Jacobian matrix - incorrect result calculation

[stansy]

Hi,

The 'sjacobi_rci_c.c' contained in the examples MKL library returns the error results:

x f1 f2

0.0000 0.0000 0.0000

0.0000 1.#INF 1.#INF

1.#INF 1.#INF 1.#INF

650751168283084220000000000000.0000 1.#INF 1.#INF

df(y0)= -431602080.0000 -431602080.0000 -431602080.0000 -431602080.0000

df(y1)= -431602080.0000 -431602080.0000 -431602080.0000 -431602080.0000

df(y2)= -431602080.0000 -431602080.0000 -431602080.0000 -431602080.0000

df(y3)= -431602080.0000 -431602080.0000 -431602080.0000 -431602080.0000

This problem does not occur for double-precision calculations:

x f1 f2

10.0000 0.0000 110.0000

10.0000 -2.2361 0.0000

10.0000 0.0000 100.0000

10.0000 0.0000 0.0000

df(y0)= 1.0000 0.0000 0.0000 0.0000

df(y1)= 10.0000 0.0000 -20.0000 0.0000

df(y2)= 0.0000 2.2361 40.0000 0.0000

df(y3)= 0.0000 -2.2361 0.0000 0.0000

I do not really know how to correct this example.

Thanks,

Stan

[Chao_Y_Intel]

Stan,

Could you provide some more information with the problem:  Which version of the Intel MKL are you using now?   what is the OS, and how are the Intel MKL linked there?

Regards,
Chao
 

[stansy]

Hi Chao,
I have done more tests, but still do not understand these erroneous results.

Incorrect results occur for configuration:
Windows 7; Intel(R) Math Kernel Library Version 10.3.6 Product Build 20110809 for Intel(R) 64 architecture applications; VS2010

Following compilations are working great:
Windows 7; Intel(R) Math Kernel Library Version 10.3.10 Product Build 20120412 for Intel(R) 64 architecture applications; Intel Parallel Studio XE 2011 + VS2010
Scientific Linux release 5.8 (Boron); Intel(R) Math Kernel Library Version 10.3.11 Product Build 20120606 for Intel(R) 64 architecture applications; Intel(R) C Intel(R) 64 Compiler XE for applications running on Intel(R) 64, Version 12.1.0.233 Build 20110811

Regards,
Stan

[Chao_Y_Intel]

Stan,

Could you check the new MKL 11.0.x release? When I tested here with new release,  I could not see this problem.

Regards,
Chao

[Alexander_K_Intel2]

Hi Stan,

Could you provide linkline to reproduce your issue? Look like the problem caused by mexed types/interfaces but we can't reproduce it on our side.

With best regards,

Alexander Kalinkin