I'm trying to get a library I wrote to run as part of a xenomai task.  The 
library  uses the Intel MKL (math kernel) library.  MKL  uses the SSE 
instruction set.  Whenever an MKL function is called,  I get a segmentation 
fault.  The fault always occurs in the MKL function, get_N_Cores.  I created a 
small test program that consists of a main and one function.  The function 
makes one MKL call.   It runs without crashing as a linux program (call the 
function directly), and crashes in get_N_cores when run as a xenomai task 
(function called via rt_task_start).  The xenomai task is running in user space.

Intel support said "Just my guess however, MKL_get_N_Cores function tries to 
recognize CPU topology, but if Xenomai framework changes it via shadowing some 
CPU parameters (for example, CPU affinity) then MKL might be confused".  Is his 
guess correct?  Is there a fix to this problem?

Thanks,

Bruce
_______________________________________________
Xenomai mailing list
[email protected]
http://www.xenomai.org/mailman/listinfo/xenomai

Reply via email to