Resurrection of whiteknight's eval_pmc branch

[Andrew Whitworth]

I haven't looked at gerdr's particular patch. My original formulation
was intended to be a drop-in replacement for the Eval PMC.

Because the PackfileView PMC version avoids nested runloops for :init
and :load Subs, the new version should be faster. I had clocked it at
one point to demonstrate a non-negligible performance bump for
compilation. As gerdr points out, the size of the bump may vary (but
I've never seen the new version run more slowly).

--Andrew Whitworth


On Sun, Feb 17, 2013 at 12:12 PM, Patrick R. Michaud <pmichaud at pobox.com> wrote:
> On Sun, Feb 17, 2013 at 05:32:14PM +0100, Gerhard R. wrote:
>> He opened a corresponding Rakudo pull request 10 months ago, but was
>> largely ignored because it (slightly) complicated the NQP side of
>> things without providing additional value.
>>
>> I now added a --target=pbc frontend to both NQP and Rakudo (see
>> https://github.com/rakudo/rakudo/pull/102 ) so a two-step build
>> process of modules is no longer necessary.
>
> Is there an updated pull request for NQP?  The old request
> at https://github.com/perl6/nqp/pull/36 can no longer be
> automatically merged.
>
> In general, there seem to be a lot of references to ".main_sub"
> in the NQP patch that I'm thinking shouldn't be there, or should
> be called something else.
>
> Looking at Rakudo #102, I'm a little concerned that the patch seems
> to assume that $precomp.main_sub() is an equivalent replacement
> to the old $precomp[0]... if I understand the eval_pmc branch
> correctly, that's not necessarily the case.
>
> Pm
> _______________________________________________
> http://lists.parrot.org/mailman/listinfo/parrot-dev