On May 7, 12:20 pm, Alberto Valverde <[EMAIL PROTECTED]> wrote: > BTW: Thanks for the detailed profile stats analysis! Very > enlightening... Any pointers to good reads for further learning?
http://wiki.python.org/moin/PythonSpeed But there are a few additional things I have learned while developing CherryPy and Dejavu: * Benchmark your code. Distribute a benchmark with your code if you can. Use it to pinpoint what changesets caused slowdowns. * Profile your code. Distribute a profiler with your code if you can. * Trace your code (with pdb or PyConquer [1]); the profiler output only goes so far, since it is a summary of call behavior. * Attack the largest potential benefits first; work your way from biggest bottleneck down. * Make those who need extra functionality pay for it (instead of making everyone pay). For example, every CP 2 filter made everyone pay a performance penalty whether they had it turned on or not; CP 3 tools, in contrast, have no cost if you don't use them. Be religious about what "extra functionality" really means (more than you think). * Pure-Python modules from the standard library are almost always more general-purpose than you need them to be. Rewrite them from scratch (if profiling dictates!). For example, CP's wsgiserver saw a big speedup (10%+ overall) by ditching the builtin mimetools module when reading request headers, and inlining HTTP-specific header logic. Robert Brewer System Architect Amor Ministries [EMAIL PROTECTED] [1] http://projects.amor.org/misc/wiki/PyConquer --~--~---------~--~----~------------~-------~--~----~ You received this message because you are subscribed to the Google Groups "TurboGears" group. To post to this group, send email to [email protected] To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/turbogears?hl=en -~----------~----~----~----~------~----~------~--~---
