floraachy
/
OpenBLAS
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Improve workaround and keep compilers from optimizing it out

This commit is contained in:
Martin Kroeker 2023-04-13 18:07:52 +02:00 committed by GitHub
parent 108a21e47a
commit 3727672a74
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 4 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
kernel/arm64/dznrm2_thunderx2t99.c
View File

@ -27,7 +27,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#include "common.h" #include "common.h"
#include <float.h>
#include <arm_neon.h> #include <arm_neon.h>
#if defined(SMP) #if defined(SMP)
@ -344,6 +344,7 @@ FLOAT CNAME(BLASLONG n, FLOAT *x, BLASLONG inc_x)
FLOAT dummy_alpha[2]; FLOAT dummy_alpha[2];
#endif #endif
FLOAT ssq, scale; FLOAT ssq, scale;
volatile FLOAT sca;
if (n <= 0 || inc_x <= 0) return 0.0; if (n <= 0 || inc_x <= 0) return 0.0;
@ -404,7 +405,8 @@ FLOAT CNAME(BLASLONG n, FLOAT *x, BLASLONG inc_x)
#else #else
nrm2_compute(n, x, inc_x, &ssq, &scale); nrm2_compute(n, x, inc_x, &ssq, &scale);
#endif #endif
if (fabs(scale) <1.e-300) return 0.; sca = fabs(scale);
if (sca < DBL_MIN) return 0.;
ssq = sqrt(ssq) * scale; ssq = sqrt(ssq) * scale;
return ssq; return ssq;