COMPMID-2764: Add support for QASYMM8_SIGNED in NEConvolutionLayer.

Change-Id: I8fbbd2e399f48968337a60147098d04f27c2d1c0
Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com>
Reviewed-on: https://review.mlplatform.org/c/2402
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>

7 files changed, 130 insertions, 44 deletions

diff --git a/arm_compute/core/NEON/kernels/NECol2ImKernel.h b/arm_compute/core/NEON/kernels/NECol2ImKernel.h
index f02858e7d9..9858d4fd56 100644
--- a/arm_compute/core/NEON/kernels/NECol2ImKernel.h
+++ b/arm_compute/core/NEON/kernels/NECol2ImKernel.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2017-2018 ARM Limited.
+ * Copyright (c) 2017-2019 ARM Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -72,7 +72,7 @@ public:
 
     /** Set the input and output of the kernel.
      *
-     * @param[in]  input          The input tensor to convert. Data types supported: U8/S8/QASYMM8/U16/S16/F16/U32/S32/F32
+     * @param[in]  input          The input tensor to convert. Data types supported: Any
      * @param[out] output         The output tensor. 3 lower dimensions represent a single output [width, height, OFM],
      *                            while the rest represent batch of outputs. Data types supported: Same as @p input
      * @param[in]  convolved_dims Output convolved dimensions.
@@ -80,7 +80,7 @@ public:
     void configure(const ITensor *input, ITensor *output, const Size2D &convolved_dims);
     /** Static function to check if given info will lead to a valid configuration of @ref NECol2ImKernel
      *
-     * @param[in] input          The input tensor to convert. Data types supported: U8/S8/QASYMM8/U16/S16/F16/U32/S32/F32
+     * @param[in] input          The input tensor to convert. Data types supported: Any
      * @param[in] output         The output tensor. 3 lower dimensions represent a single output [width, height, OFM],
      *                           while the rest represent batch of outputs. Data types supported: Same as @p input
      * @param[in] convolved_dims Output convolved dimensions.
diff --git a/arm_compute/core/NEON/kernels/NEGEMMLowpOffsetContributionOutputStageKernel.h b/arm_compute/core/NEON/kernels/NEGEMMLowpOffsetContributionOutputStageKernel.h
index dadc5c221b..ac17b2efa5 100644
--- a/arm_compute/core/NEON/kernels/NEGEMMLowpOffsetContributionOutputStageKernel.h
+++ b/arm_compute/core/NEON/kernels/NEGEMMLowpOffsetContributionOutputStageKernel.h
@@ -37,13 +37,14 @@ class ITensor;
  * This kernel takes a final int32 accumulator value (the output of @ref NEGEMMLowpMatrixMultiplyKernel),
  * and adds to it the offset contribution of matrix A and matrix B in-place.
  *
- * The output stage can perform either QuantizeDownInt32ToUint8Scale or QuantizeDownInt32ToUint8ScaleByFixedPoint.
+ * The output stage can perform either QuantizeDownInt32ToUint8Scale or QuantizeDownInt32ToUint8ScaleByFixedPoint for Uint8.
+ * The output stage can perform either QuantizeDownInt32ToInt8Scale or QuantizeDownInt32ToInt8ScaleByFixedPoint for Int8.
  *
- * For QuantizeDownInt32ToUint8Scale the final result is:
+ * For QuantizeDownInt32ToUint8Scale/QuantizeDownInt32ToInt8Scale the final result is:
  *
  * ((mm_result'[i][k] + result_offset) * result_mult_int) >> result_shift
  *
- * For QuantizeDownInt32ToUint8ScaleByFixedPoint the final result is:
+ * For QuantizeDownInt32ToUint8ScaleByFixedPoint/QuantizeDownInt32ToInt8ScaleByFixedPoint the final result is:
  *
  * (FixedPointMul(mm_result'[i][k], result_fixedpoint_multiplier) >> result_shift) + result_offset_after_shift
  *
diff --git a/arm_compute/core/NEON/kernels/NEWeightsReshapeKernel.h b/arm_compute/core/NEON/kernels/NEWeightsReshapeKernel.h
index 585c707bb6..d432b731c2 100644
--- a/arm_compute/core/NEON/kernels/NEWeightsReshapeKernel.h
+++ b/arm_compute/core/NEON/kernels/NEWeightsReshapeKernel.h
@@ -75,7 +75,8 @@ public:
     /** Set the input and output of the kernel.
