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

/*******************************************************************************
* Copyright 2023 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 "gpu/generic/sycl/ref_resampling.hpp"
#include "gpu/generic/sycl/resampling_kernels.hpp"

namespace dnnl {
namespace impl {
namespace gpu {
namespace generic {
namespace sycl {

status_t ref_resampling_fwd_t::pd_t::init_conf() {
    conf_ = sycl_resampling_conf_t();

    conf_.block_size = 16;
    conf_.wg_size = 32;

    for (int i = 0; i < DNNL_MAX_NDIMS; i++) {
        conf_.dst_dims[i] = dst_md()->dims[i];
    }
    conf_.dst_ndims = dst_md()->ndims;
    auto nelems_A = memory_desc_wrapper(src_md(0)).nelems();
    conf_.work_amount = nelems_A;
    int work_per_wg = conf_.wg_size * conf_.block_size;
    int n_wgs = (nelems_A + work_per_wg - 1) / work_per_wg;
    conf_.n_thr = n_wgs * conf_.wg_size;

    conf_.src_md = xpu::sycl::md_t(src_md(0));
    conf_.dst_md = xpu::sycl::md_t(dst_md());

    conf_.alg = desc()->alg_kind;

    conf_.post_ops = sycl_post_ops_t(attr(), dst_md());
    return status::success;
}

status_t ref_resampling_fwd_t::init(impl::engine_t *engine) {
    const auto kid = ::sycl::get_kernel_id<resampling_kernel_fwd_vec_t>();
    return create_kernel(engine, kid, &kernel_);
}

status_t ref_resampling_fwd_t::execute_forward(const exec_ctx_t &ctx) const {
    parallel_for(ctx, kernel_, [&](::sycl::handler &cgh) {
        resampling_kernel_fwd_vec_t resampling_fwd_kernel(
                pd()->conf_, cgh, ctx);

        auto nelems_A = memory_desc_wrapper(pd()->src_md(0)).nelems();
        const int block_size = pd()->conf_.block_size;
        const int wg_size = pd()->conf_.wg_size;

        int work_per_wg = wg_size * block_size;
        int n_wgs = (nelems_A + work_per_wg - 1) / work_per_wg;
        int n_thr = n_wgs * wg_size;
        cgh.parallel_for(
                ::sycl::nd_range<1>(n_thr, wg_size), resampling_fwd_kernel);
    });

    return status::success;
}

status_t ref_resampling_bwd_t::pd_t::init_conf() {
    conf_ = sycl_resampling_conf_t();

    conf_.diff_src_md = xpu::sycl::md_t(diff_src_md(0));
    conf_.diff_dst_md = xpu::sycl::md_t(diff_dst_md());

    conf_.block_size = 16;
    conf_.wg_size = 32;
    conf_.dst_ndims = dst_md()->ndims;
    auto nelems_A = memory_desc_wrapper(diff_src_md(0)).nelems();
    conf_.work_amount = nelems_A;
    int work_per_wg = conf_.wg_size * conf_.block_size;
    int n_wgs = (nelems_A + work_per_wg - 1) / work_per_wg;
    conf_.n_thr = n_wgs * conf_.wg_size;
    conf_.alg = desc()->alg_kind;

    for (int i = 0; i < DNNL_MAX_NDIMS; i++) {
        conf_.dst_dims[i] = dst_md()->dims[i];
    }
    return status::success;
}

status_t ref_resampling_bwd_t::init(impl::engine_t *engine) {
    if (pd()->conf_.alg == alg_kind::resampling_nearest) {
        const auto kid = ::sycl::get_kernel_id<resampling_kernel_bwd_vec_t>();
        CHECK(create_kernel(engine, kid, &kernel_));
    } else {
        const auto kid = ::sycl::get_kernel_id<resampling_kernel_bwd_vec1_t>();
        CHECK(create_kernel(engine, kid, &kernel_));
    }
    return status::success;
}

status_t ref_resampling_bwd_t::execute_backward(const exec_ctx_t &ctx) const {
    return parallel_for(ctx, kernel_, [&](::sycl::handler &cgh) {
        auto dst_mem_arg = CTX_IN_SYCL_KERNEL_MEMORY(DNNL_ARG_DIFF_DST);
        auto src_mem_arg = CTX_OUT_SYCL_KERNEL_MEMORY(DNNL_ARG_DIFF_SRC);

        auto nelems_A = memory_desc_wrapper(pd()->diff_src_md(0)).nelems();

        resampling_kernel_bwd_vec_t resampling_bwd_kernel(
                pd()->conf_, dst_mem_arg, src_mem_arg);

        resampling_kernel_bwd_vec1_t resampling_bwd_kernel1(
                pd()->conf_, dst_mem_arg, src_mem_arg);
        const int block_size = pd()->conf_.block_size;
        const int wg_size = pd()->conf_.wg_size;

        int work_per_wg = wg_size * block_size;
        int n_wgs = (nelems_A + work_per_wg - 1) / work_per_wg;
        int n_thr = n_wgs * wg_size;
        if (pd()->conf_.alg == alg_kind::resampling_nearest) {
            cgh.parallel_for(
                    ::sycl::nd_range<1>(n_thr, wg_size), resampling_bwd_kernel);
        } else {
            cgh.parallel_for(::sycl::nd_range<1>(n_thr, wg_size),
                    resampling_bwd_kernel1);
        }
    });
}

} // namespace sycl
} // namespace generic
} // namespace gpu
} // namespace impl
} // namespace dnnl