NilRTOS - A Fast Tiny Preemptive RTOS

Giovanni Di Sirio, the author of ChibiOS/RT, has written an experimental system called Nil RTOS.

Giovanni describes Nil RTOS as "Smaller than ChibiOS/RT, so small it's almost nil".

I decided to try Nil RTOS on an Arduino Uno. The latest version of NilRTOS is here Google Code Archive - Long-term storage for Google Code Project Hosting..

This system seems ideal for small chips like the ATmega328. Nil RTOS is extremely simple so it is great for new users.

I have written a number of examples and a guide, "Understanding the Examples", to help new users.

Here is the first example new users should try. It requires about 2150 bytes of flash on an Uno.

/*
 * Example to demonstrate thread definition, semaphores, and thread sleep.
 */
#include <NilRTOS.h>

// The LED is attached to pin 13 on Arduino.
const uint8_t LED_PIN = 13;

// Declare a semaphore with an inital counter value of zero.
SEMAPHORE_DECL(sem, 0);
//------------------------------------------------------------------------------
/*
 * Thread 1, turn the LED off when signaled by thread 2.
 */
// Declare a stack with 128 bytes beyond context switch and interrupt needs.
NIL_WORKING_AREA(waThread1, 128);

// Declare the thread function for thread 1.
NIL_THREAD(Thread1, arg) {
  while (TRUE) {
    
    // Wait for signal from thread 2.
    nilSemWait(&sem);
    
    // Turn LED off.
    digitalWrite(LED_PIN, LOW);
  }
}
//------------------------------------------------------------------------------
/*
 * Thread 2, turn the LED on and signal thread 1 to turn the LED off.
 */
// Declare a stack with 128 bytes beyond context switch and interrupt needs. 
NIL_WORKING_AREA(waThread2, 128);

// Declare the thread function for thread 2.
NIL_THREAD(Thread2, arg) {

  pinMode(LED_PIN, OUTPUT);
  
  while (TRUE) {
    // Turn LED on.
    digitalWrite(LED_PIN, HIGH);
    
    // Sleep for 200 milliseconds.
    nilThdSleepMilliseconds(200);
    
    // Signal thread 1 to turn LED off.
    nilSemSignal(&sem);
    
    // Sleep for 200 milliseconds.   
    nilThdSleepMilliseconds(200);
  }
}
//------------------------------------------------------------------------------
/*
 * Threads static table, one entry per thread.  A thread's priority is
 * determined by its position in the table with highest priority first.
 * 
 * These threads start with a null argument.  A thread's name may also
 * be null to save RAM since the name is currently not used.
 */
NIL_THREADS_TABLE_BEGIN()
NIL_THREADS_TABLE_ENTRY("thread1", Thread1, NULL, waThread1, sizeof(waThread1))
NIL_THREADS_TABLE_ENTRY("thread2", Thread2, NULL, waThread2, sizeof(waThread2))
NIL_THREADS_TABLE_END()
//------------------------------------------------------------------------------
void setup() {
  // Start Nil RTOS.
  nilSysBegin();
}
//------------------------------------------------------------------------------
// Loop is the idle thread.  The idle thread must not invoke any 
// kernel primitive able to change its state to not runnable.
void loop() {
  // Not used.
}

I am trying the NilFifoDataLogger demo, but getting this when typing a char to begin, DATA.CSV is emtpy:

type any character to begin
type any character to end
type any character to begin

PS: maybe it is a high time to introduce a new item into the "Topics", ie. " RTOS" or something like that.. :stuck_out_tongue:

What line ending are you using in the serial monitor?

I have a bug in flushing all input. I used nilThdSleep(10) in setup() before NilRTOS was started:

  // throw away input
  while (Serial.read() >= 0) {
    nilThdSleep(10);
  }

Try "No line ending" or replace the nilThdSleep(10) with delay(10).

I will quickly post an update for this and another bug in NilRTOS.

Also I will post a version of SdFat with a new printField(uint16_t value, char terminator) so this code

  uint16_t value;
 ...
  file.printField(value, ',');

is three times faster than this code

  uint16_t value;
 ...
  file.print(value);
  file.print(',');

This will allow nilFifoDataLogger to write more fields at 1024 usec intervals.

Arduino Print is too slow and nilFifoDataLogger will suffer overrun errors currently.

IDE 1.5.1r2, @115k2, ide terminal:

with nilThdSleep(10); 
No line ending:
type any character to begin
type any character to end
type any character to begin
Newline:
type any character to begin
type any character to begin
CR:
type any character to begin
type any character to begin
Both NL&CR:
type any character to begin
type any character to begin

with delay(10);
any settings
type any character to begin
type any character to end
type any character to begin

DATA.CSV 0kB

with:
//#define Serial NilSerial
and No line ending:
type any character ôtype any character to begin
type any charactertype any character to begin
type any character to eîtype any character to begin
type any character totype any character to begin

pito,

I can't reproduce your problem. Please try the attached NilRTOS and SdFat libraries.

What board are you using? I tried an Uno and a 2560 Mega.

pito.zip (158 KB)

I am using my own UNO, BTooth serial, sdfatlib20121217.zip and latest nilrtos from your repo.
Your pito.zip shows

!   C:\Documents and Settings\igi\Desktop\pito.zip: Unexpected end of archive

I downloaded pito.zip and extracted the libraries with 7-zip on a Windows 7 machine.

