Project

General

Profile

Bug #535

g_pme_error: Invalid tMPI_Comm

Added by Christoph Junghans about 9 years ago. Updated about 9 years ago.

Status:
Closed
Priority:
Normal
Assignee:
Category:
mdrun
Target version:
Affected version - extra info:
Affected version:
Difficulty:
uncategorized
Close

Description

At the end of g_pme_error, there is an error:

Calculating reciprocal error part 1 ... 100%
Calculating reciprocal error part 2 ... 100%
Direct space error est. : 3.676e-04 kJ/(mol*nm)
Reciprocal sp. err. est.: nan kJ/(mol*nm)
Starting tuning ...
Calculating reciprocal error part 1 ... 100%
Calculating reciprocal error part 2 ... 100%
Direct space error est. : 2.188e-03 kJ/(mol*nm)
Reciprocal sp. err. est.: nan kJ/(mol*nm)
tMPI error: Invalid tMPI_Comm
Aborted

History

#1 Updated by Sander Pronk about 9 years ago

I just looked at that code, and it seems to be calling all sorts of parallel initialization and shutdown code that belongs in mdrun. Should this tool really be called in parallel? If so, the documentation certainly doesn't mention it.

#2 Updated by Erik Lindahl about 9 years ago

This tool does not appear to be release-ready. I suggest we pull it for 4.5.

#3 Updated by Erik Lindahl about 9 years ago

Hm. On second thought it might not be that bad after all. The error message was just because we need a wrapper around gmx_finalize() to see if the parallel environment has been initialized. Apart from that it seems to work.

I have no idea if it works in parallel yet, but since it seems to have been prepared for that and still works I see no reason to remove all that code for now.

Fixed in commit e13ae78d18a131265c67569d01f2264f516408dc.

Also available in: Atom PDF