      *
      * @param[in]  input  The input tensor to convert. Weights are 4D tensor with dimensions [kernel_x, kernel_y, IFM, OFM] if shared,
-     *                    and 5D tensor with dimensions [kernel_x, kernel_y, IFM, OFM, num_patches] if unshared. Data types supported: QASYMM8/QSYMM8_PER_CHANNEL/FP16/F32
+     *                    and 5D tensor with dimensions [kernel_x, kernel_y, IFM, OFM, num_patches] if unshared.
+     *                    Data types supported: QASYMM8/QASYMM8_SIGNED/QSYMM8_PER_CHANNEL/FP16/F32
      * @param[in]  bias   The shared biases tensor to append.  Bias is 1D tensor with dimensions [OFM] if shared and 2D tensor with
      *                    dimensions [OFM, num_patches] if unshared. Data types supported: Same as @p input
      *                    @warning Appending biases to weights reshaped matrix is not supported for quantized asymmetric types.
@@ -85,7 +86,8 @@ public:
     /** Static function to check if given info will lead to a valid configuration of @ref NEWeightsReshapeKernel
      *
      * @param[in] input  The input tensor to convert. Weights are 4D tensor with dimensions [kernel_x, kernel_y, IFM, OFM] if shared,
-     *                   and 5D tensor with dimensions [kernel_x, kernel_y, IFM, OFM,  num_patches] if unshared. Data types supported: QASYMM8/QSYMM8_PER_CHANNEL/F16/F32
+     *                   and 5D tensor with dimensions [kernel_x, kernel_y, IFM, OFM,  num_patches] if unshared.
+     *                   Data types supported: QASYMM8/QASYMM8_SIGNED/QSYMM8_PER_CHANNEL/F16/F32
      * @param[in] biases The shared biases tensor to append.  Bias is 1D tensor with dimensions [OFM] if shared and 2D tensor with
      *                   dimensions [OFM, num_patches] if unshared. Data types supported: Same as @p input
      *                   @warning Appending biases to weights reshaped matrix is not supported for quantized asymmetric types.
diff --git a/arm_compute/core/Utils.h b/arm_compute/core/Utils.h
index 366d5dcc68..590e281bb0 100644
--- a/arm_compute/core/Utils.h
+++ b/arm_compute/core/Utils.h
@@ -549,6 +549,72 @@ inline DataType get_promoted_data_type(DataType dt)
     return DataType::UNKNOWN;
 }
 
+/** Compute the mininum and maximum values a data type can take
+ *
+ * @param[in] dt Data type to get the min/max bounds of
+ *
+ * @return A tuple (min,max) with the minimum and maximum values respectively wrapped in PixelValue.
