/*******************************************************************************
* Copyright 2018 Intel Corporation
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
*     http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*******************************************************************************/

#include "mkldnn_test_common.hpp"
#include "gtest/gtest.h"

#include "mkldnn.hpp"
#include "mkldnn_debug.h"
namespace mkldnn {
using fmt = memory::format;
struct deconvolution_test_params {
    const mkldnn::engine::kind engine_kind;
    mkldnn::algorithm aalgorithm;
    test_convolution_formats_t formats;
    test_convolution_attr_t attr;
    test_convolution_sizes_t sizes;
    bool expect_to_fail;
    mkldnn_status_t expected_status;
};
template <typename data_t>
void compute_bias_fwd(const test_convolution_sizes_t &c,
    mkldnn::memory& dst, mkldnn::memory& bias) {
    data_t *bias_data = (data_t *)bias.get_data_handle();
    data_t *dst_data = (data_t *)dst.get_data_handle();

    const memory::desc bias_d = bias.get_primitive_desc().desc();
    const memory::desc dst_d = dst.get_primitive_desc().desc();

    mkldnn::impl::parallel_nd(c.mb, c.ng, c.oc / c.ng, c.oh, c.ow,
        [&](int n, int g, int oc, int oh, int ow) {
            data_t b = bias_data[map_index(bias_d, g * c.oc / c.ng + oc)];
            int oidx = n * c.oc * c.oh * c.ow
                + g * c.oc / c.ng * c.oh * c.ow
                + oc * c.oh * c.ow + oh * c.ow + ow;
            dst_data[map_index(dst_d, oidx)] += b;
        }
    );
}

template <typename data_t>
void compute_bias_bwd(const test_convolution_sizes_t &c,
    mkldnn::memory& dst, mkldnn::memory& bias) {
    data_t *bias_data = (data_t *)bias.get_data_handle();
    data_t *dst_data = (data_t *)dst.get_data_handle();

    const memory::desc bias_d = bias.get_primitive_desc().desc();
    const memory::desc dst_d = dst.get_primitive_desc().desc();

    mkldnn::impl::parallel_nd(c.ng, c.oc / c.ng, [&](int g, int oc) {
        int bidx = g * c.oc / c.ng + oc;
        bias_data[map_index(bias_d, bidx)] = 0.0;
        for (int mb = 0; mb < c.mb; ++mb) {
            for (int oh = 0; oh < c.oh; ++oh) {
                for (int ow = 0; ow < c.ow; ++ow) {
                    int oidx = mb * c.oc * c.oh * c.ow
                            + g * c.oc / c.ng * c.oh * c.ow
                            + oc * c.oh * c.ow + oh * c.ow + ow;
                    bias_data[map_index(bias_d, bidx)]
                        += dst_data[map_index(dst_d, oidx)];
                }
            }
        }
    });
}

template <typename data_t>
void transpose_wei(const test_convolution_sizes_t &c,
    mkldnn::memory& weights, mkldnn::memory& weights_tr) {

    data_t *weights_data = (data_t *)weights.get_data_handle();
    const memory::desc weights_d = weights.get_primitive_desc().desc();
    data_t *weights_tr_data = (data_t *)weights_tr.get_data_handle();
    const memory::desc weights_tr_d = weights_tr.get_primitive_desc().desc();

    mkldnn::impl::parallel_nd(c.ng, c.oc / c.ng, c.ic / c.ng, c.kh, c.kw,
        [&](int g, int oc, int ic, int kh, int kw) {
            int widx = g * c.oc / c.ng * c.ic / c.ng * c.kh * c.kw
                     + oc * c.ic / c.ng * c.kh * c.kw
                     + ic * c.kh * c.kw + kh * c.kw + kw;
            int widx_tr = g * c.oc / c.ng * c.ic / c.ng * c.kh * c.kw
                     + ic * c.oc / c.ng * c.kh * c.kw
                     + oc * c.kh * c.kw + kh * c.kw + kw;
            weights_tr_data[map_index(weights_tr_d, widx_tr)]
                 = weights_data[map_index(weights_d, widx)];
        }
    );
}

template <typename data_t>
class deconvolution_test : public
::testing::TestWithParam<deconvolution_test_params> {
private:
   std::shared_ptr<test_memory> src;
   std::shared_ptr<test_memory> weights;
   std::shared_ptr<test_memory> dst;
   std::shared_ptr<test_memory> bias;

   std::shared_ptr<memory::desc> dec_src_desc;
   std::shared_ptr<memory::desc> dec_weights_desc;
   std::shared_ptr<memory::desc> dec_bias_desc;
   std::shared_ptr<memory::desc> dec_dst_desc;

   std::shared_ptr<memory::desc> con_src_desc;
   std::shared_ptr<memory::desc> con_bias_desc;
   std::shared_ptr<memory::desc> con_dst_desc;
   std::shared_ptr<memory::desc> con_weights_desc;

   std::shared_ptr<engine> eng;
   bool with_bias;
   std::vector<int> padR;
protected:
    virtual void SetUp() {
        auto p = ::testing::TestWithParam<deconvolution_test_params>::GetParam();
        catch_expected_failures([=](){Test();}, p.expect_to_fail,
                    p.expected_status);
    }

    void Test() {
        auto p = ::testing::TestWithParam<deconvolution_test_params>::GetParam();

        ASSERT_TRUE(p.engine_kind == engine::kind::cpu);
        eng.reset(new engine(p.engine_kind, 0));

        ASSERT_EQ(p.aalgorithm, algorithm::deconvolution_direct);
        memory::data_type data_type = data_traits<data_t>::data_type;

        test_convolution_sizes_t dd = p.sizes;
        p.formats.bias_format = memory::format::format_undef;
        with_bias = p.formats.bias_format != memory::format::format_undef;

        memory::dims src_dims = {dd.mb, dd.ic, dd.ih, dd.iw};
        memory::dims dst_dims = {dd.mb, dd.oc, dd.oh, dd.ow};
        memory::dims weights_dims, c_weights_dims;
        if (dd.ng > 1) {
            weights_dims = { dd.ng, dd.oc / dd.ng, dd.ic / dd.ng, dd.kh, dd.kw };
            c_weights_dims = { dd.ng, dd.ic / dd.ng, dd.oc / dd.ng, dd.kh, dd.kw };
        } else {
            weights_dims = { dd.oc, dd.ic, dd.kh, dd.kw };
            c_weights_dims = { dd.ic, dd.oc, dd.kh, dd.kw };
        }
        memory::dims bias_dims;
        if (with_bias) bias_dims = {dd.oc};
        else bias_dims = {};

        dec_src_desc.reset(new memory::desc(src_dims, data_type,
            p.formats.src_format));
        dec_dst_desc.reset(new memory::desc(dst_dims, data_type,
            p.formats.src_format));
        dec_weights_desc.reset(new memory::desc(weights_dims, data_type,
            p.formats.weights_format));
        dec_bias_desc.reset(new memory::desc(bias_dims, data_type,
            p.formats.bias_format));

        con_src_desc.reset(new memory::desc(dst_dims, data_type,
            p.formats.src_format));
        con_dst_desc.reset(new memory::desc(src_dims, data_type,
            p.formats.src_format));
        con_weights_desc.reset(new memory::desc(c_weights_dims, data_type,
            p.formats.weights_format));

        src.reset(new test_memory(*dec_src_desc, *eng));
        weights.reset(new test_memory(*dec_weights_desc, *eng));
        bias.reset(new test_memory(*dec_bias_desc, *eng));
        dst.reset(new test_memory(*dec_dst_desc, *eng));

        padR = {
            right_padding(dd.oh, dd.ih, dd.kh, dd.padh, dd.strh, dd.dilh),
            right_padding(dd.ow, dd.iw, dd.kw, dd.padw, dd.strw, dd.dilw)
        };
        Forward();
        BackwardData();
        BackwardWeights();
    }
    void Forward() {
        auto aprop_kind =  prop_kind::forward;
        deconvolution_test_params p =
            ::testing::TestWithParam<deconvolution_test_params>::GetParam();
        auto conv_src = test_memory(*con_src_desc, *eng);
        auto conv_dst = src;
        test_convolution_sizes_t dd = p.sizes;

        fill_data<data_t>(src->get_size() / sizeof(data_t),
            (data_t *)src->get().get_data_handle());

