Page 1 of 1

multitasking on pic 18f45k22

Posted: 24 Jul 2011 17:25
by relinquished
salve a voi del forum di supporto,
sul web ho trovato molte informazioni sul multitasking dei pic, ma essendo in assembly non ci capisco niente , quindi mi kiedevo se era posibile creare una libreria apposita per generare un multitasking tramite delle istruzzioni su mikrobasic

Re: multitasking on pic 18f45k22

Posted: 24 Jul 2011 17:26
by relinquished
prejudice to you the support forum,
on the web I found a lot of information on the pic of multitasking, but being in the assembly do not understand anything, so I was kiedevo if demand is possible to create a special library to generate a multi-tasking by the Instructions of mikroBASIC

Re: multitasking on pic 18f45k22

Posted: 26 Jul 2011 01:48
by bpislife
Good translate doesn't seem to be working all that well. PIC mcu's can't handle multi-tasking, but clever programming can make it seems as if it is. What are you trying to do and maybe we can help you that way.

Google translated from English to Italian:

Tradurre bene non sembra funzionare tanto bene. PIC MCU non in grado di gestire il multi-tasking, ma di programmazione intelligente può fare sembra che se lo è. Cosa stai cercando di fare e forse ci può aiutare in questo modo.

Re: multitasking on pic 18f45k22

Posted: 26 Jul 2011 08:46
by relinquished
inefetti, the translator of google is a little cmq is misfiring, you meant I want to do ie, turning on the web I found several SOLUTIONS asm that generates a charge of multitasking (which nn and multitasking, but in reality and an alternate who runs monotasking processes

in practice, and that's what I do \ \ \

There are basically two ways to implement multi-tasking: cooperative and preemptive multitasking. In the first are the programs that, naturally, give control to the system as soon as they completed the transaction in progress, the second is that the scheduler to the programs when the time allotted and transfers control from one to another.

The cooperative multitasking uses less computing resources and has (almost) switching delay for the task change, and also does not require any dedicated hardware structure, which makes it possible to implement it on any computer, on the other hand is very vulnerable to errors in programs (typically a program to crash the entire system drops) and the isolation between processes is very weak. It is the model used by Mac OS 9, Windows 3.0 and earlier.
The preemptive multitasking CPU needs to implement in hardware or in the levels of privilege for the execution of code and logic specific to the context switch, changing the task performed by the scheduler. Since the program interruption is arbitrary, the task change the operating system is forced to save all or almost all the CPU registers and load them with those saved by the task that takes over, losing a lot of time. Faced with these increasing demands, the preemptive multitasking system provides enhanced security and a much larger (virtual) immunity to system crashes caused by errors in programs. It is the model used by modern operating systems.
"

How to implement a management task in systems with very small resources? Cosa'è the logic of the context switch? Really need the levels of privilege? These and others are all questions that arises when we want to implement a uKernel. But to better understand what we want to achieve is perhaps easier to imagine a typical control system that we want to become multi-tasking ....

To facilitate our work better I make up an application. Let's say a client of mine asked me to develop an embedded system that plans to implement a keyboard of 11 keys and a 2x16 alphanumeric LCD display. Our client wants to display the keys pressed on the display.

We can immediately identify my two main processes.
1) Read keyboard.
2) Check LCD.

The hardware implementation at the moment we are not interested because uKernel must be released from the implementation of the card.
Reading the keyboard is easy to see how a process is asynchronous to the system (asynchronous process by defining all the processes that occur in a time not determined). The asynchronous processes need to activate a triggering event in this case we can associate with the button press.

The LCD control may be interpreted as a synchronous process. (Defining a synchronous process all the processes that occur in a given time). Imagine the state of our way to update the display every 100 msec.

From these two considerations will have to implement our uKernel logic for the context switch (which still does not know what it is ...) able to manage these two events and times.


So we can imagine that the structure of our task will become more or less like this


@ EVENT TASK_ID, EVOS_KEY_PRESSED


logical management of the key


END_TASK




@ WAIT_R TIMEOS_UPDATE_LCD


update logic of the display


END_TASK


THE CORE


Before going on the application must describe the functions of the core uKernel. We must decide to implement the type of operating system
first described by wikipedia.
Cooperative or Preempitive? Well actually we have already answered the question by describing the task as overwritten. The type of system that will be implementing a cooperative. @ @ WAIT_R EVENT and references that are in fact the core of uKernel use to call up the process and my task will be benevolent performing their functions as soon as possible. Preempitive a system requires some features that do not have the PIC 16Cxx and then immediately forget his existence.
The management of privilege levels is a beautiful feature of multitasking systems that allowed a task requiring much time as possible to the operating system. This type of request is specific preemptive systems because it is arbitrary suspension of the program and then the task remains poor not only beg the operating system to give him more time hoping that can give him, in the case of cooperative systems gives a **** to ask for once took the spring when it seems ...



Task Tables
Since the task need to manage events and times my heart has need 2 tables and call Event_Table Time_Table. Each element of this table is associated with a task.
If the task wants to be awakened by an event she requires with the @ EVENT type of event if it wants to be awakened in time with the task requires a time that will WAIT_R @ lsb as the operating system tick.


Switch Contest
To speak of the contest is now necessary to speak of the switch we are using micro which is precisely the PIC 16C5x

THE PIC
The pic is a microcontroller 16C5x would call essential. The only 33 instructions to learn is a facility that pays for itself in the internal structure of the micro cumbersome. The pages of memory is ROM and RAM hardware stack pointer (not readable) the lack of interrupts inappettibile would have made any application. Instead, the OTP can be programmed to avoid the costly and dangerous procedure of masking the micro markets have paved the way for making the fortune of Microchip's embedded.
The Pic and a working register dependence, this means that all operations of the internal registers must pass through this one working register. Unlike the rich alternatives such as the Z80 instruction that had more than 150 with this solution are able to heavily reduce the number of instructions inside (... and so much expensive silicon) but in some cases creating a nagging neck bottlenecks in the implementation of functions.
A bit 'of numbers to understand what we're talking about. The microcontroller is a 28-pin 2K bytes of RAM and ROM 72 at the time was a fast 20 MHz to 40 MHz increased hours



In light of the internal structure of the micro now describe exactly what action is brought typically Contest switch. The Contest is the switch function that saves the context of my tasks in memory and then restore it the next time it is invoked. The context is so supportive that the logs used in micro-processing program. The PIC, however, unlike the 8051s or indeed the z80 has no special registers (except the working register), but has only general purpouse global registers. This requires developing a micro uKernel with this, the task must be careful in the use of memory. When we see the 8051 uKernel understand this better, and so in our case does not implement it by ignoring it.



After this short description of the most savvy have already wrinkled her nose reading that the pic has a hardware stack pointer, and already understand that this is the main reason that does not allow the pic to implement uKernel preempitive. But the problem does not change in the case of the above tasks. How do I create a system that allows my uKernel to know how to draw my task?
We need a new table that we call TABLE_RETURN.
This new table does not lie unlike the other ram but resides in ROM.


Implementation of macro


The assembler of the microchip is very powerful and has an efficient preprocessor for the creation of macros needed for our purpose.


;------------------------------------------------- ----------------
; @ MACRO EVENT
; COMMENT TASK is waiting to wake up events
; ENTRY
; EXIT
;------------------------------------------------- ----------------
@ MACRO EVENT TASK_N, WAKE_UP
LOCAL RETURN, MEMO


BCF FSR, 5; point to bank 0
BCF FSR, 6; point to bank 0


MOVLW IND_JMP; saves the return address of an indirect
MEMO
MOVWF RET_ADD TASK_N +; in the table of returns
ORG IND_JMP
GOTO RETURN; instruction to return indirect low P.
ORG MEMO +1; resume the PC from where it was
IND_JMP
September IND_JMP +1, I update the table pointer within
MOVLW WAKE_UP; loads the wake-up mask
MOVWF TASK_N + COND, COND save in the table to return


;**** The toggle mask and wake up 'reset uploading ret_add
GOTO P_TASK; laugh 'control to sist. op.
RETURN
@ SET_BANK; sets the bench after returning to the task


ENDM


;------------------------------------------------- ----------------
; MACRO @ WAIT_R
; COMMENT TASK is waiting for the REC TIME
; ENTRY
; EXIT
;------------------------------------------------- ----------------
@ MACRO WAIT_R TASK_N, REG
LOCAL RETURN, MEMO


BCF FSR, 5; point to bank 0
BCF FSR, 6; point to bank 0


MOVLW (IND_JMP | 10000000b)

; Save the return address indirect and not the flag
MEMO
MOVWF RET_ADD TASK_N +; in the table of returns


ORG IND_JMP
GOTO RETURN; instruction to return indirect low P.
ORG MEMO +1; resume the PC from where it was


