CMSIS-NN: clang format include files

clang format files in the NN/include folder

Change-Id: I69cf711c00c65fb907e049ab1852290dd28f8800
diff --git a/CMSIS/NN/Include/arm_nn_tables.h b/CMSIS/NN/Include/arm_nn_tables.h
index f4bd7f9..35dfc3b 100644
--- a/CMSIS/NN/Include/arm_nn_tables.h
+++ b/CMSIS/NN/Include/arm_nn_tables.h
@@ -42,15 +42,15 @@
 extern const q7_t tanhTable_q7[256];
 extern const q15_t tanhTable_q15[256];
 
-  /**
-   * @brief 2-way tables for various activation functions
-   *
-   * 2-way table, H table for value larger than 1/4
-   * L table for value smaller than 1/4, H table for remaining
-   * We have this only for the q15_t version. It does not make
-   * sense to have it for q7_t type
-   */
+/**
+ * @brief 2-way tables for various activation functions
+ *
+ * 2-way table, H table for value larger than 1/4
+ * L table for value smaller than 1/4, H table for remaining
+ * We have this only for the q15_t version. It does not make
+ * sense to have it for q7_t type
+ */
 extern const q15_t sigmoidHTable_q15[192];
 extern const q15_t sigmoidLTable_q15[128];
 
-#endif                          /*  ARM_NN_TABLES_H */
+#endif /*  ARM_NN_TABLES_H */
diff --git a/CMSIS/NN/Include/arm_nn_types.h b/CMSIS/NN/Include/arm_nn_types.h
index e90b400..206af07 100644
--- a/CMSIS/NN/Include/arm_nn_types.h
+++ b/CMSIS/NN/Include/arm_nn_types.h
@@ -28,7 +28,6 @@
  * Target Processor:  Cortex-M cores
  * -------------------------------------------------------------------- */
 
-
 #ifndef _ARM_NN_TYPES_H
 #define _ARM_NN_TYPES_H
 
@@ -37,21 +36,22 @@
 /** CMSIS-NN object to contain the width and height of a tile */
 typedef struct
 {
-    int32_t w;  /**< Width */
-    int32_t h;  /**< Height */
+    int32_t w; /**< Width */
+    int32_t h; /**< Height */
 } cmsis_nn_tile;
 
 /** CMSIS-NN object used for the function context. */
 typedef struct
 {
-    void *buf;      /**< Pointer to a buffer needed for the optimization */
-    int32_t size;   /**< Buffer size */
+    void *buf;    /**< Pointer to a buffer needed for the optimization */
+    int32_t size; /**< Buffer size */
 } cmsis_nn_context;
 
 /** CMSIS-NN object to contain the dimensions of the tensors */
 typedef struct
 {
-    int32_t n; /**< Generic dimension to contain either the batch size or output channels. Please refer to the function documentation for more information */
+    int32_t n; /**< Generic dimension to contain either the batch size or output channels.
+                     Please refer to the function documentation for more information */
     int32_t h; /**< Height */
     int32_t w; /**< Width */
     int32_t c; /**< Input channels */
@@ -81,39 +81,39 @@
 /** CMSIS-NN object for the convolution layer parameters */
 typedef struct
 {
-    int32_t             input_offset;   /**< Zero value for the input tensor */
-    int32_t             output_offset;  /**< Zero value for the output tensor */
-    cmsis_nn_tile       stride;
-    cmsis_nn_tile       padding;
-    cmsis_nn_tile       dilation;
+    int32_t input_offset;  /**< Zero value for the input tensor */
+    int32_t output_offset; /**< Zero value for the output tensor */
+    cmsis_nn_tile stride;
+    cmsis_nn_tile padding;
+    cmsis_nn_tile dilation;
     cmsis_nn_activation activation;
 } cmsis_nn_conv_params;
 
 /** CMSIS-NN object for Depthwise convolution layer parameters */
 typedef struct
 {
-    int32_t             input_offset;   /**< Zero value for the input tensor */
-    int32_t             output_offset;  /**< Zero value for the output tensor */
-    int32_t             ch_mult;        /**< Channel Multiplier. ch_mult * in_ch = out_ch */
-    cmsis_nn_tile       stride;
-    cmsis_nn_tile       padding;
-    cmsis_nn_tile       dilation;
+    int32_t input_offset;  /**< Zero value for the input tensor */
+    int32_t output_offset; /**< Zero value for the output tensor */
+    int32_t ch_mult;       /**< Channel Multiplier. ch_mult * in_ch = out_ch */
+    cmsis_nn_tile stride;
+    cmsis_nn_tile padding;
+    cmsis_nn_tile dilation;
     cmsis_nn_activation activation;
 } cmsis_nn_dw_conv_params;
 /** CMSIS-NN object for pooling layer parameters */
 typedef struct
 {
-    cmsis_nn_tile       stride;
-    cmsis_nn_tile       padding;
+    cmsis_nn_tile stride;
+    cmsis_nn_tile padding;
     cmsis_nn_activation activation;
 } cmsis_nn_pool_params;
 
 /** CMSIS-NN object for Fully Connected layer parameters */
 typedef struct
 {
-    int32_t             input_offset;   /**< Zero value for the input tensor */
-    int32_t             filter_offset;   /**< Zero value for the filter tensor */
-    int32_t             output_offset;  /**< Zero value for the output tensor */
+    int32_t input_offset;  /**< Zero value for the input tensor */
+    int32_t filter_offset; /**< Zero value for the filter tensor */
+    int32_t output_offset; /**< Zero value for the output tensor */
     cmsis_nn_activation activation;
 } cmsis_nn_fc_params;
 
@@ -121,12 +121,10 @@
 typedef struct
 {
     int32_t rank;
-    int32_t input_offset; /**< Zero value for the input tensor */
+    int32_t input_offset;  /**< Zero value for the input tensor */
     int32_t output_offset; /**< Zero value for the output tensor */
     cmsis_nn_activation input_activation;
     cmsis_nn_activation output_activation;
 } cmsis_nn_svdf_params;
 
 #endif // _ARM_NN_TYPES_H
-
-
diff --git a/CMSIS/NN/Include/arm_nnfunctions.h b/CMSIS/NN/Include/arm_nnfunctions.h
index 7680c9d..7f77e9a 100644
--- a/CMSIS/NN/Include/arm_nnfunctions.h
+++ b/CMSIS/NN/Include/arm_nnfunctions.h
@@ -57,11 +57,12 @@
    * - Legacy functions supporting ARM's internal symmetric quantization(8 bits).
    * - Functions that support TensorFlow Lite framework with symmetric quantization(8 bits).
    *
-   * The legacy functions can be identified with their suffix of _q7 or _q15 and are no new development is done there. The article in [2] describes in detail
-   * how to run a network using the legacy functions.
+   * The legacy functions can be identified with their suffix of _q7 or _q15 and are no new development is done there.
+   * The article in [2] describes in detail how to run a network using the legacy functions.
    *
-   * The functions supporting TensorFlow Lite framework is identified by the _s8 suffix and can be invoked from TFL micro. The functions are bit exact to
-   * TensorFlow Lite. Refer to the TensorFlow's documentation in [3] on how to run a TensorFlow Lite model using optimized CMSIS-NN kernels.
+   * The functions supporting TensorFlow Lite framework is identified by the _s8 suffix and can be invoked from TFL
+   * micro. The functions are bit exact to TensorFlow Lite. Refer to the TensorFlow's documentation in [3] on how to run
+   * a TensorFlow Lite model using optimized CMSIS-NN kernels.
    *
    * Block Diagram
    * --------
@@ -86,12 +87,13 @@
    * Define macro ARM_MATH_MVEI, If the silicon supports M-Profile Vector Extension.
 
