Understanding Performance of Parallel Scientific Simulation Codes using Open|SpeedShop
Conference
·
OSTI ID:1035294
Conclusions of this presentation are: (1) Open SpeedShop's (OSS) is convenient to use for large, parallel, scientific simulation codes; (2) Large codes benefit from uninstrumented execution; (3) Many experiments can be run in a short time - might need multiple shots e.g. usertime for caller-callee, hwcsamp for HW counters; (4) Decent idea of code's performance is easily obtained; (5) Statistical sampling calls for decent number of samples; and (6) HWC data is very useful for micro-analysis but can be tricky to analyze.
- Research Organization:
- Lawrence Livermore National Lab. (LLNL), Livermore, CA (United States)
- Sponsoring Organization:
- USDOE
- DOE Contract Number:
- W-7405-ENG-48
- OSTI ID:
- 1035294
- Report Number(s):
- LLNL-CONF-512249; TRN: US201204%%529
- Resource Relation:
- Conference: Presented at: SC'11 International Conference for High Performance Computing, Networking, Storage and Analysis, Seattle, WA, United States, Nov 12 - Nov 18, 2011
- Country of Publication:
- United States
- Language:
- English
Similar Records
Open | SpeedShop: An Open Source Infrastructure for Parallel Performance Analysis
Building a Community Infrastructure for Scalable On-Line Performance Analysis Tools around Open|Speedshop
Building a Community Infrastructure for Scalable On-Line Performance Analysis Tools around Open|SpeedShop
Journal Article
·
Tue Jan 01 00:00:00 EST 2008
· Scientific Programming
·
OSTI ID:1035294
+3 more
Building a Community Infrastructure for Scalable On-Line Performance Analysis Tools around Open|Speedshop
Technical Report
·
Mon Jun 30 00:00:00 EDT 2014
·
OSTI ID:1035294
Building a Community Infrastructure for Scalable On-Line Performance Analysis Tools around Open|SpeedShop
Technical Report
·
Thu Dec 19 00:00:00 EST 2013
·
OSTI ID:1035294
+2 more