Re: MPS: native comp

[Andrea Corallo <acorallo@gnu.org>]

Gerd Möllmann <gerd.moellmann@gmail.com> writes:

> Gerd Möllmann <gerd.moellmann@gmail.com> writes:
>
>> Andrea Corallo <acorallo@gnu.org> writes:
>>
>>>> I'd personally prefer to trace comp units in the usual way for Lisp
>>>> objects, i.e. the fix_comp_unit.
>>>>
>>>> Main reason is simplicity. The tracing function is super simple, we only
>>>> have to identify where in the comp unit data structure, or dylib,
>>>> references exist that have to be traced. Registers/stack are not a
>>>> problem (see above).
>>>>
>>>> Other reasons, not so important, but anyway, is reducing the number of
>>>> roots. Every root adds to what MPS has to do. And ambig roots make
>>>> objects immovable, which is somewhat against the spirit of a copying
>>>> collector.
>>>>
>>>> Just my personal opinion, of course.
>>>
>>> Well other than spirit and possibly simplicity we should evaluate if
>>> (and in case how much) is the cost to have this objects movable vs
>>> non-movable.
>>
>> FYI:
>>
>> I have ambig roots for the relocs, as an experiment. That brings us
>> further. Bootstrap.pdmp is built successfully, and some .elns are
>> successfully built, too. It fails eventually in one file:
>>
>>   ./lisp.h:1631: Emacs fatal error: assertion failed: STRINGP (a)
>>   gmake[1]: *** [Makefile:285: ../lisp/files.elc] Abort trap: 6
>>   gmake: *** [Makefile:925: ../lisp/files.elc] Error 2
>>
>> and it's super slow. The reasons for that I don't know, but I know for a
>> fact that I didn't do anything for CUs in the loaded dump, which is
>> probably not helpful.
>>
>> I think I let this ripe a bit. Maybe I find the reference in macroexp,
>> and then don't need the roots. But I'll push it.
>
> Good news, to close the evening, letting it simmer without thinking much
> helped again. It's indeed simple: There are references we don't trace,
> as one would naively expect.
>
> - comp_unit
> - ephemral relocs

I think is not correct to add ephemral relocs to the compilation unit.
They are not reachable by the GC from that path on purpose, so that
after the load procedure they get GCed.

If MPS is scanning the stack as our GC does it should just work.  Are we
sure this modification was necessary?

  Andrea