From: bgleich AT blitz DOT chemie DOT uni-ulm DOT de (Bernhard Gleich)
Subject: Re: long timer interrupt
Newsgroups: comp.os.msdos.djgpp
References: <35df0db3 DOT 0 AT news DOT uni-ulm DOT de> <35e34ca3 DOT 7193856 AT news DOT image DOT dk>
NNTP-Posting-Host: tsunami.chemie.uni-ulm.de
Message-ID: <35e19579.0@news.uni-ulm.de>
Date: 24 Aug 1998 18:31:53 +0100
Lines: 52
To: djgpp AT delorie DOT com
DJ-Gateway: from newsgroup comp.os.msdos.djgpp
Precedence: bulk

Søren Egmose (egmose AT image DOT dk) wrote:
: In artikel <35df0db3 DOT 0 AT news DOT uni-ulm DOT de>,
: On 22 Aug 1998 20:28:03 +0100, bgleich AT blitz DOT chemie DOT uni-ulm DOT de
: (Bernhard Gleich) wrote:

: >Hi all.
: >
: >Mayby someone can suggest a smart solution to the following problem:
: >
: >I am doing some data aquisition and real time processing using DJGPP
: >and ALLEGRO. Most time, my real time restrictions are not that tight
: >and I have plenty of time to do the data displaying and storing
: >things. But some time I have to react very fast to an event. I can
: >predict this event to about 0.01s accuracy if it is still a long time
: >to go. (If I am close to the event, the accuracy of prediction becomes
: >better.) If the data, which indicate the event, are read, the program
: >should react within 1e-4 seconds (faster is better).
: >
: >My solution now is to call a function that estimates how long it will
: >take to the event. If it is close it waits for the event and reacts to
: >it. Naturally this function has to be be called often enough which
: >becomes very painful in a still growing program.
: >
: >A better solution would be to call this function by a timer
: >interrupt. There is only one problem. As the function waits a not
: >known time, some interrupts might be lost. So I could no more rely on
: >the clock.
: >
: >Is there some easy solution for my problem, like counting the lost
: >interrupts by the dely routine and adding them afterwards to system
: >clock.
: >
: >Some solution without reprograming the timer chip myself would be
: >prefered.
: >
: Why can't you use the input interrupts?

Because I want to react, if the read data have a special value. This
does not generate an interrupt. The data are read by DMA with 40MB/s,
so it is impossible to generate an interrupt for each word of data
(and would not be useful as I have to average some values).


: -- 
: mvh.
: Søren

... del ...


                                       Gru"s
                                        Bernhard Gleich