TheRock/rocm-libraries/projects/miopen/driver/mloGroupNormHost.hpp

/*******************************************************************************
 *
 * MIT License
 *
 * Copyright (c) 2023 Advanced Micro Devices, Inc.
 *
 * Permission is hereby granted, free of charge, to any person obtaining a copy
 * of this software and associated documentation files (the "Software"), to deal
 * in the Software without restriction, including without limitation the rights
 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
 * copies of the Software, and to permit persons to whom the Software is
 * furnished to do so, subject to the following conditions:
 *
 * The above copyright notice and this permission notice shall be included in all
 * copies or substantial portions of the Software.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
 * SOFTWARE.
 *
 *******************************************************************************/
#ifndef MLO_GROUPNORMHOST_H_
#define MLO_GROUPNORMHOST_H_

#include <miopen/tensor.hpp>

////////////////////////////////////////////////////////////
//
///////////////////////////////////////////////////////////
template <typename Tgpu, typename Tcheck>
int32_t mloGroupNormForwardRunHost(miopenTensorDescriptor_t inputDesc,
                                   Tgpu* input,
                                   Tgpu* weight,
                                   Tgpu* bias,
                                   Tcheck* outputhost,
                                   Tcheck* meanhost,
                                   Tcheck* rstdhost,
                                   uint64_t num_groups,
                                   float eps,
                                   miopenNormMode_t mode,
                                   bool use_multithread)
{
    auto dims = miopen::deref(inputDesc).GetLengths();

    size_t numel             = miopen::deref(inputDesc).GetElementSize();
    size_t numel_per_channel = numel / dims[0] / dims[1];
    size_t num_channels      = dims[1];

    size_t outer_size = dims[0] * num_groups;
    size_t inner_size = numel / outer_size;

    size_t min_grain = use_multithread ? 8 : outer_size;

    miopen::par_for(outer_size, min_grain, [&](size_t o) {
        Tcheck pmean = 0.0f;
        Tcheck pvar  = 0.0f;
        for(size_t i = 0; i < inner_size; i++)
        {
            Tcheck tmp = static_cast<Tcheck>(input[o * inner_size + i]);
            pmean += tmp;
            pvar += tmp * tmp;
        }

        pmean        = pmean / inner_size;
        pvar         = pvar / inner_size - pmean * pmean;
        Tcheck prstd = 1.0f / sqrt(pvar + eps);

        meanhost[o] = pmean;
        rstdhost[o] = prstd;

        for(size_t i = 0; i < inner_size; i++)
        {
            size_t idx     = o * inner_size + i;
            size_t c       = (idx / numel_per_channel) % num_channels;
            Tcheck pweight = mode ? static_cast<Tcheck>(weight[c]) : 1;
            Tcheck pbias   = mode ? static_cast<Tcheck>(bias[c]) : 0;

            outputhost[idx] = (static_cast<Tcheck>(input[idx]) - pmean) * prstd * pweight + pbias;
        }
    });

    return 0;
}

#endif