#include "ccv.h"

#include "ccv_internal.h"

#include "nnc/ccv_nnc.h"

#include "nnc/ccv_nnc_easy.h"

#include "nnc/ccv_nnc_internal.h"

#ifdef USE_OPENMP

#include <omp.h>

#endif

#ifdef USE_DISPATCH

#include <dispatch/dispatch.h>

#endif

static int _ccv_nnc_conv_transpose_forw(const ccv_nnc_cmd_t cmd, const ccv_nnc_hint_t hint, const int flags, ccv_nnc_tensor_t* const* const inputs, const int input_size, ccv_nnc_tensor_t* const* const outputs, const int output_size, ccv_nnc_stream_context_t* const stream_context)

{

  assert(input_size >= 2);

  const ccv_nnc_tensor_view_t* a = (ccv_nnc_tensor_view_t*)inputs[0];

  const ccv_nnc_tensor_t* w = inputs[1];

  assert(CCV_IS_TENSOR_CONTIGUOUS(w));

  const ccv_nnc_tensor_t* bias = input_size > 2 ? inputs[2] : 
00
;

  assert(!bias || !CCV_IS_TENSOR_VIEW(bias));

  assert(output_size == 1);

  ccv_nnc_tensor_view_t* b = (ccv_nnc_tensor_view_t*)outputs[0];

  const int a_nd = ccv_nnc_tensor_nd(a->info.dim);

  assert(a_nd == CCV_NNC_MAX_DIM + 1 || a_nd == CCV_NNC_MAX_DIM + 2);

  const int* adim = (a_nd == CCV_NNC_MAX_DIM + 1) ? 
a->info.dim6
 : 
a->info.dim + 13
;

  const int b_nd = ccv_nnc_tensor_nd(b->info.dim);

  assert(b_nd == CCV_NNC_MAX_DIM + 1 || b_nd == CCV_NNC_MAX_DIM + 2);

  const int* bdim = (b_nd == CCV_NNC_MAX_DIM + 1) ? 
b->info.dim6
 : 
b->info.dim + 13
;

  const int groups = cmd.info.convolution_transpose.groups;

  assert(cmd.info.convolution_transpose.count % groups == 0);

  const int group_size = cmd.info.convolution_transpose.count / groups;

  // Make sure the weights output dimension matches the network convolution kernels

  int astride[CCV_NNC_MAX_DIM_ALLOC];

  ccv_nnc_tensor_view_get_stride(a, astride);

  int bstride[CCV_NNC_MAX_DIM_ALLOC];

  ccv_nnc_tensor_view_get_stride(b, bstride);

  assert(!bias || bias->info.dim[0] == cmd.info.convolution_transpose.count);

  const int batch_size = (a_nd == CCV_NNC_MAX_DIM + 2) ? 
a->info.dim[0]3
 : 
16
;

  const int dilation[CCV_NNC_MAX_DIM] = {

    ccv_max(cmd.info.convolution_transpose.dilation[0], 1),

    ccv_max(cmd.info.convolution_transpose.dilation[1], 1)

  };

  if (a->info.format == CCV_TENSOR_FORMAT_NHWC)

  {

    // Make sure the weights dimension matches the network dimension

    assert(w->info.dim[1] == cmd.info.size.dim[0]);

    assert(w->info.dim[2] == cmd.info.size.dim[1]);

    assert(w->info.dim[CCV_NNC_MAX_DIM + 1] * groups == cmd.info.convolution_transpose.count);

    const int wdim[CCV_NNC_MAX_DIM] = {

      (w->info.dim[1] - 1) * dilation[0] + 1,

      (w->info.dim[2] - 1) * dilation[1] + 1

    };

    assert(w->info.dim[0] == adim[CCV_NNC_MAX_DIM]);

    assert(b->info.format == CCV_TENSOR_FORMAT_NHWC);

    const int channel_size = w->info.dim[CCV_NNC_MAX_DIM + 1];

    const int input_channel_size = w->info.dim[0] / groups;

    const int hwc = w->info.dim[1] * w->info.dim[2] * channel_size;

    assert(bdim[CCV_NNC_MAX_DIM] == cmd.info.convolution_transpose.count);

    
parallel_for5
(idx, cmd.info.convolution_transpose.count * batch_size) {

      int c;

      const int bidx = idx / cmd.info.convolution_transpose.count;

      const int k = idx % cmd.info.convolution_transpose.count;

      const int gidx = k / group_size;

      float* ap = a->data.f32 + bidx * astride[0];

      float* bp = b->data.f32 + bidx * bstride[0] + k;

      // kernel weight for one dim.

      float* wp = w->data.f32 + (k % group_size);

      float biasval = bias ? bias->data.f32[k] : 
00
;

      // This block will be cause in each for-loop, therefore, you can use it to generate some temporary variables.

