Condensing bootstraps

[Remi Forax]

----- Original Message -----
> From: "Brian Goetz" <brian.goetz at oracle.com>
> To: "leyden-dev" <leyden-dev at openjdk.org>
> Sent: Wednesday, August 2, 2023 7:38:24 PM
> Subject: Condensing bootstraps

> I've spent some time thinking about how to approach condensing `indy`
> sites.  My conclusion is that the "obvious" approaches are not really
> optimal, so I've sketched out a path to bringing the solution further
> into the programming model here:
> 
> https://openjdk.org/projects/leyden/notes/04-condensing-bootstraps

I will take the problem from the other side, what is the result that a condensing bootstrap method should return.

For a long time, we have used the bytecodes more as an AST than as just a direct executable code.
I think the same is true for condensing bootstrap methods, there is no point of re-inventing a new AST when we can output bytecodes.

The only thing we may need is an offline bootstrap method returning a ClassDesc/MethodDesc pointing to the generated bytecode classfile, so solution 3 with no OfflineIndyLinkage seems to fit to that idea.

Rémi