    * - ARM_MATH_AUTOVECTORIZE
-   *  Used in conjucture with ARM_MATH_MVEI to let the compiler auto vectorize for the functions that uses inline assembly.
-   *  It does not affect functions that use C or intrinsics.
+   *  Used in conjucture with ARM_MATH_MVEI to let the compiler auto vectorize for the functions that uses inline
+   *  assembly. It does not affect functions that use C or intrinsics.
    * - ARM_MATH_BIG_ENDIAN:
    *
-   * Define macro ARM_MATH_BIG_ENDIAN to build the library for big endian targets. This is supported only for the legacy functions i.e, functions targetted at
-   * TensorFlow Lite do not support big endianness. By default library builds for little endian targets.
+   * Define macro ARM_MATH_BIG_ENDIAN to build the library for big endian targets. This is supported only for the legacy
+   * functions i.e, functions targetted at TensorFlow Lite do not support big endianness. By default library builds for
+   * little endian targets.
    *
    * - ARM_NN_TRUNCATE:
    *
@@ -107,7 +109,8 @@
    * -# improve validation
    * -# improve code readability
    *
-   * The upcoming API interface change will be based on "struct" and only affect the TensorFlowLite micro compliant APIs [4] (functions with _s8 suffix)
+   * The upcoming API interface change will be based on "struct" and only affect the TensorFlowLite micro compliant
+   * APIs [4] (functions with _s8 suffix)
    *
    * Below you can find a snapshot of how the new API interface will look like (names can change)
    *
@@ -146,7 +149,8 @@
    * [1] CMSIS-NN: Efficient Neural Network Kernels for Arm Cortex-M CPUs https://arxiv.org/abs/1801.06601
    *
    * [2] Converting a Neural Network for Arm Cortex-M with CMSIS-NN
-   *     https://developer.arm.com/solutions/machine-learning-on-arm/developer-material/how-to-guides/converting-a-neural-network-for-arm-cortex-m-with-cmsis-nn/single-page
+   *
+   https://developer.arm.com/solutions/machine-learning-on-arm/developer-material/how-to-guides/converting-a-neural-network-for-arm-cortex-m-with-cmsis-nn/single-page
    * [3] https://www.tensorflow.org/lite/microcontrollers/library
    *
    * [4] https://github.com/ARM-software/CMSIS_5/tree/develop/CMSIS/NN#legacy-vs-tfl-micro-compliant-apis
@@ -161,31 +165,28 @@
 #ifndef _ARM_NNFUNCTIONS_H
 #define _ARM_NNFUNCTIONS_H
 
-#include "arm_nn_types.h"
 #include "arm_math_types.h"
+#include "arm_nn_types.h"
 
 #define USE_INTRINSIC
 
 //#define ARM_NN_TRUNCATE /* This config the rounding model to floor or round to the nearest int */
 
 #ifdef __cplusplus
-extern    "C"
-{
+extern "C" {
 #endif
 
 /**
  * @brief Struct for specifying activation function types
  *
  */
-typedef enum
-{
+typedef enum {
     ARM_SIGMOID = 0,
-                /**< Sigmoid activation function */
+    /**< Sigmoid activation function */
     ARM_TANH = 1,
-             /**< Tanh activation function */
+    /**< Tanh activation function */
 } arm_nn_activation_type;
 
-
 /**
  * @defgroup NNConv Convolution Functions
  *
@@ -203,174 +204,363 @@
  *
  */
 
-  /**
-   * @brief s8 convolution layer wrapper function with the main purpose to call the optimal kernel available in cmsis-nn to perform the convolution.
-   *
-   * @param[in, out] ctx            Function context that contains the additional buffer if required by the implementation.
-                                    arm_convolve_wrapper_s8_get_buffer_size will return the buffer_size if required
-   * @param[in]      conv_params    Convolution parameters (e.g. strides, dilations, pads,...).
-   *                                Range of conv_params->input_offset  : [-127, 128]
-   *                                Range of conv_params->output_offset : [-128, 127]
-   * @param[in]      quant_params   Per-channel quantization info.
-   *                                It contains the multiplier and shift values to be applied to each output channel
-   * @param[in]      input_dims     Input (activation) tensor dimensions. Format: [N, H, W, C_IN]
-   * @param[in]      input_data     Input (activation) data pointer. Data type: int8
-   * @param[in]      filter_dims    Filter tensor dimensions. Format: [C_OUT, HK, WK, C_IN] where HK and WK are the spatial filter dimensions
-   * @param[in]      filter_data    Filter data pointer. Data type: int8
-   * @param[in]      bias_dims      Bias tensor dimensions. Format: [C_OUT]
-   * @param[in]      bias_data      Bias data pointer. Data type: int32
-   * @param[in]      output_dims    Output tensor dimensions. Format: [N, H, W, C_OUT]
-   * @param[out]     output_data    Output data pointer. Data type: int8
-   *
-   * @return     The function returns either
-   *                  <code>ARM_MATH_SIZE_MISMATCH</code> if argument constraints fail. or,
-   *                  <code>ARM_MATH_SUCCESS</code> on successful completion.
-   *
-   */
-    arm_status arm_convolve_wrapper_s8(const cmsis_nn_context* ctx,
-                                       const cmsis_nn_conv_params* conv_params,
-                                       const cmsis_nn_per_channel_quant_params* quant_params,
-                                       const cmsis_nn_dims* input_dims,
-                                       const q7_t *input_data,
-                                       const cmsis_nn_dims* filter_dims,
-                                       const q7_t *filter_data,
-                                       const cmsis_nn_dims* bias_dims,
-                                       const int32_t *bias_data,
-                                       const cmsis_nn_dims* output_dims,
-                                       q7_t *output_data);
+/**
+ * @brief s8 convolution layer wrapper function with the main purpose to call the optimal kernel available in cmsis-nn
+ *        to perform the convolution.
+ *
+ * @param[in, out] ctx            Function context that contains the additional buffer if required by the function.
+                                  arm_convolve_wrapper_s8_get_buffer_size will return the buffer_size if required
+ * @param[in]      conv_params    Convolution parameters (e.g. strides, dilations, pads,...).
+ *                                Range of conv_params->input_offset  : [-127, 128]
+ *                                Range of conv_params->output_offset : [-128, 127]
+ * @param[in]      quant_params   Per-channel quantization info.
+ *                                It contains the multiplier and shift values to be applied to each output channel
+ * @param[in]      input_dims     Input (activation) tensor dimensions. Format: [N, H, W, C_IN]
+ * @param[in]      input_data     Input (activation) data pointer. Data type: int8
+ * @param[in]      filter_dims    Filter tensor dimensions. Format: [C_OUT, HK, WK, C_IN] where HK and WK are the
+ *                                spatial filter dimensions
+ * @param[in]      filter_data    Filter data pointer. Data type: int8
+ * @param[in]      bias_dims      Bias tensor dimensions. Format: [C_OUT]
+ * @param[in]      bias_data      Bias data pointer. Data type: int32
+ * @param[in]      output_dims    Output tensor dimensions. Format: [N, H, W, C_OUT]
+ * @param[out]     output_data    Output data pointer. Data type: int8
+ *
+ * @return     The function returns either
+ *                  <code>ARM_MATH_SIZE_MISMATCH</code> if argument constraints fail. or,
+ *                  <code>ARM_MATH_SUCCESS</code> on successful completion.
+ *
+ */
+arm_status arm_convolve_wrapper_s8(const cmsis_nn_context *ctx,
+                                   const cmsis_nn_conv_params *conv_params,
+                                   const cmsis_nn_per_channel_quant_params *quant_params,
+                                   const cmsis_nn_dims *input_dims,
+                                   const q7_t *input_data,
+                                   const cmsis_nn_dims *filter_dims,
+                                   const q7_t *filter_data,
+                                   const cmsis_nn_dims *bias_dims,
+                                   const int32_t *bias_data,
+                                   const cmsis_nn_dims *output_dims,
+                                   q7_t *output_data);
 
