1 files changed, 148 insertions, 0 deletions
diff --git a/compiler/enco/frontend/caffe/src/PoolingSpec.cpp b/compiler/enco/frontend/caffe/src/PoolingSpec.cpp
new file mode 100644
index 000000000..36216a2da
--- /dev/null
+++ b/compiler/enco/frontend/caffe/src/PoolingSpec.cpp
@@ -0,0 +1,148 @@
+/*
+ * Copyright (c) 2018 Samsung Electronics Co., Ltd. All Rights Reserved
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *    http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include "PoolingSpec.h"
+#include "PaddingUtils.h"
+
+#include <map>
+#include <cassert>
+
+PoolingSpec::PoolingSpec(const ::caffe::PoolingParameter &param) : _param(param)
+{
+  // DO NOTHING
+}
+
+PoolingMethod PoolingSpec::method(void) const
+{
+  if (!_param.has_pool())
+  {
+    // Default pooling method is MAX
+    // Reference: http://caffe.berkeleyvision.org/tutorial/layers/pooling.html
+    return PoolingMethod::Max;
+  }
+
+  std::map<::caffe::PoolingParameter_PoolMethod, PoolingMethod> methods;
+
+  // NOTE STOCHASTIC Pooling is not supported, yet
+  // TODO Support STOCHASTIC Pooling
+  methods[::caffe::PoolingParameter_PoolMethod_MAX] = PoolingMethod::Max;
+  methods[::caffe::PoolingParameter_PoolMethod_AVE] = PoolingMethod::Avg;
+
+  assert(_param.has_pool());
+  return methods.at(_param.pool());
+}
+
+uint32_t PoolingSpec::window_height(void) const
+{
+  // NOTE Global pooling is not supported, yet
+  // TODO Support global pooling
+  assert(!_param.global_pooling());
+
+  if (_param.has_kernel_h())
+  {
+    return _param.kernel_h();
+  }
+
+  assert(_param.has_kernel_size());
+  return _param.kernel_size();
+}
+
+uint32_t PoolingSpec::window_width(void) const
+{
+  // NOTE Global pooling is not supported, yet
+  // TODO Support global pooling
+  assert(!_param.global_pooling());
+
+  if (_param.has_kernel_w())
+  {
+    return _param.kernel_w();
+  }
+
+  assert(_param.has_kernel_size());
+  return _param.kernel_size();
+}
+
+uint32_t PoolingSpec::vertical_pad(void) const
+{
+  // NOTE The input of Pooling SHOULD BE a rank-4 tensor.
+  // Reference: PoolingLayer<Dtype>::Reshape in pooling_layer.cpp
+  auto raw_padding = build_raw_padding().with(_param);
+  auto spatial_padding = build_spatial_padding(2 /* SPATIAL RANK */).with(raw_padding);
+  return spatial_padding.value(0 /* H */);
+}
+
+uint32_t PoolingSpec::horizontal_pad(void) const
+{
+  // NOTE The input of Pooling SHOULD BE a rank-4 tensor.
+  // Reference: PoolingLayer<Dtype>::Reshape in pooling_layer.cpp
+  auto raw_padding = build_raw_padding().with(_param);
+  auto spatial_padding = build_spatial_padding(2 /* SPATIAL RANK */).with(raw_padding);
+  return spatial_padding.value(1 /* W */);
+}
+
+uint32_t PoolingSpec::vertical_stride(void) const
+{
+  if (_param.has_stride_h())
+  {
+    return _param.stride_h();
+  }
+
+  if (_param.has_stride())
+  {
+    return _param.stride();
+  }
+
+  return 1;
+}
+
+uint32_t PoolingSpec::horizontal_stride(void) const
+{
+  if (_param.has_stride_w())
+  {
+    return _param.stride_w();
+  }
+
+  if (_param.has_stride())
+  {
+    return _param.stride();
+  }
+
+  return 1;
+}
+
+nncc::core::ADT::tensor::Shape PoolingSpec::ofm_shape(void) const
+{
+  nncc::core::ADT::tensor::Shape res;
+
+  // NOTE Caffe supports only pooling over rank-4 tensor
+  assert(_ifm_shape.rank() == 4);
+  res.resize(4);
+
+  // N (= the number of bacths) SHOULD be same
+  res.dim(0) = _ifm_shape.dim(0);
+  // C (= the number of chaanels) SHOULD be same
+  res.dim(1) = _ifm_shape.dim(1);
+
+  // H and W are derived from IFM, Window, and Padding
+  const auto effective_input_height = _ifm_shape.dim(2) + 2 * vertical_pad() - window_height();
+  const auto effective_input_width = _ifm_shape.dim(3) + 2 * horizontal_pad() - window_width();
+  // TODO Remove the following asserts
+  assert(effective_input_height % vertical_stride() == 0);
+  assert(effective_input_width % horizontal_stride() == 0);
+  res.dim(2) = effective_input_height / vertical_stride() + 1;
+  res.dim(3) = effective_input_width / horizontal_stride() + 1;
+  return res;
+}