Differences

This shows you the differences between two versions of the page.

--- cluster:94 [2011/01/25 15:19] – hmeij
+++ cluster:94 [2011/01/25 16:39] (current) – hmeij
@@ Line 2: / Line 2: @@
 **[[cluster:0|Back]]**
+  * JAC and Factor_IX are two sample programs included with Amber (one memory intensive, one IO intensive, forgot what is what).
+  * "1g6r" is a program from Surjit Dixit that should scale really well in his opinion.
 ====== Swallowtail ======
@@ Line 20: / Line 23: @@
 Above some old benchmark runs against swallowtail's infiniband switch.
-Below new benchmarks with IPoIB.
+Below new benchmarks with IPoIB. "ptile" is the number of jobs slots to allocate per node, so with 16 requested and ptile=1 sixteen individual hosts will be selected and //all// MPI traffic between hosts will go via the Voltaire infiniband switch.
 ^  Amber  ^  What  ^  Host  ^  N  ^  JAC  ^  Factor_IX  ^ ptile= ^
@@ Line 29: / Line 32: @@
 |  9+openmpi-1.2+intel-9  |  Sander.MPI  |    |  16  |  0m42s  |  2m07s  |  1  |
-^  Amber  ^  What  ^  Host  ^  N  ^  1g6r  ^
+^  Amber  ^  What  ^  Host  ^  N  ^  1g6r  ^ ptile= ^
-|  9+openmpi-1.2+intel-9  |  pmemd  |  n1  |  02  |  2m55s  |
+|  9+openmpi-1.2+intel-9  |  pmemd  |  n1  |  02  |  2m40s  |    |
-|  9+openmpi-1.2+intel-9  |  pmemd  |  n1  |  04  |  1m32s  |
+|  9+openmpi-1.2+intel-9  |  pmemd  |  n1  |  04  |  1m23s  |    |
-|  9+openmpi-1.2+intel-9  |  pmemd  |  n1  |  08  |  1m03s  |
+|  9+openmpi-1.2+intel-9  |  pmemd  |  n1  |  08  |  0m47s  |    |
+|  9+openmpi-1.2+intel-9  |  pmemd  |    |  16  |  0m43s  |  8  |
+|  9+openmpi-1.2+intel-9  |  pmemd  |    |  16  |  0m39s  |  1  |