COMPMID-2754: Add support for QASYMM8_SIGNED in NE kernels/functions.

Kernels/Functions extended support:
- NEBatchToSpaceLayerKernel/NEBatchToSpaceLayer
- NEChannelShuffleLayerKernel/NEChannelShuffleLayer
- NECol2ImKernel/NECol2Im
- NEConvertFullyConnectedWeightsKernel/NEConvertFullyConnectedWeights
- NECopyKernel/NECopy
- NEConvolutionLayerReshapeWeights
- NEDepthToSpaceLayerKernel/NEDepthToSpaceLayer
- NEFlattenLayerKernel/NEFlattenLayer
- NEFillBorderKernel
- NEFullyConnectedLayerReshapeWeights
- NEGatherKernel/NEGather
- NEGEMMInterleave4x4Kernel
- NEGEMMTranspose1xWKernel
- NEIm2ColKernel/NEIm2Col
- NEMemsetKernel
- NEPadLayerKernel/NEPadLayer
- NEPermuteKernel/NEPermute
- NEReverseKernel/NEReverse
- NEReorgLayerKernel/NEReorgLayer
- NEReshapeLayerKernel/NEReshapeLayer
- NESplit
- NESlice
- NEStridedSliceKernel/NEStridedSlice
- NESpaceToBatchLayerKernel/NESpaceToBatchLayer
- NESpaceToDepthLayerKernel/NESpaceToDepthLayerKernel
- NEStackLayerKernel/NEStackLayer
- NETileKernel/NETile
- NETransposeKernel/NETranspose
- NEWidthConcatenateLayerKernel/NEHeightConcatenateLayer
- NEHeightConcatenateLayerKernel/NEHeightConcatenateLayer
- NEDepthConcatenateLayerKernel/NEDepthConcatenateLayer
- NEBathConcatenateLayerKernel/NEBatchConcatenateLayer

Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com>
Change-Id: Ia070332ad4c4dbced2541dc46f7f2f3a86833b65
Reviewed-on: https://review.mlplatform.org/c/2442
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>

37 files changed, 155 insertions, 127 deletions

diff --git a/src/core/NEON/kernels/NEBatchConcatenateLayerKernel.cpp b/src/core/NEON/kernels/NEBatchConcatenateLayerKernel.cpp
index cfa98fb19a..3e250f5d2e 100644
--- a/src/core/NEON/kernels/NEBatchConcatenateLayerKernel.cpp
+++ b/src/core/NEON/kernels/NEBatchConcatenateLayerKernel.cpp
@@ -66,6 +66,16 @@ void batch_concat(const ITensor *in, ITensor *out, int batch_offset, const Windo
         },
         input, output);
     }
+    else if(dt == DataType::QASYMM8_SIGNED && input_qinfo != output_qinfo)
+    {
+        execute_window_loop(window, [&](const Coordinates &)
+        {
+            const auto in_ptr  = reinterpret_cast<const int8_t *>(input_ptr + input.offset());
+            const auto out_ptr = reinterpret_cast<int8_t *>(output_ptr + output.offset());
+            vst1q_s8(out_ptr, vquantize_signed(vdequantize(vld1q_s8(in_ptr), input_qinfo), output_qinfo));
+        },
+        input, output);
+    }
     else
     {
         execute_window_loop(window, [&](const Coordinates &)
@@ -102,10 +112,7 @@ Status validate_arguments(const ITensorInfo *input, unsigned int batch_offset, c
 {
     ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input, output);
     //Note: ARM_COMPUTE_RETURN_ERROR_ON_CPU_F16_UNSUPPORTED(input) is not needed here as this kernel doesn't use NEON FP16 instructions.
-    ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::U8, DataType::S8, DataType::QASYMM8,
-                                                         DataType::U16, DataType::S16,
-                                                         DataType::U32, DataType::S32,
-                                                         DataType::F16, DataType::F32);
+    ARM_COMPUTE_RETURN_ERROR_ON(input->data_type() == DataType::UNKNOWN);
     ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input, output);
 
     ARM_COMPUTE_RETURN_ERROR_ON(input->dimension(Window::DimX) != output->dimension(Window::DimX));
@@ -138,6 +145,7 @@ void NEBatchConcatenateLayerKernel::configure(const ITensor *input, unsigned int
         case DataType::S8:
         case DataType::U8:
         case DataType::QASYMM8:
+        case DataType::QASYMM8_SIGNED:
             _func = &batch_concat<uint8_t>;
             break;
         case DataType::S16:
diff --git a/src/core/NEON/kernels/NEBatchToSpaceLayerKernel.cpp b/src/core/NEON/kernels/NEBatchToSpaceLayerKernel.cpp
index 6211abcad0..d167cbb05d 100644
--- a/src/core/NEON/kernels/NEBatchToSpaceLayerKernel.cpp
+++ b/src/core/NEON/kernels/NEBatchToSpaceLayerKernel.cpp
@@ -43,6 +43,7 @@ Status validate_arguments(const ITensorInfo *input, const ITensorInfo *block_inf
     ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input, block_info, output);
     ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(block_info, 1, DataType::S32);
     ARM_COMPUTE_RETURN_ERROR_ON(input->num_dimensions() > 4);
+    ARM_COMPUTE_RETURN_ERROR_ON(input->data_type() == DataType::UNKNOWN);
 
     // Validate output if initialized
     if(output->total_size() != 0)
diff --git a/src/core/NEON/kernels/NEChannelShuffleLayerKernel.cpp b/src/core/NEON/kernels/NEChannelShuffleLayerKernel.cpp
index b2b0dbd789..45f6fd99fe 100644
--- a/src/core/NEON/kernels/NEChannelShuffleLayerKernel.cpp
+++ b/src/core/NEON/kernels/NEChannelShuffleLayerKernel.cpp
@@ -39,12 +39,7 @@ namespace
 Status validate_arguments(const ITensorInfo *input, const ITensorInfo *output, unsigned int num_groups)
 {
     // Note: ARM_COMPUTE_RETURN_ERROR_ON_CPU_F16_UNSUPPORTED(input) is not needed here as this kernel doesn't use NEON FP16 instructions.
-    ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input,
-                                                         1,
-                                                         DataType::U8, DataType::S8, DataType::QASYMM8,
-                                                         DataType::U16, DataType::S16,
-                                                         DataType::U32, DataType::S32,
-                                                         DataType::F16, DataType::F32);
+    ARM_COMPUTE_RETURN_ERROR_ON(input->data_type() == DataType::UNKNOWN);
     ARM_COMPUTE_RETURN_ERROR_ON_DATA_LAYOUT_NOT_IN(input, DataLayout::NCHW, DataLayout::NHWC);
 
