[9fans] rio Peculiarities

[9fans] rio Peculiarities

[Akshat Kumar]

Having a window which is `current' and `visible' but not on `top'
gives rise to some effects of concern.  For example, scrolling this
current window is cause for massive load on the system; if there are
windows on top, making them current requires hiding the current one
(clicking on them is ineffective).  My use for being in such a
situation is, e.g., scrolling some acme body in the background while
doing something with an rc window on the foreground (since my acme
window takes up a large part of the screen).


Not particularly important, but an interesting case nonetheless,
ak

Re: [9fans] rio Peculiarities

[john]

> Having a window which is `current' and `visible' but not on `top'
> gives rise to some effects of concern.  For example, scrolling this
> current window is cause for massive load on the system; if there are
> windows on top, making them current requires hiding the current one
> (clicking on them is ineffective).  My use for being in such a
> situation is, e.g., scrolling some acme body in the background while
> doing something with an rc window on the foreground (since my acme
> window takes up a large part of the screen).
>
>
> Not particularly important, but an interesting case nonetheless,
> ak

How exactly are you making a window current without bringing it to the
top?


John

Re: [9fans] rio Peculiarities

[erik quanstrom]

On Fri Feb 27 16:46:34 EST 2009, akumar@mail.nanosouffle.net wrote:
> Having a window which is `current' and `visible' but not on `top'
> gives rise to some effects of concern.  For example, scrolling this
> current window is cause for massive load on the system; if there are
> windows on top, making them current requires hiding the current one
> (clicking on them is ineffective).  My use for being in such a
> situation is, e.g., scrolling some acme body in the background while
> doing something with an rc window on the foreground (since my acme
> window takes up a large part of the screen).

take a look at the pat patch or the older mtrr patch.

- erik

Re: [9fans] rio Peculiarities

[Akshat Kumar]

[-- Attachment #1: Type: text/plain, Size: 42 bytes --]

echo current > /dev/wsys/%d/wctl


ak

[-- Attachment #2: Type: message/rfc822, Size: 3115 bytes --]

From: john@csplan9.rit.edu
To: 9fans@9fans.net
Subject: Re: [9fans] rio Peculiarities
Date: Fri, 27 Feb 2009 16:51:34 -0500
Message-ID: <79a364e5a34578cd32f858082fc93216@csplan9.rit.edu>

> Having a window which is `current' and `visible' but not on `top'
> gives rise to some effects of concern.  For example, scrolling this
> current window is cause for massive load on the system; if there are
> windows on top, making them current requires hiding the current one
> (clicking on them is ineffective).  My use for being in such a
> situation is, e.g., scrolling some acme body in the background while
> doing something with an rc window on the foreground (since my acme
> window takes up a large part of the screen).
>
>
> Not particularly important, but an interesting case nonetheless,
> ak

How exactly are you making a window current without bringing it to the
top?


John

Re: [9fans] rio Peculiarities

[Akshat Kumar]

[-- Attachment #1: Type: text/plain, Size: 122 bytes --]

I tried the pat patch just now, and it doesn't seem to improve the
load or change the other effect I mentioned.


ak

[-- Attachment #2: Type: message/rfc822, Size: 3061 bytes --]

From: erik quanstrom <quanstro@quanstro.net>
To: 9fans@9fans.net
Subject: Re: [9fans] rio Peculiarities
Date: Fri, 27 Feb 2009 17:48:40 -0500
Message-ID: <31ec63c8fe239f38e1b5b53cf8215c8b@quanstro.net>

On Fri Feb 27 16:46:34 EST 2009, akumar@mail.nanosouffle.net wrote:
> Having a window which is `current' and `visible' but not on `top'
> gives rise to some effects of concern.  For example, scrolling this
> current window is cause for massive load on the system; if there are
> windows on top, making them current requires hiding the current one
> (clicking on them is ineffective).  My use for being in such a
> situation is, e.g., scrolling some acme body in the background while
> doing something with an rc window on the foreground (since my acme
> window takes up a large part of the screen).

take a look at the pat patch or the older mtrr patch.

- erik

Re: [9fans] rio Peculiarities

[erik quanstrom]

> I tried the pat patch just now, and it doesn't seem to improve the
> load or change the other effect I mentioned.
>

the pat patch is very effective for me in many situations.
however, there are some paths through the draw code
that are not optimized, or, still require large reads from
the frame buffer.  and a few cards (especially nvidia
ones) have other performance problems that seem unrelated.
i get about a 10x speedup from the pat patch for most
operations.  bringing a partially obscured large window to
the foreground is not one of them.

note to self: raise the priority of getting to that intel
gma driver.

- erik