-  /**
-   * @brief Get the required buffer size for arm_convolve_wrapper_s8
-   *
-   * @param[in]      conv_params    Convolution parameters (e.g. strides, dilations, pads,...).
-   *                                Range of conv_params->input_offset  : [-127, 128]
-   *                                Range of conv_params->output_offset : [-128, 127]
-   * @param[in]      input_dims     Input (activation) dimensions. Format: [N, H, W, C_IN]
-   * @param[in]      filter_dims    Filter dimensions. Format: [C_OUT, HK, WK, C_IN] where HK and WK are the spatial filter dimensions
-   * @param[in]      output_dims    Output tensor dimensions. Format: [N, H, W, C_OUT]
-   *
-   * @return         The function returns  required buffer size(bytes)
-   *
-   */
-    int32_t arm_convolve_wrapper_s8_get_buffer_size(const cmsis_nn_conv_params* conv_params,
-                                                    const cmsis_nn_dims* input_dims,
-                                                    const cmsis_nn_dims* filter_dims,
-                                                    const cmsis_nn_dims* output_dims);
+/**
+ * @brief Get the required buffer size for arm_convolve_wrapper_s8
+ *
+ * @param[in]      conv_params    Convolution parameters (e.g. strides, dilations, pads,...).
+ *                                Range of conv_params->input_offset  : [-127, 128]
+ *                                Range of conv_params->output_offset : [-128, 127]
+ * @param[in]      input_dims     Input (activation) dimensions. Format: [N, H, W, C_IN]
+ * @param[in]      filter_dims    Filter dimensions. Format: [C_OUT, HK, WK, C_IN] where HK and WK are the spatial
+ *                                filter dimensions
+ * @param[in]      output_dims    Output tensor dimensions. Format: [N, H, W, C_OUT]
+ *
+ * @return         The function returns  required buffer size(bytes)
+ *
+ */
+int32_t arm_convolve_wrapper_s8_get_buffer_size(const cmsis_nn_conv_params *conv_params,
+                                                const cmsis_nn_dims *input_dims,
+                                                const cmsis_nn_dims *filter_dims,
+                                                const cmsis_nn_dims *output_dims);
 
-  /**
-   * @brief Basic s8 convolution function
-   * @param[in, out] ctx            Function context that contains the additional buffer if required by the implementation.
-                                    arm_convolve_s8_get_buffer_size will return the buffer_size if required
-   * @param[in]      conv_params    Convolution parameters (e.g. strides, dilations, pads,...).
-   *                                Range of conv_params->input_offset  : [-127, 128]
-   *                                Range of conv_params->output_offset : [-128, 127]
-   * @param[in]      quant_params   Per-channel quantization info.
-   *                                It contains the multiplier and shift values to be applied to each output channel
-   * @param[in]      input_dims     Input (activation) tensor dimensions. Format: [N, H, W, C_IN]
-   * @param[in]      input_data     Input (activation) data pointer. Data type: int8
-   * @param[in]      filter_dims    Filter tensor dimensions. Format: [C_OUT, HK, WK, C_IN] where HK and WK are the spatial filter dimensions
-   * @param[in]      filter_data    Filter data pointer. Data type: int8
-   * @param[in]      bias_dims      Bias tensor dimensions. Format: [C_OUT]
-   * @param[in]      bias_data      Optional bias data pointer. Data type: int32
-   * @param[in]      output_dims    Output tensor dimensions. Format: [N, H, W, C_OUT]
-   * @param[out]     output_data    Output data pointer. Data type: int8
+/**
+ * @brief Basic s8 convolution function
+ * @param[in, out] ctx            Function context that contains the additional buffer if required by the function.
+                                  arm_convolve_s8_get_buffer_size will return the buffer_size if required
+ * @param[in]      conv_params    Convolution parameters (e.g. strides, dilations, pads,...).
+ *                                Range of conv_params->input_offset  : [-127, 128]
+ *                                Range of conv_params->output_offset : [-128, 127]
+ * @param[in]      quant_params   Per-channel quantization info.
+ *                                It contains the multiplier and shift values to be applied to each output channel
+ * @param[in]      input_dims     Input (activation) tensor dimensions. Format: [N, H, W, C_IN]
+ * @param[in]      input_data     Input (activation) data pointer. Data type: int8
+ * @param[in]      filter_dims    Filter tensor dimensions. Format: [C_OUT, HK, WK, C_IN] where HK and WK are the
+ *                                spatial filter dimensions
+ * @param[in]      filter_data    Filter data pointer. Data type: int8
+ * @param[in]      bias_dims      Bias tensor dimensions. Format: [C_OUT]
+ * @param[in]      bias_data      Optional bias data pointer. Data type: int32
+ * @param[in]      output_dims    Output tensor dimensions. Format: [N, H, W, C_OUT]
+ * @param[out]     output_data    Output data pointer. Data type: int8
 
-   * @return     The function returns <code>ARM_MATH_SUCCESS</code>
-   *
-   * @details
-   *    1. Supported framework: TensorFlow Lite micro
-   *    2. q7 is used as data type eventhough it is s8 data. It is done so to be consistent with existing APIs.
-   *    3. Additional memory is required for optimization. Refer to argument 'ctx' for details.
-   *
-   */
-    arm_status arm_convolve_s8(const cmsis_nn_context* ctx,
-                               const cmsis_nn_conv_params* conv_params,
-                               const cmsis_nn_per_channel_quant_params* quant_params,
-                               const cmsis_nn_dims* input_dims,
-                               const q7_t *input_data,
-                               const cmsis_nn_dims* filter_dims,
-                               const q7_t *filter_data,
-                               const cmsis_nn_dims* bias_dims,
-                               const int32_t *bias_data,
-                               const cmsis_nn_dims* output_dims,
-                               q7_t *output_data);
+ * @return     The function returns <code>ARM_MATH_SUCCESS</code>
+ *
+ * @details
+ *    1. Supported framework: TensorFlow Lite micro
+ *    2. q7 is used as data type eventhough it is s8 data. It is done so to be consistent with existing APIs.
+ *    3. Additional memory is required for optimization. Refer to argument 'ctx' for details.
+ *
+ */
+arm_status arm_convolve_s8(const cmsis_nn_context *ctx,
+                           const cmsis_nn_conv_params *conv_params,
+                           const cmsis_nn_per_channel_quant_params *quant_params,
+                           const cmsis_nn_dims *input_dims,
+                           const q7_t *input_data,
+                           const cmsis_nn_dims *filter_dims,
+                           const q7_t *filter_data,
+                           const cmsis_nn_dims *bias_dims,
+                           const int32_t *bias_data,
+                           const cmsis_nn_dims *output_dims,
+                           q7_t *output_data);
 
