why did Ubuntu turn off kernel preemption in Edgy?

Tobias Wolf towolf at gmail.com
Wed Nov 29 19:15:19 GMT 2006

Am Mittwoch, den 29.11.2006, 17:43 +0000 schrieb Matthew Garrett:
> On Wed, Nov 29, 2006 at 12:27:25PM -0500, Phillip Susi wrote:
> > No, this is the beat again.  You are being shown images that show the 
> > wheel more than 180 degrees out of phase with the previous and next 
> > images, which is the same as if it were going backwards at a slower 
> > rate.  If the wheel is spinning at 23 rpm and the film is 24 fps, then 
> > you get a 1 Hz beat.
> If the human eye is unable to distinguish movement at faster than the 
> equivalent of 24 frames per second, then you'd see exactly the same 
> thing in real life. You don't, which indicates that faster movement is 
> still perceivable. In computer games, the difference between 30 frames 
> per second and 60 frames per second is quite easy to distinguish.

You do. There is pretty much the same aliasing (wagon wheels) in the
visual system. The cinematic industry didn’t choose their FPS
arbitrarily, they just underestimated it somewhat.
Google for Reichhardt Detectors and aliasing to find some references.
This model integrates over time to get rid of noise, i.e., objects
jumping around in visual space. It has been argued that temporal
aliasing is due to this integration.
Flicker fusion frequency in high contrast regimes like screens is around
that but in general depends heavily on the circumstances (luminance,
contrast in-/decrement and others).
I guess increasing FPS in games amounts to something different than
perceptual flicker reduction.

> In reality, there is no concept of "frames per second" in the human 
> visual system. The eye receptors are sufficiently sensitive that a 
> single photon will produce a visible response[1]. The fact that we 
> perceive 16 frames per second as smooth motion is an artifact of 
> higher-level processing, not an indication that we're unable to perceive 
> flicker above 16Hz - there's no well defined lower bound on the minimum 
> period of time a stimulus has to appear for before it reaches higher 
> levels of consciousness, and it varies between individuals. Even with 
> continuous flickering (a worst case scenario for determining whether 
> a stimulous is continuous or not), some people are able to perceive 
> 120Hz flicker.

Fortunately you don’t watch your screen with rods. Temporal integration
in cones is 10-15ms. That makes 66-100 Hz. The Critical Fusion Frequency
peaks around 50 Hz at very high contrast in most subjects. Different
integration times are at work after the retina, quite simply, for the
totally different reasons see the above.

> > What are you talking about?  The light source in the projector is always 
> > on and 24 film frames pass in front of it each second.  Where is the 
> > double illumination?
> If that were the case, you'd see the frames moving vertically. You 
> don't. There's a shutter between the light and the film, as described in 
> http://en.wikipedia.org/wiki/Movie_projector . On modern projectors, the 
> shutter opens more than once per frame.
> [1] Sadly the spontaneous firing rate is higher than this, so a single 
> photon is somewhere below the noise threshold
> -- 
> Matthew Garrett | mjg59 at srcf.ucam.org
Tobias Wolf

More information about the ubuntu-devel mailing list