Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

lockAdd: case of 2D plane in 3D #3700

Merged
merged 1 commit into from
Jan 22, 2024
Merged

lockAdd: case of 2D plane in 3D #3700

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
60 changes: 37 additions & 23 deletions Src/Base/AMReX_BaseFab.H
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3330,15 +3330,25 @@ BaseFab<T>::lockAdd (const BaseFab<T>& src, const Box& srcbox, const Box& destbo

Array4<T> const& d = this->array();
Array4<T const> const& s = src.const_array();
auto const& dlo = destbox.smallEnd();
#if (AMREX_SPACEDIM == 3)
auto const& dhi = destbox.bigEnd();
#endif
auto const& slo = srcbox.smallEnd();
auto const offset = slo - dlo;
auto const lenx = srcbox.length(0);
auto const& dlo = amrex::lbound(destbox);
auto const& dhi = amrex::ubound(destbox);
auto const& len = amrex::length(destbox);
auto const& slo = amrex::lbound(srcbox);
Dim3 const offset{slo.x-dlo.x, slo.y-dlo.y, slo.z-dlo.z};

int planedim;
int nplanes;
int plo;
if (len.z == 1) {
planedim = 1;
nplanes = len.y;
plo = dlo.y;
} else {
planedim = 2;
nplanes = len.z;
plo = dlo.z;
}

auto const nplanes = srcbox.length(AMREX_SPACEDIM-1);
auto* mask = (bool*) amrex_mempool_alloc(sizeof(bool)*nplanes);
for (int ip = 0; ip < nplanes; ++ip) {
mask[ip] = false;
Expand All @@ -3348,27 +3358,31 @@ BaseFab<T>::lockAdd (const BaseFab<T>& src, const Box& srcbox, const Box& destbo
int planes_left = nplanes;
while (planes_left > 0) {
AMREX_ASSERT(mm < nplanes);
auto const m = mm + dlo[AMREX_SPACEDIM-1];
auto const m = mm + plo;
int ilock = m % OpenMP::nlocks;
if (ilock < 0) { ilock += OpenMP::nlocks; }
auto* lock = &(OpenMP::omp_locks[ilock]);
if (omp_test_lock(lock))
{
for (int n = 0; n < numcomp; ++n)
{
#if (AMREX_SPACEDIM == 3)
for (int j = dlo[1]; j <= dhi[1]; ++j)
{
IntVect div(dlo[0], j, m);
#elif (AMREX_SPACEDIM == 2)
{
IntVect div(dlo[0], m);
#endif
auto * pdst = d.ptr(div ,n+destcomp);
auto const* psrc = s.ptr(div+offset,n+srccomp);
auto lo = dlo;
auto hi = dhi;
if (planedim == 1) {
lo.y = m;
hi.y = m;
} else {
lo.z = m;
hi.z = m;
}

for (int n = 0; n < numcomp; ++n) {
for (int k = lo.z; k <= hi.z; ++k) {
for (int j = lo.y; j <= hi.y; ++j) {
auto * pdst = d.ptr(dlo.x,j ,k ,n+destcomp);
auto const* psrc = s.ptr(slo.x,j+offset.y,k+offset.z,n+ srccomp);
#pragma omp simd
for (int ii = 0; ii < lenx; ++ii) {
pdst[ii] += psrc[ii];
for (int ii = 0; ii < len.x; ++ii) {
pdst[ii] += psrc[ii];
}
}
}
}
Expand Down
Loading