[MITgcm-support] ADWRITE package error "tapeFileCounter > tapeMaxCounter"

[Martin Losch]

Hi Dan,

to avoid the "relocation truncated to fit” error you can try to use a different memory model for your compiler. For example for ifort this means adding the option -mcmodel=medium (or even large) and there may be other options depending on your compiler.

I assume that you use ALLOW_AUTODIFF_WHTAPEIO and ALLOW_WHIO_3D. In that case and if you are feeling bold you can modify line 36 in MDSIO_BUFF_WH.h:
      PARAMETER (nWh=30*Nr)
to increase nWh (which is then copied to tapeMaxCounter). I assume that this will also increase you memory footprint.

Martin


> On 29. May 2018, at 14:41, Dan Jones <dcjones.work at gmail.com> wrote:
> 
> Hi all,
> 
> I am attempting to run an adjoint model using the up-to-date checkpoint 67b and TAF. It compiles without any errors, but I get the following run-time error:
> 
> (PID.TID 0000.0001) ADWRITE: tapeFileCounter > tapeMaxCounter   691  690
> (PID.TID 0000.0001) for file tapelev3_45_the_main_loop_gptrnm1
> 
> I have tried changing the checkpoint levels in tamc.h, but I have not yet found the right combination. I want to run the model for 7300 timesteps, so I estimate that if nchklev_1=10, then nchklev_2 = nchklev_3 = sqrt( Number of steps / nchklev_1 ) + 1 = sqrt(7300/10) + 1 = 28. If I try to increase nchklev_1, then I get the compile-time error "relocation truncated to fit", which I usually interpret as "this executable is too big, so make nchklev_1 smaller". 
> 
> I have uploaded the adjoint-relevant portions of my code to github. The adjoint settings are largely based on the ECCOv4-r2 setup: 
> 
> https://github.com/DanJonesOcean/OxygenAdjoint/tree/d282f20c58c77b1e2c73c0303d49867ae15e54ea
> 
> Thanks in advance for any help!
> 
> Best,
> Dan 
> _______________________________________________
> MITgcm-support mailing list
> MITgcm-support at mitgcm.org
> http://mailman.mitgcm.org/mailman/listinfo/mitgcm-support