/*******************************************************************************
* Copyright 2020-2024 Intel Corporation
* Copyright 2020-2022 Codeplay Software Limited
*
* 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 "gpu/nvidia/cudnn_lrn.hpp"
#include "gpu/nvidia/stream.hpp"
#include "gpu/nvidia/sycl_cuda_scoped_context.hpp"
#include "xpu/sycl/buffer_memory_storage.hpp"

#include "xpu/sycl/memory_storage_helper.hpp"

namespace dnnl {
namespace impl {
namespace gpu {
namespace nvidia {

status_t cudnn_lrn_fwd_t::execute(const exec_ctx_t &ctx) const {

    if (memory_desc_wrapper(pd()->src_md()).has_zero_dim())
        return status::success;

    nvidia::stream_t *cuda_stream
            = utils::downcast<nvidia::stream_t *>(ctx.stream());

    return cuda_stream->interop_task([&](::sycl::handler &cgh) {
        auto arg_src = CTX_IN_SYCL_MEMORY(DNNL_ARG_SRC);
        auto arg_dst = CTX_OUT_SYCL_MEMORY(DNNL_ARG_DST);
        auto arg_wrksp = pd()->is_training()
                ? CTX_OUT_SYCL_MEMORY(DNNL_ARG_WORKSPACE)
                : arg_dst;

        compat::host_task(cgh, [=, this](const compat::interop_handle &ih) {
            auto &sycl_engine = *utils::downcast<nvidia::engine_t *>(
                    cuda_stream->engine());
            auto sc = cuda_sycl_scoped_context_handler_t(sycl_engine);
            auto handle = cuda_stream->get_cudnn_handle();

            void *src_ = arg_src.get_native_pointer(ih);
            void *dst_ = arg_dst.get_native_pointer(ih);
            void *ws_ = arg_wrksp.get_native_pointer(ih);

            std::vector<void *> args {src_, dst_, ws_};
            pd()->lrn_impl_->execute(handle, args);
        });
    });
}

status_t cudnn_lrn_bwd_t::execute(const exec_ctx_t &ctx) const {
    if (memory_desc_wrapper(pd()->src_md()).has_zero_dim())
        return status::success;

    nvidia::stream_t *cuda_stream
            = utils::downcast<nvidia::stream_t *>(ctx.stream());

    return cuda_stream->interop_task([&](::sycl::handler &cgh) {
        auto arg_src = CTX_IN_SYCL_MEMORY(DNNL_ARG_SRC);
        auto arg_diff_dst = CTX_IN_SYCL_MEMORY(DNNL_ARG_DIFF_DST);
        auto arg_diff_src = CTX_OUT_SYCL_MEMORY(DNNL_ARG_DIFF_SRC);
        auto arg_diff_ws = CTX_IN_SYCL_MEMORY(DNNL_ARG_WORKSPACE);

        compat::host_task(cgh, [=, this](const compat::interop_handle &ih) {
            std::vector<void *> args;
            auto &sycl_engine = *utils::downcast<nvidia::engine_t *>(
                    cuda_stream->engine());
            auto sc = cuda_sycl_scoped_context_handler_t(sycl_engine);
            auto handle = cuda_stream->get_cudnn_handle();

            args.push_back(arg_src.get_native_pointer(ih));
            args.push_back(arg_diff_ws.get_native_pointer(ih));
            args.push_back(arg_diff_src.get_native_pointer(ih));
            args.push_back(arg_diff_dst.get_native_pointer(ih));

            pd()->lrn_impl_->execute(handle, args);
        });
    });
}

} // namespace nvidia
} // namespace gpu
} // namespace impl
} // namespace dnnl