Question

我移植了这段代码：

    if(_layersCount > 1)
    {
        for(int i=_layersCount-2;i>=0;i--)
        {
            for(int j=0;j<_neuronsPerLayerCount[i];j++) // cuda kernel
            {
                localGradients[indexByLayerAndNeuron(i, j)] = 0;

                for(int k=0;k<_neuronsPerLayerCount[i+1];k++)
                {
                    localGradients[indexByLayerAndNeuron(i, j)] += _neuronsInputsWeights[indexByLayerNeuronAndInput(i+1, k, j)]
                                                                    * localGradients[indexByLayerAndNeuron(i+1, k)];
                }

                localGradients[indexByLayerAndNeuron(i, j)] *= derivatives[indexByLayerAndNeuron(i, j)];
            }
        }
    }

到CUDA：

    if(_layersCount > 1)
    {
        for(int i=_layersCount-2;i>=0;i--)
        {
            // calculateLocalGradientsForAnotherLayers
            blocksCount = floor((double) _neuronsPerLayerCount[i] / threads.x) + 1;
            blocks = dim3(blocksCount, 1);

            calculateLocalGradientsForAnotherLayers <<<blocks, threads>>> (deviceLocalGradients, _neuronsInputsWeights, deviceDerivatives, _neuronsPerLayerCount[i], _neuronsInPreviousLayers[i], _neuronsInPreviousLayers[i+1], _neuronsPerLayerCount[i+1], _inputsInPreviousLayers[i], _inputsInCurrentLayer[i]);
        }
    }

calculateLocalGradientsForAnotherLayers内核：

__global__ void calculateLocalGradientsForAnotherLayers(double * localGradients, double * neuronsInputsWeights, double * derivatives, int neuronsCount, int neuronsInPreviousLayers, int neuronsInPreviousLayersWithCurrent, int neuronsInNextLayer, int inputsInPreviousLayers, int inputsInCurrentLayer)
{
    int idx = blockIdx.x * blockDim.x + threadIdx.x;

    if(idx < neuronsCount)
    {
        int neuron = neuronsInPreviousLayers + idx;

        localGradients[neuron] = 0;

        // this to Kernel, then reduce localGradients.
        for(int k=0;k<neuronsInNextLayer;k++)
        {
            localGradients[neuron] += neuronsInputsWeights[inputsInPreviousLayers + k*inputsInCurrentLayer + idx]
                                                            * localGradients[neuronsInPreviousLayersWithCurrent + k];
        }

        localGradients[neuron] *= derivatives[neuron];
    }
}

但我看到结果与第二个小数位的差异。为什么错误如此之大？除此之外，所有内核都运行良好。

我的GPU是NV GF555M。它支持双精度。

Answer 1

我发现了问题。而不是行：

calculateLocalGradientsForAnotherLayers <<<blocks, threads>>> (deviceLocalGradients, _neuronsInputsWeights, deviceDerivatives, _neuronsPerLayerCount[i], _neuronsInPreviousLayers[i], _neuronsInPreviousLayers[i+1], _neuronsPerLayerCount[i+1], _inputsInPreviousLayers[i], _inputsInCurrentLayer[i]);

应写道：

calculateLocalGradientsForAnotherLayers <<<blocks, threads>>> (deviceLocalGradients, _neuronsInputsWeights, deviceDerivatives, _neuronsPerLayerCount[i], _neuronsInPreviousLayers[i], _neuronsInPreviousLayers[i+1], _neuronsPerLayerCount[i+1], _inputsInPreviousLayers[i+1], _inputsInCurrentLayer[i+1]);

Answer 2

在内核主体中，需要在localGradients数组上进行某种块间同步：

for(int k=0;k<neuronsInNextLayer;k++)
        {
            localGradients[neuron] += neuronsInputsWeights[inputsInPreviousLayers + k*inputsInCurrentLayer + idx]
                                                            * localGradients[neuronsInPreviousLayersWithCurrent + k];
        }

并发读/写访问可能会破坏localGradients元素的实际值。由于读/写没有同步，您可能会看到随机结果。

为什么这个CUDA内核会给原始代码带来不同的结果？

2 个答案: