[Fwd: Re: Wondering about cirris and stdvga]

[Fwd: Re: Wondering about cirris and stdvga]

[Dario Faggioli]

-------- Forwarded Message --------
From: Adam Jackson <ajax@redhat.com>
Subject: Re: Wondering about cirris and stdvga
Date: Fri, 18 Nov 2016 15:23:59 -0500
Message-Id: <1479500639.16973.3.camel@redhat.com>

On Fri, 2016-11-18 at 11:26 -0800, Stefano Stabellini wrote:

> 
> I am no Xorg maintainer but back in the XenClient days we did some
> perf
> analysis and it turned out that the bitblt operations supported by
> Cirrus are actually very slow to emulate in QEMU.

Yep. Turns out the "acceleration" is still going to be running on a
CPU, so you can either take a hypercall for every operation, or not.

The X drivers for a while now have detected virtual cirrus devices and
defaulted to a shadow framebuffer instead, partly for this reason and
partly to hide the absolutely terrible 24bpp framebuffer format from
applications.

- ajax
-- 
<<This happens because I choose it to happen!>> (Raistlin Majere)
-----------------------------------------------------------------
Dario Faggioli, Ph.D, http://about.me/dario.faggioli
Senior Software Engineer, Citrix Systems R&D Ltd., Cambridge (UK)

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
https://lists.xen.org/xen-devel

[Fwd: Re: Wondering about cirris and stdvga]

[Dario Faggioli]

-------- Forwarded Message --------
From: Adam Jackson <ajax@redhat.com>
Subject: Re: Wondering about cirris and stdvga
Date: Fri, 18 Nov 2016 15:19:08 -0500
Message-Id: <1479500348.16973.1.camel@redhat.com>

On Fri, 2016-11-18 at 13:32 -0500, Konrad Rzeszutek Wilk wrote:
> 
> On Fri, Nov 18, 2016 at 06:08:57PM +0000, Ian Jackson wrote:
> > 
> > 
> > > 
> > > performance wise, that cirrus is broken and impossible to
> > > fix (because it is the hardware that it's emulating that was
> > > broken),
> > > that stdvga enables better screen resolution in guests, etc.
> > 
> > The "broken and impossible to fix" sounds like FUD TBH.

The "fear" bit, perhaps; it is arguably frightening that we continue to
emulate in 2016 a chip that was inadequate in 1996. In my opinion about
cirrus being broken, however, I am neither uncertain nor in doubt.

> 
> > 
> > I wouldn't want to change the default just because Wayland is
> > broken.
> > Wayland should be fixed.

You should want to change the default because it is bad, even for X.

> 
> CCing Adam Jackson who I hope can enlighten us on the technical
> parts of Xorg.
> 
> I am really interested in knowing the technical merits
> of the stdvga vs cirrus and I hope to understand what the Xorg
> maintainer(s) have in mind.

The cirrus chip that qemu happens to emulate is trash. It can't do
32bpp, it can't do modes larger than 2048 wide, it can't do more than
4M of VRAM. We could of course extend the qemu emulation to be more
capable; doing so would require updating every guest to know about
these enhancements, though.

stdvga, as it has none of the above problems, is _today_ uniformly
better than cirrus. It is still perhaps less awesome than one would
like (no idea how multihead is expected to work for example), but there
is no world in which stdvga is worse than cirrus.

- ajax
-- 
<<This happens because I choose it to happen!>> (Raistlin Majere)
-----------------------------------------------------------------
Dario Faggioli, Ph.D, http://about.me/dario.faggioli
Senior Software Engineer, Citrix Systems R&D Ltd., Cambridge (UK)

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
https://lists.xen.org/xen-devel

Re: [Fwd: Re: Wondering about cirris and stdvga]

[Dario Faggioli]

[-- Attachment #1.1: Type: text/plain, Size: 2818 bytes --]

> On Fri, 2016-11-18 at 15:19 -0500, Adam Jackson wrote:
> > On Fri, 2016-11-18 at 13:32 -0500, Konrad Rzeszutek Wilk wrote:
> > On Fri, Nov 18, 2016 at 06:08:57PM +0000, Ian Jackson wrote:
> > > > performance wise, that cirrus is broken and impossible to
> > > > fix (because it is the hardware that it's emulating that was
> > > > broken),
> > > > that stdvga enables better screen resolution in guests, etc.
> > > 
> > > The "broken and impossible to fix" sounds like FUD TBH.
> 
> The "fear" bit, perhaps; it is arguably frightening that we continue
> to
> emulate in 2016 a chip that was inadequate in 1996. In my opinion
> about
> cirrus being broken, however, I am neither uncertain nor in doubt.
> 
Thanks Konrad for poking Adam, and thanks Adam for chiming in so
quickly.

> > > I wouldn't want to change the default just because Wayland is
> > > broken.
> > > Wayland should be fixed.
> 
> You should want to change the default because it is bad, even for X.
> 
In principle, I'm all with Ian's, and I think that's true for pretty
much everyone.

And in fact, I was really hesitant to start a discussion like this
basing on: 1) another project's bug and 2) rumors (as Ian says, and
he's right again) and incomplete knowledge.

Then, poking around a bit more, and talking with Konrad, we had the
feeling that those that were only rumors to us (well, to me), may
actually be more than just that... which indeed seems to be what Adam
is teaching us here. :-)

> > CCing Adam Jackson who I hope can enlighten us on the technical
> > parts of Xorg.
> > 
> > I am really interested in knowing the technical merits
> > of the stdvga vs cirrus and I hope to understand what the Xorg
> > maintainer(s) have in mind.
> 
> The cirrus chip that qemu happens to emulate is trash. It can't do
> 32bpp, it can't do modes larger than 2048 wide, it can't do more than
> 4M of VRAM. We could of course extend the qemu emulation to be more
> capable; doing so would require updating every guest to know about
> these enhancements, though.
> 
> stdvga, as it has none of the above problems, is _today_ uniformly
> better than cirrus. It is still perhaps less awesome than one would
> like (no idea how multihead is expected to work for example), but
> there
> is no world in which stdvga is worse than cirrus.
> 
Ok. These seems very good reasons to me.

Then of course I'm no maintainer, so it's someone else's call...
Hopefully, we at least gathered some useful insights. :-)

Thanks and Regards,
Dario
-- 
<<This happens because I choose it to happen!>> (Raistlin Majere)
-----------------------------------------------------------------
Dario Faggioli, Ph.D, http://about.me/dario.faggioli
Senior Software Engineer, Citrix Systems R&D Ltd., Cambridge (UK)

[-- Attachment #1.2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 819 bytes --]

[-- Attachment #2: Type: text/plain, Size: 127 bytes --]

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
https://lists.xen.org/xen-devel

[Fwd: Re: Wondering about cirris and stdvga]

[Dario Faggioli]

-------- Forwarded Message --------
From: Stefano Stabellini <sstabellini@kernel.org>
Subject: Re: Wondering about cirris and stdvga
Date: Fri, 18 Nov 2016 11:26:07 -0800
Mailer: Alpine 2.10 (DEB 1266 2009-07-14)
Message-Id: <alpine.DEB.2.10.1611181123310.3290@sstabellini-ThinkPad-
X260>

On Fri, 18 Nov 2016, Konrad Rzeszutek Wilk wrote:
> 
> On Fri, Nov 18, 2016 at 06:08:57PM +0000, Ian Jackson wrote:
> > 
> > Dario Faggioli writes ("Wondering about cirris and stdvga"):
> > > 
> > > I've also read around that these days, e.g., stdvga is at least
> > > as good
> > > as cirrus,
> > 
> > You pasted a url in IRC which seemed to be arguing that but
> > actually
> > failed to make the case.  (Sorry, don't have it to hand now.)
> 
> This one?
> https://www.kraxel.org/blog/2014/10/qemu-using-cirrus-considered-harm
> ful/
> > 
> > 
> > > 
> > > performance wise, that cirrus is broken and impossible to
> > > fix (because it is the hardware that it's emulating that was
> > > broken),
> > > that stdvga enables better screen resolution in guests, etc.
> > 
> > The "broken and impossible to fix" sounds like FUD TBH.
> > 
> > I wouldn't want to change the default just because Wayland is
> > broken.
> > Wayland should be fixed.
> 
> CCing Adam Jackson who I hope can enlighten us on the technical
> parts of Xorg.
> 
> I am really interested in knowing the technical merits
> of the stdvga vs cirrus and I hope to understand what the Xorg
> maintainer(s) have in mind.
 
