[En-Nut-Discussion] MultiThreading
Rich Wellner
ethernut at objenv.com
Thu Mar 1 16:08:01 CET 2007
Assuming that things occur in the order you list, when thread 2 executes
its NutSleep then control will be passed back to thread 1.
rw2
MUHAMMAD AZEEM AZAM wrote:
> when THREAD 1 has
> NutSleep(0)
> when its argument is Zero THREAD 1 will give up the CPU.
>
> when THREAD 2 has
> NutSleep(125)
> when its argument is 125ms THREAD 2 will sleep for 125 ms and CPU control is transferred to some other THREAD ?
>
> Gervasio Varela <gervarela at picandocodigo.com> wrote:
> Hi Muhammad,
>
> Nut/OS has cooperative multithreading, so, differently to Linux and other
> tipical OSes, it can't take out of the CPU any thread to put another. Is the
> executing thread the responsible of leaving the CPU in order that the OS
> could place other threads "on it".
>
> NutSleep() is used to accomplish this (leave the CPU), if you do a
> NutSleep(125) the thread is stopped AT LEAST 125 ms. But, at the moment you
> executed NutSleep(), other thread has taken over the CPU. The first thread
> doens't get back the CPU until the new (thread) do a NutSleep(), so it can
> take the CPU after 125ms or after 1000ms, if you have real time constraints,
> is your responsibility to asure that the other threads do a NutSleep()
> before 125ms.
>
> If you want that a computing thread separated from the networking code gets
> all the "idle" time, I think that you should use short NutSleep() times in
> all two threads. The computing thread should leave the CPU regularly (the
> code between NutSleep() should be little engough) to gurantee that the
> networking code process all the received request as early as posible, and
> the networking code should leave the CPU as soon as it checked if there are
> some request available.
>
> If you have litle network load, the networking thread should realease the
> CPU in few ms (the check of new request should be fast) and the computing
> thread should be using the majority of CPU time (but it should leave the CPU
> frequently to let the networking code test if there was new request...)
>
> I hope that this help you, and I'm open to suggestions, because I'm working
> in a projet with similar requisites :D.
>
> See you.
>
> PDT: Sorry for my english.
>
> On 01/03/07, MUHAMMAD AZEEM AZAM wrote:
>
>> I have a TCP Server for Microcontroller. i want that my micro controller
>> should be able to fullfil another task. Not only networking.
>> This means that i should create anther task (or thread) on the controller.
>> I want that this thread should have as much computing time of the
>> controller as
>> possible.
>> How to achieve this.
>> ---------------------------------------
>> I am confused in this concept, please clear it to me
>> Let suppose that
>> I have only 2 threads (other than main and idle).
>> Thread 1
>> Thread 2
>> in thread 1 i have a NutSleep(125), this means my thread has been slept
>> for 125ms and then the control will be transferred to Second thread. If
>> second thread has a NutSleep(500)
>> ,means thread 2 has been slept for 500ms .
>> But what after 125ms contol is transferred back to thred1 or what ?
>>
>> Thanks for ur help
>> Bye
>>
>>
>>
>> ---------------------------------
>> Be a PS3 game guru.
>> Get your game face on with the latest PS3 news and previews at Yahoo!
>> Games.
>> _______________________________________________
>> http://lists.egnite.de/mailman/listinfo/en-nut-discussion
>>
>>
>
>
>
>
More information about the En-Nut-Discussion
mailing list