aboutsummaryrefslogtreecommitdiff
path: root/arm_compute/runtime/CL/CLTensorAllocator.h
blob: f7800d39f86cad746039859bafa38a0d4baab660 (plain)
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
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
/*
 * Copyright (c) 2016-2019 ARM Limited.
 *
 * SPDX-License-Identifier: MIT
 *
 * Permission is hereby granted, free of charge, to any person obtaining a copy
 * of this software and associated documentation files (the "Software"), to
 * deal in the Software without restriction, including without limitation the
 * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
 * sell copies of the Software, and to permit persons to whom the Software is
 * furnished to do so, subject to the following conditions:
 *
 * The above copyright notice and this permission notice shall be included in all
 * copies or substantial portions of the Software.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
 * SOFTWARE.
 */
#ifndef __ARM_COMPUTE_CLTENSORALLOCATOR_H__
#define __ARM_COMPUTE_CLTENSORALLOCATOR_H__

#include "arm_compute/runtime/CL/CLArray.h"
#include "arm_compute/runtime/CL/CLMemory.h"
#include "arm_compute/runtime/ITensorAllocator.h"

#include "arm_compute/core/CL/CLTypes.h"
#include "arm_compute/core/CL/OpenCL.h"

#include <cstdint>

namespace arm_compute
{
class CLTensor;
template <typename>
class MemoryGroupBase;
/** Memory Group in OpenCL */
using CLMemoryGroup = MemoryGroupBase<CLTensor>;

/** Basic implementation of a CL memory tensor allocator. */
class CLTensorAllocator : public ITensorAllocator
{
public:
    /** Default constructor.
     *
     * @param[in] owner (Optional) Owner of the allocator.
     */
    CLTensorAllocator(CLTensor *owner = nullptr);
    /** Prevent instances of this class from being copied (As this class contains pointers) */
    CLTensorAllocator(const CLTensorAllocator &) = delete;
    /** Prevent instances of this class from being copy assigned (As this class contains pointers) */
    CLTensorAllocator &operator=(const CLTensorAllocator &) = delete;
    /** Allow instances of this class to be moved */
    CLTensorAllocator(CLTensorAllocator &&) = default;
    /** Allow instances of this class to be moved */
    CLTensorAllocator &operator=(CLTensorAllocator &&) = default;

    /** Interface to be implemented by the child class to return the pointer to the mapped data.
     *
     * @return pointer to the mapped data.
     */
    uint8_t *data();
    /** Interface to be implemented by the child class to return the pointer to the CL data.
     *
     * @return pointer to the CL data.
     */
    const cl::Buffer &cl_data() const;
    /** Wrapped quantization info data accessor
     *
     * @return A wrapped quantization info object.
     */
    CLQuantization quantization() const;

    /** Enqueue a map operation of the allocated buffer on the given queue.
     *
     * @param[in,out] q        The CL command queue to use for the mapping operation.
     * @param[in]     blocking If true, then the mapping will be ready to use by the time
     *                         this method returns, else it is the caller's responsibility
     *                         to flush the queue and wait for the mapping operation to have completed before using the returned mapping pointer.
     *
     * @return The mapping address.
     */
    uint8_t *map(cl::CommandQueue &q, bool blocking);
    /** Enqueue an unmap operation of the allocated buffer on the given queue.
     *
     * @note This method simply enqueue the unmap operation, it is the caller's responsibility to flush the queue and make sure the unmap is finished before
     *       the memory is accessed by the device.
     *
     * @param[in,out] q       The CL command queue to use for the mapping operation.
     * @param[in]     mapping The cpu mapping to unmap.
     */
    void unmap(cl::CommandQueue &q, uint8_t *mapping);

    /** Allocate size specified by TensorInfo of OpenCL memory.
     *
     * @note: The tensor must not already be allocated when calling this function.
     *
     */
    void allocate() override;

    /** Free allocated OpenCL memory.
     *
     * @note The tensor must have been allocated when calling this function.
     *
     */
    void free() override;
    /** Import an existing memory as a tensor's backing memory
     *
     * @warning memory should have been created under the same context that ACL uses.
     * @warning memory is expected to be aligned with the device requirements.
     * @warning tensor shouldn't be memory managed.
     * @warning ownership of memory is not transferred.
     * @warning memory must be writable in case of in-place operations
     * @warning padding should be accounted by the client code.
     * @note buffer size will be checked to be compliant with total_size reported by ITensorInfo.
     *
     * @param[in] buffer Buffer to be used as backing memory
     *
     * @return An error status
     */
    Status import_memory(cl::Buffer buffer);
    /** Associates the tensor with a memory group
     *
     * @param[in] associated_memory_group Memory group to associate the tensor with
     */
    void set_associated_memory_group(CLMemoryGroup *associated_memory_group);

protected:
    /** Call map() on the OpenCL buffer.
     *
     * @return A pointer to the beginning of the tensor's allocation.
     */
    uint8_t *lock() override;
    /** Call unmap() on the OpenCL buffer. */
    void unlock() override;

private:
    static const cl::Buffer _empty_buffer;

private:
    CLMemoryGroup *_associated_memory_group; /**< Registered memory manager */
    CLMemory       _memory;                  /**< OpenCL memory */
    uint8_t       *_mapping;                 /**< Pointer to the CPU mapping of the OpenCL buffer. */
    CLTensor      *_owner;                   /**< Owner of the allocator */
    CLFloatArray   _scale;                   /**< Scales array in case of quantized per channel data type */
    CLInt32Array   _offset;                  /**< Offsets array in case of quantized per channel data type */
};
} // namespace arm_compute
#endif /* __ARM_COMPUTE_CLTENSORALLOCATOR_H__ */