Thomas :<div class="gmail_quote"><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div class="im"><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<br>
</blockquote></div>
I attached my data to this mail. For the largest matrix, umfpack failed after allocating 4 GB of memory. I have not tried to figure out what's the problem there. As you can see, for these matrices the distributed solvers are</blockquote>
<div> </div><div> umfpack is a sequential package. 4GB+ likely exceeds memory capability of single core in your machine.</div><div><br></div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
slower by a factor of 2 or 3 compared to umfpack. For all solvers, I have used the standard parameters, so I have not played around with the permutation strategies and such things. This may be also the reason why superlu is much slower than superlu_dist with just one core as it makes use of different col and row permutation strategies.</blockquote>
<div> </div><div>The data on superlu_dist and mumps look reasonable to me. The poor parallel performance</div><div>is likely due to the multicore machine being used.</div><div>Try these runs on a machine that is desirable for distributed computing. See</div>
<div><a href="http://www.mcs.anl.gov/petsc/documentation/faq.html#computers">http://www.mcs.anl.gov/petsc/documentation/faq.html#computers</a>.</div><div><br></div><div>Hong</div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<div class="im"><br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
However, umpack will not work on a distributed memory machine.<br>
My personal preference is to use superlu_dist in parallel. In my<br>
experience using it as a coarse grid solver for multigrid, I find it<br>
much more reliable than mumps. However, when mumps works, its is<br>
typically slightly faster than superlu_dist. Again, not by a large<br>
amount - never more than a factor of 2 faster.<br>
</blockquote></div>
In my codes I also make use of the distributed direct solvers for the coarse grid problems. I just wanted to make some tests how far away these solvers are from the sequential counterparts.<span class="HOEnZb"><font color="#888888"><br>
<br>
Thomas</font></span><div class="HOEnZb"><div class="h5"><br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<br>
The failure rate using mumps is definitely higher (in my experience)<br>
when running on large numbers of cores compared to superlu_dist. I've<br>
never got to the bottom as to why it fails.<br>
<br>
Cheers,<br>
Dave<br>
<br>
<br>
On 15 May 2012 09:25, Thomas Witkowski<<a href="mailto:thomas.witkowski@tu-dresden.de" target="_blank">thomas.witkowski@tu-<u></u>dresden.de</a>> wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
I made some comparisons of using umfpack, superlu, superlu_dist and mumps to<br>
solve systems with sparse matrices arising from finite element method. The<br>
size of the matrices range from around 50000 to more than 3 million<br>
unknowns. I used 1, 2, 4, 8 and 16 nodes to make the benchmark. Now, I<br>
wonder that in all cases the sequential umfpack was the fastest one. So even<br>
with 16 cores, superlu_dist and mumps are slower. Can anybody of you confirm<br>
this observation? Are there any other parallel direct solvers around which<br>
are more efficient?<br>
<br>
Thomas<br>
</blockquote></blockquote>
<br>
</div></div></blockquote></div><br>