ----------------------------------------------------------------------
// set Grid parameters
// ----------------------------------------------------------------------
/* ----------------------------------------------------------------------
maxdist = max distance outside proc subdomain a particle can be
used to determine extent of ghost cells
------------------------------------------------------------------------- */
void Grid2d::set_distance(double distance)
{
maxdist = distance;
}
/* ----------------------------------------------------------------------
# of grid cells beyond an owned grid cell that caller accesses
can be different in lo vs hi direction
------------------------------------------------------------------------- */
void Grid2d::set_stencil_grid(int lo, int hi)
{
stencil_grid_lo = lo;
stencil_grid_hi = hi;
}
/* ----------------------------------------------------------------------
# of grid cells beyond a particle's grid cell that caller accesses
can be different in lo vs hi direction
------------------------------------------------------------------------- */
void Grid2d::set_stencil_atom(int lo, int hi)
{
stencil_atom_lo = lo;
stencil_atom_hi = hi;
}
/* ----------------------------------------------------------------------
shift_grid = offset within grid cell of position of grid point
0.5 = cell center (default), 0.0 = lower-left corner of cell
used to decide which proc owns a grid cell (grid pt within subdomain)
------------------------------------------------------------------------- */
void Grid2d::set_shift_grid(double shift)
{
shift_grid = shift;
}
/* ----------------------------------------------------------------------
shift_atom = offset added to atoms when caller maps them to grid cells
0.5 = half a grid cell, 0.0 = no offset
used to compute maximum possible ghost extents
use of lo/hi allows max ghost extent on each side to be different
------------------------------------------------------------------------- */
void Grid2d::set_shift_atom(double shift_lo, double shift_hi)
{
shift_atom_lo = shift_lo;
shift_atom_hi = shift_hi;
}
/* ----------------------------------------------------------------------
enable extra grid cells in Y
factor = muliplication factor on box size Y and thus grid size
factor > 1.0 when grid extends beyond Y box size (3.0 = tripled in size)
only enable yextra if factor > 1.0
default zextra = 0, factor = 1.0 (no extra grid cells in Y)
------------------------------------------------------------------------- */
void Grid2d::set_yfactor(double factor)
{
if (factor == 1.0)
yextra = 0;
else
yextra = 1;
yfactor = factor;
}
/* ----------------------------------------------------------------------
set IDs of proc neighbors used in uniform local owned/ghost comm
must be called BEFORE setup_comm() to override
the processor neighbors stored by extract_comm_info()
------------------------------------------------------------------------- */
void Grid2d::set_proc_neighs(int pxlo, int pxhi, int pylo, int pyhi)
{
procxlo = pxlo;
procxhi = pxhi;
procylo = pylo;
procyhi = pyhi;
}
/* ----------------------------------------------------------------------
set allocation dimensions of caller grid used by indices() to setup pack/unpack
must be called BEFORE setup_comm() to override
the caller grid size used in indices()
------------------------------------------------------------------------- */
void Grid2d::set_caller_grid(int fxlo, int fxhi, int fylo, int fyhi)
{
fullxlo = fxlo;
fullxhi = fxhi;
fullylo = fylo;
fullyhi = fyhi;
}
// ----------------------------------------------------------------------
// retrieve Grid parameters
// ----------------------------------------------------------------------
int Grid2d::identical(Grid2d *grid2)
{
int inxlo2, inxhi2, inylo2, inyhi2;
int outxlo2, outxhi2, outylo2, outyhi2;
grid2->get_bounds_owned(inxlo2, inxhi2, inylo2, inyhi2);
grid2->get_bounds_ghost(outxlo2, outxhi2, outylo2, outyhi2);
int flag = 0;
if ((inxlo != inxlo2) || (inxhi != inxhi2) || (inylo != inylo2) || (inyhi != inyhi2)) flag = 1;
if ((outxlo != outxlo2) || (outxhi != outxhi2) || (outylo != outylo2) || (outyhi != outyhi2))
flag = 1;
int flagall;
MPI_Allreduce(&flag, &flagall, 1, MPI_INT, MPI_SUM, gridcomm);
if (flagall) return 0;
return 1;
}
/* ---------------------------------------------------------------------- */
void Grid2d::get_size(int &nxgrid, int &nygrid)
{
nxgrid = nx;
nygrid = ny;
}
/* ---------------------------------------------------------------------- */
void Grid2d::get_bounds_owned(int &xlo, int &xhi, int &ylo, int &yhi)
{
xlo = inxlo;
xhi = inxhi;
ylo = inylo;
yhi = inyhi;
}
/* ---------------------------------------------------------------------- */
void Grid2d::get_bounds_ghost(int &xlo, int &xhi, int &ylo, int &yhi)
{
xlo = outxlo;
xhi = outxhi;
ylo = outylo;
yhi = outyhi;
}
// ----------------------------------------------------------------------
// define owned and ghost grid cells
// also store comm and grid partitioning info
// ----------------------------------------------------------------------
/* ----------------------------------------------------------------------
setup grid partition for each proc = owned + ghost cells
return:
i xy lohi = portion of global grid this proc owns, 0 <= index < N
o xy lohi = owned + ghost grid cells in all directions
for periodic dims, o indices can be < 0 or >= N
for non-periodic dims, o indices will be >= 0 and < N
since no grid comm is done across non-periodic boundaries
------------------------------------------------------------------------- */
void Grid2d::setup_grid(int &ixlo, int &ixhi, int &iylo, int &iyhi, int &oxlo, int &oxhi, int &oylo,
int &oyhi)
{
// owned grid cells = those whose grid point is within proc subdomain
// shift_grid = 0.5 for grid point at cell center, 0.0 for lower-left corner
double fraclo, frachi;
if (comm->layout != Comm::LAYOUT_TILED) {
fraclo = comm->xsplit[comm->myloc[0]];
frachi = comm->xsplit[comm->myloc[0] + 1];
partition_grid(nx, fraclo, frachi, shift_grid, 0, inxlo, inxhi);
fraclo = comm->ysplit[comm->myloc[1]];
frachi = comm->ysplit[comm->myloc[1] + 1];
partition_grid(ny, fraclo, frachi, shift_grid, yextra, inylo, inyhi);
} else {
fraclo = comm->mysplit[0][0];
frachi = comm->mysplit[0][1];
partition_grid(nx, fraclo, frachi, shift_grid, 0, inxlo, inxhi);
fraclo = comm->mysplit[1][0];
frachi = comm->mysplit[1][1];
partition_grid(ny, fraclo, frachi, shift_grid, yextra, inylo, inyhi);
}
// extend owned grid bounds with ghost grid cells in each direction
ghost_grid();
// additional intialization
// other constructor invokes this directly
initialize();
// return values
ixlo = inxlo;
ixhi = inxhi;
iylo = inylo;
iyhi = inyhi;
oxlo = outxlo;
oxhi = outxhi;
oylo = outylo;
oyhi = outyhi;
}
/* ----------------------------------------------------------------------
additional one-time setup common to both constructors
---------------------------------------------------------------------- */
void Grid2d::initialize()
{
// error check on size of grid stored by this proc
bigint total = (bigint) (outxhi - outxlo + 1) * (outyhi - outylo + 1);
if (total > MAXSMALLINT) error->one(FLERR, "Too many owned+ghost grid2d points");
// default = caller grid is allocated to ghost grid
// used when computing pack/unpack lists in indices()
// these values can be overridden using set_caller_grid()
fullxlo = outxlo;
fullxhi = outxhi;
fullylo = outylo;
fullyhi = outyhi;
// initialize data structs
nswap = maxswap = 0;
swap = nullptr;
nsend = nrecv = ncopy = 0;
send = nullptr;
recv = nullptr;
copy = nullptr;
requests = nullptr;
requests_remap = nullptr;
xsplit = ysplit = zsplit = nullptr;
grid2proc = nullptr;
rcbinfo = nullptr;
nsend_remap = nrecv_remap = self_remap = 0;
send_remap = nullptr;
recv_remap = nullptr;
// store info about Comm decomposition needed for remap operation
// two Grid instances will exist for duration of remap
// each must know Comm decomp at time Grid instance was created
extract_comm_info();
}
/* ----------------------------------------------------------------------
partition a global regular grid into one brick-shaped sub-grid per proc
if grid point is inside my sub-domain I own it,
this includes sub-domain lo boundary but excludes hi boundary
ngrid = extent of global grid in a dimension
indices into the global grid range from 0 to Ngrid-1 in that dim
shift determines position of grid pt within grid cell
shift = 0.5 for cell center, 0.0 for lower-left corner
extra = 0 if grid exactly covers the simulation box
extra = 1 if grid extends beyond the +y boundary by yfactor
effectively maps proc partitions to the box-size subset of the grid
lo/hi = inclusive lo/hi bounds for brick of global grid cells I own
lo grid index = first grid pt >= fraclo*bound
hi grid index = last grid pt < frachi*bound
if proc owns no grid cells in a dim, then inlo > inhi
special case: 2 procs share boundary which a grid point is exactly on
2 if test equalties ensure a consistent decision as to which proc owns it
------------------------------------------------------------------------- */
void Grid2d::partition_grid(int ngrid, double fraclo, double frachi, double shift, int extra,
int &lo, int &hi)
{
if (extra == 0) {
lo = static_cast<int>(fraclo * ngrid);
while (lo + shift < fraclo * ngrid) lo++;
hi = static_cast<int>(frachi * ngrid);
while (hi + shift >= frachi * ngrid) hi--;
} else {
lo = static_cast<int>(fraclo * ngrid / yfactor);
while (lo + shift < fraclo * ngrid / yfactor) lo++;
hi = static_cast<int>(frachi * ngrid / yfactor);
while (hi + shift >= frachi * ngrid / yfactor) hi--;
}
}
// clang-format off
/* ----------------------------------------------------------------------
extend ghost grid cells in each direction beyond owned grid
indices into the global grid range from 0 to N-1 in each dim
ghost cell indices for periodic systems can be < 0 or >= N
------------------------------------------------------------------------- */
void Grid2d::ghost_grid()
{
double *prd,*boxlo,*sublo,*subhi;
int triclinic = domain->triclinic;
if (triclinic == 0) {
prd = domain->prd;
boxlo = domain->boxlo;
sublo = domain->sublo;
subhi = domain->subhi;
} else {
prd = domain->prd_lamda;
boxlo = domain->boxlo_lamda;
sublo = domain->sublo_lamda;
subhi = domain->subhi_lamda;
}
// for triclinic, maxdist = different value in each orthogonal direction
double dist[3] = {0.0,0.0,0.0};
if (triclinic == 0) dist[0] = dist[1] = dist[2] = maxdist;
else MathExtra::tribbox(domain->h,maxdist,&dist[0]);
// lo/hi = min/max index of global grid cells my owned atoms can be mapped to
// includes effects of maxdist and shift_atom settings
// lo/hi can be further extended by stencil_atom and stencil_grid settings
// all those settings are set by caller
// ghost cell layers needed in each dim/dir = max of two extension effects
// OFFSET allows generation of negative indices with static_cast
// out xyz lo/hi = index range of owned + ghost cells
// if yextra, ny and effective prd[1] are both larger, so dyinv is the same
double dxinv = nx / prd[0];
double dyinv = ny / prd[1];
if (yextra) dyinv = ny / (prd[1] * yfactor);
int lo, hi;
lo = static_cast<int>((sublo[0]-dist[0]-boxlo[0]) * dxinv + shift_atom_lo + OFFSET) - OFFSET;
hi = static_cast<int>((subhi[0]+dist[0]-boxlo[0]) * dxinv + shift_atom_hi + OFFSET) - OFFSET;
outxlo = MIN(lo - stencil_atom_lo, inxlo - stencil_grid_lo);
outxhi = MAX(hi + stencil_atom_hi, inxhi + stencil_grid_hi);
lo = static_cast<int>((sublo[1]-dist[1]-boxlo[1]) * dyinv + shift_atom_lo + OFFSET) - OFFSET;
hi = static_cast<int>((subhi[1]+dist[1]-boxlo[1]) * dyinv + shift_atom_hi + OFFSET) - OFFSET;
outylo = MIN(lo - stencil_atom_lo, inylo - stencil_grid_lo);
outyhi = MAX(hi + stencil_atom_hi, inyhi + stencil_grid_hi);
// if yextra = 1:
// adjust grid boundaries for processors at +y end,
// to include added empty grid cells between periodically repeating slabs
// in this case:
// want grid data forward communicated from +y proc to -y proc, but not vice versa
// want grid data reverse communicated from -y proc to +y proc, but not vice versa
// this is accomplished by inyhi = outyhi on +y end (no ghost cells)
// also ensure no other procs use ghost cells beyond +y limit
if (yextra) {
if (comm->layout != Comm::LAYOUT_TILED) {
if (comm->myloc[1] == comm->procgrid[1]-1) inyhi = outyhi = ny - 1;
} else {
if (comm->mysplit[1][1] == 1.0) inyhi = outyhi = ny - 1;
}
outyhi = MIN(outyhi,ny-1);
}
// limit out xyz lo/hi indices to global grid for non-periodic dims
// if yextra = 1, treat y-dimension as if periodic
int *periodicity = domain->periodicity;
if (!periodicity[0]) {
outxlo = MAX(0,outxlo);
outxhi = MIN(nx-1,outxhi);
}
if (!periodicity[1] && !yextra) {
outylo = MAX(0,outylo);
outyhi = MIN(ny-1,outyhi);
}
}
/* ----------------------------------------------------------------------
store copy of info from Comm class about processor partitioning
used when a remap is performed between two Grid instances, old and new
------------------------------------------------------------------------- */
void Grid2d::extract_comm_info()
{
// for non TILED layout:
// proc xyz lohi = my 4 neighbor procs in this MPI_Comm
// these proc IDs can be overridden by caller using set_proc_neighs()
// xyz split = copy of 1d vectors in Comm
// grid2proc = copy of 3d array in Comm
if (comm->layout != Comm::LAYOUT_TILED) {
procxlo = comm->procneigh[0][0];
procxhi = comm->procneigh[0][1];
procylo = comm->procneigh[1][0];
procyhi = comm->procneigh[1][1];
xsplit = new double[comm->procgrid[0]+1];
ysplit = new double[comm->procgrid[1]+1];
memcpy(xsplit, comm->xsplit, sizeof(double) * (comm->procgrid[0]+1));
memcpy(ysplit, comm->ysplit, sizeof(double) * (comm->procgrid[1]+1));
memory->create(grid2proc,comm->procgrid[0],comm->procgrid[1],comm->procgrid[2],
"grid3d:grid2proc");
memcpy(&grid2proc[0][0][0],&comm->grid2proc[0][0][0],
sizeof(int) * comm->procgrid[0] * comm->procgrid[1] * comm->procgrid[2]);
}
// for TILED layout:
// create RCB tree of grid partitioning info for grid decomp
