I dont think the fact its not in the  kernel is a big reason either ...
worst case  you could write a kernel module or driver and pass it the data
to do the same thing ... less elegant  but would work .

Jonothan Singularity had like 8 different types of GCs did they learn
anything about switching collectors ?
That said who is to say that the Azul collector is not 50% slower.,.. there
is very little information about this.. Im sure you can build some
syncronised GC with safe spots with minimal pauses but can you do it and
not trash cache hits and  generating lots of context switches.. Such a GC
is still usefull  for a large amount of data... but it becomes attractive
to use a 2nd allocator for other things and communicating between them
which is what David was saying  in terms of taking the load of the CG. I
can see it as say managing the gen 2 ,,

Re concurrent GCs yes i prefer to use the term pausless. there are
optomizations though in the .NET concurent GC the .NET one does it in
parralel for gen 2  ( they state that gen 0 and 1 collections are so fast
that its pointless)  . And since no new items can enter Gen 2 ( only
references from earlier colelction) its not that hard . Also these types of
collectors run much better in the real world where there are more low work
times then micro benches ...where there is no down time and once a
collection is required you stop the processing . David had a 3D C# game he
was going to try on this collector and was wondering how it went.

I also think the art of memory management is still usefull with GC .. but
little used , where in C it is often  used .. if you test it and create too
many objects reduce it .. eg if you have a non array linked list reuse your
nodes , if you have vertex buffers or buffer[]  reuse them .




On Tue, Jul 16, 2013 at 12:03 PM, Jonathan S. Shapiro <[email protected]>wrote:

> All:
>
> Just so it's clearly said, David and I agree that there is no single GC
> algorithm known that is ideal for all cases, and that there exist
> pathological cases for all known GC algorithms (just as there do for
> managed storage). I'm less concerned about the kernel changes required by
> the Azul collector and its friends than David is. The Linux crew was not
> helpful in admitting those changes, and the Azul crew didn't do a good job
> of getting buy-in. Both of those are solvable problems.
>
>
> Jonathan
>
> _______________________________________________
> bitc-dev mailing list
> [email protected]
> http://www.coyotos.org/mailman/listinfo/bitc-dev
>
>
_______________________________________________
bitc-dev mailing list
[email protected]
http://www.coyotos.org/mailman/listinfo/bitc-dev

Reply via email to