-  /**
-   * @brief Get the required buffer size for s8 convolution function
-   *
-   * @param[in]       input_dims            Input (activation) tensor dimensions. Format: [N, H, W, C_IN]
-   * @param[in]       filter_dims           Filter tensor dimensions. Format: [C_OUT, HK, WK, C_IN] where HK and WK are the spatial filter dimensions
-   * @return          The function returns  required buffer size(bytes)
-   *
-   */
-    int32_t arm_convolve_s8_get_buffer_size(const cmsis_nn_dims* input_dims,
-                                            const cmsis_nn_dims* filter_dims);
+/**
+ * @brief Get the required buffer size for s8 convolution function
+ *
+ * @param[in]       input_dims            Input (activation) tensor dimensions. Format: [N, H, W, C_IN]
+ * @param[in]       filter_dims           Filter tensor dimensions. Format: [C_OUT, HK, WK, C_IN] where HK and WK are
+ *                                        the spatial filter dimensions
+ * @return          The function returns  required buffer size(bytes)
+ *
+ */
+int32_t arm_convolve_s8_get_buffer_size(const cmsis_nn_dims *input_dims, const cmsis_nn_dims *filter_dims);
 
-  /**
-   * @brief Basic Q7 convolution function
-   * @param[in]       Im_in       pointer to input tensor
-   * @param[in]       dim_im_in   input tensor dimension
-   * @param[in]       ch_im_in    number of input tensor channels
-   * @param[in]       wt          pointer to kernel weights
-   * @param[in]       ch_im_out   number of filters, i.e., output tensor channels
-   * @param[in]       dim_kernel  filter kernel size
-   * @param[in]       padding     padding sizes
-   * @param[in]       stride      convolution stride
-   * @param[in]       bias        pointer to bias
-   * @param[in]       bias_shift  amount of left-shift for bias
-   * @param[in]       out_shift   amount of right-shift for output
-   * @param[in,out]   Im_out      pointer to output tensor
-   * @param[in]       dim_im_out  output tensor dimension
-   * @param[in,out]   bufferA     pointer to buffer space for input
-   * @param[in,out]   bufferB     pointer to buffer space for output
-   * @return     The function returns <code>ARM_MATH_SUCCESS</code>
-   *
-   */
-    arm_status arm_convolve_HWC_q7_basic(const q7_t * Im_in,
-                                         const uint16_t dim_im_in,
-                                         const uint16_t ch_im_in,
-                                         const q7_t * wt,
-                                         const uint16_t ch_im_out,
-                                         const uint16_t dim_kernel,
-                                         const uint16_t padding,
-                                         const uint16_t stride,
-                                         const q7_t * bias,
-                                         const uint16_t bias_shift,
-                                         const uint16_t out_shift,
-                                         q7_t * Im_out,
-                                         const uint16_t dim_im_out,
-                                         q15_t * bufferA,
-                                         q7_t * bufferB);
+/**
+ * @brief Basic Q7 convolution function
+ * @param[in]       Im_in       pointer to input tensor
+ * @param[in]       dim_im_in   input tensor dimension
+ * @param[in]       ch_im_in    number of input tensor channels
+ * @param[in]       wt          pointer to kernel weights
+ * @param[in]       ch_im_out   number of filters, i.e., output tensor channels
+ * @param[in]       dim_kernel  filter kernel size
+ * @param[in]       padding     padding sizes
+ * @param[in]       stride      convolution stride
+ * @param[in]       bias        pointer to bias
+ * @param[in]       bias_shift  amount of left-shift for bias
+ * @param[in]       out_shift   amount of right-shift for output
+ * @param[in,out]   Im_out      pointer to output tensor
+ * @param[in]       dim_im_out  output tensor dimension
+ * @param[in,out]   bufferA     pointer to buffer space for input
+ * @param[in,out]   bufferB     pointer to buffer space for output
+ * @return     The function returns <code>ARM_MATH_SUCCESS</code>
+ *
+ */
+arm_status arm_convolve_HWC_q7_basic(const q7_t *Im_in,
+                                     const uint16_t dim_im_in,
+                                     const uint16_t ch_im_in,
+                                     const q7_t *wt,
+                                     const uint16_t ch_im_out,
+                                     const uint16_t dim_kernel,
+                                     const uint16_t padding,
+                                     const uint16_t stride,
+                                     const q7_t *bias,
+                                     const uint16_t bias_shift,
+                                     const uint16_t out_shift,
+                                     q7_t *Im_out,
+                                     const uint16_t dim_im_out,
+                                     q15_t *bufferA,
+                                     q7_t *bufferB);
 
