This repository has been archived by the owner on Dec 6, 2023. It is now read-only.
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathREADME
146 lines (105 loc) · 6.45 KB
/
README
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
--------------------------------------------------------------------------------
Multitasking essentials
--------------------------------------------------------------------------------
A small introduction into how to implement both cooperative and preemptive
multitasking in RISC-V assembly.
The examples are intended for a GD32VF103 microcontroller on a Longan Nano
board which includes an active-low RGB LED on PC13, PA1, and PA2.
--------------------------------------------------------------------------------
Multitasking is the idea to run multiple routines virtually "in parallel"
by quickly executing these in turn on a single processor, sharing the
processing time between individual tasks.
For the tasks to run "as if they were on their own", one needs a routine
for the context switch that saves and subsequently restores all required state
necessary for running a task.
There are two fundamental concepts to build multitasking on:
* In cooperative multitasking, each task releases the processor voluntarily
by calling the context switch directly at suitable places.
* In preemptive multitasking, the context switch is triggered periodically
by a timer interrupt.
The code required for these is very similar, and both are presented here,
but they have different strengths and weaknesses.
--------------------------------------------------------------------------------
Cooperative multitasking allows very efficient usage of processing resources,
as each task can release execution whenever it needs to wait for something,
and atomicity of read-modify-write operations to peripheral registers and
global data structures is guranteed, as context switches only happen at
defined places in the code, mostly in delays, loops that wait for something
like slow IO, or in breaks of very long calculations like when one calculates
multiple fast fourier transforms in a row. Usually, all what is necessary is
embedding a call to the context switch in strategically choosen low-level
routines.
One can check the performance by adding a task that just pulses an output
pin which can be observed by an oscilloscope or with a slow moving-coil
type current meter that is set to full scale by a variable resistor when no
other tasks are running.
The interrupt layer keeps clear, so one can non-critical functions like
user interface and logging in tasks, and handle the cricital hard realtime
parts in interrupts. As no hardware support is necessary beyond enough RAM to
run cooperative multitasking, it can be used under almost all circumstances
in embedded targets.
On the other hand, cooperative multitasking requires all tasks to behave well,
one task that clogs the CPU will crash the whole system, and the time of
the next context switch depends on the individual tasks.
--------------------------------------------------------------------------------
Preemptive multitasking assigns each task a fixed time slot for execution,
and transfers control to the next task automatically. A task that clogs the
processor will not halt the whole system, and each task will get a fixed share
of the available execution time in a deterministic way. Within a task, the code
does not need to know about the existence of the multitasker, as execution is
passed in and out of the task without further assistance. This can be further
combined with memory protection to isolate tasks from each other.
On the other hand, preemptive multitasking requires hardware support from the
interrupt layer, which is therefore not available solely for critical hard
realtime handlers anymore.
More advanced implementations may offer a way for a task to releasing execution
voluntarily if there is no work to be done in the current time slot, but in
simple implementations, the processing time allocated for a task is wasted if
it has to wait for something.
The biggest problem, however, is the complete loss of atomicity for access to
common periherals or data structures, as the execution of everything else
- from the viewpoint of one task - happens like "within" an interrupt handler
that can occour at any point in time. This requires very careful design to
share common resources without nasty side effects.
--------------------------------------------------------------------------------
For short, cooperative multitasking is effective in small embedded setups
in which one has complete control over all tasks that are guranteed to
behave nicely, and where performance is the most appreciated goal.
Preemptive multitasking is used in almost all widely known (desktop) operating
systems that need to handle a variety of tasks, of which some could be rogue
or faulty. The problem with loss of atomicity is usually solved by carefully
designed interfaces or hardware abstraction layers / device drivers.
--------------------------------------------------------------------------------
Despite these important differences, the actual implementation of both
ways is very similar. Compare the assembler source files yourself!
For every task, there is a memory area large enough to contain the saved
state of the registers, the stack, and a bit of task housekeeping information.
The task areas are arranged in a ring, with a pointer from each task structure
to the next one, with the last one in the ring pointing back to the first one.
Each task also has a flag whether it is active or not, as ring insertion or
deletion is not atomic under all circumstances, so if one wants to activate
or deactivate a task by a hardware event, a flag is provided.
For brevity of the examples, tasks must not return, but in a more enhanced
implementation, the return of a task might result in an error message or
simply in the removal of that particular task from the ring.
Questions are welcome!
--------------------------------------------------------------------------------
How to run:
--------------------------------------------------------------------------------
If you have connected your
Longan Nano board to
a USB-serial cable with 3.3V logic levels
3V3 or 5V -- VCC (measure the voltage present here)
R0 -- TXD
T0 -- RXD
GND -- GND
you can flash it if you first
* Press and hold BOOT0 button
* Press and release Reset button
* Release BOOT0 button
and then execute one of
stm32loader -e -w -v -p /dev/ttyUSB0 multitask-cooperative.bin
stm32loader -e -w -v -p /dev/ttyUSB0 multitask-preemptive.bin
on your host machine.
Press and release Reset button to start blinking.
--------------------------------------------------------------------------------