#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

// Shared methods.

#include "../_ccv_nnc_cpu_ref.h"

void _ccv_nnc_mul_forw_cpu_ref(const float p, ccv_nnc_tensor_view_t* const a, ccv_nnc_tensor_view_t* const b, ccv_nnc_tensor_view_t* const c)

{

  if (b == 0)

  {

    if (p == 1)

    {

      _ccv_nnc_tensor_transfer_cpu_ref_f32(a, c);

      return;

    } else if (p == 0) {

      ccv_nnc_tensor_zero(c);

      return;

    }

    // Assuming this is float 32.

    int dim[CCV_NNC_MAX_DIM_ALLOC];

    int astride[CCV_NNC_MAX_DIM_ALLOC];

    int cstride[CCV_NNC_MAX_DIM_ALLOC];

    assert(ccv_nnc_tensor_nd(a->info.dim) <= CCV_NNC_MAX_DIM + 2);

    assert(ccv_nnc_tensor_nd(c->info.dim) <= CCV_NNC_MAX_DIM + 2);

    ccv_nnc_tensor_view_get_dim(a, dim);

    assert(ccv_nnc_tensor_view_check_dim(c, dim));

    int x;

    if (!CCV_IS_TENSOR_VIEW(a) && !CCV_IS_TENSOR_VIEW(c))

    {

      // Super optimal case, just do one for-loop for sum.

      const int tensor_count = ccv_nnc_tensor_count(a->info);

      for (x = 0; x < tensor_count; 
x++8.39M
)

        c->data.f32[x] = p * a->data.f32[x];

      return;

    }

    assert
(CCV_NNC_MAX_DIM == 2)2
; // Need to change this logic for CCV_NNC_MAX_DIM == other number.

    ccv_nnc_tensor_view_get_stride(a, astride);

    ccv_nnc_tensor_view_get_stride(c, cstride);

    int i[CCV_NNC_MAX_DIM + 2];

    float* const ap = a->data.f32;

    float* const cp = c->data.f32;

    const int count = dim[2] * dim[3];

    if (astride[2] == dim[3] && cstride[2] == dim[3])

    {

      // Special casing if the ainc[3] is the same as dim[3]

      for (i[0] = 0; i[0] < dim[0]; 
i[0]++2
)

      {

        float* ap0 = ap + i[0] * astride[0];

        float* cp0 = cp + i[0] * cstride[0];

        for (i[1] = 0; i[1] < dim[1]; 
i[1]++2
)

        {

          for (x = 0; x < count; 
x++2
)

            cp0[x] = p * ap0[x];

          ap0 += astride[1];

          cp0 += cstride[1];

        }

      }

      return;

    }

    // Non-optimal case, need to do skip copy.

    for (i[0] = 0; i[0] < dim[0]; i[0]++)

    {

      float* const ap0 = ap + i[0] * astride[0];

      float* const cp0 = cp + i[0] * cstride[0];

      for (i[1] = 0; i[1] < dim[1]; i[1]++)

      {

        float* ap1 = ap0 + i[1] * astride[1];

        float* cp1 = cp0 + i[1] * cstride[1];

        for (i[2] = 0; i[2] < dim[2]; i[2]++)

        {

          for (x = 0; x < dim[3]; x++)

            cp1[x] = p * ap1[x];

          ap1 += astride[2];

          cp1 += cstride[2];

        }

      }

    }

    return;

  }

  int cdim[CCV_NNC_MAX_DIM_ALLOC];

  assert(ccv_nnc_tensor_nd(a->info.dim) <= CCV_NNC_MAX_DIM + 2);

  assert(ccv_nnc_tensor_nd(b->info.dim) <= CCV_NNC_MAX_DIM + 2);

  ccv_nnc_tensor_view_get_dim(a, cdim); // Fill in cdim first.

  ccv_nnc_tensor_view_get_broadcast_dim(b, cdim);

  assert(ccv_nnc_tensor_view_check_broadcast_dim(a, cdim));

  assert(ccv_nnc_tensor_view_check_broadcast_dim(b, cdim));

  const int a_check_dim = ccv_nnc_tensor_view_check_dim(a, cdim);

  const int b_check_dim = ccv_nnc_tensor_view_check_dim(b, cdim);

  if (p == 1 && 
a_check_dim10.4k
 && 
b_check_dim10.4k
)

  {

    _ccv_nnc_ewprod_forw_cpu_ref((ccv_nnc_tensor_view_t*[]){

      a, b

    }, 2, &c, 1);

    return;

  } else 
if (33
p == 033
) {

    ccv_nnc_tensor_zero(c);

    return;

  }

  // Assuming this is float 32.

  int adim[CCV_NNC_MAX_DIM_ALLOC];

  int bdim[CCV_NNC_MAX_DIM_ALLOC];

  ccv_nnc_tensor_view_get_dim(a, adim);

  ccv_nnc_tensor_view_get_dim(b, bdim);

  int astride[CCV_NNC_MAX_DIM_ALLOC];

  int bstride[CCV_NNC_MAX_DIM_ALLOC];

  int cstride[CCV_NNC_MAX_DIM_ALLOC];

  assert(ccv_nnc_tensor_nd(c->info.dim) <= CCV_NNC_MAX_DIM + 2);

  assert(ccv_nnc_tensor_view_check_dim(c, cdim));

  int x;

  if (!CCV_IS_TENSOR_VIEW(a) && !CCV_IS_TENSOR_VIEW(b) && !CCV_IS_TENSOR_VIEW(c) && a_check_dim && 
b_check_dim27
)

  {

    const int tensor_count = ccv_nnc_tensor_count(a->info);

    // Super optimal case, just do one for-loop for sum.

    for (x = 0; x < tensor_count; 
x++90
)

      c->data.f32[x] = p * a->data.f32[x] * b->data.f32[x];

    return;

  }

  assert
(CCV_NNC_MAX_DIM == 2)24
; // Need to change this logic for CCV_NNC_MAX_DIM == other number.

  ccv_nnc_tensor_view_get_stride(a, astride);

  ccv_nnc_tensor_view_get_stride(b, bstride);

  ccv_nnc_tensor_view_get_stride(c, cstride);

  int i[CCV_NNC_MAX_DIM + 2];

  float* const ap = a->data.f32;

  float* const bp = b->data.f32;

  float* const cp = c->data.f32;

  const int count = cdim[2] * cdim[3];

  if (astride[2] == cdim[3] && 
bstride[2] == cdim[3]19
 && 
cstride[2] == cdim[3]8
 && 
adim[2] == cdim[2]8
 && 
bdim[2] == cdim[2]7
)

  {

    // Special casing if the ainc[3] is the same as dim[3]

    for (i[0] = 0; i[0] < cdim[0]; i[0]++)

    {

      float* const ap0 = adim[0] == 1 ? ap : ap + i[0] * astride[0];

      float* const bp0 = bdim[0] == 1 ? bp : bp + i[0] * bstride[0];

      float* cp0 = cp + i[0] * cstride[0];

      for (i[1] = 0; i[1] < cdim[1]; i[1]++)

      {

        float* const ap1 = adim[1] == 1 ? ap0 : ap0 + i[1] * astride[1];

        float* const bp1 = bdim[1] == 1 ? bp0 : bp0 + i[1] * bstride[1];

        for (x = 0; x < count; x++)

          cp0[x] = p * ap1[x] * bp1[x];

        cp0 += cstride[1];

      }

    }

    return;

  }

  // Non-optimal case, need to do skip copy and handle broadcasting.

  
for (i[0] = 0; 24
i[0] < cdim[0]; 
i[0]++130
)

  {

    float* const ap0 = adim[0] == 1 ? 
ap8
 : 
ap + i[0] * astride[0]122
;

    float* const bp0 = bdim[0] == 1 ? 
bp18
 : 
bp + i[0] * bstride[0]112
;

    float* const cp0 = cp + i[0] * cstride[0];

    for (i[1] = 0; i[1] < cdim[1]; 
i[1]++492
)

