Panda 3.0 Software Release

Panda 3.0 was released in 2001. It is based on our previous release Panda 2.1. Like Panda 2.1, it is written in C++, is designed to support application programs on distributed (shared) memory parallel architectures, supports server-directed collective I/O (SDIO) and reduced-message implementation of server-directed I/O (RMSDIO). But it was also modified and improved a lot from the previous release. Changes from Panda 2.1 and new features in Panda 3.0 include:

• New class hierarchies - 3 different classes for arrays (Array, ArrayInMemory, and ArrayOnDisk), previous MPIFS class and I/O related classes in Panda 2.1 are divided into smaller ones, new classes (File, Opt, Repository, etc.)
• Better and clearer user interfaces
• New data distribution methods - HPF CYCLIC(k) and irregular distribution using "gen-block"
• Hook-up for an I/O optimizer
• No support for ArrayGroup and Shared_IO any more

Panda 3.0 has been tested on an IBM SP2 and its performance is currently being measured. You may obtain the documentation and the source code of Panda 3.0 here.

• No user's guide! But both Rocpanda and Panda 2.1 have good ones.
• Panda 3.0 source code (2.7 MB compressed)

• /bin - executables
• /conf - configuration files
• /obj - compiled obj code *.o files
• /lib - libraries (libpanda.a)
• /doc - documentation
• /scripts - sample script files
• /src - Panda 3.0 alpha source code
• /apps - sample Panda application source code