qemu-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Qemu-devel] [PATCH] Add JPEG encoding to VNC server


From: Anthony Liguori
Subject: Re: [Qemu-devel] [PATCH] Add JPEG encoding to VNC server
Date: Fri, 31 Jul 2009 10:35:36 -0500
User-agent: Thunderbird 2.0.0.21 (X11/20090320)

Jamie Lokier wrote:
Alexander Graf wrote:
While this might sound like it renders the whole implementation useless, it
does make sense to implement it nevertheless. I have some ideas to implement
progressive encodings for video.

So when we'd detect that one region is updated a lot in a short
about of time with content that zlib can't really handle well, we'd
just send a really low quality JPEG first and then send the update
after a timer if the region wasn't updated within that timeframe.

Detecting video regions may also be possible if virtual video overlay
hardware can be offered.  It would probably require a special guest
driver.  virtio-video :-)

vmware-vga can do it already.

I don't want to send JPEG, I'd rather send YUV. The nice thing about an overlay is that you get the yuv data directly and you get small update regions (which is a feature of vmware-vga). The nice thing is that you get pretty close to mpeg encoding with that. If a blit is used for tile motion, that's even better.

Also, hardware scaling could be coded effectively too.

 Then again the trend, if you believe in the
<video> element demos in web browsers, is likely towards OpenGL-driven
video frame rendering because it's more versatile.

Sending JPEG quality first and a lossless update after is an
interesting idea, even for non-video.  On a slow link, seeing boxes
and buttons appear quickly might be quicker to click through a GUI, as
you can see where to click even if the text and decorations are a bit
fuzzy for a moment.  Especially if you already know what to expect.

Is it possible to use any of the lossless or "near-lossless" JPEG
encodings, and would that be any more efficient than VNC's usual
lossless encoding?

near-lossless is going to result in very large images. I'm quite sure it's worse than even hextile.

OTOH, zlib compressed raw tiles is probably pretty close to as good as hextile and it's quite easy to do. Alex has already done it once for the zlib encoding :-)

Regards,

Anthony Liguori




reply via email to

[Prev in Thread] Current Thread [Next in Thread]