    {

      float* const ap1 = adim[1] == 1 ? 
ap08
 : 
ap0 + i[1] * astride[1]484
;

      float* const bp1 = bdim[1] == 1 ? 
bp0300
 : 
bp0 + i[1] * bstride[1]192
;

      float* cp1 = cp0 + i[1] * cstride[1];

      for (i[2] = 0; i[2] < cdim[2]; 
i[2]++1.94k
)

      {

        float* const ap2 = adim[2] == 1 ? 
ap17
 : 
ap1 + i[2] * astride[2]1.93k
;

        float* const bp2 = bdim[2] == 1 ? 
bp11.68k
 : 
bp1 + i[2] * bstride[2]258
;

        if (adim[3] == 1)

          
for (x = 0; 10
x < cdim[3]; 
x++19
)

            cp1[x] = p * ap2[0] * bp2[x];

        else if (bdim[3] == 1)

          
for (x = 0; 1.28k
x < cdim[3]; 
x++12.9k
)

            cp1[x] = p * ap2[x] * bp2[0];

        else

          
for (x = 0; 650
x < cdim[3]; 
x++3.92k
)

            cp1[x] = p * ap2[x] * bp2[x];

        cp1 += cstride[2];

      }

    }

  }

}

static int _ccv_nnc_mul_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);

  _ccv_nnc_mul_forw_cpu_ref(cmd.info.blas.a[0], (ccv_nnc_tensor_view_t*)inputs[0], (ccv_nnc_tensor_view_t*)inputs[1], (ccv_nnc_tensor_view_t*)outputs[0]);

  return CCV_NNC_EXEC_SUCCESS;

}

static int _ccv_nnc_mul_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)

{

  int gdim[CCV_NNC_MAX_DIM_ALLOC];

  int no_broadcasting = 1;

  if (outputs[0])

  {

    assert(input_size >= 3 && inputs[2]);

    ccv_nnc_tensor_view_get_dim((ccv_nnc_tensor_view_t*)outputs[0], gdim);

    ccv_nnc_tensor_view_get_broadcast_dim((ccv_nnc_tensor_view_t*)inputs[2], gdim);

    no_broadcasting = no_broadcasting && (ccv_nnc_tensor_view_check_dim((ccv_nnc_tensor_view_t*)outputs[0], gdim) && 
ccv_nnc_tensor_view_check_dim((ccv_nnc_tensor_view_t*)inputs[2], gdim)2.11k
);

  }

  if (no_broadcasting && 
output_size > 14.11k
 && 
outputs[1]4.11k
)

  {

    assert(inputs[1]);

    ccv_nnc_tensor_view_get_dim((ccv_nnc_tensor_view_t*)inputs[1], gdim);

    ccv_nnc_tensor_view_get_broadcast_dim((ccv_nnc_tensor_view_t*)outputs[1], gdim);

    no_broadcasting = no_broadcasting && (ccv_nnc_tensor_view_check_dim((ccv_nnc_tensor_view_t*)inputs[1], gdim) && 
ccv_nnc_tensor_view_check_dim((ccv_nnc_tensor_view_t*)outputs[1], gdim)4.11k
);

  }

  if (no_broadcasting)

  {

    if (outputs[0])

    {

      if (inputs[0] == 0)

        _ccv_nnc_mul_forw_cpu_ref(cmd.info.blas.a[0], (ccv_nnc_tensor_view_t*)inputs[2], 0, (ccv_nnc_tensor_view_t*)outputs[0]);

      else

        _ccv_nnc_mul_forw_cpu_ref(cmd.info.blas.a[0], (ccv_nnc_tensor_view_t*)inputs[0], (ccv_nnc_tensor_view_t*)inputs[2], (ccv_nnc_tensor_view_t*)outputs[0]);

    }

    if (output_size > 1 && outputs[1])

    {

      if (inputs[0] == 0)