+ */
+inline std::tuple<PixelValue, PixelValue> get_min_max(DataType dt)
+{
+    PixelValue min(0);
+    PixelValue max(0);
+    switch(dt)
+    {
+        case DataType::U8:
+        case DataType::QASYMM8:
+        {
+            min = PixelValue(std::numeric_limits<uint8_t>::lowest());
+            max = PixelValue(std::numeric_limits<uint8_t>::max());
+            break;
+        }
+        case DataType::S8:
+        case DataType::QSYMM8:
+        case DataType::QASYMM8_SIGNED:
+        case DataType::QSYMM8_PER_CHANNEL:
+        {
+            min = PixelValue(std::numeric_limits<int8_t>::lowest());
+            max = PixelValue(std::numeric_limits<int8_t>::max());
+            break;
+        }
+        case DataType::U16:
+        case DataType::QASYMM16:
+        {
+            min = PixelValue(std::numeric_limits<uint16_t>::lowest());
+            max = PixelValue(std::numeric_limits<uint16_t>::max());
+            break;
+        }
+        case DataType::S16:
+        case DataType::QSYMM16:
+        {
+            min = PixelValue(std::numeric_limits<int16_t>::lowest());
+            max = PixelValue(std::numeric_limits<int16_t>::max());
+            break;
+        }
+        case DataType::U32:
+        {
+            min = PixelValue(std::numeric_limits<uint32_t>::lowest());
+            max = PixelValue(std::numeric_limits<uint32_t>::max());
+            break;
+        }
+        case DataType::S32:
+        {
+            min = PixelValue(std::numeric_limits<int32_t>::lowest());
+            max = PixelValue(std::numeric_limits<int32_t>::max());
+            break;
+        }
+        case DataType::F32:
+        {
+            min = PixelValue(std::numeric_limits<float>::lowest());
+            max = PixelValue(std::numeric_limits<float>::max());
+            break;
+        }
+        default:
+            ARM_COMPUTE_ERROR("Undefined data type!");
+    }
+    return std::make_tuple(min, max);
+}
+
 /** Return true if the given format has horizontal subsampling.
  *
  * @param[in] format Format to determine subsampling.
@@ -1054,6 +1120,23 @@ inline bool is_data_type_quantized_asymmetric(DataType dt)
     }
 }
 
+/** Check if a given data type is of asymmetric quantized signed type
+ *
+ * @param[in] dt Input data type.
+ *
+ * @return True if data type is of asymmetric quantized signed type, else false.
+ */
+inline bool is_data_type_quantized_asymmetric_signed(DataType dt)
+{
+    switch(dt)
+    {
+        case DataType::QASYMM8_SIGNED:
+            return true;
+        default:
+            return false;
+    }
+}
+
 /** Check if a given data type is of symmetric quantized type
  *
  * @param[in] dt Input data type.
diff --git a/arm_compute/runtime/NEON/functions/NECol2Im.h b/arm_compute/runtime/NEON/functions/NECol2Im.h
index 64ce9944e2..613507cf6a 100644
--- a/arm_compute/runtime/NEON/functions/NECol2Im.h
+++ b/arm_compute/runtime/NEON/functions/NECol2Im.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2017-2018 ARM Limited.
+ * Copyright (c) 2017-2019 ARM Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -39,7 +39,7 @@ class NECol2Im : public INESimpleFunctionNoBorder
 public:
     /** Configure the col2im NEON kernel
      *
-     * @param[in]  input          The input tensor to convert. Data types supported: U8/S8/QASYMM8/U16/S16/F16/U32/S32/F32
+     * @param[in]  input          The input tensor to convert. Data types supported: Any
      * @param[out] output         The output tensor. 3 lower dimensions represent a single output [width, height, OFM],
      *                            while the rest represent batch of outputs. Data types supported: Same as @p input
      * @param[in]  convolved_dims Output convolved dimensions.
@@ -47,7 +47,7 @@ public:
     void configure(const ITensor *input, ITensor *output, const Size2D &convolved_dims);
     /** Static function to check if given info will lead to a valid configuration of @ref NECol2Im
      *
-     * @param[in] input          The input tensor to convert. Data types supported: U8/S8/QASYMM8/U16/S16/F16/U32/S32/F32
+     * @param[in] input          The input tensor to convert. Data types supported: Any
      * @param[in] output         The output tensor. 3 lower dimensions represent a single output [width, height, OFM],
      *                           while the rest represent batch of outputs. Data types supported: Same as @p input
      * @param[in] convolved_dims Output convolved dimensions.
diff --git a/arm_compute/runtime/NEON/functions/NEConvolutionLayer.h b/arm_compute/runtime/NEON/functions/NEConvolutionLayer.h
index 4310ab4b41..91fcef5971 100644
--- a/arm_compute/runtime/NEON/functions/NEConvolutionLayer.h
+++ b/arm_compute/runtime/NEON/functions/NEConvolutionLayer.h
@@ -80,10 +80,10 @@ public:
      *
      * @param[in]  input            Source tensor. 3 lower dimensions represent a single input [width, height, IFM],
      *                              while every optional dimension from 4 and above represent a batch of inputs.