The contents diff OK.

Edit: I tried the Windows extractor and it works too.

firefox under xp has a problem to download it, IE8 did it .. :~
but still the same issue..
I'll try with usb serial dongle..
..the same..

fyi - this is the bench from the sdfat:

Type any character to start
Free RAM: 1019
Type is FAT32
File size 5MB
Buffer size 100 bytes
Starting write test.  Please wait up to a minute
Write 135.23 KB/sec
Maximum latency: 54924 usec, Minimum Latency: 88 usec, Avg Latency: 733 usec

Starting read test.  Please wait up to a minute
Read 284.79 KB/sec
Maximum latency: 4656 usec, Minimum Latency: 84 usec, Avg Latency: 345 usec

Done

Type any character to start

I opened the example from your lib and did only this change:
Serial.begin(115200);
..no luck..

Again - I took vanilla demo and run @9k6 (usb dongle, your libs from pito.zip, IDE 1.5.1r2, UNO 328p @16MHz):

type any character to begin
type any character to endþtype any character to begin
type any character to endþtype any character to begin
../nilFifoDataLogger.cpp.hex 
Binary sketch size: 13,546 bytes (of a 32,256 byte maximum)

ser.jpg

pito,

I can't reproduce your problem so I don't know what to do. I only do the Serial.begin(115200) change with 1.5.1r2 on an Uno and get.

type any character to begin
type any character to end
Done
Maximum SD write latency: 65008 usec
Unused Stack: 57 454

I get the same result with all line endings.

To see the performance of printField(), run the PrintBenchmark example with the new SdFat. You should get something like this:

Test of println(uint16_t)
Time 10.09 sec
File size 128.89 KB
Write 12.78 KB/sec
Maximum latency: 62460 usec, Minimum Latency: 176 usec, Avg Latency: 497 usec

Test of printField(uint16_t, char)
Time 3.26 sec
File size 128.89 KB
Write 39.60 KB/sec
Maximum latency: 67240 usec, Minimum Latency: 60 usec, Avg Latency: 156 usec

The two files are identical but printField() runs three times faster.

Edit: Try making the stack for thread 1 larger like this:

NIL_WORKING_AREA(waThread1, 100);
Type any character to start
Free RAM: 1109
Type is FAT32
Starting print test.  Please wait.

Test of println(uint16_t)
Time 9.97 sec
File size 128.89 KB
Write 12.93 KB/sec
Maximum latency: 20828 usec, Minimum Latency: 176 usec, Avg Latency: 491 usec

Test of printField(uint16_t, char)
Time 3.15 sec
File size 128.89 KB
Write 40.92 KB/sec
Maximum latency: 18748 usec, Minimum Latency: 60 usec, Avg Latency: 150 usec

Test of println(double)
Time 15.06 sec
File size 149.00 KB
Write 9.89 KB/sec
Maximum latency: 19160 usec, Minimum Latency: 572 usec, Avg Latency: 746 usec

Done!

Type any character to start

Edit: Try making the stack for thread 1 larger like this:
Code:
NIL_WORKING_AREA(waThread1, 100);

..no luck..

Do the other NilRTOS examples work?

I can't do anything more since everything I try works fine. I don't have a working xp system.

Edit: Here is one more idea. Change nilSysBegin() to nilSysBeginNoFill(). This will prevent filling stacks when NilRTOS starts so the final message will be wrong.

Change nilSysBegin() to nilSysBeginNoFill().

..the same issue..

other stuff ie.:
nilblink blinks

nilBlinkPrint:

Count: 628299, Unused Stack: 57 28 1663
Count: 628306, Unused Stack: 57 28 1663
Count: 628304, Unused Stack: 57 28 1663
Count: 628302, Unused Stack: 57 28 1663

nilSemTest:

micros overhead = 4
sem switch tasks micros = 28
Thd1 Done
sem immediate timeout = 4
sem immediate OK = 4
Thd2 Done

It may be a problem with Serial.read(). nilFifoDataLogger is the only example with Serial.read().

I have tried many Arduino boards and two Windows 7 computers with no problem.

Load the nilTemplate1 example and put the following at the beginning of loop() like this.

void loop() {
  int c = Serial.read();
  if (c < 0) return;
  Serial.print((char)c);
  return;

Run the example and type one or two characters into the serial monitor and see if they are echoed.

Also run nilFifoDataLogger with no SD card and see if this prints:

type any character to begin
SD problem
SD errorCode: 0X1,0X0

Also run nilFifoDataLogger with no SD card and see if this prints:
Quote
type any character to begin
SD problem
SD errorCode: 0X1,0X0

type any character to begin
SD problem
SD errorCode: 0X1,0X0

Yes they are echoed (after pushing Send):

33
54
pito
Q

:slight_smile:

type any character to begin
type any character to end
Done
Unused Stack: 141 342

** overrun errors **

with:

// SD chip select pin.
const uint8_t sdChipSelect = 10; //SS;

pito,

Could you put a continue statement in thread 1 like this so no data is taken.

  while (1) {
    // Sleep until time for next data point
    nilThdSleep(intervalTicks);
    continue;

Let me know what happens.

type any character to begin
type any character to end
Done
Unused Stack: 141 445

data.csv is 0kB

No idea what is the issue, with "continue" it always works, with "//continue" it sometimes works, sometimes does not.