diff options
author | Martin Kroeker <martin@ruby.chemie.uni-freiburg.de> | 2017-07-06 17:30:12 +0200 |
---|---|---|
committer | GitHub <noreply@github.com> | 2017-07-06 17:30:12 +0200 |
commit | 49e62c0e7796ccaa773591e739628846e3d8ab06 (patch) | |
tree | e3029bf35d2aee558d7734b6cabd7230b8b3af54 | |
parent | 482015f8d6840da9617b422e758162cf7358c8b2 (diff) | |
download | openblas-49e62c0e7796ccaa773591e739628846e3d8ab06.tar.gz openblas-49e62c0e7796ccaa773591e739628846e3d8ab06.tar.bz2 openblas-49e62c0e7796ccaa773591e739628846e3d8ab06.zip |
fixed syrk_thread.c taken from wernsaar
Stride calculation fix copied from https://github.com/wernsaar/OpenBLAS/commit/88900e1
-rw-r--r-- | driver/level3/syrk_thread.c | 4 |
1 files changed, 2 insertions, 2 deletions
diff --git a/driver/level3/syrk_thread.c b/driver/level3/syrk_thread.c index 94274be72..5f40853dc 100644 --- a/driver/level3/syrk_thread.c +++ b/driver/level3/syrk_thread.c @@ -109,7 +109,7 @@ int CNAME(int mode, blas_arg_t *arg, BLASLONG *range_m, BLASLONG *range_n, int ( if (nthreads - num_cpu > 1) { di = (double)i; - width = ((BLASLONG)( sqrt(di * di + dnum) - di) + mask) & ~mask; + width = (BLASLONG)(( sqrt(di * di + dnum) - di + mask)/(mask+1)) * (mask+1); if ((width <= 0) || (width > n_to - i)) width = n_to - i; @@ -149,7 +149,7 @@ int CNAME(int mode, blas_arg_t *arg, BLASLONG *range_m, BLASLONG *range_n, int ( if (nthreads - num_cpu > 1) { di = (double)(arg -> n - i); - width = ((BLASLONG)(-sqrt(di * di + dnum) + di) + mask) & ~mask; + width = ((BLASLONG)((-sqrt(di * di + dnum) + di) + mask)/(mask+1)) * (mask+1); if ((width <= 0) || (width > n_to - i)) width = n_to - i; |