IND_JMP
September IND_JMP +1, I update the table pointer within


MOVFW REG
MOVWF TASK_N + COND, COND save in the table to return
; The toggle timer and 'set by loading ret_add
GOTO P_TASK; laugh 'control to sist. op.
RETURN
@ SET_BANK; sets the bench after returning to the task


ENDM



The implementation of the core is a little more elaborate the code is, however, the comments are autoespicativi


;------------------------------------------------- ----------------
; THE OPERATING SYSTEM!!
;------------------------------------------------- ----------------


SCHEDULER
CLRWDT; clear WDOG


;------------------------------------------------- ----------------
; AND TASK MANAGEMENT SYSTEM TIMER
;------------------------------------------------- ----------------


GEST_RTCC
MOVFW RTCC; read the current state of rtcc
XORWF RTCC_COPY, W, and if you look at 'the same as my copy
Btfsc STATUS, Z; if they are not equal 'time spent
GOTO P_TASK, then I do not update timers, I'm going to handle the I / O
INCF RTCC_COPY, Same, goes one step on my watch

C_TASK_0; control task 0
TASK set 0
Btfsc RET_ADD + TASK, 7; tests if events or time
DECF COND + TASK, SAME, it was time, task time decreases
C_TASK_1; Control Task 1
TASK 1 September
Btfsc RET_ADD + TASK, 7; tests if events or time
DECF COND + TASK, SAME, it was time, task time decreases
C_TASK_2; Control Task 2
TASK 2 September
Btfsc RET_ADD + TASK, 7; tests if events or time
DECF COND + TASK, SAME, it was time, task time decreases


;------------------------------------------------- ----------------
; START TASK
; WARNING: If you change the number of tasks also change the PART_TASK!
;------------------------------------------------- ----------------
MOVLW RET_ADD0
MOVWF FSR; point to task 0
CALL PART_TASK
INCF FSR, Same, point to Task 1
CALL PART_TASK
INCF FSR, Same, point to Task 2
CALL PART_TASK
FINE_P_TASK
P_TASK; calls SYS_OP fall here!


;------------------------------------------------- ----------------
; Function of I / O PERFORMED BY SCHEDULER
;------------------------------------------------- ----------------


Reading the keyboard ... and updating the event OS_KEY_PRESSED

GOTO SCHEDULER




Now we implement our modules that contain our task. Recall that in our case we have a contest and then switching to the task using the internal memory of the micro to their liking ....


;------------------------------------------------- ---------------
;------------------------------------------------- ---------------
TASK_0


EVENT @ 0, (1 <<OS_KEY_PRESSED)


Decoding the button pressed


GOTO TASK_0




;------------------------------------------------- ---------------
;------------------------------------------------- ---------------
TASK_1


@ WAIT_R 1, 100


Display update
GOTO TASK_0







Novità! Fai clic sulle parole riportate sopra per visualizzare le traduzioni alternative. Elimina



0.
0.
0.
0.
0.
0.
0.
0.

Re: multitasking on pic 18f45k22

Posted: 26 Jul 2011 10:34
by jpc
this can simply be solved by running the keyboardscan in an interrupt , the main program can update the display if required ( which in fact is hardly any work, most of the time you are waiting for a keypress ).
The context-saving is taken care of by the compiler, no need for anything more complex in this case.
Even on a small Pic this strategy can do many tasks virtually simultaneous if you design it such that there are no blocking sections in your code and they do nothing if there is nothing to do, e.g. the updating of LCD makes only sense if the contents have changed.
I succeed to scan keyboard, update LCD, deal with SPI-Ethernet, Uarts etc in an idle-loop of few ms only on a P18 .
The key is NOT to use delays , these are wasting precious cpu-cycles which you should make available to other tasks.
I usually run a 1 ms timer-interrupt in which i can deal with simple tasks like scanning key's etc.

Re: multitasking on pic 18f45k22

Posted: 26 Jul 2011 11:35
by p.erasmus
I agree with jpc
I fail to see the need for multitasking in such an simple and small appilcation !!

Re: multitasking on pic 18f45k22

Posted: 30 Jul 2011 18:44
by rmteo
relinquished, you are on the right path towards creating high quality, professional grade applications. Go for it!!! :D :D :D

Re: multitasking on pic 18f45k22

Posted: 30 Jul 2011 21:09
by relinquished
the problem and that the arm all have a package that is not possible for electronics hobbyists weld with ease, here is the reason why the 8 bits are often used as a hobby, because otherwise I would have already passed from one life to arm