COMPMID-2336: Rename the new generic depthwise convolution on NEON

Change-Id: I45cacf75b08bb9d867343037507e56f200ad6ac0
Signed-off-by: Gian Marco Iodice <gianmarco.iodice@arm.com>
Reviewed-on: https://review.mlplatform.org/c/1637
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Giorgio Arena <giorgio.arena@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>

7 files changed, 82 insertions, 53 deletions

diff --git a/arm_compute/core/NEON/NEKernels.h b/arm_compute/core/NEON/NEKernels.h
index 8fddec2c5f..bf77e576ad 100644
--- a/arm_compute/core/NEON/NEKernels.h
+++ b/arm_compute/core/NEON/NEKernels.h
@@ -53,7 +53,7 @@
 #include "arm_compute/core/NEON/kernels/NEDepthConvertLayerKernel.h"
 #include "arm_compute/core/NEON/kernels/NEDepthToSpaceLayerKernel.h"
 #include "arm_compute/core/NEON/kernels/NEDepthwiseConvolutionLayer3x3Kernel.h"
-#include "arm_compute/core/NEON/kernels/NEDepthwiseConvolutionLayerKernel.h"
+#include "arm_compute/core/NEON/kernels/NEDepthwiseConvolutionLayerNativeKernel.h"
 #include "arm_compute/core/NEON/kernels/NEDepthwiseIm2ColKernel.h"
 #include "arm_compute/core/NEON/kernels/NEDepthwiseVectorToTensorKernel.h"
 #include "arm_compute/core/NEON/kernels/NEDepthwiseWeightsReshapeKernel.h"
diff --git a/arm_compute/core/NEON/kernels/NEDepthwiseConvolutionLayerKernel.h b/arm_compute/core/NEON/kernels/NEDepthwiseConvolutionLayerNativeKernel.h
index 63635b3a6c..5db79f8bf7 100644
--- a/arm_compute/core/NEON/kernels/NEDepthwiseConvolutionLayerKernel.h
+++ b/arm_compute/core/NEON/kernels/NEDepthwiseConvolutionLayerNativeKernel.h
@@ -21,8 +21,8 @@
  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
  */
-#ifndef __ARM_COMPUTE_NEDEPTHWISECONVOLUTIONKERNEL_H__
-#define __ARM_COMPUTE_NEDEPTHWISECONVOLUTIONKERNEL_H__
+#ifndef __ARM_COMPUTE_NEDEPTHWISECONVOLUTIONLAYERNATIVEKERNEL_H__
+#define __ARM_COMPUTE_NEDEPTHWISECONVOLUTIONLAYERNATIVEKERNEL_H__
 
 #include "arm_compute/core/NEON/INEKernel.h"
 
@@ -31,24 +31,24 @@ namespace arm_compute
 // Forward declarations
 class ITensor;
 
