Faster

Joseph Redmon

2015-01-13 aa5996d58e68edfbefe51061856aecd549dd09c4

 src/network_gpu.c

@@ -24,6 +24,7 @@
{
    int i;
    for(i = 0; i < net.n; ++i){
        clock_t time = clock();
        if(net.types[i] == CONVOLUTIONAL){
            convolutional_layer layer = *(convolutional_layer *)net.layers[i];
            forward_convolutional_layer_gpu(layer, input);
@@ -52,12 +53,15 @@
            if(!train) continue;
            dropout_layer layer = *(dropout_layer *)net.layers[i];
            forward_dropout_layer_gpu(layer, input);
            input = layer.output_cl;
        }
        else if(net.types[i] == CROP){
            crop_layer layer = *(crop_layer *)net.layers[i];
            forward_crop_layer_gpu(layer, input);
            input = layer.output_cl;
        }
        check_error(cl);
        //printf("Forw %d %f\n", i, sec(clock() - time));
    }
}

@@ -67,6 +71,7 @@
    cl_mem prev_input;
    cl_mem prev_delta;
    for(i = net.n-1; i >= 0; --i){
        clock_t time = clock();
        if(i == 0){
            prev_input = input;
            prev_delta = 0;
@@ -98,6 +103,8 @@
            softmax_layer layer = *(softmax_layer *)net.layers[i];
            backward_softmax_layer_gpu(layer, prev_delta);
        }
        check_error(cl);
        //printf("Back %d %f\n", i, sec(clock() - time));
    }
}

@@ -138,7 +145,8 @@
        softmax_layer layer = *(softmax_layer *)net.layers[i];
        return layer.output_cl;
    } else if(net.types[i] == DROPOUT){
        return get_network_output_cl_layer(net, i-1);
        dropout_layer layer = *(dropout_layer *)net.layers[i];
        return layer.output_cl;
    }
    return 0;
}
@@ -161,6 +169,7 @@
        softmax_layer layer = *(softmax_layer *)net.layers[i];
        return layer.delta_cl;
    } else if(net.types[i] == DROPOUT){
        if(i == 0) return 0;
        return get_network_delta_cl_layer(net, i-1);
    }
    return 0;
@@ -192,6 +201,7 @@
    }
    else if(net.types[i] == CONNECTED){
        connected_layer layer = *(connected_layer *)net.layers[i];
        cl_read_array(layer.output_cl, layer.output, layer.outputs*layer.batch);
        return layer.output;
    }
    else if(net.types[i] == MAXPOOL){