Reply to post: Re: @ST The problem probably wasn't the software...

NASA dusts off FORTRAN manual, revives 20-year-old data on Ganymede

ST Silver badge
Devil

Re: @ST The problem probably wasn't the software...

> [ ... ] although you will have to recreate/port the old very specific libraries used to work on the data

Yeah. There's that. Porting N number of old FORTRAN libaries from DEC FORTRAN to ... Intel X86_64 I guess? How long would that take?

Then there's validation. Making sure that the numbers coming out of the VAX libraries are exactly the same as the numbers coming out of Intel. Which, at a minimum, means running the libraries' test harness both on the VAX and on Intel, and then comparing the results.

Chances are the numbers won't match 100%. IEEE rounding, flush-to-zero, denormals.

Next step: if the test harness result numbers don't match, now what? Where's the error? Do you debug on the VAX or do you debug on Intel? Probably on both if you really want to track down the error.

Some people here make it sound like it's just a matter of ftp'ing the data sets, then just a simple recompile, then everything just works. Speaking from experience: it never works that way in real life.

Have you ever seen DEC FORTRAN code from the late '80's? I have, 20 years ago. I'm still trying to unsee it.

POST COMMENT House rules

Not a member of The Register? Create a new account here.

  • Enter your comment

  • Add an icon

Anonymous cowards cannot choose their icon

Biting the hand that feeds IT © 1998–2019