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
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282

/*******************************************************************************
* Copyright 2020 Intel Corporation
* Copyright 2020-2024 FUJITSU LIMITED
* Copyright 2025 Arm Ltd. and affiliates
*
* 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 CPU_AARCH64_JIT_SVE_CONVOLUTION_HPP
#define CPU_AARCH64_JIT_SVE_CONVOLUTION_HPP

#include "common/c_types_map.hpp"
#include "common/dnnl_thread.hpp"
#include "common/memory_tracking.hpp"
#include "common/primitive.hpp"
#include "common/utils.hpp"

#include "cpu/aarch64/cpu_barrier.hpp"
#include "cpu/aarch64/cpu_reducer.hpp"
#include "cpu/cpu_convolution_pd.hpp"

#include "cpu/aarch64/jit_sve_conv_kernel.hpp"

namespace dnnl {
namespace impl {
namespace cpu {
namespace aarch64 {

template <impl::data_type_t src_type, impl::data_type_t wei_type = src_type,
        impl::data_type_t dst_type = src_type, cpu_isa_t isa = isa_undef>
struct jit_sve_convolution_fwd_t : public primitive_t {
    struct pd_t : public cpu_convolution_fwd_pd_t {
        using cpu_convolution_fwd_pd_t::cpu_convolution_fwd_pd_t;

        DECLARE_COMMON_PD_T(JIT_IMPL_NAME_HELPER("jit:", isa, ""),
                jit_sve_convolution_fwd_t);

        status_t init(engine_t *engine) {
#if defined(DNNL_AARCH64_USE_ACL)
            if (attr()->fpmath_.mode_ == fpmath_mode::bf16) {
                // prefer ACL to jit for fpmath_mode::bf16 if available
                // since it supports lower precision calculation
                return status::unimplemented;
            }
#endif

            bool ok = true && mayiuse(isa) && is_fwd()
                    && set_default_alg_kind(alg_kind::convolution_direct)
                    && expect_data_types(src_type, wei_type, dst_type, dst_type,
                            data_type::undef)
                    && attr()->has_default_values(
                            primitive_attr_t::skip_mask_t::post_ops, dst_type)
                    && !has_zero_dim_memory();
            if (!ok) return status::unimplemented;

            status_t status = jit_sve_conv_fwd_kernel_t<isa>::init_conf(jcp_,
                    *desc(), src_md_, weights_md_, dst_md_, bias_md_, *attr(),
                    dnnl_get_max_threads());
            if (status != status::success) return status;

            auto scratchpad = scratchpad_registry().registrar();
            jit_sve_conv_fwd_kernel_t<isa>::init_scratchpad(scratchpad, jcp_);

            return status;
        }

        jit_conv_conf_t jcp_ = utils::zero<decltype(jcp_)>();
    };

    jit_sve_convolution_fwd_t(const pd_t *apd) : primitive_t(apd) {}

    using src_data_t = typename prec_traits_t<src_type>::type;
    using wei_data_t = typename prec_traits_t<wei_type>::type;
    using dst_data_t = typename prec_traits_t<dst_type>::type;

    status_t init(engine_t *engine) override {
        CHECK(safe_ptr_assign(kernel_,
                new jit_sve_conv_fwd_kernel_t<isa>(pd()->jcp_, *pd()->attr())));
        return kernel_->create_kernel();
    }

    status_t execute(const exec_ctx_t &ctx) const override {
        if (pd()->ndims() == 3)
            execute_forward_1d(ctx);
        else if (pd()->ndims() == 4)
            execute_forward_2d(ctx);
        else if (pd()->ndims() == 5)
            execute_forward_3d(ctx);
        else
            assert(false);

        if (pd()->wants_zero_pad_dst()) ctx.zero_pad_output(DNNL_ARG_DST);

        return status::success;
    }

private:
    void prepare_padded_bias(const dst_data_t *&bias,
            const memory_tracking::grantor_t &scratchpad) const;
    void execute_forward_1d(const exec_ctx_t &ctx) const;
    void execute_forward_2d(const exec_ctx_t &ctx) const;
    void execute_forward_3d(const exec_ctx_t &ctx) const;
    const pd_t *pd() const { return (const pd_t *)primitive_t::pd().get(); }

    std::unique_ptr<jit_sve_conv_fwd_kernel_t<isa>> kernel_;
};

template <impl::data_type_t diff_dst_type,
        impl::data_type_t wei_type = diff_dst_type,
        impl::data_type_t diff_src_type = diff_dst_type,
        cpu_isa_t isa = isa_undef>
struct jit_sve_convolution_bwd_data_t : public primitive_t {
    struct pd_t : public cpu_convolution_bwd_data_pd_t {
        using cpu_convolution_bwd_data_pd_t::cpu_convolution_bwd_data_pd_t;

        DECLARE_COMMON_PD_T(JIT_IMPL_NAME_HELPER("jit:", isa, ""),
                jit_sve_convolution_bwd_data_t);

        status_t init(engine_t *engine) {
            bool ok = true && desc()->prop_kind == prop_kind::backward_data
                    && set_default_alg_kind(alg_kind::convolution_direct)
                    && expect_data_types(diff_src_type, wei_type,
                            data_type::undef, diff_dst_type, data_type::undef)
                    && attr()->has_default_values() && !has_zero_dim_memory();
            if (!ok) return status::unimplemented;

            status_t status
                    = jit_sve_conv_bwd_data_kernel_f32_t<isa>::init_conf(jcp_,
                            *desc(), diff_src_md_, weights_md_, diff_dst_md_,
                            dnnl_get_max_threads());
            if (status != status::success) return status;

            auto scratchpad = scratchpad_registry().registrar();
            jit_sve_conv_bwd_data_kernel_f32_t<isa>::init_scratchpad(
                    scratchpad, jcp_);