     const unsigned int channels = input->dimension(get_data_layout_dimension_index(input->data_layout(), DataLayoutDimension::CHANNEL));
diff --git a/src/core/NEON/kernels/NECol2ImKernel.cpp b/src/core/NEON/kernels/NECol2ImKernel.cpp
index cea8782354..4f1b1d8ca2 100644
--- a/src/core/NEON/kernels/NECol2ImKernel.cpp
+++ b/src/core/NEON/kernels/NECol2ImKernel.cpp
@@ -43,6 +43,7 @@ namespace
 Status validate_arguments(const ITensorInfo *input, const ITensorInfo *output, const Size2D &convolved_dims)
 {
     //Note: ARM_COMPUTE_RETURN_ERROR_ON_CPU_F16_UNSUPPORTED(input) is not needed here as this kernel doesn't use NEON FP16 instructions.
+    ARM_COMPUTE_RETURN_ERROR_ON(input->data_type() == DataType::UNKNOWN);
 
     // Validate configured output
     if(output->total_size() != 0)
diff --git a/src/core/NEON/kernels/NEConvertFullyConnectedWeightsKernel.cpp b/src/core/NEON/kernels/NEConvertFullyConnectedWeightsKernel.cpp
index b6d166d30e..ab6636018f 100644
--- a/src/core/NEON/kernels/NEConvertFullyConnectedWeightsKernel.cpp
+++ b/src/core/NEON/kernels/NEConvertFullyConnectedWeightsKernel.cpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2018 ARM Limited.
+ * Copyright (c) 2018-2019 ARM Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -26,8 +26,8 @@
 #include "arm_compute/core/Helpers.h"
 #include "arm_compute/core/Types.h"
 
-using namespace arm_compute;
-
+namespace arm_compute
+{
 NEConvertFullyConnectedWeightsKernel::NEConvertFullyConnectedWeightsKernel()
     : _input(nullptr), _output(nullptr), _factor1(0), _factor2(0)
 {
@@ -66,12 +66,9 @@ void NEConvertFullyConnectedWeightsKernel::configure(const ITensor *input, ITens
 Status NEConvertFullyConnectedWeightsKernel::validate(const ITensorInfo *input, const ITensorInfo *output, const TensorShape &original_input_shape,
                                                       DataLayout data_layout)
 {
+    ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input);
     //Note: ARM_COMPUTE_RETURN_ERROR_ON_CPU_F16_UNSUPPORTED(input) is not needed here as this kernel doesn't use NEON FP16 instructions.
-    ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1,
-                                                         DataType::U8, DataType::S8, DataType::QASYMM8,
-                                                         DataType::U16, DataType::S16,
-                                                         DataType::U32, DataType::S32,
-                                                         DataType::F16, DataType::F32);
+    ARM_COMPUTE_RETURN_ERROR_ON(input->data_type() == DataType::UNKNOWN);
     ARM_COMPUTE_RETURN_ERROR_ON(input->num_dimensions() != 2);
     ARM_COMPUTE_RETURN_ERROR_ON(input->dimension(1) != original_input_shape.total_size_lower(3));
     ARM_COMPUTE_RETURN_ERROR_ON(data_layout == DataLayout::UNKNOWN);
@@ -124,3 +121,4 @@ void NEConvertFullyConnectedWeightsKernel::run(const Window &window, const Threa
             break;
     }
 }
+} // namespace arm_compute
diff --git a/src/core/NEON/kernels/NECopyKernel.cpp b/src/core/NEON/kernels/NECopyKernel.cpp
index 83f3dded4f..6bf49549e2 100644
--- a/src/core/NEON/kernels/NECopyKernel.cpp
+++ b/src/core/NEON/kernels/NECopyKernel.cpp
@@ -38,6 +38,7 @@ namespace
 Status validate_arguments(const ITensorInfo *input, const ITensorInfo *output, const PaddingList &padding = PaddingList())
 {
     ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input, output);
+    ARM_COMPUTE_RETURN_ERROR_ON(input->data_type() == DataType::UNKNOWN);
     ARM_COMPUTE_RETURN_ERROR_ON(padding.size() > 4);
 
     // Validate output if initialized
diff --git a/src/core/NEON/kernels/NEDepthConcatenateLayerKernel.cpp b/src/core/NEON/kernels/NEDepthConcatenateLayerKernel.cpp
index 4377006f28..56ab11415c 100644
--- a/src/core/NEON/kernels/NEDepthConcatenateLayerKernel.cpp
+++ b/src/core/NEON/kernels/NEDepthConcatenateLayerKernel.cpp
@@ -66,6 +66,16 @@ void depth_concat(const ITensor *in, ITensor *out, int depth_offset, const Windo
         },
         input, output);
     }
+    else if(dt == DataType::QASYMM8_SIGNED && input_qinfo != output_qinfo)
+    {
+        execute_window_loop(window, [&](const Coordinates &)
+        {
+            const auto in_ptr  = reinterpret_cast<const int8_t *>(input_ptr + input.offset());
+            const auto out_ptr = reinterpret_cast<int8_t *>(output_ptr + output.offset());
+            vst1q_s8(out_ptr, vquantize_signed(vdequantize(vld1q_s8(in_ptr), input_qinfo), output_qinfo));
+        },
+        input, output);
+    }
     else
     {
         execute_window_loop(window, [&](const Coordinates &)
@@ -102,7 +112,7 @@ Status validate_arguments(const ITensorInfo *input, unsigned int depth_offset, c
 {
     ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input, output);
     //Note: ARM_COMPUTE_RETURN_ERROR_ON_CPU_F16_UNSUPPORTED(input) is not needed here as this kernel doesn't use NEON FP16 instructions.
-    ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::QASYMM8, DataType::F16, DataType::F32);
+    ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::QASYMM8, DataType::QASYMM8_SIGNED, DataType::F16, DataType::F32);
     ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input, output);
 
