GC performance - was Re: [gclist] GC topics

[Darius Blasband]

> 
> >>
> >> I think that 20% is an acceptable overhead for GC. I also read
> >> somewhere that it is usually less (about 5%?) .
> >>
> >
> >I guess that the difference (between 5% and 20%) might come from the fact
> >that we do not generate native code directly from YAFL, but generate
> >intermediate C instead.
> 
> Are the two of you talking about the same thing?
> 
>   In my experience, measurements of 5% or less usually refer to the time spent
>   in garbage collection itself.  Measurements of 20% or higher usually refer
>   to the maintenance overhead of supporting the garbage collection protocol
>   (e.g. setting up run-time type tags, implementing read/write barriers,
>    following indirection pointers...)
> 

I don't know where the 5% came from, my 20% did indeed refer to what you
call "maintenance overhead".

Darius