whiteknight/vtable_overrides branch

Jonathan "Duke" Leto jonathan at leto.net
Sat Mar 5 20:05:52 UTC 2011


Howdy,

On Sat, Mar 5, 2011 at 5:37 AM, Andrew Whitworth <wknight8111 at gmail.com> wrote:
> In the whiteknight/vtable_overrides branch I have added the ability to
> override the isa, isa_pmc, and can vtables. Previously these things
> were not able to be overridden from PIR. I wouldn't necessarily call
> these operations extremely common, but there are places where the
> ability to override these behaviors can be important.

This is really awesome. Thanks for hacking on this.

>
> All tests pass for me on that branch, including new tests I've added
> just for the purpose. However, this does add a certain amount of
> overhead by needing to lookup potential vtable overrides first before
> carrying on with normal behavior. I suspect this is not going to be a
> huge expenditure, but it is something and we need to be aware of it.
>
> I would like some help both testing and benchmarking this branch
> against master. I don't know what kind of benchmark represents a
> "normal" use-case for isa, isa_pmc, and can vtables. I assume NQP,
> Rakudo, and other HLLs make heavier use of these things than most of
> the Parrot test suite does. If the results don't look too bad I would
> like to discuss merging this branch. If we do see some unacceptable
> slowdowns, we can scrap this or try to improve upon it.

Perhaps just running a Rakudo spectest and perhaps the Lua test suite
on master and this branch and comparing the results?

Not the best kind of benchmark, but better than nothing.

Duke

-- 
Jonathan "Duke" Leto
jonathan at leto.net
http://leto.net


More information about the parrot-dev mailing list