#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_tanh_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 == 1);

  const ccv_nnc_tensor_t* a = inputs[0];

  assert(CCV_IS_TENSOR_CONTIGUOUS(a));

  assert(output_size == 1);

  ccv_nnc_tensor_t* b = outputs[0];

  assert(CCV_IS_TENSOR_CONTIGUOUS(b));

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

  int i;

  for (i = 0; i < CCV_NNC_MAX_DIM_ALLOC && a->info.dim[i] > 0; 
i++9
)

    { assert(a->info.dim[i] == b->info.dim[i]); }

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

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

  for (i = 0; i < count; 
i++2.41k
)

    bp[i] = tanh(ap[i]);

  return CCV_NNC_EXEC_SUCCESS;

}

static int _ccv_nnc_tanh_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)

{

  assert(input_size == 3);

  assert(output_size == 1);

  const ccv_nnc_tensor_t* g = inputs[0];

  const ccv_nnc_tensor_t* b = inputs[2];

  assert(CCV_IS_TENSOR_CONTIGUOUS(b));

  ccv_nnc_tensor_t* h = outputs[0];

  assert(CCV_IS_TENSOR_CONTIGUOUS(h));

  const int count = ccv_nnc_tensor_count(b->info);

  int i;

  for (i = 0; i < CCV_NNC_MAX_DIM_ALLOC && g->info.dim[i] > 0; 
i++5
)

    { assert(h->info.dim[i] == b->info.dim[i]); }

  if (g)

  {

    assert(CCV_IS_TENSOR_CONTIGUOUS(g));

    
for (i = 0; 3
i < CCV_NNC_MAX_DIM_ALLOC && g->info.dim[i] > 0; 
i++5
)

      { assert(g->info.dim[i] == h->info.dim[i]); }

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

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

    float* const hp = h->data.f32;

    for (i = 0; i < count; 
i++2.01k
)

      hp[i] = gp[i] * (1 - bp[i] * bp[i]);

  } else {

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

    float* const hp = h->data.f32;

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

      hp[i] = 1 - bp[i] * bp[i];

  }

  return CCV_NNC_EXEC_SUCCESS;

}

REGISTER_COMMAND_BACKEND(CCV_NNC_TANH_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_tanh_forw;

}

REGISTER_COMMAND_BACKEND(CCV_NNC_TANH_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;

  registry->tensor_datatypes = CCV_32F;

  registry->tensor_memory = CCV_TENSOR_CPU_MEMORY;

  registry->algorithms = 1;

  registry->exec = _ccv_nnc_tanh_back;

}