[haiku-commits] Re: r41864 - haiku/trunk/src/add-ons/kernel/drivers/input/usb_hid
- From: "Michael Lotz" <mmlr@xxxxxxxx>
- To: haiku-commits@xxxxxxxxxxxxx
- Date: Mon, 13 Jun 2011 14:29:53 +0200
> On Thu, 02 Jun 2011 11:34:04 +1200, <mmlr@xxxxxxxx> wrote:
>
> > Author: mmlr
> > Date: 2011-06-02 01:34:03 +0200 (Thu, 02 Jun 2011)
> > New Revision: 41864
> > Changeset: https://dev.haiku-os.org/changeset/41864
> >
> > Modified:
> > haiku/trunk/src/add-
> > ons/kernel/drivers/input/usb_hid/MouseProtocolHandler.cpp
> > haiku/trunk/src/add-
> > ons/kernel/drivers/input/usb_hid/MouseProtocolHandler.h
> > Log:
> > Since we never actually read more than one mouse_movement at a time
> > and
> > we only
> > ever wait for reports on demand, there's no need for buffering at
> > all.
> > This removes some unnecessary copying and ring buffer overhead, the
> > ring
> > buffer
> > will now in fact not be created at all.
> >
>
> just updated my system and my usb mouse hangs under heavy load now,
> i.e.
> when compiling the haiku source the mouse stagnates. Any relation to
> this
> changes?
No. As I already explained in the other thread, the process was already
fully synchronous before. It would just needlessly put the data into a
ring buffer and then read it out again making it:
ioctl
wait for report
process report
write processed data into ring buffer
read processed data from ring buffer
return that data
The ring buffer write and read have now been replaced by directly
providing the wait/process method with the ioctl buffer so it can be
filled directly. It doesn't change anything on the process itself. If
anything, the latency of the reads would now be smaller and the mouse
resolution better.
Regards
Michael
Other related posts:
