readme

mpi并行计算的c++代码 可用vc或gcc编译通过 可以用来搭建并行计算试验环境

----------------------------------------------------------------------

Mellanox has contributed substantially to the code for the ib channel.
To our original implementation based on the Infiniband send/receive 
model, Mellanox added code to use the RDMA/poll model to reduce 
latency, code to align buffers to cache line sizes, and code to handle 
registration failures for large messages by falling back on the eager 
message protocol.

The code contributed by Mellanox does not appear verbatim, but has 
been massaged and merged into the existing MPICH2 code.  We thank 
Mellanox for their contributions, which have substantially improved 
the performance of this channel.  Mellanox has made this contribution 
available under the same conditions as we Make MPICH2 available (see 
the COPYRIGHT file in the top-level directory: "Permission is hereby 
granted to use, reproduce, prepare derivative works, and to 
redistribute to others."

-----------------------------------------------------------------------