From c3e881847a856ad122b322ecf35b9b38b84ec363 Mon Sep 17 00:00:00 2001
From: liuqi <liuqi10@xiaomi.com>
Date: Wed, 13 Sep 2017 10:51:52 +0800
Subject: [PATCH] Add conv2d neon api.

---
 mace/kernels/conv_2d.h                |  9 ++++
 mace/kernels/neon/conv_2d_neon.cc     | 64 +++++++++++++++++++++++++++
 mace/kernels/neon/conv_2d_neon_3x3.cc | 22 +++++++++
 mace/kernels/neon/conv_2d_neon_base.h | 24 ++++++++++
 4 files changed, 119 insertions(+)
 create mode 100644 mace/kernels/neon/conv_2d_neon.cc
 create mode 100644 mace/kernels/neon/conv_2d_neon_3x3.cc
 create mode 100644 mace/kernels/neon/conv_2d_neon_base.h

diff --git a/mace/kernels/conv_2d.h b/mace/kernels/conv_2d.h
index 2e1a145a..a102ccef 100644
--- a/mace/kernels/conv_2d.h
+++ b/mace/kernels/conv_2d.h
@@ -108,6 +108,15 @@ class Conv2dFunctor {
     const int* dilations_; // [dilation_h, dilation_w]
 };
 
+template<>
+void Conv2dFunctor<DeviceType::NEON, float>::operator()(const float* input, // NCHW
+                const index_t* input_shape,
+                const float* filter, // c_out, c_in, kernel_h, kernel_w
+                const index_t* filter_shape,
+                const float* bias, // c_out
+                float* output, // NCHW
+                const index_t* output_shape);
+
 } //  namespace kernels
 } //  namespace mace
 
diff --git a/mace/kernels/neon/conv_2d_neon.cc b/mace/kernels/neon/conv_2d_neon.cc
new file mode 100644
index 00000000..c9678fd8
--- /dev/null
+++ b/mace/kernels/neon/conv_2d_neon.cc
@@ -0,0 +1,64 @@
+//
+// Copyright (c) 2017 XiaoMi All rights reserved.
+//
+
+#include <arm_neon.h>
+#include "mace/kernels/conv_2d.h"
+#include "mace/kernels/neon/conv_2d_neon_base.h"
+
+namespace mace {
+namespace kernels {
+
+static inline void ConstructInputWithPadding(const float* input, const index_t* input_shape,
+                                             const int* padding,
+                                             std::unique_ptr<float>& output,
+                                             index_t* output_shape) {
+
+}
+
+template<>
+void Conv2dFunctor<DeviceType::NEON, float>::operator()(const float* input, // NCHW
+                                                        const index_t* input_shape,
+                                                        const float* filter, // c_out, c_in, kernel_h, kernel_w
+                                                        const index_t* filter_shape,
+                                                        const float* bias, // c_out
+                                                        float* output, // NCHW
+                                                        const index_t* output_shape) {
+
+  static const bool selector[5][4] = {
+          {true, false, false, false},
+          {false, false, false, false},
+          {true, true, false, false},
+          {false, false, false, false},
+          {true, false, false, false},
+  };
+  // not implement yet
+  if (paddings_[0] != paddings_[1] || paddings_[0] > 5 ||
+          strides_[0] != strides_[1] || strides_[0] > 4 ||
+          dilations_[0] != 1 || dilations_[1] != 1 ||
+          !selector[paddings_[0]-1, strides_[0]-1]) {
+    Conv2dFunctor<DeviceType::CPU, float>(strides_, paddings_, dilations_)(
+            input,
+            input_shape,
+            filter,
+            filter_shape,
+            bias,
+            output,
+            output_shape
+    );
+  }
+  std::unique_ptr<float> padded_input;
+  index_t padded_input_shape[4];
+  ConstructInputWithPadding(input, input_shape, paddings_, padded_input, padded_input_shape);
+  Conv2dNeon<paddings_[0], paddings_[1], strides_[0], strides_[1]>(
+          padded_input.get(),
+          padded_input_shape,
+          filter,
+          bias,
+          output,
+          output_shape
+  );
+}
+
+} // namespace kernels
+} //  namespace mace
\ No newline at end of file
diff --git a/mace/kernels/neon/conv_2d_neon_3x3.cc b/mace/kernels/neon/conv_2d_neon_3x3.cc
new file mode 100644
index 00000000..4397076d
--- /dev/null
+++ b/mace/kernels/neon/conv_2d_neon_3x3.cc
@@ -0,0 +1,22 @@
+//
+// Copyright (c) 2017 XiaoMi All rights reserved.
+//
+
+#include <arm_neon.h>
+#include "mace/kernels/neon/conv_2d_neon_base.h"
+
+namespace mace {
+namespace kernels {
+
+template<>
+void Conv2dNeon<3, 3, 1, 1>(const float* input, // NCHW
+                       const index_t* input_shape,
+                       const float* filter, // c_out, c_in, kernel_h, kernel_w
+                       const float* bias, // c_out
+                       float* output, // NCHW
+                       const index_t* output_shape) {
+
+}
+
+} // namespace kernels
+} //  namespace mace
diff --git a/mace/kernels/neon/conv_2d_neon_base.h b/mace/kernels/neon/conv_2d_neon_base.h
new file mode 100644
index 00000000..4e0b5024
--- /dev/null
+++ b/mace/kernels/neon/conv_2d_neon_base.h
@@ -0,0 +1,24 @@
+//
+// Copyright (c) 2017 XiaoMi All rights reserved.
+//
+
+#ifndef MACE_KERNELS_NEON_CONV_2D_NEON_BASE_H_
+#define MACE_KERNELS_NEON_CONV_2D_NEON_BASE_H_
+
+#include "mace/core/common.h"
+
+namespace mace {
+namespace kernels {
+
+template <index_t kernel_h, index_t kernel_w, index_t stride_h, index_t stride_w>
+inline void Conv2dNeon(const float* input, // NCHW
+                           const index_t* input_shape,
+                           const float* filter, // c_out, c_in, kernel_h, kernel_w
+                           const float* bias, // c_out
+                           float* output, // NCHW
+                           const index_t* output_shape);
+
+} // namespace kernels
+} //  namespace mace
+
+#endif //  MACE_KERNELS_NEON_CONV_2D_NEON_BASE_H_
-- 
GitLab