onednn-src 0.1.13

/*******************************************************************************
* 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.
*******************************************************************************/

#ifndef GPU_INTEL_CONV_JIT_V2_BENCH_DATA_HPP
#define GPU_INTEL_CONV_JIT_V2_BENCH_DATA_HPP

#include "gpu/intel/conv/jit/v2/kernel_desc.hpp"

#include <vector>

namespace dnnl {
namespace impl {
namespace gpu {
namespace intel {
namespace conv {
namespace jit {
namespace v2 {

// Stores device times for primitive execution.
// Includes:
// - Total time, computed via profiling queries as:
//     the end of the last kernel - the start of the first kernel
// - Kernel times: individual kernel times
struct bench_time_t {
    uint64_t total = 0;
    std::vector<uint64_t> kernel_times;

    bench_time_t() = default;
    bench_time_t(uint64_t total) : total(total) {
        kernel_times.push_back(total);
    }
    template <typename IteratorT>
    bench_time_t(uint64_t total, IteratorT beg, IteratorT end) : total(total) {
        kernel_times = std::vector<uint64_t>(beg, end);
    }
    int nkernels() const { return (int)kernel_times.size(); }

    bench_time_t min(const bench_time_t &other) const {
        bench_time_t ret = *this;
        ret.total = std::min(ret.total, other.total);
        for (int i = 0; i < nkernels(); i++) {
            ret.kernel_times[i]
                    = std::min(ret.kernel_times[i], other.kernel_times[i]);
        }
        return ret;
    }
};

class bench_data_t {
public:
    int id = -1;
    kernel_desc_t kernel_desc;
    std::vector<problem_t> prbs;
    std::vector<bench_time_t> times;

    bench_data_t() = default;
    explicit bench_data_t(int id, const kernel_desc_t &kernel_desc)
        : id(id), kernel_desc(kernel_desc) {}

    int size() const { return (int)prbs.size(); }
    explicit operator bool() const { return size() > 0; }

    void add(const problem_t &prb, const bench_time_t &time) {
        prbs.push_back(prb);
        times.push_back(time);
    }

    std::string str() const;
};

class bench_data_set_t {
public:
    void add(const bench_data_t &bd) { vec_.push_back(bd); }
    int size() const { return (int)vec_.size(); }
    std::vector<bench_data_t>::const_iterator begin() const {
        return vec_.begin();
    }
    std::vector<bench_data_t>::const_iterator end() const { return vec_.end(); }
    std::vector<int> find_best_ids(int nbest) const;
    std::vector<bench_data_t> find_best(int nbest) const;

private:
    std::vector<int> find_best_idxs(int nbest) const;

    std::vector<bench_data_t> vec_;
};

} // namespace v2
} // namespace jit
} // namespace conv
} // namespace intel
} // namespace gpu
} // namespace impl
} // namespace dnnl

#endif