-/** Interface for the kernel to run a depthwise convolution on a tensor. */
-class NEDepthwiseConvolutionLayerKernel : public INEKernel
+/** Interface for the kernel to run a depthwise convolution native on a tensor. */
+class NEDepthwiseConvolutionLayerNativeKernel : public INEKernel
 {
 public:
     const char *name() const override
     {
-        return "NEDepthwiseConvolutionLayerKernel";
+        return "NEDepthwiseConvolutionLayerNativeKernel";
     }
     /** Default constructor */
-    NEDepthwiseConvolutionLayerKernel();
+    NEDepthwiseConvolutionLayerNativeKernel();
     /** Prevent instances of this class from being copied (As this class contains pointers) */
-    NEDepthwiseConvolutionLayerKernel(const NEDepthwiseConvolutionLayerKernel &) = delete;
+    NEDepthwiseConvolutionLayerNativeKernel(const NEDepthwiseConvolutionLayerNativeKernel &) = delete;
     /** Prevent instances of this class from being copied (As this class contains pointers) */
-    NEDepthwiseConvolutionLayerKernel &operator=(const NEDepthwiseConvolutionLayerKernel &) = delete;
+    NEDepthwiseConvolutionLayerNativeKernel &operator=(const NEDepthwiseConvolutionLayerNativeKernel &) = delete;
     /** Default Move Constructor. */
-    NEDepthwiseConvolutionLayerKernel(NEDepthwiseConvolutionLayerKernel &&) = default;
+    NEDepthwiseConvolutionLayerNativeKernel(NEDepthwiseConvolutionLayerNativeKernel &&) = default;
     /** Default move assignment operator */
-    NEDepthwiseConvolutionLayerKernel &operator=(NEDepthwiseConvolutionLayerKernel &&) = default;
+    NEDepthwiseConvolutionLayerNativeKernel &operator=(NEDepthwiseConvolutionLayerNativeKernel &&) = default;
     /** Initialize the function's source, destination and parameters.
      *
      * @note Supported data layouts: NHWC
@@ -64,7 +64,7 @@ public:
      */
     void configure(const ITensor *input, const ITensor *weights, const ITensor *biases, ITensor *output, const PadStrideInfo &conv_info, unsigned int depth_multiplier = 1,
                    const Size2D &dilation = Size2D(1U, 1U));
-    /** Static function to check if given info will lead to a valid configuration of @ref NEDepthwiseConvolutionLayerKernel
+    /** Static function to check if given info will lead to a valid configuration of @ref NEDepthwiseConvolutionLayerNativeKernel
      *
      * @note Supported data layouts: NHWC
      *
@@ -89,11 +89,11 @@ private:
     template <typename T, int S, bool has_biases>
     void run_depthwise(const Window &window);
 
-    /** Common signature for all the specialised depthwise convolution functions
+    /** Common signature for all the specialised depthwise convolution native functions
      *
      * @param[in] window Region on which to execute the kernel.
      */
-    using DepthwiseFunctionPtr = void (NEDepthwiseConvolutionLayerKernel::*)(const Window &window);
+    using DepthwiseFunctionPtr = void (NEDepthwiseConvolutionLayerNativeKernel::*)(const Window &window);
 
     DepthwiseFunctionPtr _func;
     BorderSize           _border_size;
@@ -106,4 +106,4 @@ private:
     Size2D               _dilation;
 };
 } // namespace arm_compute
-#endif /* __ARM_COMPUTE_NEDEPTHWISECONVOLUTIONKERNEL_H__ */
+#endif /* __ARM_COMPUTE_NEDEPTHWISECONVOLUTIONLAYERNATIVEKERNEL_H__ */
diff --git a/arm_compute/runtime/NEON/functions/NEDepthwiseConvolutionLayer.h b/arm_compute/runtime/NEON/functions/NEDepthwiseConvolutionLayer.h
index 5b0d1bafcd..87405fdb14 100644
--- a/arm_compute/runtime/NEON/functions/NEDepthwiseConvolutionLayer.h
+++ b/arm_compute/runtime/NEON/functions/NEDepthwiseConvolutionLayer.h
@@ -25,7 +25,7 @@
 #define __ARM_COMPUTE_NEDEPTHWISECONVOLUTION_H__
 