-     *                              Data types supported: QASYMM8/F16/F32.
+     *                              Data types supported: QASYMM8/QASYMM8_SIGNED/F16/F32.
      * @param[in]  weights          Weights tensor. Weights are 4D tensor with dimensions [kernel_x, kernel_y, IFM, OFM]. Data type supported: Same as @p input.
      * @param[in]  biases           Biases tensor. Shared biases supported. Biases are 1D tensor with dimensions [OFM].
-     *                              Data type supported: Should match @p input data type, except for input of QASYMM8 type where biases should be of S32 type.
+     *                              Data type supported: Should match @p input data type, except for input of QASYMM8/QASYMM8_SIGNED type where biases should be of S32 type.
      * @param[out] output           Destination tensor. 3 lower dimensions represent a single output [width, height, OFM], while the rest represent batch of outputs.
      *                              Data types supported: Same as @p input.
      * @param[in]  conv_info        Contains padding and stride information described in @ref PadStrideInfo.
@@ -101,10 +101,10 @@ public:
      *
      * @param[in] input            Source tensor. 3 lower dimensions represent a single input [width, height, IFM],
      *                             while every optional dimension from 4 and above represent a batch of inputs.
-     *                             Data types supported: QASYMM8/F16/F32.
+     *                             Data types supported: QASYMM8/QASYMM8_SIGNED/F16/F32.
      * @param[in] weights          Weights tensor. Weights are 4D tensor with dimensions [kernel_x, kernel_y, IFM, OFM]. Data type supported:Same as @p input.
      * @param[in] biases           Biases tensor. Shared biases supported. Biases are 1D tensor with dimensions [OFM].
-     *                             Data type supported: Should match @p input data type, except for input of QASYMM8 type where biases should be of S32 type.
+     *                             Data type supported: Should match @p input data type, except for input of QASYMM8/QASYMM8_SIGNED type where biases should be of S32 type.
      * @param[in] output           Destination tensor. 3 lower dimensions represent a single output [width, height, OFM], while the rest represent batch of outputs.
      *                             Data types supported: Same as @p input.
      * @param[in] conv_info        Contains padding and stride information described in @ref PadStrideInfo.
@@ -125,7 +125,7 @@ public:
      *
      * @param[in] input            Source tensor. 3 lower dimensions represent a single input [width, height, IFM],
      *                             while every optional dimension from 4 and above represent a batch of inputs.
-     *                             Data types supported: QASYMM8/F16/F32.
+     *                             Data types supported: QASYMM8/QASYMM8_SIGNED/F16/F32.
      * @param[in] weights          Weights tensor. Weights are 4D tensor with dimensions [kernel_x, kernel_y, IFM, OFM]. Data type supported:Same as @p input.
      * @param[in] output           Destination tensor. 3 lower dimensions represent a single output [width, height, OFM], while the rest represent batch of outputs.
      *                             Data types supported: Same as @p input.
@@ -149,5 +149,5 @@ private:
     std::shared_ptr<IMemoryManager> _memory_manager;
     std::unique_ptr<IFunction>      _function; /**< Function to run */
 };
-}
+} // namespace arm_compute
 #endif /* __ARM_COMPUTE_NECONVOLUTIONLAYER_H__ */
diff --git a/arm_compute/runtime/NEON/functions/NEGEMMConvolutionLayer.h b/arm_compute/runtime/NEON/functions/NEGEMMConvolutionLayer.h
index 6452fc9249..c513afa790 100644
--- a/arm_compute/runtime/NEON/functions/NEGEMMConvolutionLayer.h
+++ b/arm_compute/runtime/NEON/functions/NEGEMMConvolutionLayer.h
@@ -63,16 +63,16 @@ public:
     NEConvolutionLayerReshapeWeights &operator=(NEConvolutionLayerReshapeWeights &&) = default;
     /** Set the input and output tensors.