      int i[CCV_NNC_MAX_DIM];

      int n[CCV_NNC_MAX_DIM];

      int d[CCV_NNC_MAX_DIM];

      int m[CCV_NNC_MAX_DIM];

      int j[CCV_NNC_MAX_DIM];

      for (i[0] = 0; i[0] < bdim[0]; 
i[0]++86.0k
)

      {

        for (i[1] = 0; i[1] < bdim[1]; 
i[1]++19.2M
)

          bp[i[1] * bstride[CCV_NNC_MAX_DIM]] = biasval;

        bp += bstride[CCV_NNC_MAX_DIM - 1];

      }

      bp = b->data.f32 + bidx * bstride[0] + k;

      for (i[0] = 0; i[0] < adim[0]; 
i[0]++43.0k
)

      {

        SET_BORDER_OFFSET_SIZE_FOR(0, i, hint, wdim, bdim, n, m);

        m[0] = (m[0] + n[0] - 1) / dilation[0] + 1;

        const int n0 = (n[0] + dilation[0] - 1) / dilation[0];

        d[0] = n0 * dilation[0] - n[0];

        n[0] = n0;

        m[0] = m[0] - n[0];

        float* wpu = wp + n[0] * w->info.dim[CCV_NNC_MAX_DIM] * channel_size;

        for (i[1] = 0; i[1] < adim[1]; 
i[1]++4.81M
)

        {

          SET_BORDER_OFFSET_SIZE_FOR(1, i, hint, wdim, bdim, n, m);

          m[1] = (m[1] + n[1] - 1) / dilation[1] + 1;

          const int n1 = (n[1] + dilation[1] - 1) / dilation[1];

          d[1] = n1 * dilation[1] - n[1];

          n[1] = n1;

          m[1] = m[1] - n[1];

          float* wpz = wpu + n[1] * channel_size;

          const float* const apz = ap + i[1] * astride[CCV_NNC_MAX_DIM];

          float* bpz = bp + d[0] * bstride[CCV_NNC_MAX_DIM - 1] + (ccv_max(i[1] * hint.stride.dim[1] - hint.border.begin[1], 0) + d[1]) * bstride[CCV_NNC_MAX_DIM];

          for (j[0] = 0; j[0] < m[0]; 
j[0]++33.4M
)

          {

            for (j[1] = 0; j[1] < m[1]; 
j[1]++232M
)

            {

              float p = bpz[j[1] * dilation[1] * bstride[CCV_NNC_MAX_DIM]];

              for (c = 0; c < input_channel_size; 
c++22.3G
)

                 p += wpz[j[1] * channel_size + (c + gidx * input_channel_size) * hwc] * apz[c + gidx * input_channel_size];

              bpz[j[1] * dilation[1] * bstride[CCV_NNC_MAX_DIM]] = p;

            }

            wpz += w->info.dim[CCV_NNC_MAX_DIM] * channel_size;

            bpz += bstride[CCV_NNC_MAX_DIM - 1] * dilation[0];

          }

        }

        ap += astride[CCV_NNC_MAX_DIM - 1];

        bp += bstride[CCV_NNC_MAX_DIM - 1] * (ccv_max((i[0] + 1) * hint.stride.dim[0] - hint.border.begin[0], 0) - ccv_max(i[0] * hint.stride.dim[0] - hint.border.begin[0], 0));

      }

    } parallel_endfor

  } else 
if (4
a->info.format == CCV_TENSOR_FORMAT_NCHW4
) {

    // Make sure the weights dimension matches the network dimension

    assert(w->info.dim[1] * groups == cmd.info.convolution_transpose.count);

    assert(w->info.dim[2] == cmd.info.size.dim[0]);

    assert(w->info.dim[3] == cmd.info.size.dim[1]);

    const int wdim[CCV_NNC_MAX_DIM] = {

      (w->info.dim[2] - 1) * dilation[0] + 1,

      (w->info.dim[3] - 1) * dilation[1] + 1

    };

    assert(w->info.dim[0] == adim[0]);

    assert(b->info.format == CCV_TENSOR_FORMAT_NCHW);

    const int channel_size = w->info.dim[1];

    const int input_channel_size = w->info.dim[0] / groups;

    const int hw = w->info.dim[2] * w->info.dim[3];

    const int chw = channel_size * hw;

    assert(bdim[0] == cmd.info.convolution_transpose.count);

    
parallel_for4
(idx, cmd.info.convolution_transpose.count * batch_size) {

      int c;

      const int bidx = idx / cmd.info.convolution_transpose.count;

      const int k = idx % cmd.info.convolution_transpose.count;

      const int gidx = k / group_size;

      float* ap = a->data.f32 + bidx * astride[0];

      float* bp = b->data.f32 + bidx * bstride[0] + k * bstride[1];

      // kernel weight for one dim.

      float* wp = w->data.f32 + (k % group_size) * hw;

      float biasval = bias ? bias->data.f32[k] : 
00
;

      // This block will be cause in each for-loop, therefore, you can use it to generate some temporary variables.

