On Saturday 17 March 2007, Joerg Sonnenberger wrote:
> On Sat, Mar 17, 2007 at 04:47:44PM +0900, YAMAMOTO Takashi wrote:
> [snip]
>
> > you want to use ipi in the first place because you want to run these
> > operations on each cpus.  it doesn't make sense unless the callback
> > only operates the cpu on which it's running.
>
> Can we use this problem to introduce a MI interface for running a
> callback on a different CPU?

rmind@ said that yamt-idlelwp will have something for this, until then
as ad@ told me, there's no other way for this.

> This can be done in multiple ways, e.g. by moving the current thread
> voluntarily to a different CPU, by using a messsage thread o neach CPU
> to processes a queue etc. This needs some discussion of course, but can
> we at least do it now after it was shown that the existing MD ways are
> difficult to use correctly?

I just did what people suggested, the code is there, it works. 
EST/POWERNOW+SMP now works, what's the problem until we have the
proper API?

-- 

http://plog.xtrarom.org/
Juan RP's blog - NetBSD/pkgsrc news in Spanish