Terry Hancock wrote: > * The claim that a Human can optimize code better than > the compiler assumes a very smart and talented Human, > and/or a very dumb compiler. Compilers are getting smarter, > and since a lot more people find the need to program, the > average programmer is not that talented. I'm pretty I'm > not, for example -- I'll take my chances with the compiler.
More importantly IMO is that humans cannot effectively optimise code without profiling - and compilers are not very good at optimising poor algorithms either. Humans are *really* bad at guessing where bottlenecks are. I had a case recently where the *major* performance bottleneck in a (Java) system was ... equals() and hashCode(). These objects were being used in lots of hash maps and similar things. The absolute best performance improvement I was able to make was to cache the hash code, and reduce the number of times that the expensive parts of the equals() comparison were called. It's a lot easier to refactor sub-optimal code in Python than in most other languages. This is tied directly to the fact that it's easier (and shorter) to express many concepts/algorithms in Python than in other languages. This further contributes to it being more fun to code in Python. Tim Delaney -- http://mail.python.org/mailman/listinfo/python-list
