[MITgcm-devel] arrays too large for TAF?

[Matthew Mazloff]

http://mitgcm.org/pipermail/mitgcm-support/2013-August/008469.html


On Mar 6, 2014, at 10:10 AM, Daniel Goldberg <dngoldberg at gmail.com> wrote:

> Hi all
> 
> I encountered a curious issue using TAF, which I have not seen before -- and I am wondering if others know of this limitation (or bug), or if I should take it up with fastOpt. I am trying to run on a 500x432x20 grid (although I really only have 2 fields that are 3d, the rest are single-layer). Here is my SIZE.h:
> 
>      &           sNx = 125,
>      &           sNy = 144,
>      &           OLx =   4,
>      &           OLy =   4,
>      &           nSx =   4,
>      &           nSy =   3,
>      &           nPx =  1,
>      &           nPy =  1,
>      &           Nx  = sNx*nSx*nPx,
>      &           Ny  = sNy*nSy*nPy,
>      &           Nr  =  20)
> 
> When I compile, I get the errors (see below) which make it seem  as though i am exceeding some imposed memory requirement. Things become OK if I make the total # of cells on a cpu smaller, though i haven't exactly figured out the formula. It compiles if I use 12 cpus instead of 12 tiles for example (nSx=nSy=1, nPx=4,nPy=3). 
> 
> Everything is fine if i just compile the forward model, i.e. if i don't use TAF. Possibly this is related to something I've done in the Streamice package. But I'm wondering, is this familiar to anyone? If not, perhaps I should complain to fastOpt.
> 
> Thanks
> Dan
> 
> 
> ad_taf_output.o: In function `streamice_vel_solve_':
> ad_taf_output.f:(.text+0x44b21): relocation truncated to fit: R_X86_64_32 against `.lrodata'
> ad_taf_output.f:(.text+0x44b53): relocation truncated to fit: R_X86_64_32 against `.lrodata'
> ad_taf_output.f:(.text+0x44b85): relocation truncated to fit: R_X86_64_32 against `.lrodata'
> ad_taf_output.f:(.text+0x44bb7): relocation truncated to fit: R_X86_64_32 against `.lrodata'
> ad_taf_output.o: In function `streamice_solo_timestep_':
> ad_taf_output.f:(.text+0x607dd): relocation truncated to fit: R_X86_64_32 against `.lrodata'
> ad_taf_output.o: In function `streamice_init_varia_':
> ad_taf_output.f:(.text+0x6a705): relocation truncated to fit: R_X86_64_32 against `.lrodata'
> ad_taf_output.f:(.text+0x6a9d4): relocation truncated to fit: R_X86_64_32 against `.lrodata'
> ad_taf_output.f:(.text+0x6aa06): relocation truncated to fit: R_X86_64_32 against `.lrodata'
> ad_taf_output.f:(.text+0x6ab1f): relocation truncated to fit: R_X86_64_32 against `.lrodata'
> ad_taf_output.f:(.text+0x6ab51): relocation truncated to fit: R_X86_64_32 against `.lrodata'
> ad_taf_output.o: In function `streamice_init_variamd_':
> 
> -- 
> 
> Daniel Goldberg, PhD
> Lecturer in Glaciology
> School of Geosciences, University of Edinburgh
> Geography Building, Drummond Street, Edinburgh EH8 9XP
> 
> 
> em: Dan.Goldberg at ed.ac.uk
> web: http://ocean.mit.edu/~dgoldberg
> _______________________________________________
> MITgcm-devel mailing list
> MITgcm-devel at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-devel

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mitgcm.org/pipermail/mitgcm-devel/attachments/20140306/a5cbccf9/attachment.htm>