Re: HAVE_FAST_SOFTINTS other than x86

[Izumi Tsutsui <tsutsui%ceres.dti.ne.jp@localhost>]

> > A larger Q would be whether all the NetBSD ports should
> > use HAVE_FAST_SOFTINTS anytime soon?  If not, what
> > prevents it from happening?
> 
> As many archs as possible should be converted (is there any arch that would
> not benefit? I don't see a good reason).
> 
> I think nothing prevents it from happening but touching interrupt dispatch
> code is always kinda scary ;-)
> 
> Anyone who wants to give it a go: it is documented in a big comment block
> right at the start of src/sys/kern/kern_softint.c.

FYI, there is a document written by yamt@ (in Japanese (un)fortunately),
"Implementation of software interrupt thread on NetBSD/i386 vmlocking branch" 
http://www.valinux.co.jp/contents/tech/techlib/eos/softint/softint_01.html
---
Izumi Tsutsui