            return status::success;
        }

        jit_conv_conf_t jcp_ = utils::zero<decltype(jcp_)>();
    };

    jit_sve_convolution_bwd_data_t(const pd_t *apd) : primitive_t(apd) {}

    using diff_dst_data_t = typename prec_traits_t<diff_dst_type>::type;
    using wei_data_t = typename prec_traits_t<wei_type>::type;
    using diff_src_data_t = typename prec_traits_t<diff_src_type>::type;

    status_t init(engine_t *engine) override {
        CHECK(safe_ptr_assign(kernel_,
                new jit_sve_conv_bwd_data_kernel_f32_t<isa>(pd()->jcp_)));
        return kernel_->create_kernel();
    }

    status_t execute(const exec_ctx_t &ctx) const override {
        if (pd()->ndims() == 3)
            execute_backward_data_1d(ctx);
        else if (pd()->ndims() == 4)
            execute_backward_data_2d(ctx);
        else if (pd()->ndims() == 5)
            execute_backward_data_3d(ctx);
        else
            assert(false);
        return status::success;
    }

private:
    void execute_backward_data_1d(const exec_ctx_t &ctx) const;
    void execute_backward_data_2d(const exec_ctx_t &ctx) const;
    void execute_backward_data_3d(const exec_ctx_t &ctx) const;
    const pd_t *pd() const { return (const pd_t *)primitive_t::pd().get(); }

    std::unique_ptr<jit_sve_conv_bwd_data_kernel_f32_t<isa>> kernel_;
};

template <impl::data_type_t src_type,
        impl::data_type_t diff_dst_type = src_type,
        impl::data_type_t diff_weights_type = src_type,
        cpu_isa_t isa = isa_undef>
struct jit_sve_convolution_bwd_weights_t : public primitive_t {
    struct pd_t : public cpu_convolution_bwd_weights_pd_t {
        using cpu_convolution_bwd_weights_pd_t::
                cpu_convolution_bwd_weights_pd_t;

        DECLARE_COMMON_PD_T(JIT_IMPL_NAME_HELPER("jit:", isa, ""),
                jit_sve_convolution_bwd_weights_t);

        status_t init(engine_t *engine) {
            bool ok = true && desc()->prop_kind == prop_kind::backward_weights
                    && set_default_alg_kind(alg_kind::convolution_direct)
                    && expect_data_types(src_type, diff_weights_type,
                            diff_weights_type, diff_dst_type, data_type::undef)
                    && attr()->has_default_values() && !has_zero_dim_memory();
            if (!ok) return status::unimplemented;

            status_t status
                    = jit_sve_conv_bwd_weights_kernel_f32_t<isa>::init_conf(
                            jcp_, *desc(), src_md_, diff_weights_md_,
                            diff_bias_md_, diff_dst_md_,
                            dnnl_get_max_threads());
            if (status != status::success) return status;

            init_balancers();

            auto scratchpad = scratchpad_registry().registrar();
            jit_sve_conv_bwd_weights_kernel_f32_t<isa>::init_scratchpad(
                    scratchpad, jcp_);

            auto reducer_bia_scratchpad = memory_tracking::registrar_t(
                    scratchpad, memory_tracking::names::prefix_reducer_bia);
            reducer_bia_conf_.init_scratchpad(reducer_bia_scratchpad);

            return status;
        }

        jit_conv_conf_t jcp_ = utils::zero<decltype(jcp_)>();
        typename cpu_reducer_t<diff_weights_type, isa>::conf_t
                reducer_bia_conf_;

    private:
        void init_balancers() {
            const size_t max_buffer_size = jcp_.nthr * 3 * 5 * 5 * 16 * 16;
            if (with_bias()) {
                reducer_bia_conf_.init(reduce_balancer_t(jcp_.nthr,
                        jcp_.oc_block, jcp_.ngroups * jcp_.nb_oc, jcp_.mb,
                        max_buffer_size, true));
            }
        }
    };

    jit_sve_convolution_bwd_weights_t(const pd_t *apd) : primitive_t(apd) {}

    using src_data_t = typename prec_traits_t<src_type>::type;
    using diff_dst_data_t = typename prec_traits_t<diff_dst_type>::type;
    using diff_weights_data_t = typename prec_traits_t<diff_weights_type>::type;

    status_t init(engine_t *engine) override;

    status_t execute(const exec_ctx_t &ctx) const override {
        execute_backward_weights(ctx);
        return status::success;
    }

private:
    void execute_backward_weights(const exec_ctx_t &ctx) const;
    void prepare_scratchpad_data(const exec_ctx_t &ctx) const;
    struct thread_info_t;
    void compute_diff_weights(const thread_info_t *) const;
    void compute_diff_weights_2d(const thread_info_t *) const;
    void compute_diff_weights_3d(const thread_info_t *) const;
    void reduce_diff_weights(const thread_info_t *) const;
    void reduce_diff_weights_3d(const thread_info_t *) const;
    void compute_diff_bias(const thread_info_t *) const;
    void reduce_diff_bias(const thread_info_t *) const;

    const pd_t *pd() const { return (const pd_t *)primitive_t::pd().get(); }

    int nthr_, nthr_mb_, nthr_g_, nthr_oc_b_, nthr_ic_b_;

    jit_sve_conv_bwd_weights_kernel_f32_t<isa> *kernel_;
    cpu_accumulator_1d_t<diff_weights_type, isa> *acc_ker_;
    cpu_reducer_t<diff_weights_type, isa> *reducer_bias_;
};

} // namespace aarch64
} // namespace cpu
} // namespace impl
} // namespace dnnl

#endif

// vim: et ts=4 sw=4 cindent cino+=l0,\:4,N-s