        _ccv_nnc_mul_forw_cpu_ref(cmd.info.blas.a[0], (ccv_nnc_tensor_view_t*)inputs[1], 0, (ccv_nnc_tensor_view_t*)outputs[1]);

      else

        _ccv_nnc_mul_forw_cpu_ref(cmd.info.blas.a[0], (ccv_nnc_tensor_view_t*)inputs[0], (ccv_nnc_tensor_view_t*)inputs[1], (ccv_nnc_tensor_view_t*)outputs[1]);

    }

    return CCV_NNC_EXEC_SUCCESS;

  }

  int adim[CCV_NNC_MAX_DIM_ALLOC];

  int bdim[CCV_NNC_MAX_DIM_ALLOC];

  int astride[CCV_NNC_MAX_DIM_ALLOC];

  int bstride[CCV_NNC_MAX_DIM_ALLOC];

  int i[CCV_NNC_MAX_DIM + 2];

  int x;

  const float p = cmd.info.blas.a[0];

  // Now the case we need broadcasting.

  if (inputs[0] == 0)

  {

    if (outputs[0])

    {

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

      ccv_nnc_tensor_view_t* const b = (ccv_nnc_tensor_view_t*)inputs[2];

      ccv_nnc_tensor_view_get_dim(a, adim);

      ccv_nnc_tensor_view_get_dim(b, bdim);

      ccv_nnc_tensor_view_get_stride(a, astride);

      ccv_nnc_tensor_view_get_stride(b, bstride);

      ccv_nnc_tensor_zero(a);

      float* const ap = a->data.f32;

      float* const bp = b->data.f32;

      for (i[0] = 0; i[0] < gdim[0]; 
i[0]++3
)

      {

        float* const ap0 = adim[0] == 1 ? ap : 
ap + i[0] * astride[0]0
;

        float* const bp0 = bdim[0] == 1 ? bp : 
bp + i[0] * bstride[0]0
;

        for (i[1] = 0; i[1] < gdim[1]; 
i[1]++3
)

        {

          float* const ap1 = adim[1] == 1 ? ap0 : 
ap0 + i[1] * astride[1]0
;

          float* const bp1 = bdim[1] == 1 ? bp0 : 
bp0 + i[1] * bstride[1]0
;

          for (i[2] = 0; i[2] < gdim[2]; 
i[2]++8
)

          {

            float* const ap2 = adim[2] == 1 ? 
ap12
 : 
ap1 + i[2] * astride[2]6
;

            float* const bp2 = bdim[2] == 1 ? 
bp16
 : 
bp1 + i[2] * bstride[2]2
;

            if (adim[3] == 1)

              
for (x = 0; 4
x < gdim[3]; 
x++8
)

                ap2[0] += p * bp2[x];

            else if (bdim[3] == 1)

              for (x = 0; x < gdim[3]; x++)

                ap2[x] += p * bp2[0];

            else

              
for (x = 0; 4
x < gdim[3]; 
x++12
)

                ap2[x] += p * bp2[x];

          }

        }

      }

    }

    if (output_size > 1 && outputs[1])

    {

      ccv_nnc_tensor_view_t* const a = (ccv_nnc_tensor_view_t*)outputs[1];

      ccv_nnc_tensor_view_t* const b = (ccv_nnc_tensor_view_t*)inputs[1];

      ccv_nnc_tensor_view_get_dim(a, adim);

      ccv_nnc_tensor_view_get_dim(b, bdim);

      ccv_nnc_tensor_view_get_stride(a, astride);

      ccv_nnc_tensor_view_get_stride(b, bstride);

      ccv_nnc_tensor_zero(a);

      float* const ap = a->data.f32;

      float* const bp = b->data.f32;

      for (i[0] = 0; i[0] < gdim[0]; 
i[0]++3
)

      {

        float* const ap0 = adim[0] == 1 ? ap : 
ap + i[0] * astride[0]0
;

        float* const bp0 = bdim[0] == 1 ? bp : 
bp + i[0] * bstride[0]0
;

        for (i[1] = 0; i[1] < gdim[1]; 
i[1]++3
)