        fill_data<data_t>(weights->get_size() / sizeof(data_t),
                (data_t *)weights->get().get_data_handle());
        if (with_bias) {
            fill_data<data_t>(bias->get_size() / sizeof(data_t),
                    (data_t *)bias->get().get_data_handle());
        }

        auto weights_tr = memory({*con_weights_desc, *eng});
        transpose_wei<data_t>(dd, weights->get(), weights_tr);
        auto deconv_desc = with_bias ?
            deconvolution_forward::desc(aprop_kind,
                    algorithm::deconvolution_direct, *dec_src_desc,
                    *dec_weights_desc, *dec_bias_desc, *dec_dst_desc,
                    { dd.strh, dd.strw }, { dd.padh, dd.padw }, padR,
                    padding_kind::zero) :
                deconvolution_forward::desc(aprop_kind,
                        algorithm::deconvolution_direct, *dec_src_desc,
                        *dec_weights_desc, *dec_dst_desc, { dd.strh, dd.strw },
                        { dd.padh, dd.padw }, padR, padding_kind::zero);

        auto deconv_primitive_desc = deconvolution_forward::primitive_desc(
                deconv_desc, *eng);

        auto deconv = with_bias ?
            deconvolution_forward(deconv_primitive_desc, src->get(),
                    weights->get(), bias->get(), dst->get()) :
            deconvolution_forward(deconv_primitive_desc, src->get(),
                    weights->get(), dst->get());

        auto conv_desc = convolution_forward::desc(
                prop_kind::forward_training, algorithm::convolution_direct,
                *con_src_desc, *con_weights_desc, *con_dst_desc,
                { dd.strh, dd.strw }, { dd.padh, dd.padw }, padR,
                padding_kind::zero);

        auto conv_primitive_desc = convolution_forward::primitive_desc(
                conv_desc, *eng);

        auto conv_bwd_data_desc = convolution_backward_data::desc(
                algorithm::convolution_direct, *con_src_desc,
                *con_weights_desc, *con_dst_desc,
                { dd.strh, dd.strw }, { dd.padh, dd.padw }, padR,
                padding_kind::zero);

        auto conv_bwd_data_primitive_desc
            = convolution_backward_data::primitive_desc(
                    conv_bwd_data_desc, *eng, conv_primitive_desc);

        auto conv_bwd_data = convolution_backward_data(
                conv_bwd_data_primitive_desc,
                conv_dst->get(), weights_tr, conv_src.get());

        std::vector<primitive> pipeline;
        pipeline.push_back(deconv);
        pipeline.push_back(conv_bwd_data);
        stream(stream::kind::lazy).submit(pipeline).wait();

        if(with_bias) compute_bias_fwd<data_t>(dd, conv_src.get(), bias->get());
        compare_data<data_t>(conv_src.get(), dst->get());
    }

    void BackwardData() {
        auto p = ::testing::TestWithParam<deconvolution_test_params>::GetParam();
        auto conv_src = dst;
        auto conv_dst = test_memory(*con_dst_desc, *eng);
        test_convolution_sizes_t dd = p.sizes;

        fill_data<data_t>(weights->get_size() / sizeof(data_t),
            (data_t *)weights->get().get_data_handle());

        fill_data<data_t>(dst->get_size() / sizeof(data_t),
                (data_t *)dst->get().get_data_handle());

        auto weights_tr = memory({*con_weights_desc, *eng});
        transpose_wei<data_t>(dd, weights->get(), weights_tr);

        auto deconv_desc = deconvolution_forward::desc(prop_kind::forward_training,
                algorithm::deconvolution_direct, *dec_src_desc,
                *dec_weights_desc, *dec_dst_desc, { dd.strh, dd.strw },
                { dd.padh, dd.padw }, padR, padding_kind::zero);

        auto deconv_primitive_desc = deconvolution_forward::primitive_desc(
                deconv_desc, *eng);

        auto deconv_bwd_data_desc = deconvolution_backward_data::desc(
                algorithm::deconvolution_direct, *dec_src_desc,
                *dec_weights_desc, *dec_dst_desc,
                { dd.strh, dd.strw }, { dd.padh, dd.padw }, padR,
                padding_kind::zero);
        auto deconv_bwd_data_primitive_desc
            = deconvolution_backward_data::primitive_desc(
                    deconv_bwd_data_desc, *eng, deconv_primitive_desc);

