ivf11.1.072/73 + openmp

[yeg001]

Hi, every one. Iencountered a defect when i used ivf11.1.072/73for linux. And my firend tested it ina someedition ofivf for windows. And he also find it.

I compile the file like this,
ifort 1dshggl.for -openmp
and the outputarrayew1 is differntfrom ew2.

with out "-openmp", or other versions of ivf, such as 11.1.038, it is fine.

[pbkenned1]

It appears to be an -O2 or higher optimisation bug. I see correct results at -O0/-O1. Let me investigate, and I'll get back to you.

Patrick Kennedy
Intel Developer Support

[yeg001]

Thank you for your reply. I had tried to simplify the code, but failed. I hope it can help intel to improve openmp and fix the defect in next update.

[pbkenned1 · Accepted solution]

This has been reported to compiler engineering as defect #DPD200161914. I'll pass along updates I receive from the developers. The bug exists in Linux compiler 11.1.069 or later, and is present in v12.0 as well. The last known good Linux compiler is 11.1.064.

Patrick

[yeg001]

Thank you for your help, Patrick.

[pbkenned1]

This issueshould be fixed in Intel Fortran Compiler XE Update #2 (the V12.0 compiler),which we plan to release shortly. I'll update this thread with the package identifiers once the packages are posted in the download center.

Patrick Kennedy

[pbkenned1]

This issue is resolved in ifort XE update #2. The packages containing the fixes are:

l_fcompxe_2011.2.137 (Linux*)

m_fcompxe_2011.2.142 (Mac OS X*)

w_fcompxe_2011.2.154 (Windows*)

Patrick Kennedy