Greetings Open MPI users; we thought you&#39;d be interested in the&nbsp; <br>following announcement...<br><br>A new supercomputer, powered by Open MPI, has broken the petaflop&nbsp; <br>barrier to become the world's fastest supercomputer.&nbsp; The&nbsp; <br>
&quot;Roadrunner&quot; system was jointly developed by Los Alamos National&nbsp; <br>Laboratories and IBM.&nbsp; Roadrunner's design uses a cluster of AMD&nbsp; <br>dual-core processors coupled with computational accelerators based&nbsp; <br>
on the IBM Cell Broadband Engine.&nbsp; The cluster consists of 3,060&nbsp; <br>nodes, each of which has 2 dual-core AMD processors and associated&nbsp; <br>Cell accelerators.&nbsp; The AMD nodes are connected with 4x DDR&nbsp; <br>InfiniBand links.<br>
<br>Open MPI was used as the communications library for the 12,240&nbsp; <br>processes comprising the Linpack run which broke the Petaflop&nbsp; <br>barrier at 1.026 Petaflop/s.&nbsp; The version of Open MPI used in the&nbsp; <br>run-for-record was a pre-release version of the upcoming 1.3&nbsp; <br>
release.&nbsp; Enhancements in this release include modifications for&nbsp; <br>efficient, scalable process launch.&nbsp; As such, Open MPI was run&nbsp; <br>unmodified from a snapshot of the pre-1.3 source base (meaning:&nbsp; <br>there are no Roadrunner-specific enhancements that are unportable to&nbsp; <br>
other environments -- all Open MPI users benefit from the&nbsp; <br>scalability and performance improvements contributed by the&nbsp; <br>Roadrunner project).<br><br>--Brad Benton<br>Open MPI/Roadrunner Team<br>