Performance tuning

Performance tuning for OLPC

The OLPC is a very resource limited machine. Decreasing resource use on the OLPC translates into very tangible benefits such as more space on machine for user content, better system responsiveness, better battery life. There are two broad categories been looking at: processor use and memory use.

Processor utilization tools

• Sysprof (http://www.daimi.au.dk/~sandmann/sysprof/)
• Oprofile setup
• Python profiling
• Systemtap (http://www.mail-archive.com/devel@laptop.org/msg03436.html)
• Valgrind's callgrind tool to get call graph information (http://www.valgrind.org/)

Storage utilization

• Memphis (http://dev.laptop.org/git.do?p=projects/soc-memphis)
• Valgrind's massif tool to examine memory allocation (http://www.valgrind.org/)
• Reducing size of code (http://lovesunix.net/spaceoptimization.pdf)
• Improving shared libraries (http://people.redhat.com/drepper/dsohowto.pdf)
• Scripts to analyze file system and process memory use (http://people.redhat.com/berrange/olpc/image/)
• Scripts to analyze process memory map (http://people.redhat.com/berrange/olpc/performance/)
• Utility to analyze file sytem usage (http://www.pixelbeat.org/fslint/)
• Utility to analyze RAM usage (http://www.pixelbeat.org/scripts/ps_mem.py)

Other

• The Cairo graphics library has a trace utility upcoming, see

http://www.cairographics.org/FAQ/#performance_concerns