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
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
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180

/*******************************************************************************
* Copyright 2019 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 <assert.h>
#include "oneapi/dnnl/dnnl.h"
#include "opdesc.hpp"
#include "primitive_desc_iface.hpp"

#include "c_types_map.hpp"
#include "type_helpers.hpp"
#include "utils.hpp"

using namespace dnnl::impl;
using namespace dnnl::impl::utils;
using namespace dnnl::impl::status;
using namespace dnnl::impl::prop_kind;
using namespace dnnl::impl::alg_kind;
using namespace dnnl::impl::types;

#define VCHECK_RS(cond, msg, ...) \
    VCONDCHECK(primitive, create, check, resampling, (cond), \
            status::invalid_arguments, msg, ##__VA_ARGS__);

#define VCHECK_RS_UNIMPL(cond, msg, ...) \
    VCONDCHECK(primitive, create, check, resampling, (cond), \
            status::unimplemented, msg, ##__VA_ARGS__);

namespace {
status_t resampling_desc_init(resampling_desc_t *resampling_desc,
        prop_kind_t prop_kind, alg_kind_t alg_kind, const float *factors,
        const memory_desc_t *src_desc, const memory_desc_t *dst_desc) {
    VCHECK_RS(one_of(alg_kind, resampling_nearest, resampling_linear),
            VERBOSE_BAD_ALGORITHM);
    VCHECK_RS(src_desc, VERBOSE_NULL_ARG);
    VCHECK_RS(IMPLICATION(dst_desc == nullptr, factors), VERBOSE_NULL_ARG);
    VCHECK_RS(utils::one_of(src_desc->ndims, 3, 4, 5), VERBOSE_BAD_NDIMS, "src",
            src_desc->ndims);
    VCHECK_RS(!any_memory_desc_host_scalar(src_desc, dst_desc),
            VERBOSE_UNSUPPORTED_FORMAT_KIND);

    const bool is_fwd = one_of(prop_kind, forward_training, forward_inference);
    VCHECK_RS(IMPLICATION(is_fwd, src_desc->format_kind != format_kind::any),
            VERBOSE_UNSUPPORTED_TAG_S, "src");

    auto rd = resampling_desc_t();
    rd.primitive_kind = primitive_kind::resampling;
    rd.prop_kind = prop_kind;
    rd.alg_kind = alg_kind;

    VCHECK_RS(!memory_desc_wrapper(src_desc).has_runtime_dims_or_strides(),
            VERBOSE_RUNTIMEDIM_UNSUPPORTED);
    VCHECK_RS(IMPLICATION(dst_desc,
                      !memory_desc_wrapper(dst_desc)
                               .has_runtime_dims_or_strides()),
            VERBOSE_RUNTIMEDIM_UNSUPPORTED);

    auto fill_dst_md = [](const memory_desc_t *i_md, const float *factors,
                               memory_desc_t *o_md) {
        o_md->ndims = i_md->ndims;
        o_md->data_type = i_md->data_type;
        utils::array_copy(o_md->dims, i_md->dims, 2);
        for (int i = 0; i < o_md->ndims - 2; i++) {
            double i_dim_val = static_cast<double>(i_md->dims[2 + i]);
            o_md->dims[2 + i] = static_cast<dim_t>(i_dim_val * factors[i]);
        }
        o_md->format_kind = format_kind::any;
    };

    (prop_kind == backward_data ? rd.diff_src_desc : rd.src_desc) = *src_desc;
    if (dst_desc)
        (is_fwd ? rd.dst_desc : rd.diff_dst_desc) = *dst_desc;
    else {
        dst_desc = (is_fwd ? &rd.dst_desc : &rd.diff_dst_desc);
        fill_dst_md(
                src_desc, factors, (is_fwd ? &rd.dst_desc : &rd.diff_dst_desc));
    }

