[Qemu-devel] Re: [RFC][PATCH] Add HPET emulation to qemu (v2)

[Qemu-devel] Re: [RFC][PATCH] Add HPET emulation to qemu (v2)

[Samuel Thibault]

Beth Kon wrote some time ago:
> But for an idle guest in both cases (with and without HPET), the
> number of wakeups per second was relatively low (28).

Oh, btw, did you try to reduce the qemu window? (or disconnect the VNC
client if you are using VNC)  28 might be just because of the gui refresh
rate.  Reducing the window will lower the gui refresh rate to 2Hz.

Samuel

Re: [Qemu-devel] Re: [RFC][PATCH] Add HPET emulation to qemu (v2)

[Samuel Thibault]

Samuel Thibault, le Thu 07 Aug 2008 18:17:05 +0100, a écrit :
> Beth Kon wrote some time ago:
> > But for an idle guest in both cases (with and without HPET), the
> > number of wakeups per second was relatively low (28).
> 
> Oh, btw, did you try to reduce the qemu window?

Err, maybe it's "minimize" in correct English :)

Samuel

Re: [Qemu-devel] Re: [RFC][PATCH] Add HPET emulation to qemu (v2)

[Anthony Liguori]

Samuel Thibault wrote:
> Beth Kon wrote some time ago:
>   
>> But for an idle guest in both cases (with and without HPET), the
>> number of wakeups per second was relatively low (28).
>>     
>
> Oh, btw, did you try to reduce the qemu window? (or disconnect the VNC
> client if you are using VNC)  28 might be just because of the gui refresh
> rate.  Reducing the window will lower the gui refresh rate to 2Hz.
>   

It was with VNC (and I don't think a VNC client was connected).  Beth is 
on vacation this week but she can respond when she gets back next week.

Regards,

Anthony Liguori

> Samuel
>
>
>   

Re: [Qemu-devel] Re: [RFC][PATCH] Add HPET emulation to qemu (v2)

[Jamie Lokier]

Samuel Thibault wrote:
> Beth Kon wrote some time ago:
> > But for an idle guest in both cases (with and without HPET), the
> > number of wakeups per second was relatively low (28).
> 
> Oh, btw, did you try to reduce the qemu window? (or disconnect the VNC
> client if you are using VNC)  28 might be just because of the gui refresh
> rate.  Reducing the window will lower the gui refresh rate to 2Hz.

If the guest isn't drawing anything to video memory, and there are no
GUI events (e.g. redraw requests), why does the gui keep refreshing?

It shouldn't refresh at 2Hz either, it should just stop if there's no
video activity.  You should be able to run an idle tickless guest with
no host CPU at all.

-- Jamie

Re: [Qemu-devel] Re: [RFC][PATCH] Add HPET emulation to qemu (v2)

[Samuel Thibault]

Jamie Lokier, le Sun 10 Aug 2008 02:19:01 +0100, a écrit :
> Samuel Thibault wrote:
> > Beth Kon wrote some time ago:
> > > But for an idle guest in both cases (with and without HPET), the
> > > number of wakeups per second was relatively low (28).
> > 
> > Oh, btw, did you try to reduce the qemu window? (or disconnect the VNC
> > client if you are using VNC)  28 might be just because of the gui refresh
> > rate.  Reducing the window will lower the gui refresh rate to 2Hz.
> 
> If the guest isn't drawing anything to video memory, and there are no
> GUI events (e.g. redraw requests), why does the gui keep refreshing?

It needs to poll the keyboard activity at least, IIRC SDL doesn't
provide the X FD to poll.

Samuel

Re: [Qemu-devel] Re: [RFC][PATCH] Add HPET emulation to qemu (v2)

[Jamie Lokier]

Samuel Thibault wrote:
> Jamie Lokier, le Sun 10 Aug 2008 02:19:01 +0100, a écrit :
> > Samuel Thibault wrote:
> > > Beth Kon wrote some time ago:
> > > > But for an idle guest in both cases (with and without HPET), the
> > > > number of wakeups per second was relatively low (28).
> > > 
> > > Oh, btw, did you try to reduce the qemu window? (or disconnect the VNC
> > > client if you are using VNC)  28 might be just because of the gui refresh
> > > rate.  Reducing the window will lower the gui refresh rate to 2Hz.
> > 
> > If the guest isn't drawing anything to video memory, and there are no
> > GUI events (e.g. redraw requests), why does the gui keep refreshing?
> 
> It needs to poll the keyboard activity at least, IIRC SDL doesn't
> provide the X FD to poll.

Apparently it does.

See http://listas.apesol.org/pipermail/sdl-libsdl.org/2008-May/065133.html
which says:

    > 2. add a new API call to SDL to access the low-level descriptor to
    >   talk directly with the graphic server (X11 or other).  This is
    >   powerful in that it opens the way to all sort of manipulations
    >   of the graphic server, however it requires backend-specific
    >   code in the SDL application to do what one needs;

    SDL_GetWMInfo() can get you the Xlib Display* (have a look at what
    else is there). Make sure to read the documentation about the
    SDL_VERSION() bit, it's quite important!

    You'll probably then have to enable SDL_SysWMEvent and handle them to
    get at the X11 events, if I understand this correctly (I haven't done
    the specific thing you want, I get the Display* and hack other
    things).

The followup message says "thanks, that worked".

It should work in Windows, too, except you'll get a Windows HANDLE
which you wait on in a different way than unix.

-- Jamie