/Qimsl should trigger /Qftz-

[mecej4]

Steve,

An old IMSL issue surfaced again. I originally reported the issue to Roguewave (see https://forums.roguewave.com/showthread.php?1427-Strange-bug-(-)-in-FNL7-routine-IVOAM&p=4077&viewfull=1#post4077 ) in 2011.

I do not know if the bug has been fixed in IMSL 7.1. The problem is still present in the IMSL version distributed by Intel, and Hanson is no longer with Roguewave.

Is it out of order for me to request that in the Intel Fortran driver a customer's use of /Qimsl should trigger /Qftz-, so that IVOAM (and other such IMSL routines, I do not know which) will function correctly?

This may be a bit heavy handed, but is safe; another solution would be to have /Qftz- invoked if a USE statement for IVOAM (and other affected routines) is seen in the user's Fortran code, but it may be a lot of work to implement a fix this way.

[Steven_L_Intel1]

Given that RogueWave does not suggest /ftz- as a general option for IMSL (it isn't included in the FFLAGS variable set in fnlsetup.bat), we'd be reluctant to make a change of this nature. Besides, most IMSL users don't use /Qimsl, they use the include files to specify libraries to link, or do it manually. We don't have a mechanism to change compiler options based on USE of a module. I don't see that anyone from RogueWave has followed up on your report (Dr. Hanson suggested you go through IMSL support. If you were using the Intel-supplied IMSL at the time you could have asked us to do so.)

What I will do is send a query to my contacts in IMSL support, referencing your old thread, and get their comments.

[mecej4]

Steve Lionel (Intel) wrote:

What I will do is send a query to my contacts in IMSL support, referencing your old thread, and get their comments.

Fair enough.

The same routine, IVOAM, had another bug, unrelated to the present one, that Roguewave stated would be fixed in FNL 7.1, see https://forums.roguewave.com/showthread.php?1720-Bug-in-IVOAM&p=4844&viewfull=1#post4844 .