    /* User provided factors are used only to compute destination dimensions.
     Implementation uses true scaling factors from source to destination */
    for (int i = 0; i < src_desc->ndims - 2; i++) {
        double dst_dim_val = static_cast<double>(dst_desc->dims[2 + i]);
        double src_dim_val = static_cast<double>(src_desc->dims[2 + i]);
        rd.factors[i] = static_cast<float>(dst_dim_val / src_dim_val);
    }

    VCHECK_RS(src_desc->ndims == dst_desc->ndims,
            VERBOSE_INCONSISTENT_NDIMS_WITH_VALS, "src", "dst", src_desc->ndims,
            dst_desc->ndims);
    for (int i : {0, 1})
        VCHECK_RS(src_desc->dims[i] == dst_desc->dims[i],
                VERBOSE_INCONSISTENT_DIM, "src", i, "dst", i);

    *resampling_desc = rd;
    return success;
}

status_t resampling_attr_check(const resampling_desc_t &desc,
        const engine_t *engine, const primitive_attr_t *attr) {
    using smask_t = primitive_attr_t::skip_mask_t;

    if (attr == nullptr) return status::success;
    if (attr->has_default_values()) return status::success;

    if (one_of(desc.prop_kind, prop_kind::forward_inference,
                prop_kind::forward_training)) {
        // Check attributes
        const data_type_t dst_dt = desc.dst_desc.data_type;

        auto attr_mask = smask_t::post_ops;

        VCHECK_RS_UNIMPL(attr->has_default_values(attr_mask, dst_dt),
                VERBOSE_UNSUPPORTED_ATTR);

        // Check post-ops
        if (!attr->post_ops_.has_default_values()) {
            const auto &po = attr->post_ops_;
            using namespace primitive_kind;
            VCHECK_RS_UNIMPL(po.has_default_values({binary, eltwise, sum}),
                    VERBOSE_UNSUPPORTED_POSTOP);

            // Check sum
            VCHECK_RS_UNIMPL(po.check_sum_consistency(dst_dt, false, true),
                    VERBOSE_UNSUPPORTED_POSTOP);

            // Note: verbose support is inside the call.
            CHECK(po.validate_binary(engine->kind(), &desc.dst_desc));
        }
    } else {
        VCHECK_RS_UNIMPL(false, VERBOSE_UNSUPPORTED_ATTR);
    }

    return status::success;
}

} // namespace

status_t dnnl_resampling_forward_primitive_desc_create(
        primitive_desc_iface_t **primitive_desc_iface, engine_t *engine,
        prop_kind_t prop_kind, alg_kind_t alg_kind, const float *factors,
        const memory_desc_t *src_desc, const memory_desc_t *dst_desc,
        const primitive_attr_t *attr) {
    if (!one_of(prop_kind, forward_training, forward_inference))
        return invalid_arguments;

    auto resampling_desc = resampling_desc_t();
    CHECK(resampling_desc_init(&resampling_desc, prop_kind, alg_kind, factors,
            src_desc, dst_desc));
    CHECK(resampling_attr_check(resampling_desc, engine, attr));
    return primitive_desc_create(primitive_desc_iface, engine,
            (const op_desc_t *)&resampling_desc, nullptr, attr);
}

status_t dnnl_resampling_backward_primitive_desc_create(
        primitive_desc_iface_t **primitive_desc_iface, engine_t *engine,
        alg_kind_t alg_kind, const float *factors,
        const memory_desc_t *diff_src_desc, const memory_desc_t *diff_dst_desc,
        const primitive_desc_iface_t *hint_fwd_pd,
        const primitive_attr_t *attr) {

    auto resampling_desc = resampling_desc_t();
    CHECK(resampling_desc_init(&resampling_desc, backward_data, alg_kind,
            factors, diff_src_desc, diff_dst_desc));
    CHECK(resampling_attr_check(resampling_desc, engine, attr));
    return primitive_desc_create(primitive_desc_iface, engine,
            (const op_desc_t *)&resampling_desc, hint_fwd_pd, attr);
}
// vim: et ts=4 sw=4 cindent cino+=l0,\:4,N-s