GitHub-Proxy/eigen
1
0
Fork 0
mirror of https://gitlab.com/libeigen/eigen.git synced 2025-08-16 05:35:57 +08:00
Code Issues Packages Projects Releases Wiki Activity

Fix bug #714: the actual number of threads might be lower than the number of request ones.

Browse Source
This commit is contained in:
Gael Guennebaud 2015-02-18 15:24:05 +01:00
parent bb3e5c29cc
commit a2d9a4806a
1 changed files with 10 additions and 7 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

17
Eigen/src/Core/products/Parallelizer.h
View File

@ -125,19 +125,22 @@ void parallelize_gemm(const Functor& func, Index rows, Index cols, bool transpos
if(transpose) if(transpose)
std::swap(rows,cols); std::swap(rows,cols);
Index blockCols = (cols / threads) & ~Index(0x3);
Index blockRows = (rows / threads) & ~Index(0x7);
GemmParallelInfo<Index>* info = new GemmParallelInfo<Index>[threads]; GemmParallelInfo<Index>* info = new GemmParallelInfo<Index>[threads];
#pragma omp parallel for schedule(static,1) num_threads(threads) #pragma omp parallel num_threads(threads)
for(Index i=0; i<threads; ++i)
{ {
Index i = omp_get_thread_num();
// Note that the actual number of threads might be lower than the number of request ones.
Index actual_threads = omp_get_num_threads();
Index blockCols = (cols / actual_threads) & ~Index(0x3);
Index blockRows = (rows / actual_threads) & ~Index(0x7);
Index r0 = i*blockRows; Index r0 = i*blockRows;
Index actualBlockRows = (i+1==threads) ? rows-r0 : blockRows; Index actualBlockRows = (i+1==actual_threads) ? rows-r0 : blockRows;
Index c0 = i*blockCols; Index c0 = i*blockCols;
Index actualBlockCols = (i+1==threads) ? cols-c0 : blockCols; Index actualBlockCols = (i+1==actual_threads) ? cols-c0 : blockCols;
info[i].rhs_start = c0; info[i].rhs_start = c0;
info[i].rhs_length = actualBlockCols; info[i].rhs_length = actualBlockCols;