(OpenMP) Computation results are error

(OpenMP) Computation results are error

We got a code with OpenMP from other researchers. The code was developed on Alpha (DEC) workstation and ran on it well. We bought an Itanium machine and wanted to run the code on the new machine. When comparing the computation results between Alpha and Itanium, we find that the results from Itanium are wrong (crazy wrong).

We tried another way to check the code: compile the code with -openmp_stubs, and the resutls are the same with which on Alpha workstation.

Any suggestions?

System: RedHat EL 3.0
Compiler: l_fc_p_8.1.018
Compile command: ifort -w -r8 -openmp

2 posts / 0 new
Last post
For more complete information about compiler optimizations, see our Optimization Notice.

Hi, thanks for your reply to my former query, will try it out.

I have ported code from Alpha to P4s (not Itanium, but had simmilar problems having all answers wrong)

Firstly get the latest compiler as compatibility has improved.

secondly I believe your old code would have used Vaxlib and vms libraries. So you will need them as there are lots of non statndard features on code developed for the alpha.

As a pointer to your problem a typical compilation flag I would use on a p4-Xeon (I wish i were rich enough to own an Itanium, but being a student you really can't afford it!)

ifort -Vaxlib -vms -tpp7 -tune pn4 -xN -O3 -axW -w -ip -fpp -assume cc_omp -openmp -openmp_report -threads

all in one line.

as you are using an itanium the tpp7, pn4 should be substiituted as appropriate. Also consider using static flag -Bstatic if the libraries wont be present on the machine running the code.

I hope it works, I had a nightmare porting mine as the intel compiler at that time was somewhat incompatible with RH9. Now it is very good.

Leave a Comment

Please sign in to add a comment. Not a member? Join today