oneAPI Deep Neural Network Library (oneDNN)
Performance library for Deep Learning
1.96.0
Reduction Primitive Example

This C++ API example demonstrates how to create and execute a Reduction primitive.

/*******************************************************************************
* Copyright 2020 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 <cmath>
#include "example_utils.hpp"
using namespace dnnl;
void reduction_example(dnnl::engine::kind engine_kind) {
// Create execution dnnl::engine.
dnnl::engine engine(engine_kind, 0);
// Create dnnl::stream.
dnnl::stream engine_stream(engine);
// Tensor dimensions.
const memory::dim N = 3, // batch size
IC = 3, // channels
IH = 227, // tensor height
IW = 227; // tensor width
// Source (src) and destination (dst) tensors dimensions.
memory::dims src_dims = {N, IC, IH, IW};
memory::dims dst_dims = {1, IC, 1, 1};
// Allocate buffers.
std::vector<float> src_data(product(src_dims));
std::vector<float> dst_data(product(dst_dims));
// Initialize src tensor.
std::generate(src_data.begin(), src_data.end(), []() {
static int i = 0;
return std::cos(i++ / 10.f);
});
// Create src and dst memory descriptors and memory objects.
auto src_md = memory::desc(src_dims, dt::f32, tag::nchw);
auto dst_md = memory::desc(dst_dims, dt::f32, tag::nchw);
auto src_mem = memory(src_md, engine);
auto dst_mem = memory(dst_md, engine);
// Write data to memory object's handle.
write_to_dnnl_memory(src_data.data(), src_mem);
// Create operation descriptor.
auto reduction_d = reduction::desc(
algorithm::reduction_sum, src_md, dst_md, 0.f, 0.f);
// Create primitive descriptor.
auto reduction_pd = reduction::primitive_desc(reduction_d, engine);
// Create the primitive.
auto reduction_prim = reduction(reduction_pd);
// Primitive arguments.
std::unordered_map<int, memory> reduction_args;
reduction_args.insert({DNNL_ARG_SRC, src_mem});
reduction_args.insert({DNNL_ARG_DST, dst_mem});
// Primitive execution: Reduction (Sum).
reduction_prim.execute(engine_stream, reduction_args);
// Wait for the computation to finalize.
engine_stream.wait();
// Read data from memory object's handle.
read_from_dnnl_memory(dst_data.data(), dst_mem);
}
int main(int argc, char **argv) {
return handle_example_errors(reduction_example, engine::kind::cpu);
}
dnnl.hpp
C++ API.
dnnl::stream
An execution stream.
Definition: dnnl.hpp:975
dnnl::reduction::primitive_desc
Primitive descriptor for a reduction primitive.
Definition: dnnl.hpp:10129
dnnl::engine
An execution engine.
Definition: dnnl.hpp:859
dnnl::reduction
Reduction.
Definition: dnnl.hpp:10094
dnnl::reduction::desc
Descriptor for reduction.
Definition: dnnl.hpp:10096
dnnl::engine::kind
kind
Kinds of engines.
Definition: dnnl.hpp:864
DNNL_ARG_DST
#define DNNL_ARG_DST
A special mnemonic for destination argument for primitives that have a single destination.
Definition: dnnl_types.h:2019
dnnl::memory::data_type
data_type
Data type specification.
Definition: dnnl.hpp:1120
dnnl::memory::format_tag
format_tag
Memory format tag specification.
Definition: dnnl.hpp:1195
DNNL_ARG_SRC
#define DNNL_ARG_SRC
A special mnemonic for source argument for primitives that have a single source.
Definition: dnnl_types.h:1995
dnnl::memory::dim
dnnl_dim_t dim
Integer type for representing dimension sizes and indices.
Definition: dnnl.hpp:1102
dnnl::memory
Memory object.
Definition: dnnl.hpp:1098
dnnl::memory::dims
std::vector< dim > dims
Vector of dimensions.
Definition: dnnl.hpp:1105
dnnl::memory::desc
A memory descriptor.
Definition: dnnl.hpp:1718
dnnl
oneDNN namespace
Definition: dnnl.hpp:74
dnnl::engine::kind::cpu
@ cpu
CPU engine.