I would suggest that tkchia and perhaps some of us should talk to the
GCC devs and try again. ia16 is NOT a variant of ia32 because ia32
assumes a lot of things that aren't true for ia16. No segmentation to
deal with is a massive difference, lack of a huge pile of ia32
instructions is another. It would be far preferable for tkchia's efforts
to be in mainline GCC instead of in a niche fork; there are back-end
improvements in every major release and with our current 64K segment
limits we can use all the clever intermediate optimization we can get.
On 3/13/2019 7:00 PM, Marc-F. Lucca-Daniau wrote:
From what I understood, the IA16 back-end has been proposed to
mainstream by its initial author (Lambertsen), but that proposal was
rejected by the GCC maintainers because they considered such back-end
to be a variant of the IA32 one, while the author considered its
nature to make it distinct. I don't know if the next maintainers
(Jenner, Chia) tried again to push to mainstream ?
Current version of GCC-IA16 is based on the 6.3 release. 6.x line is
not the latest one (8.x), but is still maintained, as the 6.5 was
released on 2018-10-26. It should not be too hard to keep the current
fork in sync with such 6.x line. For next lines, I hardly see what
could be the effect on the IA16 back-end efficiency and thus our code ?
Whatever the future of this fork, we avoided to "sink with the DEV86
boat" and we are safe for the next years, because it was really in a
bad shape: DEV86 was somewhat "hijacked" by Lubomir Rintel who has not
responded since 2015 to issues & pull requests on GitHub, and,
something I consider as really unfair, did not even responded to
Robert de Bath himself to his proposal to transfer ownership to an
open organization, at least to give a chance to DEV86 to stay alive
and maybe evolve.
This was one reason why I focused last months on that DEV86 to
GCC-IA16 transition, because I did not want the ELKS project to suffer
any more of that pathetic situation.
MFLD.fr
Le 13/03/2019 à 21:08, David Given a écrit :
As someone who owns four different 8086-based laptops, with an option
on another one... good news!
Do you know if gcc-ia16 is ever going to get pushed to upstream?
On Wed, 13 Mar 2019 at 20:03 Marc-F. Lucca-Daniau <mfld.fr@xxxxxxxxx
<mailto:mfld.fr@xxxxxxxxx>> wrote:
All,
We are close to end a (very) long discussion about the next
compiler for
ELKS...
ELKS project (see https://github.com/jbruchon/elks) was tightly
coupled
to DEV86 until last year. ELKS is now finishing its transition to
a full
build chain based on GCC-IA16 (see
https://github.com/tkchia/gcc-ia16)
and will be no more linked to DEV86.
There are still many components in DEV86 related to ELKS, but they
don't
need to be maintained anymore by ELKS contributors. So please
consider
them as 'residual' (or 'historical') elements.
This transition could have not been completed without the huge
preparation work performed by /Juan Perez-Sanchez, who early
tested &
tuned ELKS on various compilers, and the very proactive support of
//TK
Chia, who currently maintains the GCC-IA16 fork. Special thanks to
both
of them./
/Now ELKS takes advantage of a modern compiler that makes room in
our
64K segment for new features, and raises plenty of new warnings to
take
care of :-)/
/Enjoy !/
/MFLD.fr/
