> MK> Yes, I know.  So if one could change the priority for the
 > MK> mouse-thread then the jerkiness would go away?  And why
 > MK> isn't the mouse set to time critical in the first place?
 >
 > I really don't think that the mouse should be considered time critical.

I don't either.  The best solotion would be if they fixed the graphics
drivers instead.


 > Scrolling is handled inside the presentation device drivers.  There is a
 > good reason for this, since many video boards these days have the
 > capability of accelerating common operations such as scrolling in
 > hardware.  Unfortunately, a side effect of this is that device drivers
 > are never pre-empted by the scheduler without their explicit yield, so
 > scrolling operates by default at apparent high priority.

But why?  If I for an example run a DOS-program and then enters "dir
c:\ /s" the program is STOPPED and isn't accepting any characters from
the modem = character loss.  I thing that it's a bad solotion.


 > MK> And it's just those DOS and OS/2-windows that gets this
 > MK> effect.  I don't have the same problem when scrolling for
 > MK> example an E.EXE window.
 >
 > PM itself handles scrolling differently than text scrolling.  This is
 > again an issue for the display driver internals.

To bad that the drivers for S3 and alot of other card is so bad then.


 > If you use a video board with an S3 chip, it will scroll text so fast
 > that you will no longer care about its priority.

I have a S3 864 with 2MB of memory and the scrolls isn't satisfactory to
me. The other day I tried a Matrox Millenium with 4MB of WRAM and the card
realy flies in 1024x768x16M.  I compared the scrolling speed in a
OS/2-window and just plain textmode scroll and the difference wasn't
notable.

The drivers worked almost flawless, the only problems I had was fullscreen
garbage when running DIVE applications and ofcourse the mouse jerkiness
when scrolling DOS and OS/2-windows.

---