We present a runtime library design based on the two-phase collective I/O technique for irregular applications and show the performance results on the Intel Paragon. We obtained up to 40MBytes/sec. application level performance on the Caltech's Intel Paragon (with 16 I/O nodes, each containing one disk) which includes on-the-fly reordering costs. We observed up to 60MBytes/sec. on the ASCI/Red Teraflops with only three I/P nodes (with RAIDs).
|Original language||English (US)|
|Title of host publication||Advances in Parallel Computing|
|Number of pages||4|
|State||Published - Jan 1 1998|
|Name||Advances in Parallel Computing|
ASJC Scopus subject areas
- Computer Science(all)