     ARM_COMPUTE_RETURN_ERROR_ON(input->dimension(Window::DimX) != output->dimension(Window::DimX));
@@ -134,6 +144,9 @@ void NEDepthConcatenateLayerKernel::configure(const ITensor *input, unsigned int
         case DataType::QASYMM8:
             _func = &depth_concat<uint8_t>;
             break;
+        case DataType::QASYMM8_SIGNED:
+            _func = &depth_concat<int8_t>;
+            break;
         case DataType::F16:
             _func = &depth_concat<uint16_t>;
             break;
diff --git a/src/core/NEON/kernels/NEDepthToSpaceLayerKernel.cpp b/src/core/NEON/kernels/NEDepthToSpaceLayerKernel.cpp
index df631c3c03..b51d0d66c5 100644
--- a/src/core/NEON/kernels/NEDepthToSpaceLayerKernel.cpp
+++ b/src/core/NEON/kernels/NEDepthToSpaceLayerKernel.cpp
@@ -41,6 +41,7 @@ namespace
 Status validate_arguments(const ITensorInfo *input, const ITensorInfo *output, int32_t block_shape)
 {
     ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input, output);
+    ARM_COMPUTE_RETURN_ERROR_ON(input->data_type() == DataType::UNKNOWN);
     ARM_COMPUTE_RETURN_ERROR_ON(input->num_dimensions() > 4);
     ARM_COMPUTE_RETURN_ERROR_ON(block_shape < 2);
 
diff --git a/src/core/NEON/kernels/NEFillBorderKernel.cpp b/src/core/NEON/kernels/NEFillBorderKernel.cpp
index 13db1659ce..75d46c61d8 100644
--- a/src/core/NEON/kernels/NEFillBorderKernel.cpp
+++ b/src/core/NEON/kernels/NEFillBorderKernel.cpp
@@ -97,11 +97,9 @@ NEFillBorderKernel::NEFillBorderKernel()
 
 void NEFillBorderKernel::configure(ITensor *tensor, BorderSize border_size, BorderMode border_mode, const PixelValue &constant_border_value)
 {
+    ARM_COMPUTE_ERROR_ON_NULLPTR(tensor);
     //Note: ARM_COMPUTE_RETURN_ERROR_ON_CPU_F16_UNSUPPORTED(input) is not needed here as this kernel doesn't use NEON FP16 instructions.
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(tensor, 1, DataType::U8, DataType::QASYMM8,
-                                                  DataType::U16, DataType::S16,
-                                                  DataType::U32, DataType::S32,
-                                                  DataType::F16, DataType::F32);
+    ARM_COMPUTE_ERROR_ON(tensor->info()->data_type() == DataType::UNKNOWN);
 
     _tensor                = tensor;
     _border_size           = border_size;
diff --git a/src/core/NEON/kernels/NEFlattenLayerKernel.cpp b/src/core/NEON/kernels/NEFlattenLayerKernel.cpp
index 4840a95540..a48601f7b0 100644
--- a/src/core/NEON/kernels/NEFlattenLayerKernel.cpp
+++ b/src/core/NEON/kernels/NEFlattenLayerKernel.cpp
@@ -42,10 +42,8 @@ namespace
 {
 Status validate_arguments(const ITensorInfo *input, const ITensorInfo *output)
 {
-    ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::U8, DataType::S8, DataType::QASYMM8,
-                                                         DataType::U16, DataType::S16,
-                                                         DataType::U32, DataType::S32,
-                                                         DataType::F16, DataType::F32);
+    ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(output);
+    ARM_COMPUTE_RETURN_ERROR_ON(input->data_type() == DataType::UNKNOWN);
     // Note: ARM_COMPUTE_RETURN_ERROR_ON_CPU_F16_UNSUPPORTED(input) is not needed here as this kernel doesn't use NEON FP16 instructions.
     ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(output);
 
diff --git a/src/core/NEON/kernels/NEGEMMInterleave4x4Kernel.cpp b/src/core/NEON/kernels/NEGEMMInterleave4x4Kernel.cpp
index a9c04824ae..8f73bdb3a3 100644
--- a/src/core/NEON/kernels/NEGEMMInterleave4x4Kernel.cpp
+++ b/src/core/NEON/kernels/NEGEMMInterleave4x4Kernel.cpp
@@ -44,6 +44,7 @@ namespace
 {
 Status validate_arguments(const ITensorInfo *input, const ITensorInfo *output)
 {
+    ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input);
     //Note: ARM_COMPUTE_RETURN_ERROR_ON_CPU_F16_UNSUPPORTED(input) is not needed here as this kernel doesn't use NEON FP16 instructions.
     ARM_COMPUTE_RETURN_ERROR_ON(input->data_type() == DataType::UNKNOWN);
     ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input, output);
diff --git a/src/core/NEON/kernels/NEGEMMTranspose1xWKernel.cpp b/src/core/NEON/kernels/NEGEMMTranspose1xWKernel.cpp
index ea3d32e628..88104f7297 100644
--- a/src/core/NEON/kernels/NEGEMMTranspose1xWKernel.cpp
+++ b/src/core/NEON/kernels/NEGEMMTranspose1xWKernel.cpp
@@ -54,10 +54,9 @@ TensorShape get_output_shape(const ITensorInfo *input)
 
 Status validate_arguments(const ITensorInfo *input, const ITensorInfo *output)
 {
+    ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input);
+    ARM_COMPUTE_RETURN_ERROR_ON(input->data_type() == DataType::UNKNOWN);
     //Note: ARM_COMPUTE_RETURN_ERROR_ON_CPU_F16_UNSUPPORTED(input) is not needed here as this kernel doesn't use NEON FP16 instructions.
-    ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::QASYMM8, DataType::QSYMM8_PER_CHANNEL, DataType::U8, DataType::S8,
-                                                         DataType::U16, DataType::S16, DataType::U32, DataType::S32,
-                                                         DataType::F16, DataType::F32);
     ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input, output);
 
     if(output->total_size() != 0)
diff --git a/src/core/NEON/kernels/NEGatherKernel.cpp b/src/core/NEON/kernels/NEGatherKernel.cpp
index 1e027b7292..0a7a8dfc21 100644
--- a/src/core/NEON/kernels/NEGatherKernel.cpp
+++ b/src/core/NEON/kernels/NEGatherKernel.cpp
@@ -52,6 +52,32 @@ void validate_indices(const ITensor *indices)
     }
 }
 
+Status validate_arguments(const ITensorInfo *input, const ITensorInfo *indices, const ITensorInfo *output, int axis)
+{
+    ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input, indices, output);
+    ARM_COMPUTE_RETURN_ERROR_ON(indices->num_dimensions() > 1);
+    ARM_COMPUTE_RETURN_ERROR_ON(input->num_dimensions() > 4);
+
+    if(axis < 0)
+    {
+        axis += input->num_dimensions();
+    }
+
+    ARM_COMPUTE_RETURN_ERROR_ON(0 > axis || axis >= static_cast<int32_t>(input->num_dimensions()));
+    ARM_COMPUTE_RETURN_ERROR_ON(input->data_type() == DataType::UNKNOWN);
+
+    if(output->total_size() != 0)
+    {
+        ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input, output);
+        ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_QUANTIZATION_INFO(input, output);
+        TensorShape output_shape = arm_compute::misc::shape_calculator::compute_gather_shape(input->tensor_shape(), indices->tensor_shape(), axis);
+        ARM_COMPUTE_RETURN_ERROR_ON(output_shape.total_size() != output->tensor_shape().total_size());
+    }
+
+    ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(indices, 1, DataType::U32, DataType::S32);
+
+    return Status{};
+}
 } // namespace
 
