The ABINIT Users Mailing List ( CLOSED )

[Xavier Gonze <gonze@pcpm.ucl.ac.be>]

Dear Zhou,

I tried all the sequential tests with Intel 10.0 some time ago, and  
not all
of them were correct at the numerical level (compilation was fine,  
execution
was fine, but ... not all results ... although many of them were fine).
Accordingly, I consider Intel 10.0 to be buggy.
So, perhaps you should stick to Intel 9.1 .

Xavier


On 08 Nov 2007, at 12:27, john Bob wrote:

> Dear Matthieu:
>     As I just said, it is really strange. I use the same mpf90 in  
> mpich to compile Abinit 5.3.5. And the sequential  and parellel  
> ones both work well. All what I have done special is the creation  
> of mpif90. The mpich1.2.7 in that cluster havenot compiled with  
> fortran 90. I  downloaded one copy of  mpich and compiled it with  
> intel fortran 9.1 ( 10.0 have tried). I havenot root privilege, I  
> didn't make install. In such a quizzical way, I got mpif90 and the  
> parelle abinip. For abinit 533 535, there isn't any problem in  
> testing and running. But for 544 the problem occured. This may be  
> coincidence. Maybe it is because i mess up the compiler 10.0 and  
> 9.1, mpich.  I really want it solved. Does there are any tool to  
> test the communicating of mpich written in Fortran90?
>
> Best Regards
>
> Zhou