Date:	Wed, 5 Apr 95 12:08 GMT
From:	"Kevin Ashley, Systems Development, ULCC" <CZIWKGA AT VMSFE DOT ULCC DOT AC DOT UK>
To:	DJGPP <DJGPP AT SUN DOT SOE DOT CLARKSON DOT EDU>
Subject:	RE: timers and smartdrive.exe

Craig Gullixson wrote:

>...During the acquisition of the second 100
>images, smartdrive started to flush the disk cache resulting in longer
>than requested exposure times.

>I suspect that the disk I/O interrupts during the timer delay loop
>results in the timer not catching when the delay count has just reached
>the desired count resulting in the longer exposure times

Yup, disk I/O in DOS is totally synchronous and everything else waits until
the I/O completes. Although smartdrive makes this painless in most
circumstances and may even eliminate many redundant operations, you
get caught in situations like this where you want real-time response and
smartdrive kicks in at an unpredictable time. The problem can affect
any heirarchical storage system which offers some element of
transparency to the user.

>I am now loading smartdrive with the /C opton to force the flushing of
>the cache during the write operation.

Yes, enabling write-through caching should give you predictable performance
even if it is slower overall, since predictability seems to be the key for you.
(Whether /C actually achieves this effect I don't know and I'm not near
a system where I can check.)

I am sure it is also possible for a program to interact with smartdrive
to force it to flush and empty its cache, since some Microsoft tools
obviously make use of this. How one actually does this I have no idea.
Perhaps someone else on this list can shed light.

------------------------------------------------------------------------------
Kevin Ashley                              K DOT Ashley AT Ulcc DOT ac DOT uk
Systems Development Group Manager      http://www.ulcc.ac.uk/staff/Kevin+Ashley
University of London Computer Centre.      ...ukc!ncdlab!K.Ashley
                      This is not a signature

- Raw text -