        {

          float* const ap1 = adim[1] == 1 ? ap0 : 
ap0 + i[1] * astride[1]0
;

          float* const bp1 = bdim[1] == 1 ? bp0 : 
bp0 + i[1] * bstride[1]0
;

          for (i[2] = 0; i[2] < gdim[2]; 
i[2]++8
)

          {

            float* const ap2 = adim[2] == 1 ? 
ap16
 : 
ap1 + i[2] * astride[2]2
;

            float* const bp2 = bdim[2] == 1 ? 
bp12
 : 
bp1 + i[2] * bstride[2]6
;

            if (adim[3] == 1)

              for (x = 0; x < gdim[3]; x++)

                ap2[0] += p * bp2[x];

            else if (bdim[3] == 1)

              
for (x = 0; 4
x < gdim[3]; 
x++8
)

                ap2[x] += p * bp2[0];

            else

              
for (x = 0; 4
x < gdim[3]; 
x++12
)

                ap2[x] += p * bp2[x];

          }

        }

      }

    }

    return CCV_NNC_EXEC_SUCCESS;

  }

  int gstride[CCV_NNC_MAX_DIM_ALLOC];

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

  ccv_nnc_tensor_view_get_dim(g, gdim);

  ccv_nnc_tensor_view_get_stride(g, gstride);

  if (outputs[0])

  {

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

    ccv_nnc_tensor_view_t* const b = (ccv_nnc_tensor_view_t*)inputs[2];

    ccv_nnc_tensor_view_get_dim(a, adim);

    ccv_nnc_tensor_view_get_dim(b, bdim);

    ccv_nnc_tensor_view_get_stride(a, astride);

    ccv_nnc_tensor_view_get_stride(b, bstride);

    ccv_nnc_tensor_zero(a);

    float* const ap = a->data.f32;

    float* const bp = b->data.f32;

    float* const gp = g->data.f32;

    for (i[0] = 0; i[0] < gdim[0]; 
i[0]++60
)

    {

      float* const ap0 = adim[0] == 1 ? 
ap2
 : 
ap + i[0] * astride[0]58
;

      float* const bp0 = bdim[0] == 1 ? 
bp4
 : 
bp + i[0] * bstride[0]56
;

      float* const gp0 = gp + i[0] * gstride[0];

      for (i[1] = 0; i[1] < gdim[1]; 
i[1]++230
)

      {

        float* const ap1 = adim[1] == 1 ? 
ap02
 : 
ap0 + i[1] * astride[1]228
;

        float* const bp1 = bdim[1] == 1 ? 
bp0134
 : 
bp0 + i[1] * bstride[1]96
;

        float* gp1 = gp0 + i[1] * gstride[1];

        for (i[2] = 0; i[2] < gdim[2]; 
i[2]++909
)

        {

          float* const ap2 = adim[2] == 1 ? 
ap11
 : 
ap1 + i[2] * astride[2]908
;

          float* const bp2 = bdim[2] == 1 ? 
bp1781
 : 
bp1 + i[2] * bstride[2]128
;

          if (adim[3] == 1)

            
for (x = 0; 4
x < gdim[3]; 
x++8
)

              ap2[0] += p * gp1[x] * bp2[x];

          else if (bdim[3] == 1)

            
for (x = 0; 641
x < gdim[3]; 
x++6.50k
)

              ap2[x] += p * gp1[x] * bp2[0];

          else

            
for (x = 0; 264
x < gdim[3]; 
x++1.36k
)

              ap2[x] += p * gp1[x] * bp2[x];

          gp1 += gstride[2];

        }

      }

    }

  }

  if (output_size > 1 && outputs[1])

