/******************************************************************************/
/*!
 * \file timlCNNConvInitialize.c
 */
/* Copyright (C) 2015 Texas Instruments Incorporated - http://www.ti.com/
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions
 * are met:
 *
 *    Redistributions of source code must retain the above copyright
 *    notice, this list of conditions and the following disclaimer.
 *
 *    Redistributions in binary form must reproduce the above copyright
 *    notice, this list of conditions and the following disclaimer in the
 *    documentation and/or other materials provided with the
 *    distribution.
 *
 *    Neither the name of Texas Instruments Incorporated nor the names of
 *    its contributors may be used to endorse or promote products derived
 *    from this software without specific prior written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 *
 ******************************************************************************/


/*******************************************************************************
 *
 * INCLUDES
 *
 ******************************************************************************/

#include "../api/timl.h"


/******************************************************************************/
/*!
 * \ingroup       cnn
 * \brief         Initialize the conv layer
 * \param[in]     layer Conv layer
 * \return        Error code
 */
/******************************************************************************/

int timlCNNConvInitialize(timlCNNLayer *layer)
{
   timlConvNeuralNetwork *cnn;
   timlCNNLayer          *prev;
   int                   dim;
   int                   deviceId;
   int                   threadId;
   int                   err;
   char *offset;
   // init
   err      = 0;
   cnn      = layer->cnn;
   prev     = layer->prev;
   dim      = 0;
   deviceId = layer->cnn->deviceId;
   threadId = layer->cnn->threadId;


   // params memory common to level 1, 2, 3
   // allocate biasMultiplier
   if (timlUtilMallocAcc((void**) &(layer->convParams.biasMultiplier), sizeof(float)*layer->row*layer->col)) {
      return ERROR_CNN_LAYER_ALLOCATION;
   }
   timlUtilVectorResetFloat(layer->convParams.biasMultiplier, layer->row*layer->col, 1.0, deviceId, threadId);

   if (layer->convParams.shared == false) {
       // allocate prevFeatureMapReshapeIndex, size = (kernelRow*kernelCol) * (row*col)
      if (timlUtilMallocAcc((void**) &(layer->convParams.prevFeatureMapReshapeIndex), sizeof(int)*layer->convParams.kernelRow*layer->convParams.kernelCol*layer->row*layer->col)) {
         return ERROR_CNN_LAYER_ALLOCATION;
      }
      timlUtilConv2ImageReshapeIndex(layer->convParams.prevFeatureMapReshapeIndex, prev->row, prev->col, layer->convParams.kernelRow, layer->convParams.kernelCol, layer->convParams.padUp, layer->convParams.padDown, layer->convParams.padLeft, layer->convParams.padRight, layer->convParams.strideX, layer->convParams.strideY, layer->convParams.type, deviceId, threadId);

      // allocate kernel
      dim = layer->convParams.inputFeatureMapChannel*layer->convParams.outputFeatureMapChannel*layer->convParams.kernelRow*layer->convParams.kernelCol;
      if (timlUtilMallocAcc((void**) &(layer->convParams.kernel), sizeof(float)*dim)) {
         return ERROR_CNN_LAYER_ALLOCATION;
      }

      // allocate bias
      if (timlUtilMallocAcc((void**) &(layer->convParams.bias), sizeof(float)*layer->channel)) {
         return ERROR_CNN_LAYER_ALLOCATION;
      }
   }

   //  level 1
   if (layer->allocatorLevel == Util_AllocatorLevel1) {
      // allocate feature map
      if (timlUtilMallocAcc((void**) &(layer->featureMap), sizeof(float)*layer->maxBatchSize*layer->row*layer->col*layer->channel)) {
         return ERROR_CNN_LAYER_ALLOCATION;
      }

      // allocate feature map delta
      if (timlUtilMallocAcc((void**) &(layer->delta), sizeof(float)*layer->maxBatchSize*layer->row*layer->col*layer->channel)) {
         return ERROR_CNN_LAYER_ALLOCATION;
      }

      // allocate prev feature map reshape, size = (prev->channel*kernelRow*kernelCol) * (row*col)
      if (timlUtilMallocAcc((void**) &(layer->convParams.prevFeatureMapReshape), sizeof(float)*layer->convParams.inputFeatureMapChannel*layer->convParams.kernelRow*layer->convParams.kernelCol*layer->row*layer->col)) {
         return ERROR_CNN_LAYER_ALLOCATION;
      }

      // if not shared
      if (layer->convParams.shared == false) {

         // allocate kernelInc
         if (timlUtilMallocAcc((void**) &(layer->convParams.kernelInc), sizeof(float)*dim)) {
            return ERROR_CNN_LAYER_ALLOCATION;
         }

         // allocate kernelGradAccum
         if (timlUtilMallocAcc((void**) &(layer->convParams.kernelGradAccum), sizeof(float)*dim)) {
            return ERROR_CNN_LAYER_ALLOCATION;
         }

         // allocate biasGradAccum
         if (timlUtilMallocAcc((void**) &(layer->convParams.biasGradAccum), sizeof(float)*layer->channel)) {
            return ERROR_CNN_LAYER_ALLOCATION;
         }

         // allocate biasInc
         if (timlUtilMallocAcc((void**) &(layer->convParams.biasInc), sizeof(float)*layer->channel)) {
            return ERROR_CNN_LAYER_ALLOCATION;
         }
      }
   }


   // level 2 only
   if (layer->allocatorLevel == Util_AllocatorLevel2) {
      // allocate feature map
       if (timlUtilMallocAcc((void**) &(layer->featureMap), sizeof(float)*layer->maxBatchSize*layer->row*layer->col*layer->channel)) {
          return ERROR_CNN_LAYER_ALLOCATION;
       }

       // allocate prev feature map reshape, size = (prev->channel*kernelRow*kernelCol) * (row*col)
       if (timlUtilMallocAcc((void**) &(layer->convParams.prevFeatureMapReshape), sizeof(float)*layer->convParams.inputFeatureMapChannel*layer->convParams.kernelRow*layer->convParams.kernelCol*layer->row*layer->col)) {
          return ERROR_CNN_LAYER_ALLOCATION;
       }
   }

   // level 3 only
   if (layer->allocatorLevel == Util_AllocatorLevel3) {

      if (layer->prev->memPoolPos == Util_MemPoolTop) {
         offset = cnn->memPool + cnn->memPoolSize - layer->forwardMemory;
         layer->memPoolPos = Util_MemPoolBottom;
         layer->convParams.prevFeatureMapReshape = offset;
         offset += sizeof(float)*layer->convParams.inputFeatureMapChannel*layer->convParams.kernelRow*layer->convParams.kernelCol*layer->row*layer->col;
         layer->featureMap = offset;
      }
      else { // allocate at the top
         offset = cnn->memPool;
         layer->memPoolPos = Util_MemPoolTop;
         layer->featureMap = offset;
         offset += sizeof(float)*layer->maxBatchSize*layer->row*layer->col*layer->channel;
         layer->convParams.prevFeatureMapReshape = offset;
      }
   }

   return err;
}