Subject: Re: com rumblings...
To: Izumi Tsutsui <firstname.lastname@example.org>
From: Garrett D'Amore <email@example.com>
Date: 06/16/2006 22:43:49
Izumi Tsutsui wrote:
> firstname.lastname@example.org wrote:
>> Its not just
>> com_arbus.c, btw, but really com_aubus.c that is most interesting.
> IIRC, you asked to introduce regmap for com_arbus.c on the first
> discission, and then mentioned it was also required for aucom.c.
> I think that might cause unfortunate disagreement at that time.
> IMHO, stride is still proper for com_arbus (and similar ones),
> and regmap is acceptable only for variants which have actually
> non-standard registers even on your current implementation.
> Of course people will be able to cut corners by the existing
> regmap on future implementation, but I doubt it's worth to
> convert existing stride ops to regmap.
I disagree. Having to implement an _entire_ bus space just to
accommodate a single chip is *stupid*. I've already converted
com_arbus.c, and it greatly shortened the code. This is significant on
this platform which has only 8MB RAM, though I've not checked to see how
much the _binary_ reduced.
The other hack for something like this is in hpcmips (com_hpcio.c).
I've left that stride in place, because I can't really test for it.
Anyway, what I have works, imposes no noticeable performance penalty,
and is easily maintainable. Is there some reason we are arguing about this?
> Just my two yen,
> Izumi Tsutsui
Garrett D'Amore, Principal Software Engineer
Tadpole Computer / Computing Technologies Division,
General Dynamics C4 Systems
Phone: 951 325-2134 Fax: 951 325-2191