[gclist] Cost of gc

[Hans Boehm]

I just posted a message to this list arguing that that was a major problem with
fully "accurate" garbage collection in mixed language environments.  This is
largely avoidable if you conservatively scan C data areas, as I think the
Java/C interface should have insisted on.  Unfortunately, it didn't.  And Sun
and Microsoft decided to insist on different, equally complex, interfaces.

This is my personal opinion.

Hans

On Dec 18,  7:10pm, Olivier Lefevre wrote:
> Subject: [gclist] Cost of gc
>
> No, it's not what you think :) I am not talking of slowdown but
> of this, which was recently posted on comp.lang.python by Guido
> van Rossum (the author of python), in reply to someone who
> advocated gc for python on the ground that "It greatly simplifies
> programming at only a small cost."
>
> > Before you get all excited, I recommend you talk to programmers who
> > have written native C/C++ code that links into a Java runtime. There
> > are costs there, too!  The costs are in terms of what you have to do
> > to keep your data from being garbage collected while you are working
> > on it.  I am told it can be quite painful.
>
> Can the gc experts who read this list comment on the above?
>
> Regards,
>
> -- O.L.
>-- End of excerpt from Olivier Lefevre



-- 
Hans-Juergen Boehm
boehm@mti.sgi.com