[Sdcc-user] status of the ROSES project

[Joris v. d. S. <vd...@pu...>]

Hi all,

I have a non-preemptive MCS-51 version of ROSES up-and-running. To measure
its speed, I have tried the following test program:

  #include "rtos.h"

  sbit red_led   = P1^0 ;
  sbit green_led = P1^1 ;

  void blink_red_led(void)
  {   while (1)
      {   red_led = 0 ;
          rtos_task_sleep(0) ;
          red_led = 1 ;
          rtos_task_sleep(0) ;
      }
  }

  void blink_green_led(void)
  {   while (1)
      {   green_led = 0 ;
          rtos_task_sleep(0) ;
          green_led = 1 ;
          rtos_task_sleep(0) ;
      }
  }

  void main(void)
  {   /* initialise the operating system */
      rtos_init() ;
      
      /* create blink tasks */
      rtos_task_create(1, blink_red_led, 100) ;
      rtos_task_create(2, blink_green_led, 100) ;
  
      /* reduce priority of main task */
      rtos_task_priority(0, 110) ;  
  }

This results in two tasks blinking LEDs at maximum speed (since the main
task has a lower priority and will never gain control again). Using an
oscilloscope I see two square waves with a frequency of 615 Hz, 90 degrees
out of phase. Since I used a 16 MHz. 80C32 board and it takes 4
task-switches per period of the square wave, it takes 542 cpu cycles to
switch between tasks. This means that the latency of the operating system
will be about half a millisecond.

I am quite happy with this result, considering that most of the code is
written in C to simplify porting the OS to other microcontroller families.
Note that this test was compiled using the Keil compiler, perhaps the SDCC
compiler generates even faster code.

I will try to have a sneek-preview release ready by the end of this weekend.

Regards,
Joris