 #include "arm_compute/core/NEON/kernels/NEDepthwiseConvolutionLayer3x3Kernel.h"
-#include "arm_compute/core/NEON/kernels/NEDepthwiseConvolutionLayerKernel.h"
+#include "arm_compute/core/NEON/kernels/NEDepthwiseConvolutionLayerNativeKernel.h"
 #include "arm_compute/core/NEON/kernels/NEDepthwiseIm2ColKernel.h"
 #include "arm_compute/core/NEON/kernels/NEDepthwiseVectorToTensorKernel.h"
 #include "arm_compute/core/NEON/kernels/NEDepthwiseWeightsReshapeKernel.h"
@@ -282,7 +282,7 @@ private:
 /** Basic function to execute a generic depthwise convolution. This function calls the following NEON kernels:
  *
  * If data type is F32 and data layout is NHWC:
- * -# @ref NEDepthwiseConvolutionLayerKernel
+ * -# @ref NEDepthwiseConvolutionLayerNativeKernel
  *
  * Otherwise:
  * -# @ref NEDepthwiseIm2ColKernel
@@ -344,7 +344,7 @@ private:
     NEDepthwiseIm2ColKernel                   _im2col_kernel;
     NEDepthwiseWeightsReshapeKernel           _weights_reshape_kernel;
     NEGEMMMatrixVectorMultiplyKernel          _v2mm_kernel;
-    NEDepthwiseConvolutionLayerKernel         _depthwise_conv_kernel;
+    NEDepthwiseConvolutionLayerNativeKernel   _depthwise_conv_kernel;
     NEDepthwiseVectorToTensorKernel           _vector_to_tensor_kernel;
     NEDirectConvolutionLayerOutputStageKernel _output_stage_kernel;
     NEFillBorderKernel                        _fill_border;
diff --git a/src/core/NEON/kernels/NEDepthwiseConvolutionLayerKernel.cpp b/src/core/NEON/kernels/NEDepthwiseConvolutionLayerNativeKernel.cpp
index feb2071d47..aafdb2e8a4 100644
--- a/src/core/NEON/kernels/NEDepthwiseConvolutionLayerKernel.cpp
+++ b/src/core/NEON/kernels/NEDepthwiseConvolutionLayerNativeKernel.cpp
@@ -21,7 +21,7 @@
  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
  */
-#include "arm_compute/core/NEON/kernels/NEDepthwiseConvolutionLayerKernel.h"
+#include "arm_compute/core/NEON/kernels/NEDepthwiseConvolutionLayerNativeKernel.h"
 
 #include "arm_compute/core/AccessWindowStatic.h"
 #include "arm_compute/core/NEON/wrapper/traits.h"
@@ -252,18 +252,18 @@ std::pair<Status, Window> validate_and_configure_window(ITensorInfo *input, ITen
     return std::make_pair(err, win);
 }
 
-NEDepthwiseConvolutionLayerKernel::NEDepthwiseConvolutionLayerKernel()
+NEDepthwiseConvolutionLayerNativeKernel::NEDepthwiseConvolutionLayerNativeKernel()
     : _func(), _border_size(0), _input(), _weights(), _biases(), _output(), _conv_info(), _depth_multiplier(1), _dilation()
 {
 }
 
-BorderSize NEDepthwiseConvolutionLayerKernel::border_size() const
+BorderSize NEDepthwiseConvolutionLayerNativeKernel::border_size() const
 {
     return _border_size;
 }
 
-void NEDepthwiseConvolutionLayerKernel::configure(const ITensor *input, const ITensor *weights, const ITensor *biases, ITensor *output,
-                                                  const PadStrideInfo &conv_info, unsigned int depth_multiplier, const Size2D &dilation)
+void NEDepthwiseConvolutionLayerNativeKernel::configure(const ITensor *input, const ITensor *weights, const ITensor *biases, ITensor *output,
+                                                        const PadStrideInfo &conv_info, unsigned int depth_multiplier, const Size2D &dilation)
 {
     ARM_COMPUTE_ERROR_ON_NULLPTR(input, weights, output);
     ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(input->info(), weights->info(), (biases != nullptr) ? biases->info() : nullptr, output->info(), conv_info, depth_multiplier, dilation));
@@ -280,7 +280,7 @@ void NEDepthwiseConvolutionLayerKernel::configure(const ITensor *input, const IT
     switch(_input->info()->data_type())
     {
         case DataType::F32:
-            _func = (biases != nullptr) ? &NEDepthwiseConvolutionLayerKernel::run_depthwise<float, 2, true> : &NEDepthwiseConvolutionLayerKernel::run_depthwise<float, 2, false>;
+            _func = (biases != nullptr) ? &NEDepthwiseConvolutionLayerNativeKernel::run_depthwise<float, 2, true> : &NEDepthwiseConvolutionLayerNativeKernel::run_depthwise<float, 2, false>;
             break;
         default:
             ARM_COMPUTE_ERROR("Data type not supported");
@@ -292,9 +292,9 @@ void NEDepthwiseConvolutionLayerKernel::configure(const ITensor *input, const IT
     INEKernel::configure(win_config.second);
 }
 