      int i[CCV_NNC_MAX_DIM];

      int n[CCV_NNC_MAX_DIM];

      int d[CCV_NNC_MAX_DIM];

      int m[CCV_NNC_MAX_DIM];

      int j[CCV_NNC_MAX_DIM];

      for (i[0] = 0; i[0] < bdim[1]; 
i[0]++43.0k
)

      {

        for (i[1] = 0; i[1] < bdim[2]; 
i[1]++9.63M
)

          bp[i[1]] = biasval;

        bp += bstride[CCV_NNC_MAX_DIM];

      }

      bp = b->data.f32 + bidx * bstride[0] + k * bstride[1];

      for (i[0] = 0; i[0] < adim[1]; 
i[0]++21.5k
)

      {

        SET_BORDER_OFFSET_SIZE_FOR(0, i, hint, wdim, bdim + 1, n, m);

        m[0] = (m[0] + n[0] - 1) / dilation[0] + 1;

        const int n0 = (n[0] + dilation[0] - 1) / dilation[0];

        d[0] = n0 * dilation[0] - n[0];

        n[0] = n0;

        m[0] = m[0] - n[0];

        float* wpu = wp + n[0] * w->info.dim[CCV_NNC_MAX_DIM + 1];

        for (i[1] = 0; i[1] < adim[2]; 
i[1]++2.40M
)

        {

          SET_BORDER_OFFSET_SIZE_FOR(1, i, hint, wdim, bdim + 1, n, m);

          m[1] = (m[1] + n[1] - 1) / dilation[1] + 1;

          const int n1 = (n[1] + dilation[1] - 1) / dilation[1];

          d[1] = n1 * dilation[1] - n[1];

          n[1] = n1;

          m[1] = m[1] - n[1];

          float* wpz = wpu + n[1];

          const float* apz = ap + i[1];

          float* bpz = bp + d[0] * bstride[CCV_NNC_MAX_DIM] + (ccv_max(i[1] * hint.stride.dim[1] - hint.border.begin[1], 0) + d[1]) * bstride[CCV_NNC_MAX_DIM + 1];

          for (j[0] = 0; j[0] < m[0]; 
j[0]++16.7M
)

          {

            for (j[1] = 0; j[1] < m[1]; 
j[1]++116M
)

            {

              float p = bpz[j[1] * dilation[1] * bstride[CCV_NNC_MAX_DIM + 1]];

              for (c = 0; c < input_channel_size; 
c++11.1G
)

                 p += wpz[j[1] + (c + gidx * input_channel_size) * chw] * apz[(c + gidx * input_channel_size) * astride[1]];

              bpz[j[1] * dilation[1] * bstride[CCV_NNC_MAX_DIM + 1]] = p;

            }

            wpz += w->info.dim[CCV_NNC_MAX_DIM + 1];

            bpz += bstride[CCV_NNC_MAX_DIM] * dilation[0];

          }

        }

        ap += astride[CCV_NNC_MAX_DIM];

        bp += bstride[CCV_NNC_MAX_DIM] * (ccv_max((i[0] + 1) * hint.stride.dim[0] - hint.border.begin[0], 0) - ccv_max(i[0] * hint.stride.dim[0] - hint.border.begin[0], 0));

      }

    } parallel_endfor

  }

  return CCV_NNC_EXEC_SUCCESS;

}

static int _ccv_nnc_conv_transpose_back(const ccv_nnc_cmd_t cmd, const ccv_nnc_hint_t hint, const int flags, ccv_nnc_tensor_t* const* const inputs, const int input_size, ccv_nnc_tensor_t* const* const outputs, const int output_size, ccv_nnc_stream_context_t* const stream_context)

{

  return CCV_NNC_EXEC_INVALID;

}

REGISTER_COMMAND_BACKEND(CCV_NNC_CONVOLUTION_TRANSPOSE_FORWARD, CCV_NNC_BACKEND_CPU_REF)(ccv_nnc_cmd_backend_registry_t* const registry)

{

  registry->tensor_formats = CCV_TENSOR_FORMAT_NHWC | CCV_TENSOR_FORMAT_NCHW;

  registry->tensor_datatypes = CCV_32F;

  registry->tensor_memory = CCV_TENSOR_CPU_MEMORY;

  registry->algorithms = 1;

  registry->exec = _ccv_nnc_conv_transpose_forw;

}

REGISTER_COMMAND_BACKEND(CCV_NNC_CONVOLUTION_TRANSPOSE_BACKWARD, CCV_NNC_BACKEND_CPU_REF)(ccv_nnc_cmd_backend_registry_t* const registry)

{

  registry->tensor_formats = CCV_TENSOR_FORMAT_NHWC;

  registry->tensor_datatypes = CCV_32F;

  registry->tensor_memory = CCV_TENSOR_CPU_MEMORY;

  registry->algorithms = 1;

  registry->exec = _ccv_nnc_conv_transpose_back;

}