AOT/Native-image support

[tisonkun]

Is it in considered that MoarVM supports AOT/native-image so one is able to compile a Perl 6/NQP program to an executable running on machine without MoarVM?

[jnthn]

No, this isn't supported today.

The most likely path to something like in the medium term is building a tool that builds an executable with a static MoarVM linked in to it, and all the precompilations of modules embedded, plus a custom module loader that can locate these.

That's probably blocked at the moment by the Rakudo extops, which rely on being loaded dynamically and would frustrate such an effort. However, those are going away, which would unblock work on such a tool.

That's still not really AOT, but it does give a "one executable to distribute" solution.

[tisonkun]

The most likely path to something like in the medium term is building a tool that builds an executable with a static MoarVM linked in to it, and all the precompilations of modules embedded, plus a custom module loader that can locate these.

Yes it's common approach of AOT at present.

I'm glad to hear that the blocker has gone, and this issue is as a proposal. There are many benefits if we implement AOT. The most exciting might be efficiency improvement, but the most important, our user base would be expanded( user always likes new things :-) ).

[tbrowder]

This is one of the accepted Perl 6 GSoC 2019 projects.

[tisonkun]

close as no longer interested in.

[AlexDaniel]

It's a topic for a GSoC project, so I don't think it can be closed…

[vhatzist]

Is there an update on the possibility of producing a stand-alone executable?

[raiph]

(Edited link in 2024 because there's something very odd going on with the overall blog link.)

@vhatzist See https://yakshavingcream.blogspot.com/2019/08/summer-in-review.html

[JJ]

I get this is still residing in an NQP branch, but there are no inmediate plans to incoporate it, right?

[jnthn]

I'm not aware of anyone actively working on it right now. However, I think easily producing a standalone image for Raku stuff still is blocked by the extops that Rakudo uses. Recently I eliminated a big chunk of the code in there, and the current dispatch work will get rid of another big chunk of it, so at least by the time somebody is up for working on this again, they'll hopefully not have the extops blocker (or it'll be much easier to deal with).