// Comm provides dim info for this proc, stored as RCBinfo.dim
// dim is -1 for proc 0, but never accessed
// RCBinfo.cut = this proc's inlo in that dim
// Allgather creates the tree of dims and cuts
if (comm->layout == Comm::LAYOUT_TILED) {
rcbinfo = (RCBinfo *)
memory->smalloc(nprocs*sizeof(RCBinfo),"grid3d:rcbinfo");
RCBinfo rcbone;
rcbone.dim = comm->rcbcutdim;
if (rcbone.dim <= 0) rcbone.cut = inxlo;
else if (rcbone.dim == 1) rcbone.cut = inylo;
MPI_Allgather(&rcbone,sizeof(RCBinfo),MPI_CHAR,
rcbinfo,sizeof(RCBinfo),MPI_CHAR,gridcomm);
}
}
// ----------------------------------------------------------------------
// setup of local owned/ghost grid comm
// ----------------------------------------------------------------------
/* ----------------------------------------------------------------------
setup communication of owned/ghost grid cells
either for brick decomp or tiled decomp
return sizes of two buffers needed for communication
nbuf1 = largest pack or unpack in any Send or Recv or Copy
nbuf2 = larget of sum of all packs or unpacks in Send or Recv
for brick comm, nbuf1 = nbuf2
for tiling comm, nbuf2 >= nbuf2
nbuf1,nbuf2 are counts of grid cells
caller converts them to message sizes for grid data it stores
------------------------------------------------------------------------- */
void Grid2d::setup_comm(int &nbuf1, int &nbuf2)
{
if (comm->layout != Comm::LAYOUT_TILED) setup_comm_brick(nbuf1,nbuf2);
else setup_comm_tiled(nbuf1,nbuf2);
}
/* ----------------------------------------------------------------------
setup owned/ghost comm for brick comm
each proc has 4 neighbors
comm pattern = series of swaps with one of those 4 procs
can be multiple swaps with same proc if ghost extent is large
swap may not be symmetric if both procs do not need same layers of ghosts
all procs perform same # of swaps in a direction, even if some don't need it
------------------------------------------------------------------------- */
void Grid2d::setup_comm_brick(int &nbuf1, int &nbuf2)
{
int nsent,sendfirst,sendlast,recvfirst,recvlast;
int sendplanes,recvplanes;
int notdoneme,notdone;
// notify 4 neighbor procs how many ghost grid planes I need from them
// ghost xy lo = # of my lower grid planes that proc xy lo needs as its ghosts
// ghost xy hi = # of my upper grid planes that proc xy hi needs as its ghosts
// if this proc is its own neighbor across periodic bounary, value is from self
int nplanes = inxlo - outxlo;
if (procxlo != me)
MPI_Sendrecv(&nplanes,1,MPI_INT,procxlo,0,
&ghostxhi,1,MPI_INT,procxhi,0,gridcomm,MPI_STATUS_IGNORE);
else ghostxhi = nplanes;
nplanes = outxhi - inxhi;
if (procxhi != me)
MPI_Sendrecv(&nplanes,1,MPI_INT,procxhi,0,
&ghostxlo,1,MPI_INT,procxlo,0,gridcomm,MPI_STATUS_IGNORE);
else ghostxlo = nplanes;
nplanes = inylo - outylo;
if (procylo != me)
MPI_Sendrecv(&nplanes,1,MPI_INT,procylo,0,
&ghostyhi,1,MPI_INT,procyhi,0,gridcomm,MPI_STATUS_IGNORE);
else ghostyhi = nplanes;
nplanes = outyhi - inyhi;
if (procyhi != me)
MPI_Sendrecv(&nplanes,1,MPI_INT,procyhi,0,
&ghostylo,1,MPI_INT,procylo,0,gridcomm,MPI_STATUS_IGNORE);
else ghostylo = nplanes;
// setup swaps = exchange of grid data with one of 6 neighobr procs
// can be more than one in a direction if ghost region extends beyond neigh proc
// all procs have same swap count, but swapsize npack/nunpack can be empty
nswap = 0;
// send own grid pts to -x processor, recv ghost grid pts from +x processor
nsent = 0;
sendfirst = inxlo;
sendlast = inxhi;
recvfirst = inxhi+1;
notdone = 1;
while (notdone) {
if (nswap == maxswap) grow_swap();
swap[nswap].sendproc = procxlo;
swap[nswap].recvproc = procxhi;
sendplanes = MIN(sendlast-sendfirst+1,ghostxlo-nsent);
swap[nswap].npack =
indices(swap[nswap].packlist,
sendfirst,sendfirst+sendplanes-1,inylo,inyhi);
if (procxlo != me)
MPI_Sendrecv(&sendplanes,1,MPI_INT,procxlo,0,
&recvplanes,1,MPI_INT,procxhi,0,gridcomm,MPI_STATUS_IGNORE);
else recvplanes = sendplanes;
swap[nswap].nunpack =
indices(swap[nswap].unpacklist,
recvfirst,recvfirst+recvplanes-1,inylo,inyhi);
nsent += sendplanes;
sendfirst += sendplanes;
sendlast += recvplanes;
recvfirst += recvplanes;
nswap++;
if (nsent < ghostxlo) notdoneme = 1;
else notdoneme = 0;
MPI_Allreduce(&notdoneme,&notdone,1,MPI_INT,MPI_SUM,gridcomm);
}
// send own grid pts to +x processor, recv ghost grid pts from -x processor
nsent = 0;
sendfirst = inxlo;
sendlast = inxhi;
recvlast = inxlo-1;
notdone = 1;
while (notdone) {
if (nswap == maxswap) grow_swap();
swap[nswap].sendproc = procxhi;
swap[nswap].recvproc = procxlo;
sendplanes = MIN(sendlast-sendfirst+1,ghostxhi-nsent);
swap[nswap].npack =
indices(swap[nswap].packlist,
sendlast-sendplanes+1,sendlast,inylo,inyhi);
if (procxhi != me)
MPI_Sendrecv(&sendplanes,1,MPI_INT,procxhi,0,
&recvplanes,1,MPI_INT,procxlo,0,gridcomm,MPI_STATUS_IGNORE);
else recvplanes = sendplanes;
swap[nswap].nunpack =
indices(swap[nswap].unpacklist,
recvlast-recvplanes+1,recvlast,inylo,inyhi);
nsent += sendplanes;
sendfirst -= recvplanes;
sendlast -= sendplanes;
recvlast -= recvplanes;
nswap++;
if (nsent < ghostxhi) notdoneme = 1;
else notdoneme = 0;
MPI_Allreduce(&notdoneme,&notdone,1,MPI_INT,MPI_SUM,gridcomm);
}
// send own grid pts to -y processor, recv ghost grid pts from +y processor
nsent = 0;
sendfirst = inylo;
sendlast = inyhi;
recvfirst = inyhi+1;
notdone = 1;
while (notdone) {
if (nswap == maxswap) grow_swap();
swap[nswap].sendproc = procylo;
swap[nswap].recvproc = procyhi;
sendplanes = MIN(sendlast-sendfirst+1,ghostylo-nsent);
swap[nswap].npack =
indices(swap[nswap].packlist,
outxlo,outxhi,sendfirst,sendfirst+sendplanes-1);
if (procylo != me)
MPI_Sendrecv(&sendplanes,1,MPI_INT,procylo,0,
&recvplanes,1,MPI_INT,procyhi,0,gridcomm,MPI_STATUS_IGNORE);
else recvplanes = sendplanes;
swap[nswap].nunpack =
indices(swap[nswap].unpacklist,
outxlo,outxhi,recvfirst,recvfirst+recvplanes-1);
nsent += sendplanes;
sendfirst += sendplanes;
sendlast += recvplanes;
recvfirst += recvplanes;
nswap++;
if (nsent < ghostylo) notdoneme = 1;
else notdoneme = 0;
MPI_Allreduce(&notdoneme,&notdone,1,MPI_INT,MPI_SUM,gridcomm);
}
// send own grid pts to +y processor, recv ghost grid pts from -y processor
nsent = 0;
sendfirst = inylo;
sendlast = inyhi;
recvlast = inylo-1;
notdone = 1;
while (notdone) {
if (nswap == maxswap) grow_swap();
swap[nswap].sendproc = procyhi;
swap[nswap].recvproc = procylo;
sendplanes = MIN(sendlast-sendfirst+1,ghostyhi-nsent);
swap[nswap].npack =
indices(swap[nswap].packlist,
outxlo,outxhi,sendlast-sendplanes+1,sendlast);
if (procyhi != me)
MPI_Sendrecv(&sendplanes,1,MPI_INT,procyhi,0,
&recvplanes,1,MPI_INT,procylo,0,gridcomm,MPI_STATUS_IGNORE);
else recvplanes = sendplanes;
swap[nswap].nunpack =
indices(swap[nswap].unpacklist,
outxlo,outxhi,recvlast-recvplanes+1,recvlast);
nsent += sendplanes;
sendfirst -= recvplanes;
sendlast -= sendplanes;
recvlast -= recvplanes;
nswap++;
if (nsent < ghostyhi) notdoneme = 1;
else notdoneme = 0;
MPI_Allreduce(&notdoneme,&notdone,1,MPI_INT,MPI_SUM,gridcomm);
