[Mono-dev] Ping on nternal call builders
kumpera at gmail.com
Tue Jul 29 10:00:48 EDT 2008
>From your email the advantage of the patch is that by replacing some icalls
with managed code we would get rid of some managed-native transitions.
But there are some side effects:
-It makes the new full AOT mode a bit more complicated as these would be
kind of wrappers that would need to be AOT'ed.
-Increases the footprint by just a bit, as we need to store the icalls
as well. The amount of extra code generated is bounded, specially since we
managed-to-native wrapper with a managed-to-managed with more code.
-JIT time should not change much, but this is only a guess.
I wonder if it is really an advantage of following this patch as, for
has 22 lines in your patch but 6 in the current setup. Of course the code in
your patch ends more
organized and I find it easier to follow, but should be interesting to rear
from the others.
Anyway, I think this patch is interesting, but here are my concerns:
-For simple methods such as OffsetToStringData the method call overhead
might be a killer. We
should make sure that these new wrappers get a change to be inlined.
-Does it work under trunk?
-Do you have performance numbers on your change? Since it changes
parts of the runtime, attaching a benchmark (or pointing to an existing one)
showing the implications is
-I'm not aware of how stack traces behave with wrappers, but are they
preserved with your patch?
2008/7/29 Kornél Pál <kornelpal at gmail.com>
> There is a pending patch:
> As far as I can remember I didn't get any feedback regarding this patch.
> Please review the patch. Thanks.
> Mono-devel-list mailing list
> Mono-devel-list at lists.ximian.com
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Mono-devel-list