topic bug in cblas_sscal ( ) ? in Intel® oneAPI Math Kernel Library

bug in cblas_sscal ( ) ?

Customer__Intel4 — Thu, 22 Mar 2012 17:42:14 GMT

I discovered that when usingcblas_sscal ( ) to scale an array to zero values, sometimes ( not every time) an entry of the array after scaling becomes nan. The command simply reads:cblas_sscal(L, 0.0, A, 1); where L is the length of the array A.

bug in cblas_sscal ( ) ?

TimP — Thu, 22 Mar 2012 22:19:28 GMT

cblas_sscal() requires a float constant. If you have a correct prototype in scope (by the supplied include files), the compiler would take care of cast from the double value you supplied.

bug in cblas_sscal ( ) ?

Chao_Y_Intel — Fri, 23 Mar 2012 05:40:57 GMT

Hi,
If input the data type is no problem, it needs a test code that could show thi problem. Feel free to post here if you have.

Thanks,
Chao

bug in cblas_sscal ( ) ?

Murat_G_Intel — Sun, 08 Apr 2012 22:52:13 GMT

If your input array A contains NaNs, then scaling a NaN value with 0.0 will still give a NaN.