-------------- Original message ---------------------- From: "Jeremy White" <[EMAIL PROTECTED]> > > >In particular, there are 366 entries in the search space. A binary search > >requires (at most) 9 probes for success and 10 probes for failure. This > >seems better than the existing algorithm. What do you think? > > > > I like the idea - in practise does this mean that win32-gui is noticeably > faster? >
I doubt that there would be any appreciable speedup since GUI_Constants is only called occassionally when GUI objects are created. However, my own personal feeling is that it is always a good idea to use the best code possible giving the best performance possible. The reason is that the converse leads to the development of many instances of poorly performing code. Although not one of them might be seen as causing a system performance issue, collectively they might have a different profile. So I dogmatically insist that nothing is too good for my friends. (Sorry for the digression. It's a sticky point with me.) Bye the bye, The algorithm was copied from Niklaus Wirth's book, Algorithms + Data Structures = Programs. I seem to recall that it might not work correctly and that at some poipnt I fixed it. I had trouble compiling in VC++ yesterday and (so) did not validate the code by calling it with each of the 366 entries. I'll try to do that this week. art
