<font size=2 face="sans-serif">Hello,</font>
<br>
<br><font size=2 face="sans-serif"> My application does not run because
each process is trying to ope the same output file. How can I specify a
separate working directory for each process so there is no conflict.</font>
<br><font size=2 face="sans-serif">Something like: mpiexec -n 1 myapp inp-file
: -n 1 -dir c:\temp\scratch1 : -n 1 -dir c:\temp\scratch2 : -n 1 -dir c:\temp\scratch3</font>
<br>
<br><font size=2 face="sans-serif">The application is an optimization that
only needs to run on the host machine. The three workers (different cores
on the same machine) only need to evaluate a simulation model embedded
in the application.</font>
<br><font size=2 face="sans-serif">The simulation model, however, needs
to produce an output file whose name cannot be changed so I just want it
to be output to a different directory for each worker.</font>
<br>
<br><font size=2 face="sans-serif">Is the first process always going to
be the master with rank 0?</font>
<br><font size=2 face="sans-serif">Do I need a copy of the application
and input file in each of the working directories?</font>
<br>
<br><font size=2 face="sans-serif">If I added myapp inp-file to each of
the worker commands, would the whole application just run separately on
all four processors or would they sync up and know that the workers only
have a specific task?</font>
<br>
<br><font size=2 face="sans-serif">mpiexec -n 1 myapp inp-file : -n 1 -dir
c:\temp\scratch1 myapp inp-file: -n 1 -dir c:\temp\scratch2 myapp inp-file:
-n 1 -dir c:\temp\scratch3 myapp inp-file</font>
<br>
<br><font size=2 face="sans-serif">I have tried this latter option with
both 3 and 6 workers (I am running on an eight core processor) but do not
seem to be getting any speedup so I'm wondering if I am doing this wrong.</font>
<br>
<br><font size=2 face="sans-serif">Thank you for your time,</font>
<br>
<br><font size=2 face="sans-serif">Zak</font>