 NEGatherKernel::NEGatherKernel()
@@ -107,9 +133,7 @@ void NEGatherKernel::gather_n_axis(const Window &window, const ThreadInfo &info)
 void NEGatherKernel::configure(const ITensor *input, const ITensor *indices, ITensor *output, int axis)
 {
     ARM_COMPUTE_ERROR_ON_NULLPTR(input, output, indices);
-    ARM_COMPUTE_ERROR_ON(indices->info()->num_dimensions() != 1);
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(indices, 1, DataType::U32, DataType::S32);
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::U8, DataType::S8, DataType::QASYMM8, DataType::U16, DataType::S16, DataType::U32, DataType::S32, DataType::F16, DataType::F32);
+    ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(input->info(), indices->info(), output->info(), axis));
 
     _input   = input;
     _indices = indices;
@@ -154,7 +178,7 @@ void NEGatherKernel::configure(const ITensor *input, const ITensor *indices, ITe
     }
     // Output auto initialization if not yet initialized
     TensorShape output_shape = arm_compute::misc::shape_calculator::compute_gather_shape(input->info()->tensor_shape(), indices->info()->tensor_shape(), _axis);
-    auto_init_if_empty(*output->info(), output_shape, 1, input->info()->data_type());
+    auto_init_if_empty(*output->info(), input->info()->clone()->set_tensor_shape(output_shape));
 
     // Create window
     Window win = calculate_max_window(*output->info(), Steps());
@@ -165,31 +189,7 @@ void NEGatherKernel::configure(const ITensor *input, const ITensor *indices, ITe
 
 Status NEGatherKernel::validate(const ITensorInfo *input, const ITensorInfo *indices, const ITensorInfo *output, int axis)
 {
-    ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input, indices, output);
-    ARM_COMPUTE_RETURN_ERROR_ON(indices->num_dimensions() > 1);
-    ARM_COMPUTE_RETURN_ERROR_ON(input->num_dimensions() > 4);
-
-    if(axis < 0)
-    {
-        axis += input->num_dimensions();
-    }
-
-    ARM_COMPUTE_RETURN_ERROR_ON(0 > axis || axis >= static_cast<int32_t>(input->num_dimensions()));
-    ARM_COMPUTE_RETURN_ERROR_ON_CPU_F16_UNSUPPORTED(input);
-    ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::U8, DataType::S8, DataType::QASYMM8,
-                                                         DataType::U16, DataType::S16,
-                                                         DataType::U32, DataType::S32, DataType::F16, DataType::F32);
-
-    if(output->total_size() != 0)
-    {
-        ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input, output);
-        ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_QUANTIZATION_INFO(input, output);
-        TensorShape output_shape = arm_compute::misc::shape_calculator::compute_gather_shape(input->tensor_shape(), indices->tensor_shape(), axis);
-        ARM_COMPUTE_RETURN_ERROR_ON(output_shape.total_size() != output->tensor_shape().total_size());
-    }
-
-    ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(indices, 1, DataType::U32, DataType::S32);
-
+    ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(input, indices, output, axis));
     return Status{};
 }
 
diff --git a/src/core/NEON/kernels/NEHeightConcatenateLayerKernel.cpp b/src/core/NEON/kernels/NEHeightConcatenateLayerKernel.cpp
index 18be1d5bb7..4f7ae8c0b1 100644
--- a/src/core/NEON/kernels/NEHeightConcatenateLayerKernel.cpp
+++ b/src/core/NEON/kernels/NEHeightConcatenateLayerKernel.cpp
@@ -36,8 +36,8 @@
 
 #include <cstdint>
 
-using namespace arm_compute;
-
+namespace arm_compute
+{
 namespace
 {
 std::pair<Status, Window> validate_and_configure_window(ITensorInfo *input, ITensorInfo *output)
@@ -58,10 +58,7 @@ Status validate_arguments(const ITensorInfo *input, unsigned int height_offset,
 {
     ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input, output);
     // Note: ARM_COMPUTE_RETURN_ERROR_ON_CPU_F16_UNSUPPORTED(input) is not needed here as this kernel doesn't use NEON FP16 instructions.
-    ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1,
-                                                         DataType::U8, DataType::S8, DataType::QASYMM8,
-                                                         DataType::U16, DataType::S16, DataType::F16,
-                                                         DataType::U32, DataType::F32);
+    ARM_COMPUTE_RETURN_ERROR_ON(input->data_type() == DataType::UNKNOWN);
     ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input, output);
     ARM_COMPUTE_RETURN_ERROR_ON(input->dimension(Window::DimX) != output->dimension(Window::DimX));
     ARM_COMPUTE_RETURN_ERROR_ON(input->dimension(Window::DimY) + height_offset > output->dimension(Window::DimY));
@@ -128,6 +125,15 @@ void NEHeightConcatenateLayerKernel::run(const Window &window, const ThreadInfo
         },
         input, output);
     }
+    else if(dt == DataType::QASYMM8_SIGNED && input_qinfo != output_qinfo)
+    {
+        execute_window_loop(window, [&](const Coordinates &)
+        {
+            vst1q_s8(reinterpret_cast<int8_t *>(output_ptr + output.offset()),
+                     vquantize_signed(vdequantize(vld1q_s8(reinterpret_cast<int8_t *>(input.ptr())), input_qinfo), output_qinfo));
+        },
+        input, output);
+    }
     else
     {
         execute_window_loop(window, [&](const Coordinates &)
@@ -140,3 +146,4 @@ void NEHeightConcatenateLayerKernel::run(const Window &window, const ThreadInfo
         input, output);
     }
 }
+} // namespace arm_compute
diff --git a/src/core/NEON/kernels/NEPadLayerKernel.cpp b/src/core/NEON/kernels/NEPadLayerKernel.cpp
index 88a1c2ec83..07229af358 100644
--- a/src/core/NEON/kernels/NEPadLayerKernel.cpp
+++ b/src/core/NEON/kernels/NEPadLayerKernel.cpp
@@ -38,6 +38,8 @@ namespace
 {
 Status validate_arguments(const ITensorInfo *input, const ITensorInfo *output, const PaddingList &paddings, const PaddingMode mode)
 {
+    ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input);
+    ARM_COMPUTE_RETURN_ERROR_ON(input->data_type() == DataType::UNKNOWN);
     ARM_COMPUTE_RETURN_ERROR_ON_MSG(mode != PaddingMode::CONSTANT, "Only constant padding mode is supported");
     ARM_COMPUTE_RETURN_ERROR_ON_MSG(paddings.size() > 4, "Padding list bigger than 4 dimensions");
     if(output->total_size() != 0)
diff --git a/src/core/NEON/kernels/NEPermuteKernel.cpp b/src/core/NEON/kernels/NEPermuteKernel.cpp
index 897b764b45..2c0db769f5 100644
--- a/src/core/NEON/kernels/NEPermuteKernel.cpp
+++ b/src/core/NEON/kernels/NEPermuteKernel.cpp
@@ -91,6 +91,7 @@ inline bool is_permutation_supported(const PermutationVector &v)
 