-  /**
-   * @brief Basic Q7 convolution function (non-square shape)
-   * @param[in]       Im_in        pointer to input tensor
-   * @param[in]       dim_im_in_x  input tensor dimension x
-   * @param[in]       dim_im_in_y  input tensor dimension y
-   * @param[in]       ch_im_in     number of input tensor channels
-   * @param[in]       wt           pointer to kernel weights
-   * @param[in]       ch_im_out    number of filters, i.e., output tensor channels
-   * @param[in]       dim_kernel_x filter kernel size x
-   * @param[in]       dim_kernel_y filter kernel size y
-   * @param[in]       padding_x    padding size x
-   * @param[in]       padding_y    padding size y
-   * @param[in]       stride_x     convolution stride x
-   * @param[in]       stride_y     convolution stride y
-   * @param[in]       bias         pointer to bias
-   * @param[in]       bias_shift   amount of left-shift for bias
-   * @param[in]       out_shift    amount of right-shift for output
-   * @param[in,out]   Im_out       pointer to output tensor
-   * @param[in]       dim_im_out_x output tensor dimension x
-   * @param[in]       dim_im_out_y output tensor dimension y
-   * @param[in,out]   bufferA      pointer to buffer space for input
-   * @param[in,out]   bufferB      pointer to buffer space for output
-   * @return     The function returns <code>ARM_MATH_SUCCESS</code>
-   */
-    arm_status arm_convolve_HWC_q7_basic_nonsquare(const q7_t * Im_in,
+/**
+ * @brief Basic Q7 convolution function (non-square shape)
+ * @param[in]       Im_in        pointer to input tensor
+ * @param[in]       dim_im_in_x  input tensor dimension x
+ * @param[in]       dim_im_in_y  input tensor dimension y
+ * @param[in]       ch_im_in     number of input tensor channels
+ * @param[in]       wt           pointer to kernel weights
+ * @param[in]       ch_im_out    number of filters, i.e., output tensor channels
+ * @param[in]       dim_kernel_x filter kernel size x
+ * @param[in]       dim_kernel_y filter kernel size y
+ * @param[in]       padding_x    padding size x
+ * @param[in]       padding_y    padding size y
+ * @param[in]       stride_x     convolution stride x
+ * @param[in]       stride_y     convolution stride y
+ * @param[in]       bias         pointer to bias
+ * @param[in]       bias_shift   amount of left-shift for bias
+ * @param[in]       out_shift    amount of right-shift for output
+ * @param[in,out]   Im_out       pointer to output tensor
+ * @param[in]       dim_im_out_x output tensor dimension x
+ * @param[in]       dim_im_out_y output tensor dimension y
+ * @param[in,out]   bufferA      pointer to buffer space for input
+ * @param[in,out]   bufferB      pointer to buffer space for output
+ * @return     The function returns <code>ARM_MATH_SUCCESS</code>
+ */
+arm_status arm_convolve_HWC_q7_basic_nonsquare(const q7_t *Im_in,
+                                               const uint16_t dim_im_in_x,
+                                               const uint16_t dim_im_in_y,
+                                               const uint16_t ch_im_in,
+                                               const q7_t *wt,
+                                               const uint16_t ch_im_out,
+                                               const uint16_t dim_kernel_x,
+                                               const uint16_t dim_kernel_y,
+                                               const uint16_t padding_x,
+                                               const uint16_t padding_y,
+                                               const uint16_t stride_x,
+                                               const uint16_t stride_y,
+                                               const q7_t *bias,
+                                               const uint16_t bias_shift,
+                                               const uint16_t out_shift,
+                                               q7_t *Im_out,
+                                               const uint16_t dim_im_out_x,
+                                               const uint16_t dim_im_out_y,
+                                               q15_t *bufferA,
+                                               q7_t *bufferB);
+
+/**
+ * @brief Basic Q15 convolution function
+ * @param[in]       Im_in       pointer to input tensor
+ * @param[in]       dim_im_in   input tensor dimension
+ * @param[in]       ch_im_in    number of input tensor channels
+ * @param[in]       wt          pointer to kernel weights
+ * @param[in]       ch_im_out   number of filters, i.e., output tensor channels
+ * @param[in]       dim_kernel  filter kernel size
+ * @param[in]       padding     padding sizes
+ * @param[in]       stride      convolution stride
+ * @param[in]       bias        pointer to bias
+ * @param[in]       bias_shift  amount of left-shift for bias
+ * @param[in]       out_shift   amount of right-shift for output
+ * @param[in,out]   Im_out      pointer to output tensor
+ * @param[in]       dim_im_out  output tensor dimension
+ * @param[in,out]   bufferA     pointer to buffer space for input
+ * @param[in,out]   bufferB     pointer to buffer space for output
+ * @return     The function returns <code>ARM_MATH_SUCCESS</code>
+ *
+ */
+arm_status arm_convolve_HWC_q15_basic(const q15_t *Im_in,
+                                      const uint16_t dim_im_in,
+                                      const uint16_t ch_im_in,
+                                      const q15_t *wt,
+                                      const uint16_t ch_im_out,
+                                      const uint16_t dim_kernel,
+                                      const uint16_t padding,
+                                      const uint16_t stride,
+                                      const q15_t *bias,
+                                      const uint16_t bias_shift,
+                                      const uint16_t out_shift,
+                                      q15_t *Im_out,
+                                      const uint16_t dim_im_out,
+                                      q15_t *bufferA,
+                                      q7_t *bufferB);
+
+/**
+ * @brief Fast Q7 convolution function
+ * @param[in]       Im_in       pointer to input tensor
+ * @param[in]       dim_im_in   input tensor dimension
+ * @param[in]       ch_im_in    number of input tensor channels
+ * @param[in]       wt          pointer to kernel weights
+ * @param[in]       ch_im_out   number of filters, i.e., output tensor channels
+ * @param[in]       dim_kernel  filter kernel size
+ * @param[in]       padding     padding sizes
+ * @param[in]       stride      convolution stride
+ * @param[in]       bias        pointer to bias
+ * @param[in]       bias_shift  amount of left-shift for bias
+ * @param[in]       out_shift   amount of right-shift for output
+ * @param[in,out]   Im_out      pointer to output tensor
+ * @param[in]       dim_im_out  output tensor dimension
+ * @param[in,out]   bufferA     pointer to buffer space for input
+ * @param[in,out]   bufferB     pointer to buffer space for output
+ * @return     The function returns either
+ * <code>ARM_MATH_SIZE_MISMATCH</code> or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking.
+ *
+ * This function is the version with full list of optimization tricks, but with
+ * some contraints:
+ *   ch_im_in is multiple of 4
+ *   ch_im_out is multiple of 2
+ */
+arm_status arm_convolve_HWC_q7_fast(const q7_t *Im_in,
