Subject: Re: Commit NetBSD/hpcmips codes
To: TAKEMURA, Shin <takemura@netbsd.org>
From: Jonathan Stone <jonathan@DSG.Stanford.EDU>
List: port-mips
Date: 09/15/1999 12:08:43
In message <00d301beff59$1c203e80$0101a8c0@shin>,
"TAKEMURA, Shin" writes:

>I'd like to commit these changes once as they are and change
>#ifdef __hpcmips__ appropriately later because change logs will be
>recorded in cvs.netbsd.org. As soda-san says, it is more efficient.

I understand that goal, but I'd rather the #ifef __hpcmips__ defines
in platform-independent, mips-specific (sys/arch/mips) get cleaned up
before the commit.  Because that's the right thing to do.

I'm happy to leave the power-saving features as _hipcmips_ for now,
since we agree they need more thought. But code which is truly not
hpcmips specific (e.g.,, the switch element for r4100 CPUS, the
cycle-counter compare reload, and the fixes for no-L2-cache machines)
should be "fixed" to work on non-HPC mips platforms before committing.
That way Soda-san's arc code can use them, without having to
define __hpcmips__. Make sense?

>I will commit changes and new files soon if I were allowed.
>Cholud I do that ?

If the new files are ready, commit them. That's your call, or the hpc
portmaster, if we have a formal portmaster.

It's only the sys/arch/mips changes which are labelled _hpcmips_ but
which impact other mips work-in-progress, that I'd like to see done
right before we commit.

I guess if it's really holding things up, lets see if Soda-san and you
and I can sort something out between us.