onednn-src 0.1.13

Source of oneAPI Deep Neural Network Library (oneDNN)
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99

/*******************************************************************************
* Copyright 2026 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.
*******************************************************************************/

#ifndef GRAPH_BACKEND_DNNL_KERNELS_GATED_MLP_PRIMITIVE_HPP
#define GRAPH_BACKEND_DNNL_KERNELS_GATED_MLP_PRIMITIVE_HPP

#include <algorithm>
#include <memory>
#include <string>
#include <utility>
#include <vector>

#include "graph/backend/dnnl/common.hpp"
#include "graph/backend/dnnl/dnnl_constant_tensor_cache.hpp"
#include "graph/backend/dnnl/dnnl_partition_impl.hpp"
#include "graph/backend/dnnl/op_executable.hpp"
#include "graph/backend/dnnl/scratchpad.hpp"
#include "graph/backend/dnnl/thread_local_cache.hpp"
#include "graph/backend/dnnl/utils.hpp"

#include "graph/backend/dnnl/passes/memory_planning.hpp"

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

template <bool quantized>
struct gated_mlp_primitive_kernel_t : public kernel_base_t {
private:
    allocator_t *g_alloc_ = nullptr;

    std::shared_ptr<subgraph_t> subgraph_;
    memory_planner_t memory_planner_;
    std::function<std::shared_ptr<execution_args_set_t>()> resource_ctor_;

public:
    gated_mlp_primitive_kernel_t() {
        thread_local_cache_t<execution_args_set_t> res_cache;
        res_cache.retain();
    }

    ~gated_mlp_primitive_kernel_t() override {
        thread_local_cache_t<execution_args_set_t> res_cache;
        res_cache.remove_if_exist(reinterpret_cast<size_t>(this));
        res_cache.release();
    }

    status_t compile_impl(const dnnl_partition_impl_t *part,
            const engine_t *engine, const std::vector<logical_tensor_t> &inputs,
            const std::vector<logical_tensor_t> &outputs) override;

    void prepare_args_set(const execution_args_set_t *res,
            const std::vector<tensor_t> &inputs,
            const std::vector<tensor_t> &outputs,
            const scratchpad_t &scratchpad);

    status_t execute_impl(const stream_t *stream,
            const std::vector<tensor_t> &inputs,
            const std::vector<tensor_t> &outputs) override;

#ifdef DNNL_WITH_SYCL
    status_t sycl_execute_impl(const stream_t *stream,
            const std::vector<tensor_t> &inputs,
            const std::vector<tensor_t> &outputs,
            const std::vector<::sycl::event> &deps,
            ::sycl::event *ret_event) override;
#endif

#if DNNL_GPU_RUNTIME == DNNL_RUNTIME_OCL
    status_t ocl_execute_impl(const stream_t *stream,
            const std::vector<tensor_t> &inputs,
            const std::vector<tensor_t> &outputs,
            const std::vector<cl_event> &deps, cl_event *ret_event) override;
#endif

    DEF_KERNEL_METHOD_STR(gated_mlp_primitive_kernel_t)
    DNNL_DISALLOW_COPY_AND_ASSIGN(gated_mlp_primitive_kernel_t)
};

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

#endif