Kathleen Weber wrote in a message to Mike Bilow:

-> KW>                   PRIORITY_DISK_IO=NO

-> Just so you know, this has been broken in all releases of Warp.  The
-> foreground process will get priority in the HPFS driver at a low leve

 KW> I haven't found any problem when using this in my
 KW> CONFIG.SYS.

It doesn't cause any problems, but it is ignored.

 KW> Packing my BBS message base takes 2 hours when it is the only 
 KW> program running and that is when it is the 'active' screen up.

I can't imagine what you are running there.  On a 486DX-33, I run Squish
SQPACKP daily at idle priority, and it takes less than a half hour.  I
carry a good chunk of the backbone and keep as many as 2000 messages in
some areas.

 KW> In the last few weeks I found I have had to do quite a bit
 KW> of work while the BBS message base is packing since I have a
 KW> backlog at work. So while it is running the pack routine, I
 KW> am open and closing all kinds of applications (OS/2, Win and
 KW> DOS based). Most of the time I have two or more programs
 KW> running (I am active in using the other open programs) while
 KW> the pack routine goes.  It still takes 2 hours flat.

Keep in mind that message base repacking is disk-bound, not CPU-bound. 
OS/2 takes most disk access requests as implicit yields anyway, so the
priority manager rarely has the chance to kick in.

 KW> With a program that would be so sensitive to any sort of
 KW> change in the CPU processor allocation it is very easy to
 KW> see if something else is taking away from it.

You can prove this with a simple loop that never yields the processor. 
Time its completion in the foreground and the background, and you will see
that being in the foreground gives it a substantial advantage regardless of
the setting in CONFIG.SYS.

 KW> I believe this does work and it really causes no harm to try it 
 KW> anyway. If it doesn't work then it can always be changed back.

True.  I use it myself -- it did work under OS/2 2.1 -- but it is easy to
demonstrate that it does not work under OS/2 Warp.
 
-- Mike


---