  {

    ccv_nnc_tensor_view_t* const a = (ccv_nnc_tensor_view_t*)outputs[1];

    ccv_nnc_tensor_view_t* const b = (ccv_nnc_tensor_view_t*)inputs[1];

    ccv_nnc_tensor_view_get_dim(a, adim);

    ccv_nnc_tensor_view_get_dim(b, bdim);

    ccv_nnc_tensor_view_get_stride(a, astride);

    ccv_nnc_tensor_view_get_stride(b, bstride);

    ccv_nnc_tensor_zero(a);

    float* const ap = a->data.f32;

    float* const bp = b->data.f32;

    float* const gp = g->data.f32;

    for (i[0] = 0; i[0] < gdim[0]; 
i[0]++61
)

    {

      float* const ap0 = adim[0] == 1 ? 
ap5
 : 
ap + i[0] * astride[0]56
;

      float* const bp0 = bdim[0] == 1 ? 
bp3
 : 
bp + i[0] * bstride[0]58
;

      float* const gp0 = gp + i[0] * gstride[0];

      for (i[1] = 0; i[1] < gdim[1]; 
i[1]++231
)

      {

        float* const ap1 = adim[1] == 1 ? 
ap0135
 : 
ap0 + i[1] * astride[1]96
;

        float* const bp1 = bdim[1] == 1 ? 
bp03
 : 
bp0 + i[1] * bstride[1]228
;

        float* gp1 = gp0 + i[1] * gstride[1];

        for (i[2] = 0; i[2] < gdim[2]; 
i[2]++911
)

        {

          float* const ap2 = adim[2] == 1 ? 
ap1781
 : 
ap1 + i[2] * astride[2]130
;

          float* const bp2 = bdim[2] == 1 ? 
bp13
 : 
bp1 + i[2] * bstride[2]908
;

          if (adim[3] == 1)

            
for (x = 0; 643
x < gdim[3]; 
x++6.50k
)

              ap2[0] += p * gp1[x] * bp2[x];

          else if (bdim[3] == 1)

            
for (x = 0; 4
x < gdim[3]; 
x++8
)

              ap2[x] += p * gp1[x] * bp2[0];

          else

            
for (x = 0; 264
x < gdim[3]; 
x++1.36k
)

              ap2[x] += p * gp1[x] * bp2[x];

          gp1 += gstride[2];

        }

      }

    }

  }

  return CCV_NNC_EXEC_SUCCESS;

}

REGISTER_COMMAND_BACKEND(CCV_NNC_MUL_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 | CCV_TENSOR_FORMAT_CHWN;

  registry->tensor_datatypes = CCV_32F;

  registry->tensor_memory = CCV_TENSOR_CPU_MEMORY;

  registry->algorithms = 1;

  registry->exec = _ccv_nnc_mul_forw;

}

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

{

  registry->tensor_formats = CCV_TENSOR_FORMAT_NHWC | CCV_TENSOR_FORMAT_NCHW | CCV_TENSOR_FORMAT_CHWN;

  registry->tensor_datatypes = CCV_32F;

  registry->tensor_memory = CCV_TENSOR_CPU_MEMORY;

  registry->algorithms = 1;

  registry->exec = _ccv_nnc_mul_back;

}

static void _ccv_nnc_scalar_mul_forw_cpu_ref_i32(const float p, ccv_nnc_tensor_view_t* const a, ccv_nnc_tensor_view_t* const c)

{

  // Assuming this is int.

  int dim[CCV_NNC_MAX_DIM_ALLOC];

  int astride[CCV_NNC_MAX_DIM_ALLOC];

  int cstride[CCV_NNC_MAX_DIM_ALLOC];

  assert(ccv_nnc_tensor_nd(a->info.dim) <= CCV_NNC_MAX_DIM + 2);

  assert(ccv_nnc_tensor_nd(c->info.dim) <= CCV_NNC_MAX_DIM + 2);

  ccv_nnc_tensor_view_get_dim(a, dim);

  assert(ccv_nnc_tensor_view_check_dim(c, dim));

  int x;

  if (CCV_IS_TENSOR_CONTIGUOUS(a) && CCV_IS_TENSOR_CONTIGUOUS(c))

  {

    // Super optimal case, just do one for-loop for sum.

    const int tensor_count = ccv_nnc_tensor_count(a->info);

    for (x = 0; x < tensor_count; 
x++6
)

      c->data.i32[x] = (int)(p * (float)a->data.i32[x]);

    return;

