Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > comp.programming > #1459

Re: [Cross post: prog+asm+c]thread problem: one write, many read

Cross-posted to 3 groups.

Show all headers | View raw

"io_x" <a@b.c.invalid> ha scritto nel messaggio
news:4f83d522$0$1385$4fafbaef@reader2.news.tin.it...
>
> "io_x" <a@b.c.invalid> ha scritto nel messaggio
> news:4f83d0ff$0$1377$4fafbaef@reader2.news.tin.it...
>> there is one buffer, only one thread write,
>> many threads read that buffer
>>
>> what is the best solution for this problem for the x86 cpu?
>> How do you useful express that using the C language? :) i'm not ot
>>
>> i think it is this: [because i believe it is possible many
>> threads can read one peace of memory in the same time...]
>>
>> thread write:
>> unsigned32bits   v=0, mem=0;
>>
>> mem=7;
>> /* wait until v is 0 */
>> Wait(&v, 0);
>> change(buffer, len)
>> mem=0;
>>
>> --------------
>>
>> threads read:
>> /* wait until mem is 0 */
>> Wait(&mem, 0);
> because here it is possible the thread write change mem...
>
>> atomic++(&v); /* if v==-1 seg fault :) */
>> leggi(buffer, len)
>> atomic--(&v);
>>
>> you see dead lock on all this? i not...
>
>
> unsigned32bits   v=0, mem=0;
>
> thread write:
> /*atomic means if it is lock not write ...*/
> atomic_move(&mem, 7);
> /* wait until v is 0 */
> Wait(&v, 0);
> change(buffer, len)
> atomic_move(&mem, 0);
>
>
> threads read:
> /* wait until mem is 0 if is 0 lock mem
>   and atomic increment v
>   unlock mem
> */
> WaitIncrement(&mem, &v); /* if v==-1 seg fault :) */
> leggi(buffer, len)
> atomic--(&v);

unsigned32bits   v=0, mem=0;

thread write:
/* atomic++mem means:
1) in the *same time* mem is increment no other thread write it
   and the thread read it, that thread read the before
   [or the pass?] value
2) if mem is lock wait [for to do increment] until it is unlock
3) if mem==-1  inc mem is seg fault
*/
atomic++mem;
wait(&v, 0);
write(buffer, len)
atomic--mem;


threads read:
/* waitLock(&mem)
   wait until mem is 0, in the same time mem is 0 lock it.
   is this operation possible?
*/
waitLock(&mem); atomic++v; unlock(&mem);
read(buffer, len)
                atomic--v;

where are deadLock and how to find them?
Do you know the right primitive for get threads go ok?

thank you

Back to comp.programming | Previous | Next — Previous in thread | Next in thread | Find similar

Thread

[Cross post: prog+asm+c]thread problem: one write, many read "io_x" <a@b.c.invalid> - 2012-04-10 08:24 +0200
  Re: [Cross post: prog+asm+c]thread problem: one write, many read "io_x" <a@b.c.invalid> - 2012-04-10 08:42 +0200
    Re: [Cross post: prog+asm+c]thread problem: one write, many read "io_x" <a@b.c.invalid> - 2012-04-10 10:34 +0200
      Re: [Cross post: prog+asm+c]thread problem: one write, many read "io_x" <a@b.c.invalid> - 2012-04-10 11:26 +0200
        Re: [Cross post: prog+asm+c]thread problem: one write, many read "io_x" <a@b.c.invalid> - 2012-04-10 12:56 +0200
          Re: [Cross post: prog+asm+c]thread problem: one write, many read "io_x" <a@b.c.invalid> - 2012-04-14 07:20 +0200
            Re: [Cross post: prog+asm+c]thread problem: one write, many read "io_x" <a@b.c.invalid> - 2012-04-15 08:04 +0200
  Re: [Cross post: prog+asm+c]thread problem: one write, many read "Dmitry A. Kazakov" <mailbox@dmitry-kazakov.de> - 2012-04-10 10:01 +0200
  Re: thread problem: one write, many read "christian.bau" <christian.bau@cbau.wanadoo.co.uk> - 2012-04-13 17:13 -0700
    Re: thread problem: one write, many read "io_x" <a@b.c.invalid> - 2012-04-14 10:56 +0200

csiph-web