      *
-     * @param[in]  weights Weights tensor. Weights are 4D tensor with dimensions [kernel_x, kernel_y, IFM, OFM]. Data type supported: QASYMM8/QSYMM8_PER_CHANNEL/F16/F32.
+     * @param[in]  weights Weights tensor. Weights are 4D tensor with dimensions [kernel_x, kernel_y, IFM, OFM]. Data type supported: QASYMM8/QASYMM8_SIGNED/QSYMM8_PER_CHANNEL/F16/F32.
      * @param[in]  biases  Biases tensor. Shared biases supported. Biases are 1D tensor with dimensions [OFM]. Data type supported: Same as @p weights.
      * @param[out] output  Destination tensor. Data types supported: Same as @p weights.
      */
     void configure(const ITensor *weights, const ITensor *biases, ITensor *output);
     /** Static function to check if given info will lead to a valid configuration of @ref NEConvolutionLayerReshapeWeights
      *
-     * @param[in] weights Weights tensor. Weights are 4D tensor with dimensions [kernel_x, kernel_y, IFM, OFM]. Data type supported: QASYMM8/QSYMM8_PER_CHANNEL/F16/F32.
-     * @param[in] biases  Biases tensor. Shared biases supported. Biases are 1D tensor with dimensions [OFM]. Data type supported: Same as @p weights.
-     * @param[in] output  Destination tensor. Data types supported: Same as @p weights.
+     * @param[in] weights Weights tensor info. Weights are 4D tensor with dimensions [kernel_x, kernel_y, IFM, OFM]. Data type supported: QASYMM8/QASYMM8_SIGNED/QSYMM8_PER_CHANNEL/F16/F32.
+     * @param[in] biases  Biases tensor info. Shared biases supported. Biases are 1D tensor with dimensions [OFM]. Data type supported: Same as @p weights.
+     * @param[in] output  Destination tensor info. Data types supported: Same as @p weights.
      *
      * @return an error status
      */
@@ -135,8 +135,8 @@ private:
  *
  * -# @ref NEIm2ColKernel
  * -# @ref NEGEMM (if the data type is FP32 or FP16)
- * -# @ref NEGEMMLowpMatrixMultiplyCore (if the data type is QASYMM8)
- * -# @ref NEGEMMLowpQuantizeDownInt32ToUint8ScaleByFixedPoint (if the data type is QASYMM8)
+ * -# @ref NEGEMMLowpMatrixMultiplyCore (if the data type is QASYMM8/QASYMM8_SIGNED)
+ * -# @ref NEGEMMLowpQuantizeDownInt32ToUint8ScaleByFixedPoint (if the data type is QASYMM8/QASYMM8_SIGNED)
  * -# @ref NEArithmeticAdditionKernel (if biases != nullptr and we have a 1x1 convolution with the NHWC data layout)
  * -# @ref NECol2ImKernel (if NCHW data layout)
  *
@@ -158,10 +158,10 @@ public:
      *
      * @param[in]  input        Source tensor. 3 lower dimensions represent a single input [width, height, IFM],
      *                          while every optional dimension from 4 and above represent a batch of inputs.
-     *                          Data types supported: QASYMM8/F16/F32.
-     * @param[in]  weights      Weights tensor. Weights are 4D tensor with dimensions [kernel_x, kernel_y, IFM, OFM]. Data type supported: QASYMM8/QSYMM8_PER_CHANNEL/F16/F32.
+     *                          Data types supported: QASYMM8/QASYMM8_SIGNED/F16/F32.