+                                    const uint16_t dim_im_in,
+                                    const uint16_t ch_im_in,
+                                    const q7_t *wt,
+                                    const uint16_t ch_im_out,
+                                    const uint16_t dim_kernel,
+                                    const uint16_t padding,
+                                    const uint16_t stride,
+                                    const q7_t *bias,
+                                    const uint16_t bias_shift,
+                                    const uint16_t out_shift,
+                                    q7_t *Im_out,
+                                    const uint16_t dim_im_out,
+                                    q15_t *bufferA,
+                                    q7_t *bufferB);
+
+/**
+ * @brief Fast Q7 convolution function (non-sqaure shape)
+ * @param[in]       Im_in        pointer to input tensor
+ * @param[in]       dim_im_in_x  input tensor dimension x
+ * @param[in]       dim_im_in_y  input tensor dimension y
+ * @param[in]       ch_im_in     number of input tensor channels
+ * @param[in]       wt           pointer to kernel weights
+ * @param[in]       ch_im_out    number of filters, i.e., output tensor channels
+ * @param[in]       dim_kernel_x filter kernel size x
+ * @param[in]       dim_kernel_y filter kernel size y
+ * @param[in]       padding_x    padding size x
+ * @param[in]       padding_y    padding size y
+ * @param[in]       stride_x     convolution stride x
+ * @param[in]       stride_y     convolution stride y
+ * @param[in]       bias         pointer to bias
+ * @param[in]       bias_shift   amount of left-shift for bias
+ * @param[in]       out_shift    amount of right-shift for output
+ * @param[in,out]   Im_out       pointer to output tensor
+ * @param[in]       dim_im_out_x output tensor dimension x
+ * @param[in]       dim_im_out_y output tensor dimension y
+ * @param[in,out]   bufferA      pointer to buffer space for input
+ * @param[in,out]   bufferB      pointer to buffer space for output
+ * @return     The function returns either
+ * <code>ARM_MATH_SIZE_MISMATCH</code> or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking.
+ *
+ * This function is the version with full list of optimization tricks, but with
+ * some contraints:
+ *   ch_im_in is multiple of 4
+ *   ch_im_out is multiple of 2
+ */
+
+arm_status arm_convolve_HWC_q7_fast_nonsquare(const q7_t *Im_in,
+                                              const uint16_t dim_im_in_x,
+                                              const uint16_t dim_im_in_y,
+                                              const uint16_t ch_im_in,
+                                              const q7_t *wt,
+                                              const uint16_t ch_im_out,
+                                              const uint16_t dim_kernel_x,
+                                              const uint16_t dim_kernel_y,
+                                              const uint16_t padding_x,
+                                              const uint16_t padding_y,
+                                              const uint16_t stride_x,
+                                              const uint16_t stride_y,
+                                              const q7_t *bias,
+                                              const uint16_t bias_shift,
+                                              const uint16_t out_shift,
+                                              q7_t *Im_out,
+                                              const uint16_t dim_im_out_x,
+                                              const uint16_t dim_im_out_y,
+                                              q15_t *bufferA,
+                                              q7_t *bufferB);
+
+/**
+ * @brief Fast Q7 version of 1x1 convolution (non-sqaure shape)
+ * @param[in]       Im_in        pointer to input tensor
+ * @param[in]       dim_im_in_x  input tensor dimension x
+ * @param[in]       dim_im_in_y  input tensor dimension y
+ * @param[in]       ch_im_in     number of input tensor channels
+ * @param[in]       wt           pointer to kernel weights
+ * @param[in]       ch_im_out    number of filters, i.e., output tensor channels
+ * @param[in]       dim_kernel_x filter kernel size x
+ * @param[in]       dim_kernel_y filter kernel size y
+ * @param[in]       padding_x    padding size x
+ * @param[in]       padding_y    padding size y
+ * @param[in]       stride_x     convolution stride x
+ * @param[in]       stride_y     convolution stride y
+ * @param[in]       bias         pointer to bias
+ * @param[in]       bias_shift   amount of left-shift for bias
+ * @param[in]       out_shift    amount of right-shift for output
+ * @param[in,out]   Im_out       pointer to output tensor
+ * @param[in]       dim_im_out_x output tensor dimension x
+ * @param[in]       dim_im_out_y output tensor dimension y
+ * @param[in,out]   bufferA      pointer to buffer space for input
+ * @param[in,out]   bufferB      pointer to buffer space for output
+ * @return     The function returns either
+ *                          <code>ARM_MATH_SIZE_MISMATCH</code> if argument constraints fail. or,
+ *                          <code>ARM_MATH_SUCCESS</code> on successful completion.
+ *
+ * This function implement convolution with 1x1 kernel size (i.e., dim_kernel_x=1
+ * and dim_kernel_y=1). It can be used for
+ * second half of MobileNets after depthwise separable convolution.
+ *
+ * This function is the version with full list of optimization tricks, but with
+ * some contraints:
+ *   ch_im_in is multiple of 4
+ *   ch_im_out is multiple of 2
+ */
+arm_status arm_convolve_1x1_HWC_q7_fast_nonsquare(const q7_t *Im_in,
                                                   const uint16_t dim_im_in_x,
                                                   const uint16_t dim_im_in_y,
                                                   const uint16_t ch_im_in,
-                                                  const q7_t * wt,
+                                                  const q7_t *wt,
                                                   const uint16_t ch_im_out,
                                                   const uint16_t dim_kernel_x,
                                                   const uint16_t dim_kernel_y,
@@ -378,647 +568,48 @@
                                                   const uint16_t padding_y,
                                                   const uint16_t stride_x,
                                                   const uint16_t stride_y,
-                                                  const q7_t * bias,
+                                                  const q7_t *bias,
                                                   const uint16_t bias_shift,
                                                   const uint16_t out_shift,
-                                                  q7_t * Im_out,
+                                                  q7_t *Im_out,
                                                   const uint16_t dim_im_out_x,
                                                   const uint16_t dim_im_out_y,
-                                                  q15_t * bufferA,
-                                                  q7_t * bufferB);
+                                                  q15_t *bufferA,
+                                                  q7_t *bufferB);
 
