[转载]FLOW-3D的高性能计算---并行(SMP / MPI)运算

标签:
转载 |
FLOW-3D
http://s12/bmiddle/616aad09g705d21d3aedb&690/
Shared-Memory Computers
FLOW-3D
SMP计算的优势
不同于分布式内存(MPI),MPI要求一个网格区块必须分配一个CPU,而SMP在计算之前,perpin文件不需要修改(与串行版本相同),唯一要注意的是需要一个SMP版的licence文件。
Another important advantage lies in the rapidly declining cost of dual-processor PCs such AMD's Opteron® and Intel's Xeon®. Add 64-bit capability, and the attractiveness of OpenMP, especially for certain types of simulations, is clear.
SMP计算的不足
使用OpenMP协议进行运算的加速效果会受限于计算机硬件和软件,由于OpenMP受到限制, FLOW-3D 的一些模型算法不能并行计算。而且, 单个计算机的最大内存容量也限制了流体问题的计算。
由于SMP方法自身的原因, 超过3或4个处理器进行计算的加速效果就会降低,如果超过8个处理器时,计算时间反而会增加。 (建议:使用SMP时,处理器的数量不要超过8个)
另外,
(建议:使用SMP时,网格数不要超过800万)
FLOW-3D/MP for Linux clusters
FLOW-3D 用户可以利用MPI版本来体验分布式内存在Linux集群的加速计算潜能。FLOW-3D/MP 3.2版是基于FLOW-3D version 9.2.1版,FLOW-3D/MP 3.2有32-bit和64-bit两个集群版本。
集群计算的优势
FLOW-3D/MP
FLOW-3D/MP的不足
计算量小的问题加速效果较小,而且充型计算时不会有加速效果。
For example, smaller problems will not scale well. Moreover, filling simulations will not scale as well as those where the entire domain is in play at all times during the simulation.
Requirements
SMP并行版本可以运行在Windows XP & Vista (32-bit and
64-bit)
为了运行FLOW-3D/MP, 用户必须提供一台集群运算服务器(a
group of similarly configured x86-32 or x86-64
processors),可以运行在Redhat Enterprise 3 or higher, Suse Enterprise 9
or higher, or Windows CCS 2003, a high speed network interconnect
such as Gigabit Ethernet or Infiniband, and a large shared NFS disk
accessible from all nodes in
cluster.