}
// ngrid = max of any forward/reverse pack/unpack grid points
int ngrid = 0;
for (int i = 0; i < nswap; i++) {
ngrid = MAX(ngrid,swap[i].npack);
ngrid = MAX(ngrid,swap[i].nunpack);
}
nbuf1 = nbuf2 = ngrid;
}
/* ----------------------------------------------------------------------
setup owned/ghost comm for tiled comm
each proc has arbitrary # of neighbors that overlap its ghost extent
identify which procs will send me ghost cells, and vice versa
may not be symmetric if both procs do not need same layers of ghosts
comm pattern = post recvs for all my ghosts, send my owned, wait on recvs
no exchanges by dimension, unlike CommTiled forward/reverse comm of particles
------------------------------------------------------------------------- */
void Grid2d::setup_comm_tiled(int &nbuf1, int &nbuf2)
{
int i,m;
double xlo,xhi,ylo,yhi;
int ghostbox[4],pbc[2];
// find overlaps of my extended ghost box with all owned boxes
// accounts for ghost box overlapping periodic boundaries
// noverlap = # of overlaps, including self
// overlap = vector of overlap info using Overlap data struct
ghostbox[0] = outxlo;
ghostbox[1] = outxhi;
ghostbox[2] = outylo;
ghostbox[3] = outyhi;
pbc[0] = pbc[1] = 0;
Overlap *overlap;
int noverlap = compute_overlap(1,ghostbox,pbc,overlap);
// send each proc an overlap message
// content: me, index of my overlap, box that overlaps with its owned cells
// ncopy = # of overlaps with myself, across a periodic boundary
// skip copy to self when non-PBC
int *proclist;
memory->create(proclist,noverlap,"grid2d:proclist");
srequest = (Request *)
memory->smalloc(noverlap*sizeof(Request),"grid2d:srequest");
int nsend_request = 0;
ncopy = 0;
for (m = 0; m < noverlap; m++) {
if (overlap[m].proc == me) {
if (overlap[m].pbc[0] == 0 && overlap[m].pbc[1] == 0) continue;
ncopy++;
} else {
proclist[nsend_request] = overlap[m].proc;
srequest[nsend_request].sender = me;
srequest[nsend_request].index = m;
for (i = 0; i < 4; i++)
srequest[nsend_request].box[i] = overlap[m].box[i];
nsend_request++;
}
}
auto irregular = new Irregular(lmp);
int nrecv_request = irregular->create_data(nsend_request,proclist,1);
auto rrequest = (Request *) memory->smalloc(nrecv_request*sizeof(Request),"grid2d:rrequest");
irregular->exchange_data((char *) srequest,sizeof(Request),(char *) rrequest);
irregular->destroy_data();
// compute overlaps between received ghost boxes and my owned box
// overlap box used to setup my Send data struct and respond to requests
send = (Send *) memory->smalloc(nrecv_request*sizeof(Send),"grid2d:send");
sresponse = (Response *) memory->smalloc(nrecv_request*sizeof(Response),"grid2d:sresponse");
memory->destroy(proclist);
memory->create(proclist,nrecv_request,"grid2d:proclist");
for (m = 0; m < nrecv_request; m++) {
send[m].proc = rrequest[m].sender;
xlo = MAX(rrequest[m].box[0],inxlo);
xhi = MIN(rrequest[m].box[1],inxhi);
ylo = MAX(rrequest[m].box[2],inylo);
yhi = MIN(rrequest[m].box[3],inyhi);
send[m].npack = indices(send[m].packlist,xlo,xhi,ylo,yhi);
proclist[m] = rrequest[m].sender;
sresponse[m].index = rrequest[m].index;
sresponse[m].box[0] = xlo;
sresponse[m].box[1] = xhi;
sresponse[m].box[2] = ylo;
sresponse[m].box[3] = yhi;
}
nsend = nrecv_request;
// reply to each Request message with a Response message
// content: index for the overlap on requestor, overlap box on my owned grid
int nsend_response = nrecv_request;
int nrecv_response = irregular->create_data(nsend_response,proclist,1);
auto rresponse = (Response *) memory->smalloc(nrecv_response*sizeof(Response),"grid2d:rresponse");
irregular->exchange_data((char *) sresponse,sizeof(Response),(char *) rresponse);
irregular->destroy_data();
delete irregular;
// process received responses
// box used to setup my Recv data struct after unwrapping via PBC
// adjacent = 0 if any box of ghost cells does not adjoin my owned cells
recv = (Recv *) memory->smalloc(nrecv_response*sizeof(Recv),"grid2d:recv");
adjacent = 1;
for (i = 0; i < nrecv_response; i++) {
m = rresponse[i].index;
recv[i].proc = overlap[m].proc;
xlo = rresponse[i].box[0] + overlap[m].pbc[0] * nx;
xhi = rresponse[i].box[1] + overlap[m].pbc[0] * nx;
ylo = rresponse[i].box[2] + overlap[m].pbc[1] * ny;
yhi = rresponse[i].box[3] + overlap[m].pbc[1] * ny;
recv[i].nunpack = indices(recv[i].unpacklist,xlo,xhi,ylo,yhi);
if (xlo != inxhi+1 && xhi != inxlo-1 &&
ylo != inyhi+1 && yhi != inylo-1) adjacent = 0;
}
nrecv = nrecv_response;
// create Copy data struct from overlaps with self
// skip copy to self when non-PBC
copy = (Copy *) memory->smalloc(ncopy*sizeof(Copy),"grid2d:copy");
ncopy = 0;
for (m = 0; m < noverlap; m++) {
if (overlap[m].proc != me) continue;
if (overlap[m].pbc[0] == 0 && overlap[m].pbc[1] == 0) continue;
xlo = overlap[m].box[0];
xhi = overlap[m].box[1];
ylo = overlap[m].box[2];
yhi = overlap[m].box[3];
copy[ncopy].npack = indices(copy[ncopy].packlist,xlo,xhi,ylo,yhi);
xlo = overlap[m].box[0] + overlap[m].pbc[0] * nx;
xhi = overlap[m].box[1] + overlap[m].pbc[0] * nx;
ylo = overlap[m].box[2] + overlap[m].pbc[1] * ny;
yhi = overlap[m].box[3] + overlap[m].pbc[1] * ny;
copy[ncopy].nunpack = indices(copy[ncopy].unpacklist,xlo,xhi,ylo,yhi);
ncopy++;
}
// set offsets for received data
int offset = 0;
for (m = 0; m < nsend; m++) {
send[m].offset = offset;
offset += send[m].npack;
}
offset = 0;
for (m = 0; m < nrecv; m++) {
recv[m].offset = offset;
offset += recv[m].nunpack;
}
// length of MPI requests vector is max of nsend, nrecv
int nrequest = MAX(nsend,nrecv);
requests = new MPI_Request[nrequest];
// clean-up
clean_overlap();
memory->destroy(proclist);
memory->sfree(srequest);
memory->sfree(rrequest);
memory->sfree(sresponse);
memory->sfree(rresponse);
// nbuf1 = largest pack or unpack in any Send or Recv or Copy
// nbuf2 = larget of sum of all packs or unpacks in Send or Recv
nbuf1 = 0;
for (m = 0; m < ncopy; m++) {
nbuf1 = MAX(nbuf1,copy[m].npack);
nbuf1 = MAX(nbuf1,copy[m].nunpack);
}
int nbufs = 0;
for (m = 0; m < nsend; m++) {
nbuf1 = MAX(nbuf1,send[m].npack);
nbufs += send[m].npack;
}
int nbufr = 0;
for (m = 0; m < nrecv; m++) {
nbuf1 = MAX(nbuf1,recv[m].nunpack);
nbufr += recv[m].nunpack;
}
nbuf2 = MAX(nbufs,nbufr);
}
// ----------------------------------------------------------------------
// query locality of forwrd/reverse grid comm
// ----------------------------------------------------------------------
/* ----------------------------------------------------------------------
check if all procs only need ghost info from adjacent procs
return 1 if yes, 0 if no
------------------------------------------------------------------------- */
int Grid2d::ghost_adjacent()
{
if (comm->layout != Comm::LAYOUT_TILED) return ghost_adjacent_brick();
return ghost_adjacent_tiled();
}
/* ----------------------------------------------------------------------
adjacent = 0 if a proc's ghost xy lohi values exceed its subdomain size
return 0 if adjacent=0 for any proc, else 1
------------------------------------------------------------------------- */
