Software Tools

SGI/Cray Research Performance Tools Update

Mary Kay Bunde
This talk will describe current offerings from Cray and SGI in application performance tools. Cray MPP tools, PAT and the MPP Apprentice, will be described, as well as SpeedShop, the SGI performance tool. Other performance tools will be touched on, but the focus will remain with those above.

Programming Environments Update

Dianna Crawford
This presentation will lay out the current plans for Programming Environments 3.0, including major purposes, initiatives and priorities; supported hardware and anticipated release timeframe.

High Performance PE 3.0 Vectorized Libm Intrinsics on the T3E

Neal Gaarder
This paper describes the new vectorized Libm capability which is in PE 3.0 on the T3E. These are selected Libm routines, such as SQRT, which operate on a ìvectorî of arguments in memory and return a ìvectorî of results in memory. These routines are typically three times faster than the scalar routines. Applications and performance will be discussed.

Library Update

Bill Harrod
This talk will focus on the new features and performance of the scientific, MPT, and FFIO libraries, as well as address migration issues from Cray to SGI platforms.

UNICOS to IRIX Applications Directions

Peter Rigsbee
The hardware and software roadmaps presented by Cray Research and Silicon Graphics show IRIX systems as the follow-on systems to traditional Cray architectures. These roadmaps, though, are at too high a level for application programmers to understand and assess the impact of these changes on their applications. This paper presents plans and on-going activities within Cray to ensure that this application migration proceeds smoothly.

How to Write Parallel Programs on the T3E Using Linda

Carlos Sosa and Nicholas Carriero
Linda is a model for parallel processing based on distributed data structures. Processors and data objects communicate by means of the tuple space (TS) or Linda memory. The interaction between the tuple space and the processors is carried out by four basic operations: out, in, eval, and rd. These operations add/remove/read data objects from/to the tuple space. Eval on the other hand also forks processes to initiate parallel execution. In this presentation we will discuss the use of Linda on the T3E.

Table of Contents | Author Index | CUG Home Page | Home (Title Page)