 Status validate_arguments(const ITensorInfo *input, const ITensorInfo *output, const PermutationVector &perm)
 {
+    ARM_COMPUTE_RETURN_ERROR_ON(input->data_type() == DataType::UNKNOWN);
     ARM_COMPUTE_RETURN_ERROR_ON_MSG(!is_permutation_supported(perm), "PermutationVector not supported.");
 
     const TensorShape output_shape = misc::shape_calculator::compute_permutation_output_shape(*input, perm);
diff --git a/src/core/NEON/kernels/NEReorgLayerKernel.cpp b/src/core/NEON/kernels/NEReorgLayerKernel.cpp
index ece5aa431c..f8a8732c64 100644
--- a/src/core/NEON/kernels/NEReorgLayerKernel.cpp
+++ b/src/core/NEON/kernels/NEReorgLayerKernel.cpp
@@ -41,11 +41,7 @@ namespace
 Status validate_arguments(const ITensorInfo *input, const ITensorInfo *output, int32_t stride)
 {
     //Note: ARM_COMPUTE_RETURN_ERROR_ON_CPU_F16_UNSUPPORTED(input) is not needed here as this kernel doesn't use NEON FP16 instructions.
-    ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1,
-                                                         DataType::U8, DataType::S8, DataType::QASYMM8,
-                                                         DataType::U16, DataType::S16,
-                                                         DataType::U32, DataType::S32,
-                                                         DataType::F16, DataType::F32);
+    ARM_COMPUTE_RETURN_ERROR_ON(input->data_type() == DataType::UNKNOWN);
     ARM_COMPUTE_RETURN_ERROR_ON(input->data_layout() == DataLayout::UNKNOWN);
 
     const size_t idx_width  = get_data_layout_dimension_index(input->data_layout(), DataLayoutDimension::WIDTH);
diff --git a/src/core/NEON/kernels/NEReshapeLayerKernel.cpp b/src/core/NEON/kernels/NEReshapeLayerKernel.cpp
index 649fba30d4..53fcfd724d 100644
--- a/src/core/NEON/kernels/NEReshapeLayerKernel.cpp
+++ b/src/core/NEON/kernels/NEReshapeLayerKernel.cpp
@@ -44,8 +44,7 @@ Status validate_arguments(const ITensorInfo *input, const ITensorInfo *output)
 {
     ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input, output);
     // Note: ARM_COMPUTE_RETURN_ERROR_ON_CPU_F16_UNSUPPORTED(input) is not needed here as this kernel doesn't use NEON FP16 instructions.
-    ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::U8, DataType::S8, DataType::QASYMM8, DataType::U16, DataType::S16,
-                                                         DataType::U32, DataType::S32, DataType::F16, DataType::F32);
+    ARM_COMPUTE_RETURN_ERROR_ON(input->data_type() == DataType::UNKNOWN);
 
     ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input, output);
     ARM_COMPUTE_RETURN_ERROR_ON(input->tensor_shape().total_size() != output->tensor_shape().total_size());
@@ -107,6 +106,7 @@ void NEReshapeLayerKernel::run(const Window &window, const ThreadInfo &info)
         case DataType::U8:
         case DataType::S8:
         case DataType::QASYMM8:
+        case DataType::QASYMM8_SIGNED:
             reshape_tensor<uint8_t>(window, _input, _output);
             break;
         case DataType::U16:
diff --git a/src/core/NEON/kernels/NEReverseKernel.cpp b/src/core/NEON/kernels/NEReverseKernel.cpp
index 99328deecd..2f584164dc 100644
--- a/src/core/NEON/kernels/NEReverseKernel.cpp
+++ b/src/core/NEON/kernels/NEReverseKernel.cpp
@@ -49,10 +49,7 @@ Status validate_arguments(const ITensorInfo *input, const ITensorInfo *output, c
 {
     ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input, output, axis);
     ARM_COMPUTE_RETURN_ERROR_ON_CPU_F16_UNSUPPORTED(input);
-    ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::U8, DataType::S8, DataType::QASYMM8,
-                                                         DataType::U16, DataType::S16,
-                                                         DataType::U32, DataType::S32,
-                                                         DataType::F16, DataType::F32);
+    ARM_COMPUTE_RETURN_ERROR_ON(input->data_type() == DataType::UNKNOWN);
     ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(axis, 1, DataType::U32);
     ARM_COMPUTE_RETURN_ERROR_ON_MSG(axis->num_dimensions() > 1, "Axis must be a 1D tensor");
     ARM_COMPUTE_RETURN_ERROR_ON_MSG(axis->dimension(0) > 4, "Only up to 4 dimensions can be reversed");
@@ -200,6 +197,7 @@ void NEReverseKernel::run(const Window &window, const ThreadInfo &info)
             run_reverse<uint16_t>(window, _input, _axis, _output);
             break;
         case DataType::QASYMM8:
+        case DataType::QASYMM8_SIGNED:
         case DataType::U8:
         case DataType::S8:
             run_reverse<uint8_t>(window, _input, _axis, _output);
diff --git a/src/core/NEON/kernels/NESpaceToBatchLayerKernel.cpp b/src/core/NEON/kernels/NESpaceToBatchLayerKernel.cpp
index f51c3940b7..ba3377f13b 100644
--- a/src/core/NEON/kernels/NESpaceToBatchLayerKernel.cpp
+++ b/src/core/NEON/kernels/NESpaceToBatchLayerKernel.cpp
@@ -41,6 +41,7 @@ namespace
 Status validate_arguments(const ITensorInfo *input, const ITensorInfo *block_info, const ITensorInfo *padddings, const ITensorInfo *output)
 {
     ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input, block_info, padddings, output);
+    ARM_COMPUTE_RETURN_ERROR_ON(input->data_type() == DataType::UNKNOWN);
     ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(block_info, 1, DataType::S32);
     ARM_COMPUTE_RETURN_ERROR_ON(input->num_dimensions() > 4);
     ARM_COMPUTE_RETURN_ERROR_ON(block_info->num_dimensions() > 1);
@@ -62,6 +63,7 @@ Status validate_arguments_static(const ITensorInfo *input, const int block_shape
                                  const ITensorInfo *output)
 {
     ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input, output);
+    ARM_COMPUTE_RETURN_ERROR_ON(input->data_type() == DataType::UNKNOWN);
     ARM_COMPUTE_RETURN_ERROR_ON(block_shape_x < 1 || block_shape_y < 1);
     ARM_COMPUTE_RETURN_ERROR_ON(input->num_dimensions() > 4);
 
