diff options
author | Erik Schnetter <schnetter@cct.lsu.edu> | 2008-01-14 14:50:00 +0000 |
---|---|---|
committer | Erik Schnetter <schnetter@cct.lsu.edu> | 2008-01-14 14:50:00 +0000 |
commit | a10a5ba686830a2947d0235cb5ce2fb22bc14f4d (patch) | |
tree | 4d676b3f8725f3ce86c1715b60fbf5d7cc1bd07c /Carpet/CarpetWeb/index.html | |
parent | 6c052dc480be9267e3ab4eb4613d8fbf833fa3ae (diff) |
CarpetWeb: Add scaling results from Ranger
darcs-hash:20080114145039-dae7b-5c7a54e9c35863199f9faaa99b32325f249520cd.gz
Diffstat (limited to 'Carpet/CarpetWeb/index.html')
-rw-r--r-- | Carpet/CarpetWeb/index.html | 37 |
1 files changed, 36 insertions, 1 deletions
diff --git a/Carpet/CarpetWeb/index.html b/Carpet/CarpetWeb/index.html index 9e4d556a4..3de34d4e9 100644 --- a/Carpet/CarpetWeb/index.html +++ b/Carpet/CarpetWeb/index.html @@ -107,6 +107,41 @@ <h2>News</h2> + <table><tr><td valign="top"> + <p><b>January 14, 2008:</b> Carpet's communication + infrastructure has been improved significantly, making Carpet + scale to at least 4,000 processors, including mesh refinement. + Using "friendly user time" + on <a + href="http://www.tacc.utexas.edu/services/userguides/ranger/">Ranger</a>, + the new 60,000 + core <a href="http://www.teragrid.org/">TeraGrid</a> + supercomputer + at <a href="http://www.tacc.utexas.edu/">TACC</a>, we measured + the benchmark results below for a numerical relativity kernel + solving the BSSN equations. These benchmarks emply a hybrid + communication scheme + combining <a href="http://www-unix.mcs.anl.gov/mpi/">MPI</a> + and + <a href="http://www.openmp.org/">OpenMP</a>, using the shared + memory capabilities of Ranger's nodes to reduce the memory + overhead of parallelisation. We are grateful for the help we + received from Ranger's support team.</p> + <p>The graph below shows weak scaling tests for both unigrid and + mesh refinement benchmarks. The problem size per core was + kept fixed, and there were 4 OpenMP threads per MPI process, + with 1 MPI process per socket. The benchmark was also run + with the PUGH driver for comparison for certain core counts. + As the graphs show, this benchmark scales near perfectly for + unigrid, and has only small variations in run time for nine + levels of mesh refinement.</p> + </td><td valign="top"> + <p><a + href="scaling-ranger/results-ranger.pdf"><img + src="scaling-ranger/results-ranger.png" width="200" + alt="Scaling graph for Ranger"/></a></p> + </td></tr></table> + <!-- <p><b>October 4, 2007:</b> Announcement: <i>The BBH factory</i><br /> <b>Physicist version:</b> The BBH factory is a glorified set of shell @@ -185,7 +220,7 @@ example application without using I/O or analysis methods.</p> </td><td valign="top"> <p><a - href="hybrid-scaling/results-wavetoy-abe.png"><img + href="hybrid-scaling/results-wavetoy-abe.pdf"><img src="hybrid-scaling/results-wavetoy-abe.png" width="200" alt="Scaling graph for Abe"/></a></p> </td></tr></table> |