  }

  assert
(CCV_NNC_MAX_DIM == 2)0
; // Need to change this logic for CCV_NNC_MAX_DIM == other number.

  ccv_nnc_tensor_view_get_stride(a, astride);

  ccv_nnc_tensor_view_get_stride(c, cstride);

  int i[CCV_NNC_MAX_DIM + 2];

  int* const ap = a->data.i32;

  int* const cp = c->data.i32;

  const int count = dim[2] * dim[3];

  if (astride[2] == dim[3] && cstride[2] == dim[3])

  {

    // Special casing if the ainc[3] is the same as dim[3]

    for (i[0] = 0; i[0] < dim[0]; i[0]++)

    {

      int* ap0 = ap + i[0] * astride[0];

      int* cp0 = cp + i[0] * cstride[0];

      for (i[1] = 0; i[1] < dim[1]; i[1]++)

      {

        for (x = 0; x < count; x++)

          cp0[x] = (int)(p * (float)ap0[x]);

        ap0 += astride[1];

        cp0 += cstride[1];

      }

    }

    return;

  }

  // Non-optimal case, need to do skip copy.

  for (i[0] = 0; i[0] < dim[0]; i[0]++)

  {

    int* const ap0 = ap + i[0] * astride[0];

    int* const cp0 = cp + i[0] * cstride[0];

    for (i[1] = 0; i[1] < dim[1]; i[1]++)

    {

      int* ap1 = ap0 + i[1] * astride[1];

      int* cp1 = cp0 + i[1] * cstride[1];

      for (i[2] = 0; i[2] < dim[2]; i[2]++)

      {

        for (x = 0; x < dim[3]; x++)

          cp1[x] = (int)(p * (float)ap1[x]);

        ap1 += astride[2];

        cp1 += cstride[2];

      }

    }

  }

}

static int _ccv_nnc_scalar_mul_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)

{

  if (inputs[0]->info.datatype == CCV_32S)

  {

    _ccv_nnc_scalar_mul_forw_cpu_ref_i32(cmd.info.blas.a[0], (ccv_nnc_tensor_view_t*)inputs[0], (ccv_nnc_tensor_view_t*)outputs[0]);

    return CCV_NNC_EXEC_SUCCESS;

  }

  _ccv_nnc_mul_forw_cpu_ref(cmd.info.blas.a[0], (ccv_nnc_tensor_view_t*)inputs[0], 0, (ccv_nnc_tensor_view_t*)outputs[0]);

  return CCV_NNC_EXEC_SUCCESS;

}

static int _ccv_nnc_scalar_mul_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)

{

  if (inputs[0])

    _ccv_nnc_mul_forw_cpu_ref(cmd.info.blas.a[0], (ccv_nnc_tensor_view_t*)inputs[0], 0, (ccv_nnc_tensor_view_t*)outputs[0]);

  else

    _ccv_nnc_tensor_set_cpu_ref_f32((ccv_nnc_tensor_view_t*)outputs[0], cmd.info.blas.a[0]);

  return CCV_NNC_EXEC_SUCCESS;

}

REGISTER_COMMAND_BACKEND(CCV_NNC_SCALAR_MUL_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 | CCV_TENSOR_FORMAT_CHWN;

  registry->tensor_datatypes = CCV_32F | CCV_32S;

  registry->tensor_memory = CCV_TENSOR_CPU_MEMORY;

  registry->algorithms = 1;

  registry->exec = _ccv_nnc_scalar_mul_forw;

}

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

{

  registry->tensor_formats = CCV_TENSOR_FORMAT_NHWC | CCV_TENSOR_FORMAT_NCHW | CCV_TENSOR_FORMAT_CHWN;

  registry->tensor_datatypes = CCV_32F;

  registry->tensor_memory = CCV_TENSOR_CPU_MEMORY;

  registry->algorithms = 1;

  registry->exec = _ccv_nnc_scalar_mul_back;

}