RE: Interesting Side-Effects Problem & Solution

Subject: RE: Interesting Side-Effects Problem & Solution
From: Avi Kivity <Avi@xxxxxxxxxxxxx>
Date: Fri, 11 Dec 1998 13:58:32 +0200
On Friday, December 11, 1998 11:59, ht@xxxxxxxxxxxxxxx
[SMTP:ht@xxxxxxxxxxxxxxx] wrote:
> There are two prices you are paying here, one you knew about and one
> you may have missed:
> 1) You've cut yourself off from the power of pattern-directed
> application:  no construction rules will get a chance once you take
> control;

Actually, you can, for complete subtrees, go back to the 'native' method
of doing things (as is done for non-element nodes).

> 2) You've serialised things:  no benefit from parallel processing is
> possible.

True - but the problem is serial in nature. There is no gain in parallel
if the amount of work increases in several orders of magintude.

> Neither of these is a killer, just worth noting.

One other thing - memory usage increases dramatically. Still have to figure
out why.
This may be a killer if it's not an implementation artifact.

"Thou shalt not travel faster than light" - Carl Sagan, Contact

 DSSSList info and archive:

Current Thread