Re: "Constant" performance question

[CRPence <CRPbottle@xxxxxxxxx>]

Vern Hamberg wrote:
> Your point is well-taken. I remember a book on VB by Gary
> Cornell, where his 3 rules of optimization were 1. Don't 2. Don't
> 3. Don't. His basic point is the same as yours - don't do stuff
> that is "cute", hence, unmaintainable. The time spent on
> microseconds is better spent elsewhere. Seems it was different in
> the beginning of computing, when resources were scant and slow.
> But not anymore, usually.

  Applicable to application programming almost exclusively, IMO. I 
do not want OS programmers to just assume that some bigger & better 
hardware exists or will come along to solve any performance concerns 
that arise with their code :-)  Unfortunately however, it seems too 
often they do :-(  If something is performed enough times, even the 
smallest numbers in duration can sum up to what is perceived as a 
"long wait".

  Back on topic... A particular language in combination with the 
compiler or translator could optimize its constant data to remain in 
a register over many instructions referencing a datum without any 
re-loading into the register.  Such optimization with data that is 
not protected from change [i.e. a variable] is not as easy in 
general, but especially without also ensuring & enforcing 
differences between debug-capable and debug-incapable code.

Regards, Chuck