diff --git a/src/core/NEON/kernels/NESpaceToDepthLayerKernel.cpp b/src/core/NEON/kernels/NESpaceToDepthLayerKernel.cpp
index 4803365013..b2ce63e549 100644
--- a/src/core/NEON/kernels/NESpaceToDepthLayerKernel.cpp
+++ b/src/core/NEON/kernels/NESpaceToDepthLayerKernel.cpp
@@ -41,6 +41,7 @@ namespace
 Status validate_arguments(const ITensorInfo *input, const ITensorInfo *output, int32_t block_shape)
 {
     ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input, output);
+    ARM_COMPUTE_RETURN_ERROR_ON(input->data_type() == DataType::UNKNOWN);
     ARM_COMPUTE_RETURN_ERROR_ON(input->num_dimensions() > 4);
 
     ARM_COMPUTE_RETURN_ERROR_ON(block_shape < 1);
diff --git a/src/core/NEON/kernels/NEStackLayerKernel.cpp b/src/core/NEON/kernels/NEStackLayerKernel.cpp
index 3447d59bcc..5deca9e595 100644
--- a/src/core/NEON/kernels/NEStackLayerKernel.cpp
+++ b/src/core/NEON/kernels/NEStackLayerKernel.cpp
@@ -45,9 +45,7 @@ Status validate_arguments(const ITensorInfo *input, unsigned int axis, unsigned
 {
     ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input, output);
     // Note: ARM_COMPUTE_RETURN_ERROR_ON_CPU_F16_UNSUPPORTED(input) is not needed here as this kernel doesn't use NEON FP16 instructions.
-    ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::QASYMM8, DataType::U8, DataType::S8,
-                                                         DataType::U16, DataType::S16, DataType::U32, DataType::S32,
-                                                         DataType::F16, DataType::F32);
+    ARM_COMPUTE_RETURN_ERROR_ON(input->data_type() == DataType::UNKNOWN);
     ARM_COMPUTE_RETURN_ERROR_ON(idx_input >= num_tensors);
     ARM_COMPUTE_RETURN_ERROR_ON(axis > input->num_dimensions());
     ARM_COMPUTE_RETURN_ERROR_ON(input->num_dimensions() > 4);
diff --git a/src/core/NEON/kernels/NEStridedSliceKernel.cpp b/src/core/NEON/kernels/NEStridedSliceKernel.cpp
index 2de49c6864..15f786a521 100644
--- a/src/core/NEON/kernels/NEStridedSliceKernel.cpp
+++ b/src/core/NEON/kernels/NEStridedSliceKernel.cpp
@@ -43,11 +43,7 @@ Status validate_arguments(const ITensorInfo *input, const ITensorInfo *output,
                           int32_t begin_mask, int32_t end_mask, int32_t shrink_axis_mask)
 {
     ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input, output);
-    ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1,
-                                                         DataType::U8, DataType::S8, DataType::QASYMM8,
-                                                         DataType::U16, DataType::S16, DataType::QASYMM16, DataType::QSYMM16,
-                                                         DataType::U32, DataType::S32,
-                                                         DataType::F16, DataType::F32);
+    ARM_COMPUTE_RETURN_ERROR_ON(input->data_type() == DataType::UNKNOWN);
 
     ARM_COMPUTE_RETURN_ERROR_ON(input->tensor_shape().num_dimensions() > 4);
     ARM_COMPUTE_RETURN_ERROR_ON(starts.num_dimensions() > input->num_dimensions());
diff --git a/src/core/NEON/kernels/NETileKernel.cpp b/src/core/NEON/kernels/NETileKernel.cpp
index dbeacfad94..98f66e8391 100644
--- a/src/core/NEON/kernels/NETileKernel.cpp
+++ b/src/core/NEON/kernels/NETileKernel.cpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2018 ARM Limited.
+ * Copyright (c) 2018-2019 ARM Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -38,6 +38,7 @@ namespace
 Status validate_arguments(const ITensorInfo *input, const ITensorInfo *output, const Multiples &multiples)
 {
     ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input, output);
+    ARM_COMPUTE_RETURN_ERROR_ON(input->data_type() == DataType::UNKNOWN);
     ARM_COMPUTE_RETURN_ERROR_ON(multiples.size() > 4);
     ARM_COMPUTE_RETURN_ERROR_ON(multiples.empty());
     ARM_COMPUTE_RETURN_ERROR_ON(std::any_of(multiples.begin(), multiples.end(), [](uint32_t e)
diff --git a/src/core/NEON/kernels/NETransposeKernel.cpp b/src/core/NEON/kernels/NETransposeKernel.cpp
index a0a8b82494..6a8e6ffeb5 100644
--- a/src/core/NEON/kernels/NETransposeKernel.cpp
+++ b/src/core/NEON/kernels/NETransposeKernel.cpp
@@ -75,10 +75,9 @@ unsigned int num_elems_processed(size_t element_size)
 
 Status validate_arguments(const ITensorInfo *input, const ITensorInfo *output)
 {
+    ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input);
     //Note: ARM_COMPUTE_RETURN_ERROR_ON_CPU_F16_UNSUPPORTED(input) is not needed here as this kernel doesn't use NEON FP16 instructions.
-    ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::U8, DataType::S8, DataType::QASYMM8, DataType::U16, DataType::S16, DataType::U32, DataType::S32,
-                                                         DataType::F16,
-                                                         DataType::F32);
+    ARM_COMPUTE_RETURN_ERROR_ON(input->data_type() == DataType::UNKNOWN);
 
     if(output->total_size() != 0)
     {
diff --git a/src/core/NEON/kernels/NEWidthConcatenateLayerKernel.cpp b/src/core/NEON/kernels/NEWidthConcatenateLayerKernel.cpp
index cafa20a1bd..e164a38708 100644
--- a/src/core/NEON/kernels/NEWidthConcatenateLayerKernel.cpp
+++ b/src/core/NEON/kernels/NEWidthConcatenateLayerKernel.cpp
@@ -36,8 +36,8 @@
 
