On 02/08/11 19:42, Chris Angelico wrote: > On Tue, Aug 2, 2011 at 6:18 PM, Thomas Jollans <t...@jollybox.de> wrote: >> I suppose it would be possible to introduce a kind of "constant >> globals" namespace that a JIT compiler could then use to optimise, but >> how much would this help? > > Surely it must help a lot; looking up names is string operations. If > "len" could be replaced with "@10794928" where 10794928 is the actual > address of the len object, then it'd be doing no work that isn't > normally done, and would go straight to the object and call it. > > But I don't really know how to go about profiling this to be sure. Any ideas?
Well, you could run a direct comparison of one function where len is global, and an identitical function where len is local (which does not do a dict lookup, am I right?) Of course the global dict lookup takes time, but I doubt that many modules look up the same name often enough for it to actually be significant. I don't really know enough about either profiling or the CPython interpreter, but I assume there's some function that's called to look up globals; you could profile the Python interpreter (don't ask me with which tool) and see how much time that function uses. Thomas -- http://mail.python.org/mailman/listinfo/python-list
