<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<meta content="text/html; charset=UTF-8" http-equiv="Content-Type">
</head>
<body bgcolor="#ffffff" text="#000000">
On 11/19/2010 02:38 PM, Jim wrote:
<blockquote cite="mid:4CE6FC5B.4060306@comcast.net" type="cite">
<meta content="text/html; charset=UTF-8" http-equiv="Content-Type">
<br>
I hate to have to tell you guys this , but !!!<br>
<br>
My EEEpc playing justintv video is much better, the hesitation is
so slight.<br>
Playing the "Transformers", I just can't get over it.<br>
<br>
</blockquote>
<br>
Sorry, but that's just your perception.<br>
<br>
This fix only affects TTYs (i.e. terminal sessions). For instance,
if you're watching a video and compiling something in a terminal,
this fix will make sure that the terminal as a whole (no matter how
many background processes it starts) shares time fairly with your
video player.<br>
In other words, if you don't have a single terminal open, you
shouldn't see any change at all.<br>
<br>
<br>
In more detail, the specific problem is this:<br>
* Kernel sees 10 CPU-hungry tasks: 1,2,3,4,5,6,7,8,9,10. Since the
kernel doesn't know any better, it considers them equals and gives
them same CPU time share (10%)<br>
<br>
* In reality, process (1) is code compiling, (2)..(9) are started by
(1) automagically, and (10) is a youtube video you're watching
(because <a class="moz-txt-link-freetext" href="http://xkcd.com/303/">http://xkcd.com/303/</a>).<br>
(10) requires at least 50% CPU time or it'll be jittery. <br>
<br>
* The solution, then, is to treat the *group* (1)..(9) as an equal
to (10), and divide the time accordingly:<br>
(10) gets 50%, while (1)..(9) get 50/9 ~= 5.6% <br>
<br>
<br>
HTH<br>
<br>
</body>
</html>