[FFmpeg-devel] [PATCH] Reimplementation of OS/2 threads model

[KO Myung-Hun]

M?ns Rullg?rd wrote:
> KO Myung-Hun <komh at chollian.net> writes:
> 
>> Hi/2.
>>
>> This patch reimplements OS/2 threads model based on Win32 threads model.
>>
>> -- 
>> KO Myung-Hun
>>
>> Using Mozilla SeaMonkey 2.0.3
>> Under OS/2 Warp 4 for Korean with FixPak #15
>> On AMD ThunderBird 1GHz with 512 MB RAM
>>
>> Korean OS/2 User Community : http://www.ecomstation.co.kr
>>
>>
>> Index: libavcodec/os2thread.c
>> ===================================================================
>> --- libavcodec/os2thread.c	(revision 22960)
>> +++ libavcodec/os2thread.c	(working copy)
>> @@ -1,5 +1,5 @@
>>  /*
>> - * Copyright (c) 2004 Michael Niedermayer <michaelni at gmx.at>
>> + * Copyright (c) 2010 KO Myung-Hun <komh at chollian.net>
>>   *
>>   * This file is part of FFmpeg.
>>   *
>> @@ -17,51 +17,120 @@
>>   * License along with FFmpeg; if not, write to the Free Software
>>   * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
>>   */
>> -//#define DEBUG
>>
>> -// Ported by Vlad Stelmahovsky
>> +#define INCL_DOS
>> +#include <os2.h>
>>
>> +#undef __STRICT_ANSI__  /* for _beginthread() */
>> +#include <stdlib.h>
>> +
>>  #include "avcodec.h"
>>
>> -#define INCL_DOS
>> -#define INCL_DOSERRORS
>> -#define INCL_DOSDEVIOCTL
>> -#include <os2.h>
>> +typedef struct Semaphore{
>> +    HEV   hev;
>> +    HMTX  hmtxWait;
>> +    HMTX  hmtxCount;
>> +    LONG  lCount;
>> +    LONG  lMax;
>> +}Semaphore;
> 
> Can we please not use Hungarian names?

Ok.

> Also, please add a space
> before { and after }.  This applies to the entire patch.
> 

w32thread.c has a that style. So I just followed it.

Anyway, maybe do you also want K&R style like Diego ? ^^

>>  typedef struct ThreadContext{
>>      AVCodecContext *avctx;
>> -    int thread;
>> -    HEV work_sem;
>> -    HEV done_sem;
>> +    TID thread;
>> +    Semaphore *work_sem;
>> +    Semaphore *job_sem;
>> +    Semaphore *done_sem;
>>      int (*func)(AVCodecContext *c, void *arg);
>> +    int (*func2)(AVCodecContext *c, void *arg, int, int);
>>      void *arg;
>> -    int ret;
>> +    int argsize;
>> +    int *jobnr;
>> +    int *ret;
>> +    int threadnr;
>>  }ThreadContext;
>>
>> +#define INFINITE  ((ULONG)SEM_INDEFINITE_WAIT)
> 
> What's with the cast?
> 

Removed.