 #include <cstdint>
 
-using namespace arm_compute;
-
+namespace arm_compute
+{
 namespace
 {
 std::pair<Status, Window> validate_and_configure_window(ITensorInfo *input, unsigned int width_offset, ITensorInfo *output)
@@ -58,10 +58,7 @@ Status validate_arguments(const ITensorInfo *input, unsigned int width_offset, c
 {
     ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input, output);
     // Note: ARM_COMPUTE_RETURN_ERROR_ON_CPU_F16_UNSUPPORTED(input) is not needed here as this kernel doesn't use NEON FP16 instructions.
-    ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1,
-                                                         DataType::U8, DataType::S8, DataType::QASYMM8,
-                                                         DataType::U16, DataType::S16, DataType::F16,
-                                                         DataType::U32, DataType::S32, DataType::F32);
+    ARM_COMPUTE_RETURN_ERROR_ON(input->data_type() == DataType::UNKNOWN);
     ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input, output);
     ARM_COMPUTE_RETURN_ERROR_ON(input->dimension(0) + width_offset > output->dimension(0));
 
@@ -128,6 +125,15 @@ void NEWidthConcatenateLayerKernel::run(const Window &window, const ThreadInfo &
         },
         input, output);
     }
+    else if(dt == DataType::QASYMM8_SIGNED && input_qinfo != output_qinfo)
+    {
+        execute_window_loop(window, [&](const Coordinates &)
+        {
+            vst1q_s8(reinterpret_cast<int8_t *>(output_ptr + output.offset()),
+                     vquantize_signed(vdequantize(vld1q_s8(reinterpret_cast<int8_t *>(input.ptr())), input_qinfo), output_qinfo));
+        },
+        input, output);
+    }
     else
     {
         execute_window_loop(window, [&](const Coordinates &)
@@ -140,3 +146,4 @@ void NEWidthConcatenateLayerKernel::run(const Window &window, const ThreadInfo &
         input, output);
     }
 }
+} // namespace arm_compute
\ No newline at end of file
diff --git a/src/runtime/NEON/functions/NECol2Im.cpp b/src/runtime/NEON/functions/NECol2Im.cpp
index 78c6bc0475..a1113bf5bd 100644
--- a/src/runtime/NEON/functions/NECol2Im.cpp
+++ b/src/runtime/NEON/functions/NECol2Im.cpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2017 ARM Limited.
+ * Copyright (c) 2017-2019 ARM Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -26,8 +26,8 @@
 #include "arm_compute/core/NEON/kernels/NECol2ImKernel.h"
 #include "support/ToolchainSupport.h"
 
-using namespace arm_compute;
-
+namespace arm_compute
+{
 void NECol2Im::configure(const ITensor *input, ITensor *output, const Size2D &convolved_dims)
 {
     auto k = arm_compute::support::cpp14::make_unique<NECol2ImKernel>();
@@ -39,3 +39,4 @@ Status NECol2Im::validate(const ITensorInfo *input, const ITensorInfo *output, c
 {
     return NECol2ImKernel::validate(input, output, convolved_dims);
 }
+} // namespace arm_compute
diff --git a/src/runtime/NEON/functions/NECopy.cpp b/src/runtime/NEON/functions/NECopy.cpp
index efa8b893aa..5f46023ecc 100644
--- a/src/runtime/NEON/functions/NECopy.cpp
+++ b/src/runtime/NEON/functions/NECopy.cpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2018 ARM Limited.
+ * Copyright (c) 2018-2019 ARM Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -28,8 +28,8 @@
 
 #include <utility>
 
-using namespace arm_compute;
-
+namespace arm_compute
+{
 void NECopy::configure(ITensor *input, ITensor *output)
 {
     auto k = arm_compute::support::cpp14::make_unique<NECopyKernel>();
@@ -41,3 +41,4 @@ Status NECopy::validate(const arm_compute::ITensorInfo *input, const arm_compute
 {
     return NECopyKernel::validate(input, output);
 }
+} // namespace arm_compute
\ No newline at end of file
diff --git a/src/runtime/NEON/functions/NEFillBorder.cpp b/src/runtime/NEON/functions/NEFillBorder.cpp
index 44e49520dc..6b7a0faa85 100644
--- a/src/runtime/NEON/functions/NEFillBorder.cpp
+++ b/src/runtime/NEON/functions/NEFillBorder.cpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2016, 2017 ARM Limited.
+ * Copyright (c) 2016-2019 ARM Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -26,8 +26,8 @@
 #include "arm_compute/core/Window.h"
 #include "arm_compute/runtime/NEON/NEScheduler.h"
 
-using namespace arm_compute;
-
+namespace arm_compute
+{
 void NEFillBorder::configure(ITensor *input, unsigned int border_width, BorderMode border_mode, const PixelValue &constant_border_value)
 {
     _border_handler.configure(input, BorderSize(border_width), border_mode, constant_border_value);
@@ -37,3 +37,4 @@ void NEFillBorder::run()
 {
     NEScheduler::get().schedule(&_border_handler, Window::DimZ);
 }
+} // namespace arm_compute
\ No newline at end of file
diff --git a/src/runtime/NEON/functions/NEFlattenLayer.cpp b/src/runtime/NEON/functions/NEFlattenLayer.cpp
index 57bef2b933..fb175a1dca 100644
--- a/src/runtime/NEON/functions/NEFlattenLayer.cpp
+++ b/src/runtime/NEON/functions/NEFlattenLayer.cpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2017-2018 ARM Limited.
+ * Copyright (c) 2017-2019 ARM Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -27,8 +27,8 @@
 #include "arm_compute/core/Size2D.h"
 #include "support/ToolchainSupport.h"
 
