Cluster Installation: Difference between revisions

← Older edit

Cluster Installation (view source)

Revision as of 14:48, 29 November 2019

511 bytes added , 29 November 2019

→‎Performance

Kay

Bureaucrats

2,652

edits

@@ Line 86: / Line 86: @@
 == Performance ==
-For the qsub command at the end of the script, it is not necessary to stick to the value of maxcpu as provided by $2 of forkxds (which is supplied by xds_par). If the cluster has nodes with more cores than the node where xds_par was started, it may be more efficient to override maxcpu, by grep-ing MAXIMUM_NUMBER_OF_PROCESSORS from XDS.INP, or by supplying it to the script from an environment variable, or to hard-code the value. If the cluster is homogeneous then this is of course no concern.
+Cluster nodes may have different numbers of processors.
+Please note that the output line
+ number of OpenMP threads used  NN
+in COLSPOT.LP and INTEGRATE.LP may be incorrect if MAXIMUM_NUMBER_OF_JOBS > 1, and the submitting node (the node that runs xds_par) has a different number of processors than the processing node(s) (the nodes that run mcolspot_par and mintegrate_par). The actual numbers of threads on the processing nodes may be obtained by
+ grep PARALLEL COLSPOT.LP
+ grep USING INTEGRATE.LP | uniq
+The algorithm that determines the number of threads used on a processing node is:
+ NB = DELPHI / OSCILLATION_RANGE   # this may be slightly adjusted by XDS if DATA_RANGE / NB is not integer
+ NCORE = number of processors in the processing node, obtained by OMP_GET_NUM_PROCS()
+ if MAXIMUM_NUMBER_OF_PROCESSORS is not specified in XDS.INP then MAXIMUM_NUMBER_OF_PROCESSORS = NCORE
+ number_of_threads = MIN( NB, NCORE, MAXIMUM_NUMBER_OF_PROCESSORS, 99 )
+This is implemented in BUILT=20191015 onwards.