ZX Time #04 17 ноября 2001

Программирование

Discussion - "Multitask. ... The realities and myths"

Discussion. Multitask. Realities and myths ...


                                KilleRam


                 - - -


  Editor's Note: I know what an uproar would cause the 
following material. However, we thus observe a

principles of our newspaper - I mean, consider and publish any, 
even the totally polar, and sometimes conflicting somewhere in 
some way incorrect, opinions ... To maintain a balance ... AND 
it does not mean that I or anyone else others to agree with the 
author of the material ... 


                 - - -


  Fuck, I'm sitting right now for his glyukavoy
Claudia and has been in some time trying to write an article 
about multitasking in response Articles DWT.



  Female, some e! # $ $ D virus zahaval
my floppy disk, paper and source code (well,
I pulled them RDM'om, and then
ALASM'ovsky (ALASM Rulezz forever)
find'er not really looking for them).


  So multi-tasking. Many people think
that this device can not be implemented on SPECCY due to the 
fact that its resources and the rate percent is not enough 
developed. Females, n e !@#$!$ to your PyTsy and Dr.#$%' e 
multitasking LINUX'A. 


  In short, enough mats to bend, to the point. I want to say 
that the algorithm is multi-tasking fully implemented on 
ALASM'e, but not optimized. However, it shows a fairly high 
rate of speed of execution of prog, suspended on the it (3 
tasks). The coefficient Brakes on average somewhere

1.1-1.17, but I know how to reduce it.


  All right, proceed to the description algorima
multitasking. Its essence is in
"Vlazhivanii" interrupt each other.
Something like this:


  1. Being implemented on the main bend
EI (with this problem, but about them later);


  2. Came to interrupt - we climb to the prog
processing; she retains full situation before the interrupt 
vector changes - and to another task.



  3. Fulfill the second task.


  4. Come and break all the same
Step 2. So until the last task after
which - on the main processor, and from
it into the main prog.


  Restrictions on the use of this algorithm is as follows:


  1. Memory is not below 512K (where else
keep the problem);


  2. Stable bus, and it will turn out
vector table that will hold
more than a problem.


  3. Style of programming will also be
special. In particular, the program must be
performed from any address;


  And here's what I want to realize - as
it is the Task Manager, which will be
distribute tasks by priority
(Very cleverly: the length and time of execution of the INT), 
split into blocks of memory (Likely to 2KB) and follow the 
boundaries of tasks and the process of information exchange 
between tasks, and much more to the smallest detail. Most 
likely it was he would be a major challenge. That's all that I 
I want to say about multi-tasking and its relizatsii on SPECCY 
... 


                 *