-using namespace arm_compute;
-
+namespace arm_compute
+{
 void NEFlattenLayer::configure(const ITensor *input, ITensor *output)
 {
     auto k = arm_compute::support::cpp14::make_unique<NEFlattenLayerKernel>();
@@ -39,4 +39,5 @@ void NEFlattenLayer::configure(const ITensor *input, ITensor *output)
 Status NEFlattenLayer::validate(const ITensorInfo *input, const ITensorInfo *output)
 {
     return NEFlattenLayerKernel::validate(input, output);
-}
\ No newline at end of file
+}
+} // namespace arm_compute
\ No newline at end of file
diff --git a/src/runtime/NEON/functions/NEGEMMTranspose1xW.cpp b/src/runtime/NEON/functions/NEGEMMTranspose1xW.cpp
index 802b94650e..cb93712da0 100644
--- a/src/runtime/NEON/functions/NEGEMMTranspose1xW.cpp
+++ b/src/runtime/NEON/functions/NEGEMMTranspose1xW.cpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2017-2018 ARM Limited.
+ * Copyright (c) 2017-2019 ARM Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -30,8 +30,8 @@
 #include "arm_compute/core/Validate.h"
 #include "support/ToolchainSupport.h"
 
-using namespace arm_compute;
-
+namespace arm_compute
+{
 void NEGEMMTranspose1xW::configure(const ITensor *input, ITensor *output)
 {
     auto k = arm_compute::support::cpp14::make_unique<NEGEMMTranspose1xWKernel>();
@@ -42,3 +42,4 @@ Status NEGEMMTranspose1xW::validate(const ITensorInfo *input, const ITensorInfo
 {
     return NEGEMMTranspose1xWKernel::validate(input, output);
 }
+} // namespace arm_compute
diff --git a/src/runtime/NEON/functions/NEGather.cpp b/src/runtime/NEON/functions/NEGather.cpp
index 078bd5ab26..428ef72d11 100644
--- a/src/runtime/NEON/functions/NEGather.cpp
+++ b/src/runtime/NEON/functions/NEGather.cpp
@@ -41,5 +41,4 @@ Status NEGather::validate(const ITensorInfo *input, const ITensorInfo *indices,
 {
     return NEGatherKernel::validate(input, indices, output, axis);
 }
-
 } // namespace arm_compute
diff --git a/src/runtime/NEON/functions/NEIm2Col.cpp b/src/runtime/NEON/functions/NEIm2Col.cpp
index 9102fca7f6..2a6972918e 100644
--- a/src/runtime/NEON/functions/NEIm2Col.cpp
+++ b/src/runtime/NEON/functions/NEIm2Col.cpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2017-2018 ARM Limited.
+ * Copyright (c) 2017-2019 ARM Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -27,8 +27,8 @@
 #include "arm_compute/runtime/NEON/NEScheduler.h"
 #include "support/ToolchainSupport.h"
 
-using namespace arm_compute;
-
+namespace arm_compute
+{
 NEIm2Col::NEIm2Col()
     : _kernel(), _y_dim(1)
 {
@@ -51,3 +51,4 @@ void NEIm2Col::run()
 {
     NEScheduler::get().schedule(&_kernel, _y_dim);
 }
+} // namespace arm_compute
\ No newline at end of file
diff --git a/src/runtime/NEON/functions/NEPermute.cpp b/src/runtime/NEON/functions/NEPermute.cpp
index 92abd03e2a..e6c41bfb03 100644
--- a/src/runtime/NEON/functions/NEPermute.cpp
+++ b/src/runtime/NEON/functions/NEPermute.cpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2018 ARM Limited.
+ * Copyright (c) 2018-2019 ARM Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -26,8 +26,8 @@
 #include "arm_compute/core/NEON/kernels/NEPermuteKernel.h"
 #include "support/ToolchainSupport.h"
 
-using namespace arm_compute;
-
+namespace arm_compute
+{
 void NEPermute::configure(const ITensor *input, ITensor *output, const PermutationVector &perm)
 {
     auto k = arm_compute::support::cpp14::make_unique<NEPermuteKernel>();
@@ -39,3 +39,4 @@ Status NEPermute::validate(const ITensorInfo *input, const ITensorInfo *output,
 {
     return NEPermuteKernel::validate(input, output, perm);
 }
+} // namespace arm_compute
diff --git a/src/runtime/NEON/functions/NEReshapeLayer.cpp b/src/runtime/NEON/functions/NEReshapeLayer.cpp
index 4600f36660..e2cd45ae6e 100644
--- a/src/runtime/NEON/functions/NEReshapeLayer.cpp
+++ b/src/runtime/NEON/functions/NEReshapeLayer.cpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2017-2018 ARM Limited.
+ * Copyright (c) 2017-2019 ARM Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -29,8 +29,8 @@
 
 #include <utility>
 
-using namespace arm_compute;
-
+namespace arm_compute
+{
 void NEReshapeLayer::configure(const ITensor *input, ITensor *output)
 {
     auto k = arm_compute::support::cpp14::make_unique<NEReshapeLayerKernel>();
@@ -45,3 +45,4 @@ Status NEReshapeLayer::validate(const ITensorInfo *input, const ITensorInfo *out
 
     return Status{};
 }
+} // namespace arm_compute
\ No newline at end of file
diff --git a/src/runtime/NEON/functions/NESelect.cpp b/src/runtime/NEON/functions/NESelect.cpp
index 509bbaa24e..c6089c8ec0 100644
--- a/src/runtime/NEON/functions/NESelect.cpp
+++ b/src/runtime/NEON/functions/NESelect.cpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2018 ARM Limited.
+ * Copyright (c) 2018-2019 ARM Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -26,8 +26,6 @@
 #include "arm_compute/core/NEON/kernels/NESelectKernel.h"
 #include "arm_compute/core/Types.h"
 
-using namespace arm_compute;
-
 namespace arm_compute
 {
 void NESelect::configure(const ITensor *c, const ITensor *x, const ITensor *y, ITensor *output)
diff --git a/src/runtime/NEON/functions/NETranspose.cpp b/src/runtime/NEON/functions/NETranspose.cpp
index b5b28e8e18..fd0d73aa7c 100644
--- a/src/runtime/NEON/functions/NETranspose.cpp
+++ b/src/runtime/NEON/functions/NETranspose.cpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2017 ARM Limited.
+ * Copyright (c) 2017-2019 ARM Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -28,8 +28,8 @@
 
 #include <utility>
 
-using namespace arm_compute;
-
+namespace arm_compute
+{
 void NETranspose::configure(const ITensor *input, ITensor *output)
 {
     auto k = arm_compute::support::cpp14::make_unique<NETransposeKernel>();
@@ -40,4 +40,5 @@ void NETranspose::configure(const ITensor *input, ITensor *output)
 Status NETranspose::validate(const ITensorInfo *input, const ITensorInfo *output)
 {
     return NETransposeKernel::validate(input, output);
-}
\ No newline at end of file
+}
+} // namespace arm_compute
\ No newline at end of file