Re: size of parse table?

Hello,


Also, today's modern machines may provide, say 64M to 128M of memory,
but if the user of our compiler is running Windows or NT, for
instance, he/she may not appreciate our using a _large_ chuck of this,
when it may limit the number of other tasks that can efficiently be
run. (Virtual memory is a lot slower, of course, than RAM and even
disk space for the VM runs out at some point, not to mention the
paging overhead :-(


And, of course, those tables not only take memory space while the
compiler is running, but take up disk space in the module as it
resides on disk. High capacity of modern disk technology is not an
execuse to waste disk real estate.


Of course, these things are not cut in stone...it must always be a
trade-off based on weighing the resources against the goals of the
compiler project :-) (And, of course, the capabilities of the tools
available to the compiler writer!)


Of course there is another axiom to consider: Make it right, then make
it smaller; Make it right, then make it faster :-)


--
Charles cbrtjr@ix.netcom.com
* http://pw2.netcom.com/~cbrtjr/wrdthing.html *


Josef Grosch <grosch@cocolab.de> wrote in message
> Our moderator wrote:
>
> > [How important is it to pack parse tables these days? It was a hot
> > topic back when your entire compiler had to fit in 64K, but now that
> > PC applications routinely contain a megabyte of garbage because the
> > programmer doesn't bother to delete unused templates and such, why
> > bother? -John]
>
> Packing parse tables is not important for parsers for small languages,
> except in embedded systems or in antique and restricted environments.
> However, it is still an issue for huge parsers, in my opinion.

Return to the comp.compilers page.
Search the comp.compilers archives again.