view w32thread.c @ 12197:fbf4d5b1b664 libavcodec

Remove FF_MM_SSE2/3 flags for CPUs where this is generally not faster than regular MMX code. Examples of this are the Core1 CPU. Instead, set a new flag, FF_MM_SSE2/3SLOW, which can be checked for particular SSE2/3 functions that have been checked specifically on such CPUs and are actually faster than their MMX counterparts. In addition, use this flag to enable particular VP8 and LPC SSE2 functions that are faster than their MMX counterparts. Based on a patch by Loren Merritt <lorenm AT u washington edu>.

/*
 * Copyright (c) 2004 Michael Niedermayer <[e-mail]>
 *
 * This file is part of FFmpeg.
 *
 * FFmpeg is free software; you can redistribute it and/or
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * version 2.1 of the License, or (at your option) any later version.
 *
 * FFmpeg is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public
 * 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

#include "avcodec.h"

#define WIN32_LEAN_AND_MEAN
#include <windows.h>
#include <process.h>

typedef struct ThreadContext{
    AVCodecContext *avctx;
    HANDLE thread;
    HANDLE work_sem;
    HANDLE job_sem;
    HANDLE done_sem;
    int (*func)(AVCodecContext *c, void *arg);
    int (*func2)(AVCodecContext *c, void *arg, int, int);
    void *arg;
    int argsize;
    int *jobnr;
    int *ret;
    int threadnr;
}ThreadContext;


static unsigned WINAPI attribute_align_arg thread_func(void *v){
    ThreadContext *c= v;

    for(;;){
        int ret, jobnr;
//printf("thread_func %X enter wait\n", (int)v); fflush(stdout);
        WaitForSingleObject(c->work_sem, INFINITE);
        // avoid trying to access jobnr if we should quit
        if (!c->func && !c->func2)
            break;
        WaitForSingleObject(c->job_sem, INFINITE);
        jobnr = (*c->jobnr)++;
        ReleaseSemaphore(c->job_sem, 1, 0);
//printf("thread_func %X after wait (func=%X)\n", (int)v, (int)c->func); fflush(stdout);
        if(c->func)
            ret= c->func(c->avctx, (uint8_t *)c->arg + jobnr*c->argsize);
        else
            ret= c->func2(c->avctx, c->arg, jobnr, c->threadnr);
        if (c->ret)
            c->ret[jobnr] = ret;
//printf("thread_func %X signal complete\n", (int)v); fflush(stdout);
        ReleaseSemaphore(c->done_sem, 1, 0);
    }

    return 0;
}

/**
 * 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.
 */
void avcodec_thread_free(AVCodecContext *s){
    ThreadContext *c= s->thread_opaque;
    int i;

    for(i=0; i<s->thread_count; i++){

        c[i].func= NULL;
        c[i].func2= NULL;
    }
    ReleaseSemaphore(c[0].work_sem, s->thread_count, 0);
    for(i=0; i<s->thread_count; i++){
        WaitForSingleObject(c[i].thread, INFINITE);
        if(c[i].thread)   CloseHandle(c[i].thread);
    }
    if(c[0].work_sem) CloseHandle(c[0].work_sem);
    if(c[0].job_sem)  CloseHandle(c[0].job_sem);
    if(c[0].done_sem) CloseHandle(c[0].done_sem);

    av_freep(&s->thread_opaque);
}

static int avcodec_thread_execute(AVCodecContext *s, int (*func)(AVCodecContext *c2, void *arg2),void *arg, int *ret, int count, int size){
    ThreadContext *c= s->thread_opaque;
    int i;
    int jobnr = 0;

    assert(s == c->avctx);

    /* note, we can be certain that this is not called with the same AVCodecContext by different threads at the same time */

    for(i=0; i<s->thread_count; i++){
        c[i].arg= arg;
        c[i].argsize= size;
        c[i].func= func;
        c[i].ret= ret;
        c[i].jobnr = &jobnr;
    }
    ReleaseSemaphore(c[0].work_sem, count, 0);
    for(i=0; i<count; i++)
        WaitForSingleObject(c[0].done_sem, INFINITE);

    return 0;
}

static int avcodec_thread_execute2(AVCodecContext *s, int (*func)(AVCodecContext *c2, void *arg2, int, int),void *arg, int *ret, int count){
    ThreadContext *c= s->thread_opaque;
    int i;
    for(i=0; i<s->thread_count; i++)
        c[i].func2 = func;
    avcodec_thread_execute(s, NULL, arg, ret, count, 0);
}

int avcodec_thread_init(AVCodecContext *s, int thread_count){
    int i;
    ThreadContext *c;
    uint32_t threadid;

    s->thread_count= thread_count;

    if (thread_count <= 1)
        return 0;

    assert(!s->thread_opaque);
    c= av_mallocz(sizeof(ThreadContext)*thread_count);
    s->thread_opaque= c;
    if(!(c[0].work_sem = CreateSemaphore(NULL, 0, INT_MAX, NULL)))
        goto fail;
    if(!(c[0].job_sem  = CreateSemaphore(NULL, 1, 1, NULL)))
        goto fail;
    if(!(c[0].done_sem = CreateSemaphore(NULL, 0, INT_MAX, NULL)))
        goto fail;

    for(i=0; i<thread_count; i++){
//printf("init semaphors %d\n", i); fflush(stdout);
        c[i].avctx= s;
        c[i].work_sem = c[0].work_sem;
        c[i].job_sem  = c[0].job_sem;
        c[i].done_sem = c[0].done_sem;
        c[i].threadnr = i;

//printf("create thread %d\n", i); fflush(stdout);
        c[i].thread = (HANDLE)_beginthreadex(NULL, 0, thread_func, &c[i], 0, &threadid );
        if( !c[i].thread ) goto fail;
    }
//printf("init done\n"); fflush(stdout);

    s->execute= avcodec_thread_execute;
    s->execute2= avcodec_thread_execute2;

    return 0;
fail:
    avcodec_thread_free(s);
    return -1;
}