onednn-src 0.1.13

Source of oneAPI Deep Neural Network Library (oneDNN)
Documentation
/*******************************************************************************
 * Copyright 2025 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 "graph/backend/dnnl/executables/host_scalar.hpp"

namespace dnnl {
namespace impl {
namespace graph {
namespace dnnl_impl {

void host_scalar_executable_t::execute(const stream &stream,
        const std::unordered_map<int, memory> &args) const {
    auto it_src = args.find(DNNL_ARG_FROM);
    auto it_dst = args.find(DNNL_ARG_TO);

    if (it_src == args.end() || it_dst == args.end()) {
        assert(!"cannot find memory for DNNL_ARG_FROM or DNNL_ARG_TO");
        return;
    }

    const memory &src_mem = it_src->second;
    const memory &dst_mem = it_dst->second;
    DNNL_HOST_SCALAR_TYPE_SWITCH(src_mem.get_desc().get_data_type(), DType, {
        const DType val = src_mem.get_host_scalar_value<DType>();
        std::memcpy(dst_mem.get_data_handle(), &val, sizeof(DType));
    });
}

#ifdef DNNL_WITH_SYCL
std::optional<::sycl::event> host_scalar_executable_t::execute_sycl(
        const stream &stream, const std::unordered_map<int, memory> &args,
        const std::vector<::sycl::event> &deps) const {
    auto it_src = args.find(DNNL_ARG_FROM);
    auto it_dst = args.find(DNNL_ARG_TO);

    if (it_src == args.end() || it_dst == args.end()) {
        // TODO(xxx): this case should not happen. We may want to convert it to
        // a verbose error.
        assert(!"cannot find memory for DNNL_ARG_FROM or DNNL_ARG_TO");
        return std::nullopt;
    }

    const memory &src_mem = it_src->second;
    const memory &dst_mem = it_dst->second;

    // get_data_handle() is blocked for host scalar memories at the C API
    // level, so we access the underlying storage pointer directly.
    void *src_ptr = const_cast<memory_t *>(src_mem.get())
                            ->memory_storage()
                            ->data_handle();
    void *dst_ptr = dst_mem.get_data_handle();
    const size_t size = src_mem.get_desc().get_size();
    auto sycl_queue = dnnl::sycl_interop::get_queue(stream);
    return sycl_queue.submit([&](::sycl::handler &cgh) {
        cgh.depends_on(deps);
        cgh.memcpy(dst_ptr, src_ptr, size);
    });
}
#endif

#if DNNL_GPU_RUNTIME == DNNL_RUNTIME_OCL
cl_event host_scalar_executable_t::execute_ocl(const stream &stream,
        const std::unordered_map<int, memory> &args,
        const std::vector<cl_event> &deps) const {
    auto it_src = args.find(DNNL_ARG_FROM);
    auto it_dst = args.find(DNNL_ARG_TO);

    if (it_src == args.end() || it_dst == args.end()) {
        assert(!"cannot find memory for DNNL_ARG_FROM or DNNL_ARG_TO");
        return {};
    }

    const memory &src_mem = it_src->second;
    const memory &dst_mem = it_dst->second;

    assert(deps.size() <= 1);
    // Passing the empty event to memcpy below causes failure.
    const bool empty = deps.empty() || deps[0] == nullptr;
    const cl_uint num = empty ? 0 : static_cast<cl_uint>(deps.size());
    const size_t size = src_mem.get_desc().get_size();
    const auto dt = src_mem.get_desc().get_data_type();
    assert(size == types::data_type_size(static_cast<impl::data_type_t>(dt)));
    cl_event e = nullptr;
    DNNL_HOST_SCALAR_TYPE_SWITCH(dt, DType, {
        const DType val = src_mem.get_host_scalar_value<DType>();
        UNUSED_STATUS(xpu::ocl::usm::memcpy(stream.get(),
                dst_mem.get_data_handle(), static_cast<const void *>(&val),
                size, num, empty ? nullptr : deps.data(), &e));
        xpu::ocl::clWaitForEvents(1, &e);
        xpu::ocl::clReleaseEvent(e);
    });
    return nullptr;
}
#endif

arg_indices_t host_scalar_executable_t::get_arg_indices(const op_t *op) {
    UNUSED(op);
    arg_indices_t args;

    args.insert({DNNL_ARG_FROM, {indices_t::type_t::input, 0}});
    args.insert({DNNL_ARG_TO, {indices_t::type_t::output, 0}});
    return args;
}

} // namespace dnnl_impl
} // namespace graph
} // namespace impl
} // namespace dnnl