<br><br><div class="gmail_quote"><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;"><br>Hello, every body. I have a question about parallel running of mdrun_mpi. I doubt that sometimes the parallel running of mdrun_mpi can not utilize the domain decomposition. <br>
This is the case when I use the batch work in the computer cluster which install the MPICH2. In the case, I use commond like ''mpirun -np 8 mdrun_mpi -deffnm *** -v'' . (The command ''mpirun -np 3 /usr/bin/mdrun_mpi -deffnm *** -v'' doesn't work) .The working didn't show any information about domain decomposition. And the computing rate is rather slow, which is only a little bit higher than single CPU.<br>
<br>However, in my person computer which install the MPICH (not MPICH 2), I use the commond like this ''mpirun -np 3 /usr/bin/mdrun_mpi -deffnm *** -v''. It would show the domain decomposition and the speed is quicked than 8 CPUS. <br>
<br>What is the problem???<br> <br clear="all"><br><br>
<br></blockquote></div><br><br>