[CP2K:9846] Re: Build failure on 32bit

[Alfio Lazzaro]

Hi Michael,
it seems from this page:

https://buildd.debian.org/status/package.php?p=cp2k

that the compilation went over. However, I see tests failing with message:

CP2K and libint were compiled with different LIBINT_MAX_AM.

Indeed, I see that you are not setting any values for that. You should add 
the flags"

-D__LIBINT_MAX_AM=6 -D__LIBDERIV_MAX_AM1=5

replacing the numbers with the values used in LIBINT compilation (see the 
CP2K INSTALL file for more info)

Alfio


Il giorno lunedì 8 gennaio 2018 14:53:40 UTC+1, Michael Banck ha scritto:
>
> Hi, 
>
> On Mon, Jan 08, 2018 at 05:42:10AM -0800, Alfio Lazzaro wrote: 
> > mp_max_memory_size is the maximum size for a MPI message, which I set to 
> > 2GB (2*2^30), i.e. 
> > 
> > mp_max_memory_size = 2*INT(2, KIND=MPI_ADDRESS_KIND)**30 
> > 
> > The error is that I have to subtract 1 on a 32bit system (Thanks Tiziano 
> to 
> > point me that)... However, this is exact what the HUGE function does, 
> > therefore I changed the code in the trunk with: 
> > 
> > -   INTEGER(KIND=MPI_ADDRESS_KIND), PARAMETER, PRIVATE :: 
> > mp_max_memory_size = 2*INT(2, KIND=MPI_ADDRESS_KIND)**30 
> > +   INTEGER(KIND=MPI_ADDRESS_KIND), PARAMETER, PRIVATE :: 
> > mp_max_memory_size = HUGE(INT(1,KIND=int_4)) 
> > 
> > Could you try it? 
>
> Thanks for the quick fix, I've included the patch in the package and 
> uploaded it now - let's see what the autobuilders say (might take a few 
> hours for them to pick it up). 
>
>
> Michael 
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20180108/1f9a77e2/attachment.htm>