-Status NEDepthwiseConvolutionLayerKernel::validate(const ITensorInfo *input, const ITensorInfo *weights, const ITensorInfo *biases, const ITensorInfo *output, const PadStrideInfo &conv_info,
-                                                   unsigned int  depth_multiplier,
-                                                   const Size2D &dilation)
+Status NEDepthwiseConvolutionLayerNativeKernel::validate(const ITensorInfo *input, const ITensorInfo *weights, const ITensorInfo *biases, const ITensorInfo *output, const PadStrideInfo &conv_info,
+                                                         unsigned int  depth_multiplier,
+                                                         const Size2D &dilation)
 {
     ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(input, weights, biases, output, conv_info, depth_multiplier, dilation));
     ARM_COMPUTE_RETURN_ON_ERROR(validate_and_configure_window(input->clone().get(), weights->clone().get(), (biases != nullptr) ? biases->clone().get() : nullptr, output->clone().get(), conv_info,
@@ -303,7 +303,7 @@ Status NEDepthwiseConvolutionLayerKernel::validate(const ITensorInfo *input, con
     return Status{};
 }
 
-void NEDepthwiseConvolutionLayerKernel::run(const Window &window, const ThreadInfo &info)
+void NEDepthwiseConvolutionLayerNativeKernel::run(const Window &window, const ThreadInfo &info)
 {
     ARM_COMPUTE_UNUSED(info);
     ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
@@ -313,7 +313,7 @@ void NEDepthwiseConvolutionLayerKernel::run(const Window &window, const ThreadIn
 }
 
 template <typename T, int S, bool has_biases>
-void NEDepthwiseConvolutionLayerKernel::run_depthwise(const Window &window)
+void NEDepthwiseConvolutionLayerNativeKernel::run_depthwise(const Window &window)
 {
     ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
     ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(INEKernel::window(), window);
diff --git a/src/runtime/NEON/functions/NEDepthwiseConvolutionLayer.cpp b/src/runtime/NEON/functions/NEDepthwiseConvolutionLayer.cpp
index c2ed901169..cdd278b2f1 100644
--- a/src/runtime/NEON/functions/NEDepthwiseConvolutionLayer.cpp
+++ b/src/runtime/NEON/functions/NEDepthwiseConvolutionLayer.cpp
@@ -947,7 +947,7 @@ Status NEDepthwiseConvolutionLayer::validate(const ITensorInfo *input, const ITe
     }
     else
     {
-        ARM_COMPUTE_RETURN_ON_ERROR(NEDepthwiseConvolutionLayerKernel::validate(input, weights, biases, output, conv_info, depth_multiplier, dilation));
+        ARM_COMPUTE_RETURN_ON_ERROR(NEDepthwiseConvolutionLayerNativeKernel::validate(input, weights, biases, output, conv_info, depth_multiplier, dilation));
     }
 
     // Validate Activation Layer
diff --git a/tests/validation/NEON/DepthwiseConvolutionLayerKernel.cpp b/tests/validation/NEON/DepthwiseConvolutionNativeLayer.cpp
index 3af835855b..a44c2dc3c9 100644
--- a/tests/validation/NEON/DepthwiseConvolutionLayerKernel.cpp
+++ b/tests/validation/NEON/DepthwiseConvolutionNativeLayer.cpp
@@ -21,7 +21,7 @@
  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
  */
-#include "arm_compute/core/NEON/kernels/NEDepthwiseConvolutionLayerKernel.h"
+#include "arm_compute/core/NEON/kernels/NEDepthwiseConvolutionLayerNativeKernel.h"
 #include "tests/NEON/Accessor.h"
 #include "tests/NEON/Helper.h"
 #include "tests/framework/Macros.h"
@@ -38,11 +38,11 @@ namespace validation
 using namespace arm_compute::misc::shape_calculator;
 
