X-Authentication-Warning: delorie.com: mail set sender to djgpp-bounces using -f
Message-ID: <002a01c644cd$45e6b5d0$36aebc3e@AMILO>
From: "Martin Jarman (Cabledoc)" <asal56 AT dsl DOT pipex DOT com>
To: "DJ Delorie" <djgpp AT delorie DOT com>
Subject: Can you help?
Date: Sat, 11 Mar 2006 05:31:13 -0000
MIME-Version: 1.0
Content-Type: text/plain;
	charset="iso-8859-1"
Content-Transfer-Encoding: 7bit
X-Priority: 3
X-MSMail-Priority: Normal
X-Mailer: Microsoft Outlook Express 6.00.2800.1506
X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2800.1506
Reply-To: djgpp AT delorie DOT com

Hello DJ Delorie

I have been working on a real-time control system for a couple of months and
have now hit a serious problem which you may be able to help with.
Pleeeease, I'm desperate!!

Really, I would just like to know if using DJGPP will get me out of this
pickle...

I am using FreeDOS in conjunction with a multifunction PCI card (digital
I/O, ADC, DAC & onboard timer).  It is **essential** that my control code
runs without any significant interruptions - and preferably none at all.
(The odd few microseconds may be OK, especially if they are deterministic).

Sparing the details, my code or psuedo code is:

GetLargeArraytoOutputToPCIcard();  // Using TCP/IP
disable();                         //disable interrupts
//Read the external clock on the PCI card :
nPreviousTime = inp(timerPortAddress);
while (bMoreLargeArrayItemsToOutput)
//Accurately synchronise to the next millisecond:
    KeepGettingTheTimeUntilItEquals_nPreviousTime_PlusOneMilliSecond();
//Every call to the external clock above is 'Logged' to
//allow analysis of the performance of this code
    outp(PortA,NextValueInOutputArray);
    // Do other inp() & outp() here (must total < 1 millisec)
   }
enable(); //enable interrupts

The good news is that the above code works very deterministically.  The
dreadful news is that I overlooked the memory limitations of DOS and cannot
allocate sufficient memory for the arrays (several Megabytes are needed).
I've tried using files on RAM disks, but the moment these are read, the
accurate timing I need is lost.  I presume this to be partly because of the
switch into / out of  Protected Mode, and perhaps also because interrupts
may be occuring whilst in PM.

So here's my question(s).  If I were to use DJGPP, can I run the all the
code between disable() and enable() in PM? (I think this is another way of
asking if inp() in outp() can be used in PM without needing to return to RM,
isn't it?)

Secondly, the above code only works well when surrounded by the disable()
and enable() functions.  I noted from some of your documentation you have
similar functions in PM.  Is it acceptable to run the loop for long periods
with interrupts disabled in PM (and be confident that won't be interrupted)?

Lastly, and if the answer to both of the above isn't "Yes", if I have to
keep switching from PM to RM and back, can I keep interrupts disabled all
the time, including thoughout the transitions between the two modes?

Well it's now 5:20 AM - time for bed I think!  If you can offer any advice,
this would be greatly appreciated.

Regards

Martin Jarman
mjarman AT cabledoc DOT co DOT uk