[Simgrid-user] SURF: Using GTNetS and network action delays

Arnaud Legrand arnaud.legrand at imag.fr
Wed Jul 16 22:06:07 CEST 2008


Hi,

Kurt Vanmechelen a écrit :
>> Yeah, gtnets needs informations about routers. Give a try to the
>> platform files in /examples/msg/gtnets.
> OK thanks, dogbone-p.xml works! The bigger platform with the r-n200-f50-s4-2-p.xml file returns :
> Invalid XML (XML input line 608, state 16): Starting tag <router> is not allowed here.
> ** SimGrid: UNCAUGHT EXCEPTION received on (0): category: unknown_err; value: 0
> ** Parse error in /home/kvanmech/workspace/simgrid/examples/msg/gtnets/r-n200-f50-s4-2-p.xml
> ** Thrown by () in this process
> [0.000000] xbt/ex.c:113: [xbt_ex/CRITICAL] Parse error in /home/kvanmech/workspace/simgrid/examples/msg/gtnets/r-n200-f50-s4-2-p.xml
> 
> **   In xbt_backtrace_current() at /home/kvanmech/workspace/simgrid/src/xbt/backtrace_linux.c:19
> **   In parse_platform_file() at /home/kvanmech/workspace/simgrid/src/surf/surfxml_parse.c:617
> Aborted
> I guess all the router tags need to be moved to the top of the file...(?).

Yes, That's needed now. We have completely rewritten the parser some 
time ago and changed the DTD. Here is what it looks like now:

<!ELEMENT platform 
((include|trace|random|cluster|set|foreach|host|router|link|trace:conn
ect)*,(include|route:multi|route|trace:connect|process)*)>

So the router tags indeed have to be before the route tags.

> Now, I have a question regarding the reported finishing times of actions with GTNetS. If I run the test program below, I notice that the reported finished times of the actions
> do not coincide with the internal clock of the simulator at each surf_solve() call, i.e. after the first call to surf_solve :
> Action : C1toS1
> Action state : 3
> Action Remains :0
> Action Finished :0.330214
> Action : C2toS2
> Action state : 1
> Action Remains :0
> Action Finished :-1
> Current Time : 0.165107
> 
> So action C1toS1 reportedly finished at 0.330214 but the current time of the simulator is 0.165107. At the end of the simulation the current 
> SURF clock is 0.324606. Is this logical? My guess is that the real ending times of the actions would be 0.165107 and 0.324606. Could you confirm this?

Mmh. I need to look more carefully into this. I have other urgent things 
to finish so leave me a few days and I'll tell you where the problem 
comes from.

Cheers,

     Arnaud




More information about the Simgrid-user mailing list