dma可以替代memcpy吗
我的程序用双缓存绘图(直接操作framebuffer)
每次需要2次memcpy操作, 屏幕数据600k
每次执行memcpy经测试需要15ms, 这样2次就30ms
在考虑执行其他任务的情况下, 速度更低
硬件是OMAPL138 (arm926 ejs)

dma不知道可不可以做到内存之间直接拷贝, 如果可以, 速度会提高多少?

还有最好有示例代码, 网上找的例子和linux内核的例子不是遍不过去就是无法执行

[解决办法]
在内核中调用EDMA的实例代码，平—M6446

C/C++ code

int edma3_memtomemcpy(struct edma_linux_params *params){    int result = 0;    int i;    unsigned int numenabled = 0;    struct paramentry_descriptor param_set;    unsigned int dma_ch;    unsigned int tcc = EDMA_TCC_ANY;    result = davinci_request_dma(EDMA_DMA_CHANNEL_ANY, "A-SYNC_DMA0",                     edma_callback, NULL,                     &dma_ch, &tcc, 0);    if (0 != result) {        DMA_PRINTK            ("\nedma3_memtomemcpytest_dma::davinci_request_dma failed for dma_ch, error:%d\n",             result);        return result;    }    davinci_set_dma_src_params(dma_ch, (unsigned long)(params->src),                   INCR, W8BIT);    davinci_set_dma_dest_params(dma_ch, (unsigned long)(params->dst),                    INCR, W8BIT);    davinci_set_dma_src_index(dma_ch, params->srcbidx, params->srccidx);    davinci_set_dma_dest_index(dma_ch, params->dstbidx, params->dstcidx);    /* AB Sync Transfer Mode */    davinci_set_dma_transfer_params(dma_ch, params->acnt, params->bcnt, params->ccnt,             params->bcnt, ABSYNC);    /* Enable the Interrupts on Channel 1 */    davinci_get_dma_params(dma_ch, &param_set);    param_set.opt |= (1 << ITCINTEN_SHIFT);    param_set.opt |= (1 << TCINTEN_SHIFT);    davinci_set_dma_params(dma_ch, &param_set);    numenabled = params->ccnt;    for (i = 0; i < numenabled; i++) {        irqraised1 = 0;        /* Now enable the transfer as many times as calculated above. */        result = davinci_start_dma(dma_ch);        if (result != 0) {            DMA_PRINTK                ("edma3_memtomemcpytest_dma: davinci_start_dma failed \n");            break;        }        /* Wait for the Completion ISR. */        while (irqraised1 == 0u) ;        /* Check the status of the completed transfer */        if (irqraised1 < 0) {            /* Some error occured, break from the FOR loop. */            DMA_PRINTK("edma3_memtomemcpytest_dma: "                   "Event Miss Occured!!!\n");            break;        }    }    davinci_stop_dma(dma_ch);    davinci_free_dma(dma_ch);    return result;}
[解决办法]
1. DMA可以用作M2M.
2. 用DMA进行传输的话，速度会比你之前提高很多，多少的话要视环境而定，你可以测出来。
   但单拿CPU拷贝和DMA拷贝的速度进行对比的话，感觉CPU要快点。