Opened 7 years ago

Last modified 5 years ago

#186 new defect - convergence

grid aborts due to element non-conservation

Reported by: peter Owned by: peter
Priority: minor Milestone: C13 branch
Component: grains Version: trunk
Keywords: Cc:

Description

The attached grid crashes with c10_branch@4322 and newmole@4322. The situation is better with c10_branch since there only 4 grid points crash:

IONIZATION PARAMETER= -1.000000 LOG
METALS= 3.250000 LOG GRAINS

IONIZATION PARAMETER= -0.500000 LOG  
METALS= 2.750000 LOG GRAINS  

IONIZATION PARAMETER= -0.500000 LOG  
METALS= 3.750000 LOG GRAINS  

IONIZATION PARAMETER= 0.000000 LOG  
METALS= 3.750000 LOG GRAINS  

On newmole the following 8 grid points crash:

IONIZATION PARAMETER= -1.000000 LOG  
METALS= 3.250000 LOG GRAINS  

IONIZATION PARAMETER= -1.000000 LOG  
METALS= 3.750000 LOG GRAINS  

IONIZATION PARAMETER= -0.500000 LOG  
METALS= 2.750000 LOG GRAINS  

IONIZATION PARAMETER= -0.500000 LOG  
METALS= 3.250000 LOG GRAINS  

IONIZATION PARAMETER= -0.500000 LOG  
METALS= 3.750000 LOG GRAINS  

IONIZATION PARAMETER= 0.000000 LOG  
METALS= 2.250000 LOG GRAINS  

IONIZATION PARAMETER= 0.000000 LOG  
METALS= 3.250000 LOG GRAINS  

IONIZATION PARAMETER= 0.000000 LOG  
METALS= 3.750000 LOG GRAINS  

In all cases the aborts were due to non-conservation of nuclei or densities exceeding MAX_DENSITY. The code was compiled with g++ 4.4.3 and MPI enabled.

Attachments (1)

pann.in (249 bytes) - added by peter 7 years ago.
input script

Download all attachments as: .zip

Change History (3)

Changed 7 years ago by peter

Attachment: pann.in added

input script

comment:2 Changed 5 years ago by Ryan Porter

Component: ionization convergencegrains
Owner: changed from nobody to peter
Priority: majorminor
Type: defect - code abortsdefect - convergence

This now finishes at r6340, albeit with a number of "PROBLEM CoolNet? derivative oscillating in initial solution" and "PROBLEM qheat did not converge grain" issues. One example of the latter is the U = -1, Metals = 2.75 point. With lgDoInitConv flipped to false in ConvInitSolution?, the former problem goes away and only the qheat issues remain. Changing component to grains, priority to minor, and type to "defect - convergence".

Note: See TracTickets for help on using tickets.