I am no Xorg maintainer but back in the XenClient days we did some perf
analysis and it turned out that the bitblt operations supported by
Cirrus are actually very slow to emulate in QEMU. stdvga is faster
because it doesn't support them. In addition stdvga supports higher
resolutions: as high as the videoram assigned to the guest can support,
while cirrus is limited.
-- 
<<This happens because I choose it to happen!>> (Raistlin Majere)
-----------------------------------------------------------------
Dario Faggioli, Ph.D, http://about.me/dario.faggioli
Senior Software Engineer, Citrix Systems R&D Ltd., Cambridge (UK)

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
https://lists.xen.org/xen-devel

[Fwd: Re: Wondering about cirris and stdvga]

[Dario Faggioli]

-------- Forwarded Message --------
From: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Subject: Re: Wondering about cirris and stdvga
Date: Fri, 18 Nov 2016 13:32:24 -0500
Mailer: Mutt/1.7.1 (2016-10-04)
Message-Id: <20161118183224.GB10560@x230.dumpdata.com>

On Fri, Nov 18, 2016 at 06:08:57PM +0000, Ian Jackson wrote:
> 
> Dario Faggioli writes ("Wondering about cirris and stdvga"):
> > 
> > I've also read around that these days, e.g., stdvga is at least as
> > good
> > as cirrus,
> 
> You pasted a url in IRC which seemed to be arguing that but actually
> failed to make the case.  (Sorry, don't have it to hand now.)

This one?
https://www.kraxel.org/blog/2014/10/qemu-using-cirrus-considered-harmfu
l/
> 
> 
> > 
> > performance wise, that cirrus is broken and impossible to
> > fix (because it is the hardware that it's emulating that was
> > broken),
> > that stdvga enables better screen resolution in guests, etc.
> 
> The "broken and impossible to fix" sounds like FUD TBH.
> 
> I wouldn't want to change the default just because Wayland is broken.
> Wayland should be fixed.

CCing Adam Jackson who I hope can enlighten us on the technical
parts of Xorg.

I am really interested in knowing the technical merits
of the stdvga vs cirrus and I hope to understand what the Xorg
maintainer(s) have in mind.

> 
> 
> There may be other reasons to change the default but none of what you
> said is convincing.  (Partly because it's mostly rumours.)
> 
> Thanks,
> Ian.
-- 
<<This happens because I choose it to happen!>> (Raistlin Majere)
-----------------------------------------------------------------
Dario Faggioli, Ph.D, http://about.me/dario.faggioli
Senior Software Engineer, Citrix Systems R&D Ltd., Cambridge (UK)

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
https://lists.xen.org/xen-devel

[Fwd: Re: Wondering about cirris and stdvga]

[Dario Faggioli]

Making the thread appear on xen-devel, as it actually was not there,
because I can't type! :-/

-------- Forwarded Message --------
From: Ian Jackson <ian.jackson@eu.citrix.com>
Subject: Re: Wondering about cirris and stdvga
Date: Fri, 18 Nov 2016 18:08:57 +0000
Mailer: VM 8.2.0b under 24.4.1 (i586-pc-linux-gnu)
Message-Id: <22575.17337.627347.934459@mariner.uk.xensource.com>

Dario Faggioli writes ("Wondering about cirris and stdvga"):
> 
> I've also read around that these days, e.g., stdvga is at least as
> good
> as cirrus,

You pasted a url in IRC which seemed to be arguing that but actually
failed to make the case.  (Sorry, don't have it to hand now.)

> 
> performance wise, that cirrus is broken and impossible to
> fix (because it is the hardware that it's emulating that was broken),
> that stdvga enables better screen resolution in guests, etc.

The "broken and impossible to fix" sounds like FUD TBH.

I wouldn't want to change the default just because Wayland is broken.
Wayland should be fixed.

There may be other reasons to change the default but none of what you
said is convincing.  (Partly because it's mostly rumours.)

Thanks,
Ian.
-- 
<<This happens because I choose it to happen!>> (Raistlin Majere)
-----------------------------------------------------------------
Dario Faggioli, Ph.D, http://about.me/dario.faggioli
Senior Software Engineer, Citrix Systems R&D Ltd., Cambridge (UK)

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
https://lists.xen.org/xen-devel