Hallo, Murray!

*** Am 20. Januar 1997 um 07:51 schrieb Murray Lesser an Thomas Seeling:

 ML>     I learned many years ago that, for best performance, one should
 ML> always put the most likely case first (even if it is the negative case).

This sounds reasonable (I learnt it, too), but modern optimizers turn e.g.
a switch into an array of function (or jump address) pointers, and some
even optimize if/else if/else if/.../else cascades this way.

In general I agree: write it readable, not "human optimized". let
the compiler do its best (there have been studies that the straight-forward
approach when coding an algorithm yields the best results after compiler
optimization, and the manually optimized code could not be optimized by the
compiler and was much worse than all other ones).

If there is still a bottleneck after all compiler optimization has been
done, start the profiler.


Tschau...Thomas

--- E3-32/1.11-32/2.51+