Shuenhoy/vtk-m
1
0
Fork 0
mirror of https://gitlab.kitware.com/vtk/vtk-m synced 2024-09-08 13:23:51 +00:00
Code Issues 2 Actions Packages Projects Releases Wiki Activity

fixes OMP reduction when using OMP_NUM_THREADS lt 4

Browse Source 
This fixes, which where triggered since in the new CI, one of the
docker runner set `OMP_NUM_THREADS=3`:

1. `UnitTestOpenMPDeviceAdapter`
2. `UnitTestMeshQualityFilter`

In the redution optimized implementation for _OpenMP_, it unrolls
the reduce loop in iterations of four elements. The last iteration
in the loop might overflow the loop end element (when it is not a
multiple of four).

This commit fixes this by setting the OpenMP unrolled reduce loop
end element to its previous closest multiple of four of the original end
element.

Signed-off-by: Vicente Adolfo Bolea Sanchez <vicente.bolea@kitware.com>
This commit is contained in:
Vicente Adolfo Bolea Sanchez 2020-04-22 20:14:34 -04:00
parent f6970314a1
commit 738c05ae87
1 changed files with 6 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
vtkm/cont/openmp/internal/FunctorsOpenMP.h
View File

@ -368,10 +368,13 @@ struct ReduceHelper
// Use the first (numThreads*2) values for initializing:
ReturnType accum = f(data[2 * tid], data[2 * tid + 1]);
vtkm::Id i = numThreads * 2;
const vtkm::Id unrollEnd = ((numVals / 4) * 4) - 4;
const vtkm::Id offset = numThreads * 2;
const vtkm::Id end = std::max(vtkm::Id(((numVals / 4) * 4) - 4), offset);
const vtkm::Id unrollEnd = end - ((end - offset) % 4);
vtkm::Id i = offset;
VTKM_OPENMP_DIRECTIVE(for schedule(static))
for (i = numThreads * 2; i < unrollEnd; i += 4)
for (i = offset; i < unrollEnd; i += 4)
{
const auto t1 = f(data[i], data[i + 1]);
const auto t2 = f(data[i + 2], data[i + 3]);