Class EReduceMaxZeroGradOp

en_ops/e_reduce_max_zero_.cc:88–133

REGISTER_KERNEL_BUILDER(Name("EReduceMaxZero").Device(DEVICE_CPU), EReduceMaxZeroOp);

class EReduceMaxZeroGradOp : public OpKernel
{
public:
    explicit EReduceMaxZeroGradOp(OpKernelConstruction *context) : OpKernel(context)
    {
        OP_REQUIRES_OK(context, context->GetAttr("eid_low", &eid_low_));
        OP_REQUIRES_OK(context, context->GetAttr("eid_high", &eid_high_));
        OP_REQUIRES_OK(context, context->GetAttr("times", &times_));
        lib = dlopen("/home/zhangyan/jhrsgx/privacy_test/privacy_tf/sgx_tf_ops/sgx.so", RTLD_LAZY);
        OP_REQUIRES(context, lib != NULL, errors::Unknown("Unable to load sgx.so!"));
    }
    void Compute(OpKernelContext *context) override
    {

        const Tensor &grad = context->input(0);
        const Tensor &input = context->input(1);

        auto grad_flat = grad.flat<float>();
        auto input_flat = input.flat<float>();

        const TensorShape &input_shape = input.shape();

        Tensor *output = NULL;
        OP_REQUIRES_OK(context, context->allocate_output(0, input_shape, &output));
        auto output_flat = output->flat<float>();

        int N = input_flat.size() / times_;
        int M = input_shape.dim_size(0) / times_;
        int C = input_shape.dim_size(1);
        int L = input_shape.dim_size(2);

        unsigned long int eid_ = (eid_high_ << 32) + eid_low_;
        typedef void (*function)(unsigned long int eid, float *input, float *grad, int N, int M, int C, int L, float *output);
        dlerror();
        function reducemax_zero_grad_kernel = (function)dlsym(lib, "reducemax_zero_grad");
        const char *dlsym_error = dlerror();
        OP_REQUIRES(context, !dlsym_error, errors::Unknown("loading of reducemax_zero_grad failed: ", dlsym_error));
        reducemax_zero_grad_kernel(eid_, (float *)input_flat.data(), (float *)grad_flat.data(), N, M, C, L, (float *)output_flat.data());
    };

private:
    void *lib;
    int64 eid_low_;
    int64 eid_high_;
    int64 times_;
};
REGISTER_KERNEL_BUILDER(Name("EReduceMaxZeroGrad").Device(DEVICE_CPU), EReduceMaxZeroGradOp);