int Grid2d::ghost_adjacent_brick()
{
adjacent = 1;
if (ghostxlo > inxhi-inxlo+1) adjacent = 0;
if (ghostxhi > inxhi-inxlo+1) adjacent = 0;
if (ghostylo > inyhi-inylo+1) adjacent = 0;
if (ghostyhi > inyhi-inylo+1) adjacent = 0;
int adjacent_all;
MPI_Allreduce(&adjacent,&adjacent_all,1,MPI_INT,MPI_MIN,gridcomm);
return adjacent_all;
}
/* ----------------------------------------------------------------------
adjacent = 0 if a proc's received ghosts were flagged
as non-adjacent in setup_tiled()
return 0 if adjacent=0 for any proc, else 1
------------------------------------------------------------------------- */
int Grid2d::ghost_adjacent_tiled()
{
int adjacent_all;
MPI_Allreduce(&adjacent,&adjacent_all,1,MPI_INT,MPI_MIN,gridcomm);
return adjacent_all;
}
// ----------------------------------------------------------------------
// forward/reverse comm of owned/ghost grid data via callbacks
// ----------------------------------------------------------------------
/* ----------------------------------------------------------------------
forward comm of my owned cells to other's ghost cells
------------------------------------------------------------------------- */
void Grid2d::forward_comm(int caller, void *ptr, int which, int nper, int nbyte,
void *buf1, void *buf2, MPI_Datatype datatype)
{
if (comm->layout != Comm::LAYOUT_TILED) {
if (caller == KSPACE)
forward_comm_brick<KSpace>((KSpace *) ptr,which,nper,nbyte,
buf1,buf2,datatype);
else if (caller == PAIR)
forward_comm_brick<Pair>((Pair *) ptr,which,nper,nbyte,
buf1,buf2,datatype);
else if (caller == FIX)
forward_comm_brick<Fix>((Fix *) ptr,which,nper,nbyte,
buf1,buf2,datatype);
} else {
if (caller == KSPACE)
forward_comm_tiled<KSpace>((KSpace *) ptr,which,nper,nbyte,
buf1,buf2,datatype);
else if (caller == PAIR)
forward_comm_tiled<Pair>((Pair *) ptr,which,nper,nbyte,
buf1,buf2,datatype);
else if (caller == FIX)
forward_comm_tiled<Fix>((Fix *) ptr,which,nper,nbyte,
buf1,buf2,datatype);
}
}
/* ----------------------------------------------------------------------
forward comm for brick decomp via list of swaps with 6 neighbor procs
------------------------------------------------------------------------- */
template < class T >
void Grid2d::
forward_comm_brick(T *ptr, int which, int nper, int /*nbyte*/,
void *buf1, void *buf2, MPI_Datatype datatype)
{
int m;
MPI_Request request;
for (m = 0; m < nswap; m++) {
if (swap[m].sendproc == me)
ptr->pack_forward_grid(which,buf2,swap[m].npack,swap[m].packlist);
else
ptr->pack_forward_grid(which,buf1,swap[m].npack,swap[m].packlist);
if (swap[m].sendproc != me) {
if (swap[m].nunpack) MPI_Irecv(buf2,nper*swap[m].nunpack,datatype,
swap[m].recvproc,0,gridcomm,&request);
if (swap[m].npack) MPI_Send(buf1,nper*swap[m].npack,datatype,
swap[m].sendproc,0,gridcomm);
if (swap[m].nunpack) MPI_Wait(&request,MPI_STATUS_IGNORE);
}
ptr->unpack_forward_grid(which,buf2,swap[m].nunpack,swap[m].unpacklist);
}
}
/* ----------------------------------------------------------------------
forward comm for tiled decomp via Send/Recv lists of each neighbor proc
------------------------------------------------------------------------- */
template < class T >
void Grid2d::
forward_comm_tiled(T *ptr, int which, int nper, int nbyte,
void *buf1, void *vbuf2, MPI_Datatype datatype)
{
int i,m,offset;
auto buf2 = (char *) vbuf2;
// post all receives
for (m = 0; m < nrecv; m++) {
offset = nper * recv[m].offset * nbyte;
MPI_Irecv((void *) &buf2[offset],nper*recv[m].nunpack,datatype,
recv[m].proc,0,gridcomm,&requests[m]);
}
// perform all sends to other procs
for (m = 0; m < nsend; m++) {
ptr->pack_forward_grid(which,buf1,send[m].npack,send[m].packlist);
MPI_Send(buf1,nper*send[m].npack,datatype,send[m].proc,0,gridcomm);
}
// perform all copies to self
for (m = 0; m < ncopy; m++) {
ptr->pack_forward_grid(which,buf1,copy[m].npack,copy[m].packlist);
ptr->unpack_forward_grid(which,buf1,copy[m].nunpack,copy[m].unpacklist);
}
// unpack all received data
for (i = 0; i < nrecv; i++) {
MPI_Waitany(nrecv,requests,&m,MPI_STATUS_IGNORE);
offset = nper * recv[m].offset * nbyte;
ptr->unpack_forward_grid(which,(void *) &buf2[offset],
recv[m].nunpack,recv[m].unpacklist);
}
}
/* ----------------------------------------------------------------------
reverse comm of my ghost cells to sum to owner cells
------------------------------------------------------------------------- */
void Grid2d::reverse_comm(int caller, void *ptr, int which, int nper, int nbyte,
void *buf1, void *buf2, MPI_Datatype datatype)
{
if (comm->layout != Comm::LAYOUT_TILED) {
if (caller == KSPACE)
reverse_comm_brick<KSpace>((KSpace *) ptr,which,nper,nbyte,
buf1,buf2,datatype);
else if (caller == PAIR)
reverse_comm_brick<Pair>((Pair *) ptr,which,nper,nbyte,
buf1,buf2,datatype);
else if (caller == FIX)
reverse_comm_brick<Fix>((Fix *) ptr,which,nper,nbyte,
buf1,buf2,datatype);
} else {
if (caller == KSPACE)
reverse_comm_tiled<KSpace>((KSpace *) ptr,which,nper,nbyte,
buf1,buf2,datatype);
else if (caller == PAIR)
reverse_comm_tiled<Pair>((Pair *) ptr,which,nper,nbyte,
buf1,buf2,datatype);
else if (caller == FIX)
reverse_comm_tiled<Fix>((Fix *) ptr,which,nper,nbyte,
buf1,buf2,datatype);
}
}
/* ----------------------------------------------------------------------
reverse comm for brick decomp via list of swaps with 6 neighbor procs
------------------------------------------------------------------------- */
template < class T >
void Grid2d::
reverse_comm_brick(T *ptr, int which, int nper, int /*nbyte*/,
void *buf1, void *buf2, MPI_Datatype datatype)
{
int m;
MPI_Request request;
for (m = nswap-1; m >= 0; m--) {
if (swap[m].recvproc == me)
ptr->pack_reverse_grid(which,buf2,swap[m].nunpack,swap[m].unpacklist);
else
ptr->pack_reverse_grid(which,buf1,swap[m].nunpack,swap[m].unpacklist);
if (swap[m].recvproc != me) {
if (swap[m].npack) MPI_Irecv(buf2,nper*swap[m].npack,datatype,
swap[m].sendproc,0,gridcomm,&request);
if (swap[m].nunpack) MPI_Send(buf1,nper*swap[m].nunpack,datatype,
swap[m].recvproc,0,gridcomm);
if (swap[m].npack) MPI_Wait(&request,MPI_STATUS_IGNORE);
}
ptr->unpack_reverse_grid(which,buf2,swap[m].npack,swap[m].packlist);
}
}
/* ----------------------------------------------------------------------
reverse comm for tiled decomp via Send/Recv lists of each neighbor proc
------------------------------------------------------------------------- */
template < class T >
void Grid2d::
reverse_comm_tiled(T *ptr, int which, int nper, int nbyte,
void *buf1, void *vbuf2, MPI_Datatype datatype)
{
int i,m,offset;
auto buf2 = (char *) vbuf2;
// post all receives
for (m = 0; m < nsend; m++) {
offset = nper * send[m].offset * nbyte;
MPI_Irecv((void *) &buf2[offset],nper*send[m].npack,datatype,
send[m].proc,0,gridcomm,&requests[m]);
}
// perform all sends to other procs
for (m = 0; m < nrecv; m++) {
ptr->pack_reverse_grid(which,buf1,recv[m].nunpack,recv[m].unpacklist);
MPI_Send(buf1,nper*recv[m].nunpack,datatype,recv[m].proc,0,gridcomm);
}
// perform all copies to self
for (m = 0; m < ncopy; m++) {
ptr->pack_reverse_grid(which,buf1,copy[m].nunpack,copy[m].unpacklist);
ptr->unpack_reverse_grid(which,buf1,copy[m].npack,copy[m].packlist);
}
// unpack all received data
for (i = 0; i < nsend; i++) {
MPI_Waitany(nsend,requests,&m,MPI_STATUS_IGNORE);
offset = nper * send[m].offset * nbyte;
ptr->unpack_reverse_grid(which,(void *) &buf2[offset],
send[m].npack,send[m].packlist);
}
}
// ----------------------------------------------------------------------
// remap comm between 2 old/new grid decomposition of owned grid data
// ----------------------------------------------------------------------
/* ----------------------------------------------------------------------
setup remap from old grid decomposition to this grid decomposition
return sizes of two buffers needed for communication
either for brick decomp or tiling decomp
nbuf1 = largest pack or unpack in any Send or Recv or Copy
nbuf2 = larget of sum of all packs or unpacks in Send or Recv
for brick comm, nbuf1 = nbuf2
for tiled comm, nbuf2 >= nbuf2
nbuf1,nbuf2 are just count of grid points
caller converts them to message size for grid data it stores
------------------------------------------------------------------------- */
void Grid2d::setup_remap(Grid2d *old, int &nremap_buf1, int &nremap_buf2)
{
int m;
int pbc[2];
int *box;
// deallocated existing remap data structs
deallocate_remap();
// overlaps of my old decomp owned box with all owned boxes in new decomp
// noverlap_old = # of overlaps, including self
// overlap_old = vector of overlap info in Overlap data struct
int oldbox[4];
old->get_bounds_owned(oldbox[0],oldbox[1],oldbox[2],oldbox[3]);
pbc[0] = pbc[1] = 0;
Overlap *overlap_old;
int noverlap_old = compute_overlap(0,oldbox,pbc,overlap_old);
// use overlap_old to construct send and copy lists
// skip overlaps that contain no grid cells
self_remap = 0;
nsend_remap = 0;
for (m = 0; m < noverlap_old; m++) {
box = overlap_old[m].box;
if (box[0] > box[1] || box[2] > box[3]) continue;
if (overlap_old[m].proc == me) self_remap =1;
else nsend_remap++;
}
send_remap = new Send[nsend_remap];
nsend_remap = 0;
for (m = 0; m < noverlap_old; m++) {
box = overlap_old[m].box;
if (box[0] > box[1] || box[2] > box[3]) continue;
if (overlap_old[m].proc == me) {
copy_remap.npack =
old->indices(copy_remap.packlist,box[0],box[1],box[2],box[3]);
} else {
send_remap[nsend_remap].proc = overlap_old[m].proc;
send_remap[nsend_remap].npack =
old->indices(send_remap[nsend_remap].packlist,
box[0],box[1],box[2],box[3]);
nsend_remap++;
}
}
// overlaps of my new decomp owned box with all owned boxes in old decomp
// noverlap_new = # of overlaps, including self
// overlap_new = vector of overlap info in Overlap data struct
int newbox[4];
get_bounds_owned(newbox[0],newbox[1],newbox[2],newbox[3]);
pbc[0] = pbc[1] = 0;
Overlap *overlap_new;
int noverlap_new = old->compute_overlap(0,newbox,pbc,overlap_new);
// use overlap_new to construct recv and copy lists
// skip overlaps that contain no grid cells
// set offsets for Recv data
nrecv_remap = 0;
for (m = 0; m < noverlap_new; m++) {
box = overlap_new[m].box;
if (box[0] > box[1] || box[2] > box[3]) continue;
if (overlap_new[m].proc != me) nrecv_remap++;
}
recv_remap = new Recv[nrecv_remap];
nrecv_remap = 0;
for (m = 0; m < noverlap_new; m++) {
box = overlap_new[m].box;
if (box[0] > box[1] || box[2] > box[3]) continue;
if (overlap_new[m].proc == me) {
copy_remap.nunpack =
indices(copy_remap.unpacklist,box[0],box[1],box[2],box[3]);
} else {
recv_remap[nrecv_remap].proc = overlap_new[m].proc;
recv_remap[nrecv_remap].nunpack =
indices(recv_remap[nrecv_remap].unpacklist,
box[0],box[1],box[2],box[3]);
nrecv_remap++;
}
}
// set offsets for received data
int offset = 0;
for (m = 0; m < nrecv_remap; m++) {
recv_remap[m].offset = offset;
offset += recv_remap[m].nunpack;
}
// length of MPI requests vector = nrecv_remap
delete [] requests_remap;
requests_remap = new MPI_Request[nrecv_remap];
// clean-up
clean_overlap();
old->clean_overlap();
// nremap_buf1 = largest pack or unpack in any Send or Recv or Copy
// nremap_buf2 = sum of all unpacks in Recv
nremap_buf1 = 0;
if (self_remap) {
nremap_buf1 = MAX(nremap_buf1,copy_remap.npack);
nremap_buf1 = MAX(nremap_buf1,copy_remap.nunpack);
}
for (m = 0; m < nsend_remap; m++)
nremap_buf1 = MAX(nremap_buf1,send_remap[m].npack);
nremap_buf2 = 0;
for (m = 0; m < nrecv_remap; m++) {
nremap_buf1 = MAX(nremap_buf1,recv_remap[m].nunpack);
nremap_buf2 += recv_remap[m].nunpack;
}
}
/* ----------------------------------------------------------------------
perform remap from old grid decomposition to this grid decomposition
pack/unpack operations are performed by caller via callbacks
------------------------------------------------------------------------- */
void Grid2d::remap(int caller, void *ptr, int which, int nper, int nbyte,
void *buf1, void *buf2, MPI_Datatype datatype)
{
if (caller == FIX) remap_style<Fix>((Fix *) ptr,which,nper,nbyte,buf1,buf2,datatype);
}
/* ------------------------------------------------------------------------- */
template < class T >
void Grid2d::remap_style(T *ptr, int which, int nper, int nbyte,
void *buf1, void *vbuf2, MPI_Datatype datatype)
{
int i,m,offset;
auto buf2 = (char *) vbuf2;
// post all receives
for (m = 0; m < nrecv_remap; m++) {
offset = nper * recv_remap[m].offset * nbyte;
MPI_Irecv((void *) &buf2[offset],nper*recv_remap[m].nunpack,datatype,
recv_remap[m].proc,0,gridcomm,&requests_remap[m]);
}
// perform all sends to other procs
for (m = 0; m < nsend_remap; m++) {
ptr->pack_remap_grid(which,buf1,send_remap[m].npack,send_remap[m].packlist);
MPI_Send(buf1,nper*send_remap[m].npack,datatype,send_remap[m].proc,0,gridcomm);
}
// perform remap to self if defined
if (self_remap) {
ptr->pack_remap_grid(which,buf1,copy_remap.npack,copy_remap.packlist);
ptr->unpack_remap_grid(which,buf1,copy_remap.nunpack,copy_remap.unpacklist);
}
// unpack all received data
for (i = 0; i < nrecv_remap; i++) {
MPI_Waitany(nrecv_remap,requests_remap,&m,MPI_STATUS_IGNORE);
offset = nper * recv_remap[m].offset * nbyte;
ptr->unpack_remap_grid(which,(void *) &buf2[offset],
recv_remap[m].nunpack,recv_remap[m].unpacklist);
}
}
// ----------------------------------------------------------------------
// grid I/O methods
// ----------------------------------------------------------------------
/* ----------------------------------------------------------------------
read grid values from a file
------------------------------------------------------------------------- */
void Grid2d::read_file(int caller, void *ptr, FILE *fp, int nchunk, int maxline)
{
if (caller == FIX)
read_file_style<Fix>((Fix *) ptr,fp,nchunk,maxline);
}
/* ----------------------------------------------------------------------
proc 0 reads one chunk of lines at a time from file
broadcast chunk buffer to other procs
call back to caller so it can process the chunk of lines
caller returns count of grid-value lines in chunk