 // Create function for NEDepthwiseConvolutionLayerKernel
-using NEDepthwiseConvolutionLayer = NESynthetizeFunctionWithZeroConstantKernelBorder<NEDepthwiseConvolutionLayerKernel>;
+using NEDepthwiseConvolutionLayerNative = NESynthetizeFunctionWithZeroConstantKernelBorder<NEDepthwiseConvolutionLayerNativeKernel>;
 
 // Fixture for NEDepthwiseConvolutionLayerKernel
 template <typename T>
-using NEDepthwiseConvolutionLayerKernelFixture = DepthwiseConvolutionLayerKernelValidationFixture<Tensor, Accessor, NEDepthwiseConvolutionLayer, T>;
+using NEDepthwiseConvolutionLayerNativeFixture = DepthwiseConvolutionLayerNativeValidationFixture<Tensor, Accessor, NEDepthwiseConvolutionLayerNative, T>;
 
 namespace
 {
@@ -52,24 +52,36 @@ RelativeTolerance<float> rel_tolerance_f32(0.001f);
 constexpr float          abs_tolerance_f32(0.0001f);
 
 /** Width values to test - Precommit */
-const auto width_values = framework::dataset::make("width", { 17U, 47U } );
+const auto width_values_precommit = framework::dataset::make("width", { 17U } );
+
+/** Width values to test - Nightly */
+const auto width_values_nightly = framework::dataset::make("width", { 53U, 47U } );
 
 /** Height values to test - Precommit */
-const auto height_values = framework::dataset::make("height", { 19U, 43U } );
+const auto height_values_precommit = framework::dataset::make("height", { 19U } );
+
+/** Height values to test - Nightly */
+const auto height_values_nightly = framework::dataset::make("height", { 39U, 43U } );
 
 /** Channel values to test - Precommit */
-const auto channel_values = framework::dataset::make("channels", { 32U, 128U });
+const auto channel_values_precommit = framework::dataset::make("channels", { 15U });
+
+/** Channel values to test - Nightly */
+const auto channel_values_nightly = framework::dataset::make("channels", { 33U, 19U });
 
 /** Batch values to test - Precommit */
-const auto batch_values = framework::dataset::make("batch", { 1U, 3U });
+const auto batch_values_precommit = framework::dataset::make("batch", { 1U, 2U });
+
+/** Batch values to test - Nightly */
+const auto batch_values_nightly = framework::dataset::make("batch", { 1U, 3U });
 
-/** Kernel size values to test - Precommit */
+/** Kernel size values to test - All */
 const auto kernel_sz_values = framework::dataset::make("kernel_size", { Size2D(3U, 5U), Size2D(5U, 3U) });
 
-/** Depth multiplier values to test - Precommit */
+/** Depth multiplier values to test - All */
 const auto depth_multiplier_values = framework::dataset::make("depth_multiplier", { 1U, 3U });
 
-/** Dilation values to test - Precommit */
+/** Dilation values to test - All */
 const auto dilation_values = framework::dataset::make("dilation", { Size2D(1U, 1U), Size2D(3U, 3U) });
 
 /** Stride values to test - All */
@@ -129,19 +141,19 @@ void validate_configuration(size_t width_value, size_t height_value, size_t chan
     ARM_COMPUTE_EXPECT(dst.info()->is_resizable(), framework::LogLevel::ERRORS);
 
     // Create and configure function
-    NEDepthwiseConvolutionLayer dwc;
+    NEDepthwiseConvolutionLayerNative dwc;
     dwc.configure(&src, &weights, &biases, &dst, conv_info, depth_multiplier_value, dilation_value);
 }
 } // namespace
 
