Shuenhoy/vtk-m
1
0
Fork 0
mirror of https://gitlab.kitware.com/vtk/vtk-m synced 2024-09-20 02:55:47 +00:00
Code Issues 2 Actions Packages Projects Releases Wiki Activity

fix OpenMP PRAGMA warnings for auto-associated looping variable types

Browse Source
This commit is contained in:
nadavi 2020-08-26 17:29:38 +00:00
parent 06579aac8c
commit 89cf93a66b
1 changed files with 18 additions and 9 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

27
vtkm/cont/openmp/internal/FunctorsOpenMP.h
View File

@ -379,9 +379,9 @@ struct ReduceHelper
// This gives faster code for floats and non-trivial types. // This gives faster code for floats and non-trivial types.
template <typename ReturnType, typename IterType, typename FunctorType> template <typename ReturnType, typename IterType, typename FunctorType>
static ReturnType DoParallelReduction(IterType data, static ReturnType DoParallelReduction(IterType data,
vtkm::Id numVals, const vtkm::Id& numVals,
int tid, const int& tid,
int numThreads, const int& numThreads,
FunctorType f, FunctorType f,
std::false_type /* isIntegral */) std::false_type /* isIntegral */)
{ {
@ -389,18 +389,24 @@ struct ReduceHelper
ReturnType accum = f(data[2 * tid], data[2 * tid + 1]); ReturnType accum = f(data[2 * tid], data[2 * tid + 1]);
const vtkm::Id offset = numThreads * 2; const vtkm::Id offset = numThreads * 2;
const vtkm::Id end = std::max(vtkm::Id(((numVals / 4) * 4) - 4), offset); const vtkm::Id end = std::max(((numVals / 4) * 4) - 4, offset);
const vtkm::Id unrollEnd = end - ((end - offset) % 4); const vtkm::Id unrollEnd = end - ((end - offset) % 4);
vtkm::Id i = offset; vtkm::Id i = offset;
// When initializing the looping iterator to a non integral type, intel compilers will
// convert the iterator type to an unsigned value
#pragma GCC diagnostic push
#pragma GCC diagnostic ignored "-Wsign-conversion"
VTKM_OPENMP_DIRECTIVE(for schedule(static)) VTKM_OPENMP_DIRECTIVE(for schedule(static))
for (i = offset; i < unrollEnd; i += 4) for (i = offset; i < unrollEnd; i += 4)
#pragma GCC diagnostic pop
{ {
const auto t1 = f(data[i], data[i + 1]); const auto t1 = f(data[i], data[i + 1]);
const auto t2 = f(data[i + 2], data[i + 3]); const auto t2 = f(data[i + 2], data[i + 3]);
accum = f(accum, t1); accum = f(accum, t1);
accum = f(accum, t2); accum = f(accum, t2);
} }
// Let the last thread mop up any remaining values as it would // Let the last thread mop up any remaining values as it would
// have just accessed the adjacent data // have just accessed the adjacent data
if (tid == numThreads - 1) if (tid == numThreads - 1)
@ -418,18 +424,21 @@ struct ReduceHelper
// hurt performance. // hurt performance.
template <typename ReturnType, typename IterType, typename FunctorType> template <typename ReturnType, typename IterType, typename FunctorType>
static ReturnType DoParallelReduction(IterType data, static ReturnType DoParallelReduction(IterType data,
vtkm::Id numVals, const vtkm::Id& numVals,
int tid, const int& tid,
int numThreads, const int& numThreads,
FunctorType f, FunctorType f,
std::true_type /* isIntegral */) std::true_type /* isIntegral */)
{ {
// Use the first (numThreads*2) values for initializing: // Use the first (numThreads*2) values for initializing:
ReturnType accum = f(data[2 * tid], data[2 * tid + 1]); ReturnType accum = f(data[2 * tid], data[2 * tid + 1]);
// Assign each thread chunks of the remaining values for local reduction // Assign each thread chunks of the remaining values for local reduction
#pragma GCC diagnostic push
#pragma GCC diagnostic ignored "-Wsign-conversion"
VTKM_OPENMP_DIRECTIVE(for schedule(static)) VTKM_OPENMP_DIRECTIVE(for schedule(static))
for (vtkm::Id i = numThreads * 2; i < numVals; i++) for (vtkm::Id i = numThreads * 2; i < numVals; i++)
#pragma GCC diagnostic pop
{ {
accum = f(accum, data[i]); accum = f(accum, data[i]);
} }