+     * @param[in]  weights      Weights tensor. Weights are 4D tensor with dimensions [kernel_x, kernel_y, IFM, OFM]. Data type supported: QASYMM8/QASYMM8_SIGNED/QSYMM8_PER_CHANNEL/F16/F32.
      * @param[in]  biases       Biases tensor. Shared biases supported. Biases are 1D tensor with dimensions [OFM].
-     *                          Data type supported: Should match @p input data type, except for input of QASYMM8 type where biases should be of S32 type.
+     *                          Data type supported: Should match @p input data type, except for input of QASYMM8/QASYMM8_SIGNED type where biases should be of S32 type.
      * @param[out] output       Destination tensor. 3 lower dimensions represent a single output [width, height, OFM], while the rest represent batch of outputs.
      *                          Data types supported: Same as @p input.
      * @param[in]  conv_info    Contains padding and stride information described in @ref PadStrideInfo.
@@ -175,13 +175,13 @@ public:
                    const Size2D &dilation = Size2D(1U, 1U), const ActivationLayerInfo &act_info = ActivationLayerInfo(), unsigned int num_groups = 1);
     /** Static function to check if given info will lead to a valid configuration of @ref NEGEMMConvolutionLayer
      *
-     * @param[in] input        Source tensor. 3 lower dimensions represent a single input [width, height, IFM],
+     * @param[in] input        Source tensor info. 3 lower dimensions represent a single input [width, height, IFM],
      *                         while every optional dimension from 4 and above represent a batch of inputs.
-     *                         Data types supported: QASYMM8/F16/F32.
-     * @param[in] weights      Weights tensor. Weights are 4D tensor with dimensions [kernel_x, kernel_y, IFM, OFM]. Data type supported: QASYMM8/QSYMM8_PER_CHANNEL/F16/F32.
-     * @param[in] biases       Biases tensor. Shared biases supported. Biases are 1D tensor with dimensions [OFM].
-     *                         Data type supported: Should match @p input data type, except for input of QASYMM8 type where biases should be of S32 type.
-     * @param[in] output       Destination tensor. 3 lower dimensions represent a single output [width, height, OFM], while the rest represent batch of outputs.
+     *                         Data types supported: QASYMM8/QASYMM8_SIGNED/F16/F32.
+     * @param[in] weights      Weights tensor info. Weights are 4D tensor with dimensions [kernel_x, kernel_y, IFM, OFM]. Data type supported: QASYMM8/QASYMM8_SIGNED/QSYMM8_PER_CHANNEL/F16/F32.
+     * @param[in] biases       Biases tensor info. Shared biases supported. Biases are 1D tensor with dimensions [OFM].
+     *                         Data type supported: Should match @p input data type, except for input of QASYMM8/QASYMM8_SIGNED type where biases should be of S32 type.
+     * @param[in] output       Destination tensor info. 3 lower dimensions represent a single output [width, height, OFM], while the rest represent batch of outputs.
      *                         Data types supported: Same as @p input.
      * @param[in] conv_info    Contains padding and stride information described in @ref PadStrideInfo.
      * @param[in] weights_info Specifies if the weights tensor has been reshaped with NEWeightsReshapeKernel. If this is not part of the fully connected layer the weights
@@ -202,24 +202,24 @@ public:
 private:
     /** Configures the appropriate matrix multiply routine
      *
-     * @param[in]  input         Input tensor. Data types supported: QASYMM8/F16/F32.
-     * @param[in]  weights       Weights tensor. Data type supported: QASYMM8/QSYMM8_PER_CHANNEL/F16/F32.
+     * @param[in]  input         Input tensor. Data types supported: QASYMM8/QASYMM8_SIGNED/F16/F32.
+     * @param[in]  weights       Weights tensor. Data type supported: QASYMM8/QASYMM8_SIGNED/QSYMM8_PER_CHANNEL/F16/F32.
      * @param[in]  biases        Biases tensor. Shared biases supported. Biases are 1D tensor with dimensions [OFM].