 TEST_SUITE(NEON)
-TEST_SUITE(DepthwiseConvolutionLayer)
+TEST_SUITE(DepthwiseConvolutionLayerNative)
 TEST_SUITE(Float)
 TEST_SUITE(FP32)
-DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, combine(combine(combine(combine(combine(combine(combine(combine(combine(combine(width_values,
-                                                                                                                                           height_values),
-                                                                                                                                           channel_values),
-                                                                                                                                           batch_values),
+DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, combine(combine(combine(combine(combine(combine(combine(combine(combine(combine(width_values_precommit,
+                                                                                                                                           height_values_precommit),
+                                                                                                                                           channel_values_precommit),
+                                                                                                                                           batch_values_precommit),
                                                                                                                                            kernel_sz_values),
                                                                                                                                            depth_multiplier_values),
                                                                                                                                            dilation_values),
@@ -154,11 +166,28 @@ width_value, height_value, channel_value, batch_value, kernel_sz_value, depth_mu
     validate_configuration(width_value, height_value, channel_value, batch_value, kernel_sz_value, depth_multiplier_value, dilation_value, stride_value, padding_valid_value, data_type_value, data_layout_value);
 }
 
-FIXTURE_DATA_TEST_CASE(RunSmall, NEDepthwiseConvolutionLayerKernelFixture<float>, framework::DatasetMode::ALL,
-                combine(combine(combine(combine(combine(combine(combine(combine(combine(combine(width_values,
-                                                                                                height_values),
-                                                                                                channel_values),
-                                                                                                batch_values),
+FIXTURE_DATA_TEST_CASE(RunSmall, NEDepthwiseConvolutionLayerNativeFixture<float>, framework::DatasetMode::ALL,
+                combine(combine(combine(combine(combine(combine(combine(combine(combine(combine(width_values_precommit,
+                                                                                                height_values_precommit),
+                                                                                                channel_values_precommit),
+                                                                                                batch_values_precommit),
+                                                                                                kernel_sz_values),
+                                                                                                depth_multiplier_values),
+                                                                                                dilation_values),
+                                                                                                stride_values),
+                                                                                                padding_valid_values),
+                                                                                                data_type_values),
+                                                                                                data_layout_values))
+{
+    // Validate output
+    validate(Accessor(_target), _reference, rel_tolerance_f32, 0.f, abs_tolerance_f32);
+}
+
+FIXTURE_DATA_TEST_CASE(RunLarge, NEDepthwiseConvolutionLayerNativeFixture<float>, framework::DatasetMode::NIGHTLY,
+                combine(combine(combine(combine(combine(combine(combine(combine(combine(combine(width_values_nightly,
+                                                                                                height_values_nightly),
+                                                                                                channel_values_nightly),
+                                                                                                batch_values_nightly),
                                                                                                 kernel_sz_values),
                                                                                                 depth_multiplier_values),
                                                                                                 dilation_values),
@@ -173,7 +202,7 @@ FIXTURE_DATA_TEST_CASE(RunSmall, NEDepthwiseConvolutionLayerKernelFixture<float>
 
 TEST_SUITE_END() // FP32
 TEST_SUITE_END() // Float
-TEST_SUITE_END() // DepthwiseConvolutionLayer
+TEST_SUITE_END() // DepthwiseConvolutionLayerNative
 TEST_SUITE_END() // NEON
 } // namespace validation
 } // namespace test
diff --git a/tests/validation/fixtures/DepthwiseConvolutionLayerFixture.h b/tests/validation/fixtures/DepthwiseConvolutionLayerFixture.h
index 30b8df9da5..a3ac49eef1 100644
--- a/tests/validation/fixtures/DepthwiseConvolutionLayerFixture.h
+++ b/tests/validation/fixtures/DepthwiseConvolutionLayerFixture.h
@@ -193,7 +193,7 @@ public:
 };
 
 template <typename TensorType, typename AccessorType, typename FunctionType, typename T>
-class DepthwiseConvolutionLayerKernelValidationFixture : public DepthwiseConvolutionLayerValidationGenericFixture<TensorType, AccessorType, FunctionType, T>
+class DepthwiseConvolutionLayerNativeValidationFixture : public DepthwiseConvolutionLayerValidationGenericFixture<TensorType, AccessorType, FunctionType, T>
 {
 public:
     template <typename...>