        auto deconv_bwd_data = deconvolution_backward_data(
                deconv_bwd_data_primitive_desc, dst->get(), weights->get(),
                src->get());

        auto conv_desc = convolution_forward::desc(
                prop_kind::forward_training, algorithm::convolution_direct,
                *con_src_desc, *con_weights_desc, *con_dst_desc,
                { dd.strh, dd.strw }, { dd.padh, dd.padw }, padR,
                padding_kind::zero);

        auto conv_primitive_desc = convolution_forward::primitive_desc(
                conv_desc, *eng);

        auto conv = convolution_forward(conv_primitive_desc, conv_src->get(),
                weights_tr, conv_dst.get());

        std::vector<primitive> pipeline;
        pipeline.push_back(deconv_bwd_data);
        pipeline.push_back(conv);
        stream(stream::kind::lazy).submit(pipeline).wait();

        compare_data<data_t>(conv_dst.get(), src->get());
    }

    void BackwardWeights() {
        auto p = ::testing::TestWithParam<deconvolution_test_params>::GetParam();
        auto conv_src = dst;
        auto conv_dst = src;
        auto conv_weights = memory({*con_weights_desc, *eng});
        test_convolution_sizes_t dd = p.sizes;

        fill_data<data_t>(src->get_size() / sizeof(data_t),
            (data_t *)src->get().get_data_handle());

        fill_data<data_t>(dst->get_size() / sizeof(data_t),
                (data_t *)dst->get().get_data_handle());

        auto deconv_desc = deconvolution_forward::desc(prop_kind::forward_training,
                algorithm::deconvolution_direct, *dec_src_desc,
                *dec_weights_desc, *dec_bias_desc, *dec_dst_desc,
                { dd.strh, dd.strw }, { dd.padh, dd.padw }, padR, padding_kind::zero);

        auto deconv_primitive_desc = deconvolution_forward::primitive_desc(
                deconv_desc, *eng);

        auto deconv_bwd_weights_desc = deconvolution_backward_weights::desc(
                algorithm::deconvolution_direct, *dec_src_desc,
                *dec_weights_desc, *dec_bias_desc, *dec_dst_desc,
                { dd.strh, dd.strw }, { dd.padh, dd.padw }, padR,
                padding_kind::zero);
        auto deconv_bwd_weights_primitive_desc
            = deconvolution_backward_weights::primitive_desc(
                    deconv_bwd_weights_desc, *eng, deconv_primitive_desc);

        auto deconv_bwd_weights = deconvolution_backward_weights(
                deconv_bwd_weights_primitive_desc, src->get(), dst->get(),
                weights->get(), bias->get());

        auto conv_desc = convolution_forward::desc(
                prop_kind::forward_training, algorithm::convolution_direct,
                *con_src_desc, *con_weights_desc, *con_dst_desc,
                { dd.strh, dd.strw }, { dd.padh, dd.padw }, padR,
                padding_kind::zero);

        auto conv_primitive_desc = convolution_forward::primitive_desc(
                conv_desc, *eng);

        auto conv_bwd_weights_desc = convolution_backward_weights::desc(
                algorithm::convolution_direct, *con_src_desc, *con_weights_desc,
                *con_dst_desc, { dd.strh, dd.strw }, { dd.padh, dd.padw },
                padR, padding_kind::zero);

        auto conv_bwd_weights_primitive_desc =
            convolution_backward_weights::primitive_desc(
                    conv_bwd_weights_desc, *eng, conv_primitive_desc);

        auto conv_bwd_weights =
            convolution_backward_weights(conv_bwd_weights_primitive_desc,
                    conv_src->get(), conv_dst->get(), conv_weights);

        std::vector<primitive> pipeline;
        pipeline.push_back(conv_bwd_weights);
        pipeline.push_back(deconv_bwd_weights);
        stream(stream::kind::lazy).submit(pipeline).wait();

        auto weights_tr = memory({*con_weights_desc, *eng});
        transpose_wei<data_t>(dd, weights->get(), weights_tr);

        compare_data<data_t>(weights_tr, conv_weights);

        if (with_bias) {
            auto ref_bias = memory({*dec_bias_desc, *eng});
            compute_bias_bwd<data_t>(dd, dst->get(), ref_bias);
            compare_data<data_t>(ref_bias, bias->get());
        }
    }
};

using deconvolution_test_float = deconvolution_test<float>;

