NAN problem

[brianlane723]

Hello. I have written a code using pdsytrd to tridiagonalize a matrix, and it has worked perfectly on one set of machines (i386 running linux), but when I compile and run it on a second set (i686-pc-linux-gnu), I encounter an error that produces NAN's (not-a-number). I've traced the problem down to occur during a call to pdlatrd, when it calls the PBLAS routdine pdgemv. Has anyone seen this happen or know what might be causing it? Thanks!

[Julie]

Brian,

Did you try to run the TRD tests from SCALAPACK/TESTING directory with your i386 and i686 machines?
If yes, do you have the same error?

Julie