[MITgcm-support] build errors for llc_1080

Ryan Abernathey ryan.abernathey at gmail.com
Tue Dec 6 16:21:51 EST 2016

Just an update on my issue.

I have built and run on yellowstone and...still crashes in the same way.

The only possible issue I can think of now is that I am not using
code_async. But my understanding was that code_async is exclusively for I/O
and shouldn't affect the model solution at all.

I guess I will now try to repeat the same thing on pleaides with and
without code_async.


On Wed, Nov 30, 2016 at 1:47 PM, Ryan Abernathey <ryan.abernathey at gmail.com>

> Ok, so I have verified the md5 checksums for all the binary inputs. AND I
> have compiled with checkpoint64p (the original one) using -debug mode (no
> optimizations).
> Starting from the pickup file or the initial conditions, I still get the
> crash on the third timestep.
> Starting to feel pretty stuck... I will try porting the whole thing to
> yellowstone, but that will take some time.
> -Ryan
> On Tue, Nov 29, 2016 at 8:02 PM, Dimitris Menemenlis <
> dmenemenlis at gmail.com> wrote:
>> Good point about compiler/machine issues.
>> Could you maybe try compiling the code with O0?
>> Sometimes optimization flags cause unpredictable trouble.
>> Later in the week or early next week, I can try to reproduce problem on
>> pleiades.
>> Dimitris Menemenlis
>> On Nov 29, 2016, at 4:54 PM, Ryan Abernathey <ryan.abernathey at gmail.com>
>> wrote:
>> I have tried with checkpoint64p, 65v, and the latest master. They all
>> fail in the same way, although I have not done this full test matrix for
>> each checkpoint.
>> I am trying to replicate your initial run on pleaides as closely as
>> possible. So I have not changed any of the initial conditions or forcing
>> files from what was originally used. It is particularly distressing that
>> the pickup file doesn't work, because that presumably represents a stable,
>> self-consistent initial condition.
>> I will try on yellowstone later this week. That might help rule out
>> certain machine / compiler dependent problems.
>> Best,
>> Ryan
>> _______________________________________________
>> MITgcm-support mailing list
>> MITgcm-support at mitgcm.org
>> http://mitgcm.org/mailman/listinfo/mitgcm-support
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mitgcm.org/pipermail/mitgcm-support/attachments/20161206/4725f48f/attachment.html>

More information about the MITgcm-support mailing list