Am Montag, 2. Juli 2007 schrieb Alan Stern:
> On Mon, 2 Jul 2007, Oliver Neukum wrote:
> 
> > > I don't think so.  For one thing, we'd be allocating fewer URBs.  For 
> > > another, the total number of submissions would be the same; they would 
> > > just be spread out in time instead of all at once.
> > 
> > But the number of interrupts would grow. The ideal number of interrupts
> > per transfer is 1. If you can avoid using more by using a bit more
> > memory, is a wińning strategy.
> 
> Currently the number of interrupts per transfer is larger than 1.  Of 
> course we can change that, but should we?  On small systems, saving a 
> little CPU time by using a lot more memory is not a win.

It is the current design goal. The sg code requests an interrupt only for
the last element of the scatter list.

> > > Actually, the best way to approach this would be to relax the guarantee
> > > that completion routines are called with interrupts disabled.  There's
> > > no real reason for that guarantee; it's just an historical remnant.
> > 
> > It speeds up execution in real interrupts, which is good. 
> 
> I don't buy that.  Leaving interrupts disabled 90% of the time would
> also speed up execution.  But it would ruin latency.
> 
> > Completion handlers might be called in a bottom half, but this
> > is a rather intrusive change.
> 
> We don't need bottom halves.  Just remove the guarantee that interrupts
> will be disabled.

OK, very well. How shall we split the audit load? I volunteer for
drivers/usb/serial

        Regards
                Oliver

-------------------------------------------------------------------------
This SF.net email is sponsored by DB2 Express
Download DB2 Express C - the FREE version of DB2 express and take
control of your XML. No limits. Just data. Click to get it now.
http://sourceforge.net/powerbar/db2/
_______________________________________________
linux-usb-devel@lists.sourceforge.net
To unsubscribe, use the last form field at:
https://lists.sourceforge.net/lists/listinfo/linux-usb-devel

Reply via email to