topic Different results using 11.2.1 vs 11.1.0 in Intel® oneAPI Math Kernel Library

Different results using 11.2.1 vs 11.1.0

xian-zhong_guous_cd- — Wed, 10 Dec 2014 01:05:31 GMT

I have the identical SPD matrix and rhs but got the different solution using mkl 11.2.1 on win vs 11.1.0 on linux.

on win

Major version: 11
Minor version: 2
Update version: 1
Product status: Product
Build: n20141023
Processor optimization: Intel(R) Streaming SIMD Extensions 4.2 (Intel(R) SSE4.2) enabled processors

on linux

Major version: 11
Minor version: 1
Update version: 0
Product status: Product
Build: n20130711
Processor optimization: Intel(R) Advanced Vector Extensions (Intel(R) AVX) Enabled Processor

A related question: we want to use intel 15.0. It seems we must use mkl11.2. Is that true?

What do you mean by "we want

Gennady_F_Intel — Wed, 10 Dec 2014 05:30:53 GMT

What do you mean by "we want to use intel 15.0"? You mean Intel Parallel Studio XE 2015 Composer Edition?

Have you used the same #of threads when called Pardiso from 11.1 and 11.2 versions?

--Gennady

xian-zhong.guos: why do you

mecej4 — Wed, 10 Dec 2014 12:18:06 GMT

xian-zhong.guos: why do you have different versions of the input (matrix and rhs) files for Linux and Windows? Are their contents, taken in pairs, the same? I have not looked at the contents of the files (since the files are large), but it should be understandable that different input files will give different output.

Yes, we want to use Intel

xian-zhong_guous_cd- — Wed, 10 Dec 2014 18:04:34 GMT

Yes, we want to use Intel Parallel Studio XE 2015 Composer Edition.

Yes, I have compared the matrix/rhs from linux and win and they are same. Sorry, I could have just uploaded one set of matrix/rhs.

They both use one thread.

xian-zhong_guous_cd- — Wed, 10 Dec 2014 18:17:49 GMT

They both use one thread.