> On Saturday 07 April 2007 08:21, YAMAMOTO Takashi wrote:
> > > On Friday 06 April 2007 03:22, YAMAMOTO Takashi wrote:
> > > > > > how about having set_curlwp() which can be overridden by MD
> > > > > > definition?
> > > > >
> > > > > Did you envision a new place where it would be used?
> > > >
> > > > - MI switching code. (namely mi_switch and lwp_startup)
> > > > - main for boot cpu.  MD cpu startup code for other cpus.
> > > >
> > > > i want to make asm code minimal.
> > >
> > > Will the possibility to provide optimised asm versions of large use
> > > functions be removed?
> > >
> > > Nick
> >
> > can "curlwp = l" be optimized a lot if it was a part of cpu_switchto?
> 
> This isn't was I was thinking about. __HAVE_MD_RUNQUEUE seems to die on 
> yamt-idlelwp. Can't it be tied to sched_4bsd?
> 
> Nick

probably it could, but i personally don't think it's worth.
even sched_4bsd will not likely stay as-is.
eg. when we implement processor sets, etc...

oops, it seems that __HAVE_MD_RUNQUEUE has been broken on trunk
wrt priority inheritance, because MD code still uses raw l_priority.
my bad.

YAMAMOTO Takashi