-     *                           Data type supported: Should match @p input data type, except for input of QASYMM8 type where biases should be of S32 type.
+     *                           Data type supported: Should match @p input data type, except for input of QASYMM8/QASYMM8_SIGNED type where biases should be of S32 type.
      * @param[out] output        Output tensor. Data types supported: Same as @p input,
-     *                           except for input of QASYMM8 type where output should be of S32 type.
+     *                           except for input of QASYMM8/QASYMM8_SIGNED type where output should be of S32 type.
      * @param[in]  act_info      (Optional) Activation layer information in case of a fused activation. Only RELU, BOUNDED_RELU and LU_BOUNDED_RELU supported.
      * @param[in]  gemm_3d_depth (Optional) Depth of GEMM 3D (Defaults to 1)
      */
     void configure_mm(const ITensor *input, const ITensor *weights, const ITensor *biases, ITensor *output, const ActivationLayerInfo &act_info = ActivationLayerInfo(), int gemm_3d_depth = 1);
     /** Static function to check if given info will lead to a valid configuration of @ref NEGEMMConvolutionLayer matrix multiply routines
      *
-     * @param[in] input         Input tensor. Data types supported: QASYMM8/F16/F32.
-     * @param[in] weights       Weights tensor. Data type supported: QASYMM8/QSYMM8_PER_CHANNEL/F16/F32.
-     * @param[in] biases        Biases tensor. Shared biases supported. Biases are 1D tensor with dimensions [OFM].
-     *                          Data type supported: Should match @p input data type, except for input of QASYMM8 type where biases should be of S32 type.
-     * @param[in] output        Output tensor. Data types supported: Same as @p input,
-     *                          except for input of QASYMM8 type where output should be of S32 type.
+     * @param[in] input         Input tensor info. Data types supported: QASYMM8/QASYMM8_SIGNED/F16/F32.
+     * @param[in] weights       Weights tensor info. Data type supported: QASYMM8/QASYMM8_SIGNED/QSYMM8_PER_CHANNEL/F16/F32.
+     * @param[in] biases        Biases tensor info. Shared biases supported. Biases are 1D tensor with dimensions [OFM].
+     *                          Data type supported: Should match @p input data type, except for input of QASYMM8/QASYMM8_SIGNED type where biases should be of S32 type.
+     * @param[in] output        Output tensor info. Data types supported: Same as @p input,
+     *                          except for input of QASYMM8/QASYMM8_SIGNED type where output should be of S32 type.
      * @param[in] act_info      (Optional) Activation layer information in case of a fused activation. Only RELU, BOUNDED_RELU and LU_BOUNDED_RELU supported.
      * @param[in] gemm_3d_depth (Optional) Depth of GEMM 3D (Defaults to 1)
      * @param[in] skip_im2col   (Optional) Flag which specifies if im2col has to be skipped. i.e. 1x1 convolution with NHWC data layout. (Default to false)
@@ -230,8 +230,8 @@ private:
                               int gemm_3d_depth = 1, bool skip_im2col = false);
     /** Static function to check if GEMM3D is supported in @ref NEGEMM or in @ref NEGEMMLowpMatrixMultiplyCore
      *
-     * @param[in] input_info    Input tensor info. Data types supported: QASYMM8/F16/F32.
-     * @param[in] weights_info  Weights tensor info. Data types supported: QASYMM8/F16/F32.
+     * @param[in] input_info    Input tensor info. Data types supported: QASYMM8/QASYMM8_SIGNED/F16/F32.
+     * @param[in] weights_info  Weights tensor info. Data types supported: QASYMM8/QASYMM8_SIGNED/F16/F32.
      * @param[in] act_info      Activation layer information in case of a fused activation. Only RELU, BOUNDED_RELU and LU_BOUNDED_RELU supported.
      * @param[in] gemm_3d_depth Depth of GEMM 3D
      * @param[in] skip_im2col   Flag which specifies if im2col has to be skipped. i.e. 1x1 convolution with NHWC data layout