-  /**
-   * @brief Basic Q15 convolution function
-   * @param[in]       Im_in       pointer to input tensor
-   * @param[in]       dim_im_in   input tensor dimension
-   * @param[in]       ch_im_in    number of input tensor channels
-   * @param[in]       wt          pointer to kernel weights
-   * @param[in]       ch_im_out   number of filters, i.e., output tensor channels
-   * @param[in]       dim_kernel  filter kernel size
-   * @param[in]       padding     padding sizes
-   * @param[in]       stride      convolution stride
-   * @param[in]       bias        pointer to bias
-   * @param[in]       bias_shift  amount of left-shift for bias
-   * @param[in]       out_shift   amount of right-shift for output
-   * @param[in,out]   Im_out      pointer to output tensor
-   * @param[in]       dim_im_out  output tensor dimension
-   * @param[in,out]   bufferA     pointer to buffer space for input
-   * @param[in,out]   bufferB     pointer to buffer space for output
-   * @return     The function returns <code>ARM_MATH_SUCCESS</code>
-   *
-   */
-    arm_status arm_convolve_HWC_q15_basic(const q15_t * Im_in,
-                                          const uint16_t dim_im_in,
-                                          const uint16_t ch_im_in,
-                                          const q15_t * wt,
-                                          const uint16_t ch_im_out,
-                                          const uint16_t dim_kernel,
-                                          const uint16_t padding,
-                                          const uint16_t stride,
-                                          const q15_t * bias,
-                                          const uint16_t bias_shift,
-                                          const uint16_t out_shift,
-                                          q15_t * Im_out,
-                                          const uint16_t dim_im_out,
-                                          q15_t * bufferA,
-                                          q7_t * bufferB);
-
-  /**
-   * @brief Fast Q7 convolution function
-   * @param[in]       Im_in       pointer to input tensor
-   * @param[in]       dim_im_in   input tensor dimension
-   * @param[in]       ch_im_in    number of input tensor channels
-   * @param[in]       wt          pointer to kernel weights
-   * @param[in]       ch_im_out   number of filters, i.e., output tensor channels
-   * @param[in]       dim_kernel  filter kernel size
-   * @param[in]       padding     padding sizes
-   * @param[in]       stride      convolution stride
-   * @param[in]       bias        pointer to bias
-   * @param[in]       bias_shift  amount of left-shift for bias
-   * @param[in]       out_shift   amount of right-shift for output
-   * @param[in,out]   Im_out      pointer to output tensor
-   * @param[in]       dim_im_out  output tensor dimension
-   * @param[in,out]   bufferA     pointer to buffer space for input
-   * @param[in,out]   bufferB     pointer to buffer space for output
-   * @return     The function returns either
-   * <code>ARM_MATH_SIZE_MISMATCH</code> or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking.
-   *
-   * This function is the version with full list of optimization tricks, but with
-   * some contraints:
-   *   ch_im_in is multiple of 4
-   *   ch_im_out is multiple of 2
-   */
-    arm_status arm_convolve_HWC_q7_fast(const q7_t * Im_in,
-                                        const uint16_t dim_im_in,
-                                        const uint16_t ch_im_in,
-                                        const q7_t * wt,
-                                        const uint16_t ch_im_out,
-                                        const uint16_t dim_kernel,
-                                        const uint16_t padding,
-                                        const uint16_t stride,
-                                        const q7_t * bias,
-                                        const uint16_t bias_shift,
-                                        const uint16_t out_shift,
-                                        q7_t * Im_out,
-                                        const uint16_t dim_im_out,
-                                        q15_t * bufferA,
-                                        q7_t * bufferB);
-
-  /**
-   * @brief Fast Q7 convolution function (non-sqaure shape)
-   * @param[in]       Im_in        pointer to input tensor
-   * @param[in]       dim_im_in_x  input tensor dimension x
-   * @param[in]       dim_im_in_y  input tensor dimension y
-   * @param[in]       ch_im_in     number of input tensor channels
-   * @param[in]       wt           pointer to kernel weights
-   * @param[in]       ch_im_out    number of filters, i.e., output tensor channels
-   * @param[in]       dim_kernel_x filter kernel size x
-   * @param[in]       dim_kernel_y filter kernel size y
-   * @param[in]       padding_x    padding size x
-   * @param[in]       padding_y    padding size y
-   * @param[in]       stride_x     convolution stride x
-   * @param[in]       stride_y     convolution stride y
-   * @param[in]       bias         pointer to bias
-   * @param[in]       bias_shift   amount of left-shift for bias
-   * @param[in]       out_shift    amount of right-shift for output
-   * @param[in,out]   Im_out       pointer to output tensor
-   * @param[in]       dim_im_out_x output tensor dimension x
-   * @param[in]       dim_im_out_y output tensor dimension y
-   * @param[in,out]   bufferA      pointer to buffer space for input
-   * @param[in,out]   bufferB      pointer to buffer space for output
-   * @return     The function returns either
-   * <code>ARM_MATH_SIZE_MISMATCH</code> or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking.
-   *
-   * This function is the version with full list of optimization tricks, but with
-   * some contraints:
-   *   ch_im_in is multiple of 4
-   *   ch_im_out is multiple of 2
-   */
-
-    arm_status arm_convolve_HWC_q7_fast_nonsquare(const q7_t * Im_in,
-                                                  const uint16_t dim_im_in_x,
-                                                  const uint16_t dim_im_in_y,
-                                                  const uint16_t ch_im_in,
-                                                  const q7_t * wt,
-                                                  const uint16_t ch_im_out,
-                                                  const uint16_t dim_kernel_x,
-                                                  const uint16_t dim_kernel_y,
-                                                  const uint16_t padding_x,
-                                                  const uint16_t padding_y,
-                                                  const uint16_t stride_x,
-                                                  const uint16_t stride_y,
-                                                  const q7_t * bias,
-                                                  const uint16_t bias_shift,
-                                                  const uint16_t out_shift,
-                                                  q7_t * Im_out,
-                                                  const uint16_t dim_im_out_x,
-                                                  const uint16_t dim_im_out_y,
-                                                  q15_t * bufferA,
-                                                  q7_t * bufferB);
-
-  /**
-   * @brief Fast Q7 version of 1x1 convolution (non-sqaure shape)
-   * @param[in]       Im_in        pointer to input tensor
-   * @param[in]       dim_im_in_x  input tensor dimension x
-   * @param[in]       dim_im_in_y  input tensor dimension y
-   * @param[in]       ch_im_in     number of input tensor channels
-   * @param[in]       wt           pointer to kernel weights
-   * @param[in]       ch_im_out    number of filters, i.e., output tensor channels
-   * @param[in]       dim_kernel_x filter kernel size x
-   * @param[in]       dim_kernel_y filter kernel size y
-   * @param[in]       padding_x    padding size x
-   * @param[in]       padding_y    padding size y
-   * @param[in]       stride_x     convolution stride x
-   * @param[in]       stride_y     convolution stride y
-   * @param[in]       bias         pointer to bias
-   * @param[in]       bias_shift   amount of left-shift for bias
-   * @param[in]       out_shift    amount of right-shift for output
-   * @param[in,out]   Im_out       pointer to output tensor
-   * @param[in]       dim_im_out_x output tensor dimension x
-   * @param[in]       dim_im_out_y output tensor dimension y
-   * @param[in,out]   bufferA      pointer to buffer space for input
-   * @param[in,out]   bufferB      pointer to buffer space for output
-   * @return     The function returns either
-   *                          <code>ARM_MATH_SIZE_MISMATCH</code> if argument constraints fail. or,
-   *                          <code>ARM_MATH_SUCCESS</code> on successful completion.
-   *
-   * This function implement convolution with 1x1 kernel size (i.e., dim_kernel_x=1
-   * and dim_kernel_y=1). It can be used for
-   * second half of MobileNets after depthwise separable convolution.
-   *
-   * This function is the version with full list of optimization tricks, but with
-   * some contraints:
-   *   ch_im_in is multiple of 4
-   *   ch_im_out is multiple of 2
-   */
-    arm_status arm_convolve_1x1_HWC_q7_fast_nonsquare(const q7_t * Im_in,
-                                                      const uint16_t dim_im_in_x,
-                                                      const uint16_t dim_im_in_y,
-                                                      const uint16_t ch_im_in,
-                                                      const q7_t * wt,
-                                                      const uint16_t ch_im_out,
-                                                      const uint16_t dim_kernel_x,
-                                                      const uint16_t dim_kernel_y,
-                                                      const uint16_t padding_x,
-                                                      const uint16_t padding_y,
-                                                      const uint16_t stride_x,
-                                                      const uint16_t stride_y,
-                                                      const q7_t * bias,
-                                                      const uint16_t bias_shift,
-                                                      const uint16_t out_shift,
-                                                      q7_t * Im_out,
-                                                      const uint16_t dim_im_out_x,
-                                                      const uint16_t dim_im_out_y,
-                                                      q15_t * bufferA,
-                                                      q7_t * bufferB);
-
-  /**
-   * @brief Fast s8 version for 1x1 convolution (non-square shape)
-   *
-   * @param[in, out] ctx            Function context that contains the additional buffer if required by the implementation.