TEST_P(deconvolution_test_float, TestDeconvolution)
{
}

#define EXPAND_FORMATS(src, weights, bias, dst) \
    { mkldnn::memory::format::src, mkldnn::memory::format::weights, \
    mkldnn::memory::format::bias, mkldnn::memory::format::dst }

#define ENGINE engine::kind::cpu
#define ALGORITHM mkldnn::deconvolution_direct

#define PARAMS(src, weights, bias, dst, ...) \
        deconvolution_test_params { ENGINE, ALGORITHM, \
            EXPAND_FORMATS(src, weights, bias, dst), {}, \
                {__VA_ARGS__} }

#define INST_TEST_CASE(str, ...) INSTANTIATE_TEST_SUITE_P( \
        str, deconvolution_test_float, ::testing::Values(__VA_ARGS__))

#define FMT_BIAS x
#define FMT_DATA_BLOCKED nChw8c
#define FMT_WEIGHTS_BLOCKED Ohwi8o

INST_TEST_CASE(SimpleSmall_NCHW,
    PARAMS(nchw, oihw, x, nchw,
        2, 1, 6, 4, 4, 4, 4, 4, 3, 3, 1, 1, 1, 1),
    PARAMS(nchw, oihw, x, nchw,
        2, 1, 6, 2, 2, 4, 4, 4, 3, 3, 0, 0, 1, 1),
    PARAMS(nhwc, oihw, x, nhwc,
        2, 1, 6, 2, 2, 4, 4, 4, 3, 3, 0, 0, 1, 1),
    PARAMS(nhwc, hwio, x, nhwc,
        2, 1, 6, 4, 4, 4, 4, 4, 3, 3, 1, 1, 1, 1),
    PARAMS(nhwc, hwio, x, nhwc,
        2, 1, 6, 2, 2, 4, 4, 4, 3, 3, 0, 0, 1, 1),
    PARAMS(nhwc, goihw, x, nhwc,
        2, 2, 6, 4, 4, 4, 4, 4, 3, 3, 0, 0, 1, 1),
    PARAMS(nhwc, hwigo, x, nhwc,
        2, 2, 6, 4, 4, 4, 4, 4, 3, 3, 1, 1, 1, 1)

);

INST_TEST_CASE(SimpleSmall_Blocked,
    PARAMS(FMT_DATA_BLOCKED, FMT_WEIGHTS_BLOCKED, FMT_BIAS, FMT_DATA_BLOCKED,
        2, 1, 32, 12, 12, 32, 13, 13, 3, 3, 0, 0, 1, 1),
    PARAMS(FMT_DATA_BLOCKED, FMT_WEIGHTS_BLOCKED, FMT_BIAS, FMT_DATA_BLOCKED,
        2, 1, 32, 4, 4, 32, 3, 3, 3, 3, 1, 1, 1, 1),
    PARAMS(FMT_DATA_BLOCKED, FMT_WEIGHTS_BLOCKED, FMT_BIAS, FMT_DATA_BLOCKED,
        2, 1, 32, 4, 4, 32, 4, 4, 3, 3, 0, 0, 1, 1),
    PARAMS(FMT_DATA_BLOCKED, FMT_WEIGHTS_BLOCKED, FMT_BIAS, FMT_DATA_BLOCKED,
        2, 1, 32, 2, 2, 32, 3, 3, 3, 3, 0, 0, 1, 1),
    PARAMS(FMT_DATA_BLOCKED, FMT_WEIGHTS_BLOCKED, FMT_BIAS, FMT_DATA_BLOCKED,
        2, 1, 32, 2, 2, 32, 2, 2, 3, 3, 1, 1, 1, 1),
    PARAMS(FMT_DATA_BLOCKED, FMT_WEIGHTS_BLOCKED, FMT_BIAS, FMT_DATA_BLOCKED,
        2, 1, 48, 13, 13, 32, 13, 13, 3, 3, 1, 1, 1, 1),
    PARAMS(FMT_DATA_BLOCKED, FMT_WEIGHTS_BLOCKED, FMT_BIAS, FMT_DATA_BLOCKED,
        2, 1, 48, 11, 11, 32, 13, 13, 3, 3, 0, 0, 1, 1)
);


}

