latence of RT-Thread vs. eCOS

[GoogleCodeExporter]

Perhaps to use instead of "rt_hw_interrupt_enable" and 
"rt_hw_interrupt_disable" switching on/off inner scheduler of RT-Thread ?

Original issue reported on code.google.com by ed...@bk.ru on 25 Jul 2011 at 10:34

[GoogleCodeExporter]

Could you please give more details?

Original comment by mbb...@gmail.com on 26 Jul 2011 at 2:37

[GoogleCodeExporter]

For example, in function rt_sem_release(...) is present critical sectionЖ

/* disable interrupt */
temp = rt_hw_interrupt_disable();

    RT_DEBUG_LOG(RT_DEBUG_IPC,
        ("thread %s releases sem:%s, which value is: %d\n", rt_thread_self()->name,
        ((struct rt_object*)sem)->name, sem->value));

    if ( !rt_list_isempty(&sem->parent.suspend_thread) )
    {
        /* resume the suspended thread */
        rt_ipc_list_resume(&(sem->parent.suspend_thread));
        need_schedule = RT_TRUE;
    }
    else sem->value ++; /* increase value */

/* enable interrupt */
rt_hw_interrupt_enable(temp);

Between functiones rt_hw_interrupt_disable() and rt_hw_interrupt_enable() is 
very much the code, which grow the latency of interrupt. 

Original comment by ed...@bk.ru on 27 Jul 2011 at 7:57

[GoogleCodeExporter]

in eCOS the same semaphore is realized thus:

void Cyg_Binary_Semaphore::post()
{
    // Prevent preemption
    Cyg_Scheduler::lock();

    CYG_INSTRUMENT_BINSEM( POST, this, 0 );

    state = true;

    if( !queue.empty() ) {

        // The queue is non-empty, so grab the next
        // thread from it and wake it up. The waiter
        // will clear the flag.

        Cyg_Thread *thread = queue.dequeue();

        thread->set_wake_reason( Cyg_Thread::DONE );

        thread->wake();

        CYG_INSTRUMENT_BINSEM( WAKE, this, thread );
    }

    // Unlock the scheduler and maybe switch threads
    Cyg_Scheduler::unlock();

}

Here the critical section is disposed between Cyg_Scheduler::lock() and 
Cyg_Scheduler::unlock(), wherewith the latency of interrupts very low. 

Original comment by ed...@bk.ru on 27 Jul 2011 at 12:09