Intel® oneAPI Math Kernel Library
Ask questions and share information with other developers who use Intel® Math Kernel Library.

Incorrect results or crashes found in MKL 11.0 Update 3 LAPACK; Patch available soon

Todd_R_Intel
Employee
402 Views

Problems have been found in an LAPACK internal function in Intel MKL 11.0 update 3 (available in Intel® Composer XE 2013 update 3) which can cause many other LAPACK functions to crash or provide incorrect results. The problem is related to an operation that divides work among threads for a certain class of LAPACK functions that are bandwidth limited, but can still benefit from some parallelism. The problem occurs when a thread count greater than 4 and not divisible by 4 is chosen. For a list of all affected functions, please refer to our knowledgebase article on this topic.

A patch is available from the knowledgebase article or the download pages at the Intel registration center site.

 Workaround: explicitly set the number of threads to 1,2,3, or 4n (where n is an integer greater than 1) 

0 Kudos
4 Replies
YoungTaek_Oh
Beginner
402 Views

Workaround does not work for 4n (n > 1). http://software.intel.com/en-us/forums/topic/390568

0 Kudos
Todd_R_Intel
Employee
402 Views

Are you using a system with less than 8 physical (as opposed to HT virtual cores)? I see you reference "i7-990x" which has six cores, so if this a single socket system, Intel MKL is probably dialling your 8 threads setting back to 6 threads and you're back into the problem case.

I'll double check, but it seems the the extra information I should add on this workaround is that 4n should be less than the number of 'physical' cores on your system.

0 Kudos
Todd_R_Intel
Employee
402 Views

We've uploaded a patch that you can access from the knowledgebase article or the download pages at the Intel registration center site.

0 Kudos
Gennady_F_Intel
Moderator
402 Views

the fix of the problem available in 11.0 update 4 which released the last week.

0 Kudos
Reply