Linux Clusters Applications Tutorial
This four hour tutorial will describe how to port and optimize codes to Linux clusters, like the ones that have been built at Alliance sites, and also those planned as part of the TeraGrid project.
The attendees will get step by step instructions on how to build their codes on a Linux cluster, set up batch scripts for job submission, move data to and from the cluster. The impact of the memory hierarchy on performance will discussed in quantitative terms. Tips on the effective use of the compiler, and workarounds/troubleshooting, I/O issues, optimized libraries will be a focus of the tutorial. Attendees will learn about tools for getting hardware counter information and performance analysis of their codes. Finally, some users will describe their experience porting their codes to Linux clusters, and what kind of performance they are getting.
|Introduction to Porting Codes to Linux Clusters
- Architecture overview, connecting, file transfer, compiling, linking, PBS/batch, VMI/mpich, storage
- Application example
- Compiler tips, interpreting listings and reports (flags, PGO)
- 32- vs. 64-bit issues
- Available Libraries: HDF, MKL, Parallel Numerical Libraries
- Kernels/benchmarks, impact of memory hierarchy