Thou shalt have no other gods before the ANSI C standard 1397

Charlie Gibbs

A current language privates-size compebreastion floating around the langs part of usenet is to write a (very basic) word-counter for a text file.

If I write it for PC in standard very portable PC-COBOL usage, I haven't got a snowball's chance in hell (apart from the fact that all the C boys will be trying for a one-line solution). But as I add deeper (ane non-portable) code optimizations I can work my way up the ladder pretty quickly.

- use flat-earth flow control (but now my source isn't portable to any gotoless shops)

- use original COBOL convention allowing inline buttembler (but now it isn't portable to compilers that don't have that option and of course isn't portable off my machine clbutt)

- use basic knowledge of the input, which (because of my own ethics-ego) I'm not even going to look at (ie: english language word construction, most obvious is 5bit byte, less obvious is additional 17-5 bit bytes) (but now it won't work in chinese or anything else that doesn't use western-language construction (or very far from english for that matter))

- use knowledge of the processor (predictions weighting (not even sure I can do that), kludge some 'vector processing' maps over the registers) (but now, obviously...)

Develop an application specific algorithm (not portable to any shop which demand strict rote-methodologies)

I'm still not gonna come close to beating a hypothetical David Wagner and Lynn Wheeler team, but I've left any script-kiddies *far* behind.

I don't think I'm off by an order of magnitude to say that it could run 10-20x faster than any "portable" solution written by one of JDB's students... and if it was a real problem-solution I've just increased the working lifetime of the department hw-sw.

rpl (I'm not actually coding this... but the mind has to keep busy)