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

/*******************************************************************************
* Copyright 2022 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/intel/jit/ir/slm_reduce_builder.hpp"

#include <algorithm>

#include "gpu/intel/jit/ir/reduce.hpp"
#include "gpu/intel/jit/ir/send_builder.hpp"
#include "gpu/intel/jit/utils/utils.hpp"

namespace dnnl {
namespace impl {
namespace gpu {
namespace intel {
namespace jit {

slm_reduce_builder_t::slm_reduce_builder_t(ir_context_t &ir_ctx,
        const grid_info_t &tg_grid, const expr_t &reg_buf,
        const layout_t &reg_layout, const tile_coord_t &thr_tile_coord,
        dim_idx_t dim)
    : ir_ctx_(&ir_ctx)
    , tg_grid_(tg_grid)
    , reg_buf_(reg_buf)
    , reg_layout_(reg_layout)
    , thr_tile_coord_(thr_tile_coord)
    , dim_(dim) {
    gpu_assert((dim_ != dim_idx::invalid) && (dim_ <= 2));
    gpu_assert(tg_grid_.dim(dim_) > 1);

    tmp_reg_buf_
            = ir_ctx.create_tmp_var(dsl::type_t::byte(dsl::type::attr_t::ptr));
    slm_buf_ = ir_ctx.create_tmp_var(
            dsl::type_t::byte(dsl::type::attr_t::ptr), "reduce_slm");
    tg_ndims_ = (dim_ != dim_idx_t(2)) ? dim_ + 1 : tg_grid_.ndims();

    build();
}

void slm_reduce_builder_t::build() {
    auto ndims = reg_layout_.ndims();

    // Create SLM layout to store all intermediate buffers from the thread
    // group.
    layout_t slm_layout(reg_layout_.type(), reg_layout_.blocks(),
            reg_layout_.offset(), ndims + tg_ndims_);
    for (int i = tg_ndims_ - 1; i >= 0; i--) {
        slm_layout = slm_layout.with_block({ndims + i, tg_grid_.dim(i)});
    }

    slm_buf_size_ = into<int>(size_bytes(slm_layout));

    // Write thread tile to SLM.
    tile_t write_tile;
    coord_t write_start;
    tile_t reg_tile = reg_layout_.tile();
    for (size_t i = 0; i < ndims; i++)
        write_tile[i] = reg_tile[i];
    for (int i = tg_ndims_ - 1; i >= 0; i--) {
        write_start[ndims + i] = tg_grid_.idx(i);
    }
    auto write = make_access_builder(*ir_ctx_,
            view_t(slm_layout.sub(write_tile, write_start)), slm_buf_, reg_buf_,
            send_op_t::store, send_address_t::slm);
    store_stmt_ = write.stmt();

    auto &write_layout = write.reg_layout();
    gpu_assert(write_layout.is_equal_normalized(reg_layout_))
            << "Incompatible layouts.";

    // Redistribute the layout to read/reduce all k-axis tiles from every
    // thread.
    grid_info_t full_grid = tg_grid_.sub_grid({dim_});
    grid_info_t split_grid;
    auto local_thr_tile_coord = split(reg_layout_, full_grid, &split_grid);
    reg_layout_ = reg_layout_.sub(local_thr_tile_coord.tile);

    if (split_grid.elems() != full_grid.elems()) {
        for (dim_idx_t i = 0; i < full_grid.ndims(); i++) {
            if (split_grid.dim(i) == full_grid.dim(i)) continue;
            auto cond = full_grid.idx(i) < split_grid.dim(i);
            if (reduce_cond_.is_empty())
                reduce_cond_ = std::move(cond);
            else
                reduce_cond_ &= cond;
        }
    }

    tile_t read_tile;
    coord_t read_start;
    tile_t slm_tile = slm_layout.tile();
    for (size_t i = 0; i < ndims; i++) {
        read_tile[i] = local_thr_tile_coord.tile[i];
        read_start[i] = local_thr_tile_coord.coord[i];
        auto cond = read_start[i] < slm_tile[i];
        if (reduce_cond_.is_empty())
            reduce_cond_ = std::move(cond);
        else
            reduce_cond_ &= cond;
    }
    read_tile[ndims + dim_] = tg_grid_.dim(dim_);
    for (dim_idx_t i = 0; i < tg_ndims_; i++) {
        read_start[ndims + i] = (i == dim_) ? 0 : tg_grid_.idx(i);
    }
    auto read = make_access_builder(*ir_ctx_,
            view_t(slm_layout.sub(read_tile, read_start)), slm_buf_,
            tmp_reg_buf_, send_op_t::load, send_address_t::slm);

    load_stmt_ = load_stmt_.append(
            funcs::zero_out(reg_buf_, size_bytes(reg_layout_)));
    load_stmt_ = load_stmt_.append(read.stmt());

    tmp_reg_buf_size_ = std::max(tmp_reg_buf_size_, read.reg_buf_size());

    auto &read_layout = read.reg_layout();
    load_stmt_ = load_stmt_.append(create_reduce_stmt(read_layout, reg_layout_,
            tmp_reg_buf_, reg_buf_, tile_t(), reduction_mask()));

    allocs_.push_back(
            alloc_t::make(slm_buf_, slm_buf_size_, alloc_kind_t::slm));
    allocs_.push_back(
            alloc_t::make(tmp_reg_buf_, tmp_reg_buf_size_, alloc_kind_t::grf));

    if (reduce_cond_) load_stmt_ = if_t::make(reduce_cond_, load_stmt_);
    thr_tile_coord_ = thr_tile_coord_.sub(local_thr_tile_coord);
}

} // namespace jit
} // namespace intel
} // namespace gpu
} // namespace impl
} // namespace dnnl