[antlr-interest] philosophy about translation

Jim Idle jimi at intersystems.com
Tue Oct 31 12:58:49 PST 2006



-----Original Message-----
From: antlr-interest-bounces at antlr.org [mailto:antlr-interest-bounces at antlr.org] On Behalf Of Andy Tripp 

> Sometimes they do, but sometimes they don't. Compilers never require 
> in-depth domain expertise. 

I think that in order to write very efficient code they do. This is especially true of higher level languages given that optimizations must be conservative (always generate working code). Ignoring algorithms, you can still write code that executes slower written one way than another because of the way it causes the execution level code to be generated. Optimizers get better all the time, but there are always limits of one sort or another. Without in-depth domain knowledge you are doomed to randomness.

> I know almost nothing about byte-code generation, yet I use javac every minute or two. 

That makes you, and the guys that did the byte code generation for Java ;-). The problem with Java is that for a lot of real world stuff it just isn’t useable, with the VM crashing, garbage (collection)? and so on just getting in the way. as we get further way from the machine programming gets sloppier because as well as being easy to write code it is much easier to write sloppy code. Sure you don’t need to malloc anything, but that's why you need 3,243,242MB of memory and things like confluence crash whereas twiki doesn't. ;-)

Jim

-- 
No virus found in this outgoing message.
Checked by AVG Free Edition.
Version: 7.1.409 / Virus Database: 268.13.19/507 - Release Date: 10/31/2006
 


More information about the antlr-interest mailing list