[CIG-MC] Installing ASPECT on Cray XC30

[Marine Lasbleis]

Dear Timo and Wolfgang, 

I tried with the flag  -D ASPECT_USE_FP_EXCEPTIONS=OFF, and ASPECT is indeed running!
(by the way, which would be the best example to use from the cookbook to check how well aspect is running? I tried the convection-box. I want to use ASPECT for inner core, so I’ll be also running the inner core cookbook soon)

amongst other things we also tried: 
- same than before, with apron -n 1 (so monoproc), same error. 
- static installation (following subsection  3.3.5 Compiling a static ASPECT executable)
It did not work on our cluster (not sure why, but building of trilinos failed)


So, is there any reason the -D ASPECT_USE_FP_EXCEPTIONS=OFF should not be used? 
"Do you get the same error if you run with one processor? If so, do you know how to generate a backtrace in a debugger to figure out where this is happening?” 
> Yes, same error. No, I don’t know how to do a backtrace. If you think I should try, please let me know! (either with explanation on how to do it, or I’ll have a look at internet)

I will investigate a little bit more how well the runs are, and trying a couple of examples to be sure it seems to work. I’ll install it on other sessions of the cluster as well. 

Thank you! 
And thank you so much for the candi.sh script adapted to Cray :-) I tried installing ASPECT on the cluster (and on my mac) a couple of years ago, but with no success. Also, as we are a couple of people here in ELSI this summer wanting to use ASPECT, I installed it with docker on another colleague’s mac, and it seems to be working correctly (at first, I gave her my old virtual machine from a summer workshop in 2014). It’s likely that you’ll get questions from people in Nantes (France) willing to use ASPECT on their own cluster as well. 

Best, 
Marine





Marine Lasbleis
 
=====
ELSI Research Scientist
Earth Life Science Institute
Tokyo Institute of Technology
Tokyo, Japan
+81 70 1572 5070
marine.lasbleis at elsi.jp <mailto:marine.lasbleis at elsi.jp>

https://members.elsi.jp/~marine.lasbleis/ <https://members.elsi.jp/~marine.lasbleis/> 
http://elsi.jp/en/ <http://elsi.jp/en/> 
 

> On 12 Jul 2017, at 03:10, Wolfgang Bangerth <bangerth at tamu.edu> wrote:
> 
> On 07/10/2017 06:00 PM, Marine Lasbleis wrote:
>> [0]PETSC ERROR: [1]PETSC ERROR: [3]PETSC ERROR: [2]PETSC ERROR: ------------------------------------------------------------------------
>> [0]PETSC ERROR: ------------------------------------------------------------------------
>> ------------------------------------------------------------------------
>> [2]PETSC ERROR: ------------------------------------------------------------------------
>> [1]PETSC ERROR: [3]PETSC ERROR: Caught signal number 8 FPE: Floating Point Exception,probably divide by zero
>> [1]PETSC ERROR: Try option -start_in_debugger or -on_error_attach_debugger
>> Caught signal number 8 FPE: Floating Point Exception,probably divide by zero
> 
> ASPECT uses signaling floating point NaNs to track down errors, but as Matt already mentioned, the error is only caught -- not generated -- by PETSc.
> 
> Do you get the same error if you run with one processor? If so, do you know how to generate a backtrace in a debugger to figure out where this is happening?
> 
> (I'll mention that on some systems with some compilers, we have seen these sorts of floating point exceptions come out of the standard C library -- that's highly annoying because it means that the system is defeating our ways of debugging problems. In that case, you can just disable floating point exceptions when you run ASPECT's cmake script. I think there is a blurb about that in the readme file. But before you do that, I'd still be interested to see whether you can get a backtrace :-) )
> 
> Best
> W.
> 
> -- 
> ------------------------------------------------------------------------
> Wolfgang Bangerth          email:                 bangerth at colostate.edu
>                           www: http://www.math.colostate.edu/~bangerth/
> 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.geodynamics.org/pipermail/cig-mc/attachments/20170712/5242e102/attachment.html>