------------------------------------------------------------------------- */
template < class T >
void Grid2d::read_file_style(T *ptr, FILE *fp, int nchunk, int maxline)
{
auto buffer = new char[nchunk * maxline];
bigint ntotal = (bigint) nx * ny;
bigint nread = 0;
while (nread < ntotal) {
int mychunk = MIN(ntotal - nread, nchunk);
int eof = utils::read_lines_from_file(fp, mychunk, maxline, buffer, me, world);
if (eof) error->all(FLERR, "Unexpected end of grid data file");
nread += ptr->unpack_read_grid(nchunk,buffer);
}
delete [] buffer;
}
/* ----------------------------------------------------------------------
write grid values to a file
------------------------------------------------------------------------- */
void Grid2d::write_file(int caller, void *ptr, int which,
int nper, int nbyte, MPI_Datatype datatype)
{
if (caller == FIX)
write_file_style<Fix>((Fix *) ptr, which, nper, nbyte, datatype);
}
/* ----------------------------------------------------------------------
proc 0 reads one chunk of lines at a time from file
broadcast chunk buffer to other procs
call back to caller so it can process the chunk of lines
caller returns count of grid-value lines in chunk
------------------------------------------------------------------------- */
template < class T >
void Grid2d::write_file_style(T *ptr, int which,
int nper, int nbyte, MPI_Datatype datatype)
{
// maxsize = max size of grid data owned by any proc
int mysize = (inxhi-inxlo+1) * (inyhi-inylo+1);
mysize *= nper;
int maxsize;
MPI_Allreduce(&mysize,&maxsize,1,MPI_INT,MPI_MAX,world);
// pack my grid data via callback to caller
char *onebuf;
if (me == 0) memory->create(onebuf,maxsize*nbyte,"grid3d:onebuf");
else memory->create(onebuf,mysize*nbyte,"grid3d:onebuf");
ptr->pack_write_grid(which,onebuf);
// ping each proc for its grid data
// call back to caller with each proc's grid data
int tmp;
int bounds[4];
if (me == 0) {
MPI_Status status;
MPI_Request request;
for (int iproc = 0; iproc < nprocs; iproc++) {
if (iproc) {
MPI_Irecv(onebuf,maxsize,datatype,iproc,0,world,&request);
MPI_Send(&tmp,0,MPI_INT,iproc,0,world);
MPI_Wait(&request,&status);
MPI_Recv(bounds,4,MPI_INT,iproc,0,world,&status);
} else {
bounds[0] = inxlo;
bounds[1] = inxhi;
bounds[2] = inylo;
bounds[3] = inyhi;
}
ptr->unpack_write_grid(which,onebuf,bounds);
}
} else {
MPI_Recv(&tmp,0,MPI_INT,0,0,world,MPI_STATUS_IGNORE);
MPI_Rsend(onebuf,mysize,datatype,0,0,world);
bounds[0] = inxlo;
bounds[1] = inxhi;
bounds[2] = inylo;
bounds[3] = inyhi;
MPI_Send(bounds,4,MPI_INT,0,0,world);
}
// clean up
memory->destroy(onebuf);
}
// ----------------------------------------------------------------------
// overlap methods for brick and tiled RCB decompositions
// overlap = overlap of owned or owned+ghost box with all boxes of a decomposition
// for owned/ghost grid comm, called only by tiled decomposition
// brick decomp uses one or more comm passes with neigh procs
// like forward/reverse comm for atoms
// for remap, called by both brick and tiled decompositions
// ----------------------------------------------------------------------
/* ----------------------------------------------------------------------
compute list of overlaps between box and the owned grid boxes of all procs
ghostflag = 1 if box includes ghost grid pts, called by setup_tiled()
ghostflag = 0 if box has no ghost grid pts, called by setup_remap()
layout != LAYOUT_TILED is only invoked by setup_remap()
for brick decomp of Grid, done using xyz split + grid2proc copied from Comm
for tiled decomp of Grid, done via recursive box drop on RCB tree
box = 4 integers = (xlo,xhi,ylo,yhi)
box can be owned cells or owned + ghost cells
pbc = flags for grid periodicity in each dim
if box includes ghost cells, it can overlap PBCs (only for setup_tiled)
each lo/hi value may extend beyond 0 to N-1 into another periodic image
return # of overlaps including with self, caller handles self overlaps as needed
return list of overlaps
for setup_tiled() this is what box_drop() computes
entire box for each overlap
caller will determine extent of overlap using PBC info
for setup_remap(), return extent of overlap (no PBC info involved)
use proc_box_uniform() or tiled() and MAX/MIN to determine this
------------------------------------------------------------------------- */
int Grid2d::compute_overlap(int ghostflag, int *box, int *pbc, Overlap *&overlap)
{
int obox[4];
memory->create(overlap_procs,nprocs,"grid2d:overlap_procs");
noverlap_list = maxoverlap_list = 0;
overlap_list = nullptr;
// skip overlap check if box contains no grid cells
if (box[0] > box[1] || box[2] > box[3]) {
overlap = overlap_list;
return noverlap_list;
}
// test obox against appropriate layout
if (comm->layout != Comm::LAYOUT_TILED) {
// find comm->procgrid indices in each dim for box bounds
int iproclo = proc_index_uniform(box[0],nx,shift_grid,0,xsplit);
int iprochi = proc_index_uniform(box[1],nx,shift_grid,0,xsplit);
int jproclo = proc_index_uniform(box[2],ny,shift_grid,1,ysplit);
int jprochi = proc_index_uniform(box[3],ny,shift_grid,1,ysplit);
// compute extent of overlap of box with with each proc's obox
for (int j = jproclo; j <= jprochi; j++)
for (int i = iproclo; i <= iprochi; i++) {
partition_grid(nx,xsplit[i],xsplit[i+1],shift_grid,0,obox[0],obox[1]);
partition_grid(ny,ysplit[j],ysplit[j+1],shift_grid,yextra,obox[2],obox[3]);
if (noverlap_list == maxoverlap_list) grow_overlap();
overlap_list[noverlap_list].proc = grid2proc[i][j][0];
overlap_list[noverlap_list].box[0] = MAX(box[0],obox[0]);
overlap_list[noverlap_list].box[1] = MIN(box[1],obox[1]);
overlap_list[noverlap_list].box[2] = MAX(box[2],obox[2]);
overlap_list[noverlap_list].box[3] = MIN(box[3],obox[3]);
noverlap_list++;
}
} else {
box_drop(box,pbc);
// compute extent of overlap of box with with each proc's obox
if (ghostflag == 0) {
for (int m = 0; m < noverlap_list; m++) {
obox[0] = 0;
obox[1] = nx-1;
obox[2] = 0;
obox[3] = ny-1;
partition_tiled(overlap_list[m].proc,0,nprocs-1,obox);
overlap_list[m].box[0] = MAX(box[0],obox[0]);
overlap_list[m].box[1] = MIN(box[1],obox[1]);
overlap_list[m].box[2] = MAX(box[2],obox[2]);
overlap_list[m].box[3] = MIN(box[3],obox[3]);
}
}
}
overlap = overlap_list;
return noverlap_list;
}
/* ----------------------------------------------------------------------
deallocate data created by recursive overlap computation
------------------------------------------------------------------------- */
void Grid2d::clean_overlap()
{
memory->destroy(overlap_procs);
memory->sfree(overlap_list);
}
/* ----------------------------------------------------------------------
recursively split a box until it doesn't overlap any periodic boundaries
box = 4 integers = (xlo,xhi,ylo,yhi)
each lo/hi value may extend beyonw 0 to N-1 into another periodic image
pbc = flags in each dim of which periodic image the caller box was in
when a box straddles a periodic bounadry, split it in two
when a box does not straddle, drop it down RCB tree
add all the procs it overlaps with to Overlap list
