Ja - folgende Info direkt aus dem Internet:
In Pro/MECHANICA STRUCTURE/THERMAL 2001 parallel processing is enabled again. Parallel processing allows the engine job to run on multiple CPUs at the same time, with the goal of reducing the elapsed time for the job.
For MEC/S/T 2001, the following platforms support parallel processing:
i486_nt
ibm_rs6000
sgi_elf4
sgi_mips4
sun4_solaris
To request parallel processing on these platforms set the environment variable MEC_NUM_THREADS. Example:
Unix, csh: setenv MEC_NUM_THREADS n
Unix, ksh: export MEC_NUM_THREADS=n
Windows: in DOS Shell: set MEC_NUM_THREADS=n
or in the #Control Panel, #System, Environment
where n is the number of idle CPUs on the system. If the Structure/Thermal engine is the only job running on the machine, the number of processors can be selected automatically by setting MEC_NUM_THREADS to "-1". Example for csh:
setenv MEC_NUM_THREADS -1
If MEC_NUM_THREADS is set greater than the number of idle CPUs, the parallel tasks may interfere with each other, and cause performance to drop.
The number of parallel processors requested can be verified at the top of the summary file <study>.rpt. For four-way parallel processing, the <study>.rpt file looks like this:
------------------------------------------------------------
Pro/MECHANICA STRUCTURE Version 23.3(311)
Summary for Design Study "anlys1"
Sat Jan 6, 2001 01:48:12
------------------------------------------------------------
Run Settings
Memory allocation for block solver: 1000.0
Parallel Processing Status
Parallel task limit for current run: 4
Parallel task limit for current platform: 64
Number of processors detected automatically: 4
Pro/MECHANICA STRUCTURE Model Summary
. . .
If a significant portion of the job truly ran in parallel mode, the CPU time for the job (or a job step) will be greater than the elapsed time. This can be verified in the log file <study>.stt. For example, the equation solver step for this job may print a message similar to:
Begin Equation Solve, Pass 1
Sat Jan 6, 2001 01:49:47
Elapsed Time (sec): 147.52
CPU Time (sec): 145.29
Memory Usage (kb): 411200
Work Dir Disk Usage (kb): 356353
Step Elapsed Time (sec): 65.68
Step CPU Time (sec): 115.59
Note that this job step accumulated about 116 seconds of CPU time during about 66 seconds of elapsed time. CPU time can exceed elapsed time only if several processors are running in parallel.
When requesting parallel processing, and CPU time remained less than elapsed time for all job steps, the job may not have run very long in parallel mode. For example, the model may have been too small to allow parallel mode to operate efficiently. Parallel performance may also be low if the computer system is busy running several different applications, or the memory and I/O system is not fast enough to supply parallel CPUs.
Parallel performance is likely to be best for:
o Large models which have a chunky overall shape.
o Computer systems with large RAM, fast swap and disk I/O systems, and several idle CPUs.
Increasing RAM, and improving swap and disk I/O are likely to improve performance even for jobs that do not benefit greatly from parallel processing alone.
Eine Antwort auf diesen Beitrag verfassen (mit Zitat/Zitat des Beitrags) IP
Foren auf CAD.de (alle Foren)
| 