>> +static Semaphore *CreateSemaphore(void *v1, LONG lCount, LONG lMax, void *v2){
>> +    Semaphore *sem;
>> +
>> +    sem = av_mallocz(sizeof(Semaphore));
>> +    if (!sem)
>> +        return NULL;
>> +
>> +    DosCreateEventSem(NULL, &sem->hev, 0, lCount > 0 ? TRUE : FALSE );
>> +    DosCreateMutexSem(NULL, &sem->hmtxWait, 0, FALSE );
>> +    DosCreateMutexSem(NULL, &sem->hmtxCount, 0, FALSE );
>> +
>> +    sem->lCount = lCount;
>> +    sem->lMax   = lMax;
>> +
>> +    return sem;
>> +}
>> +
>> +static void ReleaseSemaphore(Semaphore *sem, LONG lInc, PLONG plPrevCount){
>> +    DosRequestMutexSem(sem->hmtxCount, SEM_INDEFINITE_WAIT);
>> +
>> +    if (sem->lCount + lInc <= sem->lMax){
>> +        sem->lCount += lInc;
>> +        DosPostEventSem(sem->hev);
>> +    }
>> +
>> +    DosReleaseMutexSem(sem->hmtxCount);
>> +}
>> +
>> +static void WaitSemaphore(Semaphore *sem, ULONG ulTimeout){
>> +    DosRequestMutexSem(sem->hmtxWait, SEM_INDEFINITE_WAIT);
>> +
>> +    DosWaitEventSem(sem->hev, ulTimeout);
>> +
>> +    DosRequestMutexSem(sem->hmtxCount, SEM_INDEFINITE_WAIT);
>> +
>> +    sem->lCount--;
>> +    if (sem->lCount == 0){
>> +        ULONG ulCount;
>> +
>> +        DosResetEventSem(sem->hev, &ulCount);
>> +    }
>> +
>> +    DosReleaseMutexSem(sem->hmtxCount);
>> +
>> +    DosReleaseMutexSem(sem->hmtxWait);
>> +}
>> +
>> +static void CloseSemaphore(Semaphore *sem){
>> +    DosCloseEventSem(sem->hev);
>> +    DosCloseMutexSem(sem->hmtxWait);
>> +    DosCloseMutexSem(sem->hmtxCount);
>> +
>> +    av_freep(&sem);
>> +}
> 
> Is this emulating something?  What's wrong with semaphore calls in the
> old code?
> 

The (event) semaphore of the old code wakes up all the waiting threads.
And it resets the status to non-signaled regardless of its post count.

So I decided to emulate the semaphore model of Win32. It can wake up
waiting threads one by one and decrease its post count.

>>  static void attribute_align_arg thread_func(void *v){
>>      ThreadContext *c= v;
>>
>>      for(;;){
>> -        //printf("thread_func %X enter wait\n", (int)v); fflush(stdout);
>> -        DosWaitEventSem(c->work_sem, SEM_INDEFINITE_WAIT);
>> -//        WaitForSingleObject(c->work_sem, INFINITE);
>> -//printf("thread_func %X after wait (func=%X)\n", (int)v, (int)c->func); fflush(stdout);
>> +        int ret, jobnr;
>> +        WaitSemaphore(c->work_sem, INFINITE);
>> +        // avoid trying to access jobnr if we should quit
>> +        if (!c->func && !c->func2)
>> +            break;
>> +        WaitSemaphore(c->job_sem, INFINITE);
>> +        jobnr = (*c->jobnr)++;
>> +        ReleaseSemaphore(c->job_sem, 1, 0);
>>          if(c->func)
>> -            c->ret= c->func(c->avctx, c->arg);
>> +            ret= c->func(c->avctx, (uint8_t *)c->arg + jobnr*c->argsize);
>>          else
>> -            return;
>> -        //printf("thread_func %X signal complete\n", (int)v); fflush(stdout);
>> -        DosPostEventSem(c->done_sem);
>> -//        ReleaseSemaphore(c->done_sem, 1, 0);
>> +            ret= c->func2(c->avctx, c->arg, jobnr, c->threadnr);
>> +        if (c->ret)
>> +            c->ret[jobnr] = ret;
>> +        ReleaseSemaphore(c->done_sem, 1, 0);
>>      }
>> -
>> -    return;
>>  }
>>
>>  /**
>> - * free what has been allocated by avcodec_thread_init().
>> - * must be called after decoding has finished, especially do not call while avcodec_thread_execute() is running
>> + * Free what has been allocated by avcodec_thread_init().
>> + * Must be called after decoding has finished, especially do not call while avcodec_thread_execute() is running.
>>   */
> 
> Please do spelling fixes like this separately.
> 

Ooops. It's not my intention. I just copied w32thread.c to os2thread.c,
and worked on it.

Anyway removed those parts.


-- 
KO Myung-Hun

Using Mozilla SeaMonkey 2.0.3
Under OS/2 Warp 4 for Korean with FixPak #15
On AMD ThunderBird 1GHz with 512 MB RAM

Korean OS/2 User Community : http://www.ecomstation.co.kr

-------------- next part --------------
A non-text attachment was scrubbed...
Name: os2thread.diff
Type: text/x-diff
Size: 8586 bytes
Desc: not available
URL: <http://lists.mplayerhq.hu/pipermail/ffmpeg-devel/attachments/20100426/f500bee8/attachment.diff>