------------------------------------------------------------------------- */
void Grid2d::box_drop(int *box, int *pbc)
{
int i,m;
// newbox12 and newpbc are initially copies of caller box and pbc
int newbox1[4],newbox2[4],newpbc[2];
for (i = 0; i < 4; i++) newbox1[i] = newbox2[i] = box[i];
for (i = 0; i < 2; i++) newpbc[i] = pbc[i];
// 4 if tests to see if box needs to be split across a periodic boundary
// newbox1 and 2 = new split boxes, newpbc increments current pbc
// final else is no split
int splitflag = 1;
if (box[0] < 0) {
newbox1[0] = 0;
newbox2[0] = box[0] + nx;
newbox2[1] = nx - 1;
newpbc[0]--;
} else if (box[1] >= nx) {
newbox1[1] = nx - 1;
newbox2[0] = 0;
newbox2[1] = box[1] - nx;
newpbc[0]++;
} else if (box[2] < 0) {
newbox1[2] = 0;
newbox2[2] = box[2] + ny;
newbox2[3] = ny - 1;
newpbc[1]--;
} else if (box[3] >= ny) {
newbox1[3] = ny - 1;
newbox2[2] = 0;
newbox2[3] = box[3] - ny;
newpbc[1]++;
// box is not split, drop on RCB tree
// returns np = # of procs it overlaps, including self
// returns proc_overlap = list of proc IDs it overlaps
// add each overlap to overlap list
} else {
splitflag = 0;
int np = 0;
box_drop_grid(box,0,nprocs-1,np,overlap_procs);
for (m = 0; m < np; m++) {
if (noverlap_list == maxoverlap_list) grow_overlap();
overlap_list[noverlap_list].proc = overlap_procs[m];
for (i = 0; i < 4; i++) overlap_list[noverlap_list].box[i] = box[i];
for (i = 0; i < 2; i++) overlap_list[noverlap_list].pbc[i] = pbc[i];
noverlap_list++;
}
}
// recurse with 2 split boxes
if (splitflag) {
box_drop(newbox1,pbc);
box_drop(newbox2,newpbc);
}
}
/* ----------------------------------------------------------------------
recursively drop a box down the RCB tree to find all procs it overlaps with
box = 4 integers = (xlo,xhi,ylo,yhi)
each lo/hi value ranges from 0 to N-1 in a dim, N = grid size in that dim
box is guaranteed to be wholly within the global domain
return Np = # of procs, plist = proc IDs
------------------------------------------------------------------------- */
void Grid2d::box_drop_grid(int *box, int proclower, int procupper,
int &np, int *plist)
{
// end recursion when partition is a single proc
// add proclower to plist
if (proclower == procupper) {
plist[np++] = proclower;
return;
}
// drop box on each side of cut it extends beyond
// use < and >= criteria so does not include a box it only touches
// procmid = 1st processor in upper half of partition
// = location in tree that stores this cut
// cut = index of first grid cell in upper partition
// dim = 0,1,2 dimension of cut
int procmid = proclower + (procupper - proclower) / 2 + 1;
int dim = rcbinfo[procmid].dim;
int cut = rcbinfo[procmid].cut;
if (box[2*dim] < cut) box_drop_grid(box,proclower,procmid-1,np,plist);
if (box[2*dim+1] >= cut) box_drop_grid(box,procmid,procupper,np,plist);
}
// ----------------------------------------------------------------------
// miscellaneous methods
// ----------------------------------------------------------------------
/* ----------------------------------------------------------------------
grow list of swaps by DELTA
------------------------------------------------------------------------- */
void Grid2d::grow_swap()
{
maxswap += DELTA;
swap = (Swap *) memory->srealloc(swap,maxswap*sizeof(Swap),"grid2d:swap");
}
/* ----------------------------------------------------------------------
grow list of overlaps by DELTA
------------------------------------------------------------------------- */
void Grid2d::grow_overlap()
{
maxoverlap_list += DELTA;
overlap_list = (Overlap *)
memory->srealloc(overlap_list,maxoverlap_list*sizeof(Overlap),"grid2d:overlap");
}
/* ----------------------------------------------------------------------
deallocate remap data structs
------------------------------------------------------------------------- */
void Grid2d::deallocate_remap()
{
for (int i = 0; i < nsend_remap; i++)
memory->destroy(send_remap[i].packlist);
delete [] send_remap;
for (int i = 0; i < nrecv_remap; i++)
memory->destroy(recv_remap[i].unpacklist);
delete [] recv_remap;
if (self_remap) {
memory->destroy(copy_remap.packlist);
memory->destroy(copy_remap.unpacklist);
}
}
/* ----------------------------------------------------------------------
create 1d list of offsets into 2d array section (xlo:xhi,ylo:yhi)
assume caller's 2d array is allocated as
(fullxlo:fullxhi,fullylo:fullyhi)
------------------------------------------------------------------------- */
int Grid2d::indices(int *&list, int xlo, int xhi, int ylo, int yhi)
{
int nmax = (xhi-xlo+1) * (yhi-ylo+1);
memory->create(list,nmax,"grid2d:indices");
if (nmax == 0) return 0;
int nx = (fullxhi-fullxlo+1);
int n = 0;
int ix,iy;
for (iy = ylo; iy <= yhi; iy++)
for (ix = xlo; ix <= xhi; ix++)
list[n++] = (iy-fullylo)*nx + (ix-fullxlo);
return nmax;
}
/* ----------------------------------------------------------------------
find the comm->procgrid index = which proc owns the igrid index
igrid = grid index (0 to N-1) in dim
n = # of grid points in dim
shift determines position of grid pt within grid cell
shift = 0.5 for cell center, 0.0 for lower-left corner
dim = which dimension (0,1)
split = comm->x/y/z split for fractional bounds of each proc domain
------------------------------------------------------------------------- */
int Grid2d::proc_index_uniform(int igrid, int n, double shift, int dim, double *split)
{
int lo,hi;
double fraclo,frachi;
// loop over # of procs in this dime
// compute the grid bounds for that proc
// if igrid falls within those bounds, return m = proc index
// same logic as in partition_grid()
int m;
for (m = 0; m < comm->procgrid[dim]; m++) {
fraclo = split[m];
frachi = split[m+1];
lo = static_cast<int> (fraclo * n);
while (lo+shift < fraclo*n) lo++;
hi = static_cast<int> (frachi * n);
if (hi+shift >= frachi*n) hi--;
if (igrid >= lo && igrid <= hi) break;
}
return m;
}
/* ----------------------------------------------------------------------
compute the grid box owned by proc within tiled decomposition
performed recursively until proclower = procupper = proc
return box = lo/hi bounds of proc's box in 2 dims
------------------------------------------------------------------------- */
void Grid2d::partition_tiled(int proc, int proclower, int procupper, int *box)
{
// end recursion when partition is a single proc
if (proclower == procupper) return;
// split processor partition
// procmid = 1st processor in upper half of partition
// = location in tree that stores this cut
// cut = index of first grid cell in upper partition
// dim = 0,1 dimension of cut
int procmid = proclower + (procupper - proclower) / 2 + 1;
int dim = rcbinfo[procmid].dim;
int cut = rcbinfo[procmid].cut;
// adjust box to reflect which half of partition the proc is in
if (proc < procmid) {
box[2*dim+1] = cut-1;
partition_tiled(proc,proclower,procmid-1,box);
} else {
box[2*dim] = cut;
partition_tiled(proc,procmid,procupper,box);
}
}
Reference in New Issue View Git Blame Copy Permalink