arpack/eigs problems (was: Re: moving toward a 3.0 release)

[John W. Eaton]

On 29-Sep-2006, Joe Koski wrote:

| on 9/29/06 1:53 AM, David Bateman at address@hidden wrote:
| 
| > Thanks for the ideas Joe, I think I'll definitely try the -r8 option and
| > see if it help..
| 
| One more comment, and then I'll shut up.
| 
| With old code, which this probably isn't, the old Fortran practice of
| relying on variables being unchanged when you reenter to a subroutine can
| bite you. I think the g77 flag for saving all variables is -fno-automatic,
| but just placing a blanket SAVE instead of a SAVE j,etc. in the code can
| give the same effect. Not an efficient or elegant solution, but it often
| works. We resorted to that fix in one routine of our local CFD code, when,
| after much head scratching, we couldn't figure out which variable wasn't
| being properly SAVEd.

I guessed that some variable that should have been saved was not, so I
tried -fno-automatic and it did not avoid the crash.

jwe