Re: gc issues

[Michael Livshin]

Mikael Djurfeldt <address@hidden> writes:

> The original idea was, as you said earlier, to stop all threads during
> GC.  Furthermore, the idea was not to stop them at any point but at a
> well-defined point, like SCM_TICK or some other macro which we can
> spread out through the Guile source.  (The original idea was to do it
> in SCM_ALLOW_INTS.)
> 
> If we do this, we are guaranteed at GC time that newly initialized
> cells are in a consistent state.

and this is fine, except I'm not sure whether this guarantee is really
worth the cost of the additional syncronisation.

it seems to me that this is an issue that is better decided by
implementing both variants and measuring them than by speculating on a 
mailing list.

you surely thought much more about the subject than me, though...

-- 
programmer, n:
        A red eyed, mumbling mammal capable of conversing with inanimate
        monsters.