Re: hacking on the "heap" implementation in gc.c

On Nov 1, 2007, at 2:01 PM, Hugh Sasse wrote:

> Would you consider renaming to gc_heap.[ch]?
> But IANM (I am not Matz), so I don't know if this matters REALLY.

If these changes ever get seriously considered, I'm flexible on all  
points of style.  gc_heap sounds good, and I'll make that change,  
along with some much needed cleanup (learning as I go)

>> Some preliminary results and my patches are available here:
>> http://lloydforge.org/projects/ruby
>
> Took me a while to spot the patches, I was looking for blue links.
> Could you put dates on them and mark which are obsolete?  It might be
> useful to update the text you say needs updating on there.

Indeed. the site is a mess.  I will reorganize it in the next day or  
two.  I'm a terrible web developer.

>>
>> What I would love is criticism on the approach, and some  
>> performance testing
>
> I don't know when I'll get chance to do that, but I'll try in due  
> course.
> I have some interesting tests based on Tom Copeland's problem

I can generate patched binaries and post them on the site along with  
the redesign, if this makes things easier for you.  For all I know  
these changes might be terrible in everything other than the handful  
of cases I've tried :)

What os/arch are you on?

>
> http://tomcopeland.blogs.com/juniordeveloper/2007/09/tracking-down- 
> a.html
>
> - I tried Yaml structures and a couple of other things to stress that.
> I'll send you what I have off list if you wish.  I suspect you have  
> better
> memory tools than me.

I would love whatever tests you have, and would integrate them into  
my suite.

best,
lloyd

>
>>
>> hope this is interesting,
>> lloyd
>
>         Thank you,
>         Hugh
>