-                                    arm_convolve_1x1_s8_fast_get_buffer_size will return the buffer_size if required
-   * @param[in]      conv_params    Convolution parameters (e.g. strides, dilations, pads,...).
-   *                                Range of conv_params->input_offset  : [-127, 128]
-   *                                Range of conv_params->output_offset : [-128, 127]
-   * @param[in]      quant_params   Per-channel quantization info.
-   *                                It contains the multiplier and shift values to be applied to each output channel
-   * @param[in]      input_dims     Input (activation) tensor dimensions. Format: [N, H, W, C_IN]
-   * @param[in]      input_data     Input (activation) data pointer. Data type: int8
-   * @param[in]      filter_dims    Filter tensor dimensions. Format: [C_OUT, 1, 1, C_IN]
-   * @param[in]      filter_data    Filter data pointer. Data type: int8
-   * @param[in]      bias_dims      Bias tensor dimensions. Format: [C_OUT]
-   * @param[in]      bias_data      Optional bias data pointer. Data type: int32
-   * @param[in]      output_dims    Output tensor dimensions. Format: [N, H, W, C_OUT]
-   * @param[out]     output_data    Output data pointer. Data type: int8
-   *
-   * @return     The function returns either
-   *                  <code>ARM_MATH_SIZE_MISMATCH</code> if argument constraints fail. or,
-   *                  <code>ARM_MATH_SUCCESS</code> on successful completion.
-   *
-   * @details
-   *   - Supported framework : TensorFlow Lite Micro
-   *   - The following constrains on the arguments apply
-   *      -# input_dims->c is a multiple of 4
-   *      -# conv_params->padding.w = conv_params->padding.h = 0
-   *      -# conv_params->stride.w = conv_params->stride.h = 1
-   *
-   */
-    arm_status arm_convolve_1x1_s8_fast(const cmsis_nn_context* ctx,
-                                        const cmsis_nn_conv_params* conv_params,
-                                        const cmsis_nn_per_channel_quant_params* quant_params,
-                                        const cmsis_nn_dims* input_dims,
-                                        const q7_t *input_data,
-                                        const cmsis_nn_dims* filter_dims,
-                                        const q7_t *filter_data,
-                                        const cmsis_nn_dims* bias_dims,
-                                        const int32_t *bias_data,
-                                        const cmsis_nn_dims* output_dims,
-                                        q7_t *output_data);
-
-  /**
-   * @brief Get the required buffer size for arm_convolve_1x1_s8_fast
-   *
-   * @param[in]       input_dims            Input (activation) dimensions
-   * @return          The function returns the required buffer size in bytes
-   *
-   */
-    int32_t arm_convolve_1x1_s8_fast_get_buffer_size(const cmsis_nn_dims* input_dims);
-
-  /**
-   * @brief 1xn convolution
-   *
-   * @param[in, out] ctx            Function context that contains the additional buffer if required by the implementation.
-                                    arm_convolve_1_x_n_s8_get_buffer_size will return the buffer_size if required
-   * @param[in]      conv_params    Convolution parameters (e.g. strides, dilations, pads,...).
-   *                                Range of conv_params->input_offset  : [-127, 128]
-   *                                Range of conv_params->output_offset : [-128, 127]
-   * @param[in]      quant_params   Per-channel quantization info.
-   *                                It contains the multiplier and shift values to be applied to each output channel
-   * @param[in]      input_dims     Input (activation) tensor dimensions. Format: [N, H, W, C_IN]
-   * @param[in]      input_data     Input (activation) data pointer. Data type: int8
-   * @param[in]      filter_dims    Filter tensor dimensions. Format: [C_OUT, 1, WK, C_IN] where WK is the horizontal spatial filter dimension
-   * @param[in]      filter_data    Filter data pointer. Data type: int8
-   * @param[in]      bias_dims      Bias tensor dimensions. Format: [C_OUT]
-   * @param[in]      bias_data      Optional bias data pointer. Data type: int32
-   * @param[in]      output_dims    Output tensor dimensions. Format: [N, H, W, C_OUT]
-   * @param[out]     output_data    Output data pointer. Data type: int8
-   *
-   * @return     The function returns either
-   *                  <code>ARM_MATH_SIZE_MISMATCH</code> if argument constraints fail. or,
-   *                  <code>ARM_MATH_SUCCESS</code> on successful completion.
-   *
-   * @details
-   *   - Supported framework : TensorFlow Lite Micro
-   *   - The following constrains on the arguments apply
-   *      -# input_dims->n equals 1
-   *      -# ouput_dims->w is a multiple of 4
-   *      -# Explicit constraints(since it is for 1xN convolution)
-   *      -## input_dims->h equals 1
-   *      -## output_dims->h equals 1
-   *      -## filter_dims->h equals 1
-   *@todo  Remove constraint on output_dims->w to make the function generic.
-   *
-   */
-   arm_status arm_convolve_1_x_n_s8(const cmsis_nn_context* ctx,
-                                    const cmsis_nn_conv_params* conv_params,
-                                    const cmsis_nn_per_channel_quant_params* quant_params,
-                                    const cmsis_nn_dims* input_dims,
-                                    const q7_t *input_data,
-                                    const cmsis_nn_dims* filter_dims,
-                                    const q7_t *filter_data,
-                                    const cmsis_nn_dims* bias_dims,
-                                    const int32_t *bias_data,
-                                    const cmsis_nn_dims* output_dims,
-                                    q7_t *output_data);
-
-  /**
-   * @brief Get the required additional buffer size for 1xn convolution
-   *
-   * @param[in]       input_dims            Input (activation) tensor dimensions. Format: [N, H, W, C_IN]
-   * @param[in]       filter_dims           Filter tensor dimensions. Format: [C_OUT, 1, WK, C_IN] where WK is the horizontal spatial filter dimension
-   * @return          The function returns  required buffer size(bytes)
-   *
-   */
-    int32_t arm_convolve_1_x_n_s8_get_buffer_size(const cmsis_nn_dims* input_dims,
-                                                  const cmsis_nn_dims* filter_dims);
-
-  /**
-   * @brief Q7 version of convolution for RGB image
-   * @param[in]       Im_in       pointer to input tensor
-   * @param[in]       dim_im_in   input tensor dimension
-   * @param[in]       ch_im_in    number of input tensor channels
-   * @param[in]       wt          pointer to kernel weights
-   * @param[in]       ch_im_out   number of filters, i.e., output tensor channels
-   * @param[in]       dim_kernel  filter kernel size
-   * @param[in]       padding     padding sizes
-   * @param[in]       stride      convolution stride
-   * @param[in]       bias        pointer to bias
-   * @param[in]       bias_shift  amount of left-shift for bias
-   * @param[in]       out_shift   amount of right-shift for output
-   * @param[in,out]   Im_out      pointer to output tensor
-   * @param[in]       dim_im_out  output tensor dimension
-   * @param[in,out]   bufferA     pointer to buffer space for input
-   * @param[in,out]   bufferB     pointer to buffer space for output
-   * @return     The function returns either
-   * <code>ARM_MATH_SIZE_MISMATCH</code> or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking.
-   *
-   * This kernel is written exclusively for convolution with ch_im_in
-   * equals 3. This applies on the first layer of CNNs which has input
-   * image with RGB format.
-   */
-
-    arm_status arm_convolve_HWC_q7_RGB(const q7_t * Im_in,
-                                       const uint16_t dim_im_in,
-                                       const uint16_t ch_im_in,
-                                       const q7_t * wt,
-                                       const uint16_t ch_im_out,
-                                       const uint16_t dim_kernel,
-                                       const uint16_t padding,
-                                       const uint16_t stride,
-                                       const q7_t * bias,
-                                       const uint16_t bias_shift,
-                                       const uint16_t out_shift,
-                                       q7_t * Im_out,
-                                       const uint16_t dim_im_out,
-                                       q15_t * bufferA,
-                                       q7_t * bufferB);
-
-  /**
-   * @brief Fast Q15 convolution function
-   * @param[in]       Im_in       pointer to input tensor
-   * @param[in]       dim_im_in   input tensor dimension
-   * @param[in]       ch_im_in    number of input tensor channels
-   * @param[in]       wt          pointer to kernel weights
-   * @param[in]       ch_im_out   number of filters, i.e., output tensor channels
-   * @param[in]       dim_kernel  filter kernel size
-   * @param[in]       padding     padding sizes
-   * @param[in]       stride      convolution stride
-   * @param[in]       bias        pointer to bias
-   * @param[in]       bias_shift  amount of left-shift for bias
-   * @param[in]       out_shift   amount of right-shift for output
-   * @param[in,out]   Im_out      pointer to output tensor
-   * @param[in]       dim_im_out  output tensor dimension
-   * @param[in,out]   bufferA     pointer to buffer space for input
-   * @param[in,out]   bufferB     pointer to buffer space for output
-   * @return     The function returns either
-   * <code>ARM_MATH_SIZE_MISMATCH</code> or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking.
-   *
-   * This function is the version with full list of optimization tricks, but with
-   * some contraints:
-   *   ch_im_in is multiple of 2
-   *   ch_im_out is multiple of 2
-   */
-
-    arm_status arm_convolve_HWC_q15_fast(const q15_t * Im_in,
-                                         const uint16_t dim_im_in,
-                                         const uint16_t ch_im_in,
-                                         const q15_t * wt,
-                                         const uint16_t ch_im_out,
-                                         const uint16_t dim_kernel,
-                                         const uint16_t padding,
-                                         const uint16_t stride,
-                                         const q15_t * bias,
-                                         const uint16_t bias_shift,
-                                         const uint16_t out_shift,
-                                         q15_t * Im_out,
-                                         const uint16_t dim_im_out,
-                                         q15_t * bufferA,
-                                         q7_t * bufferB);
-
-  /**
-   * @brief Fast Q15 convolution function (non-sqaure shape)
-   * @param[in]       Im_in        pointer to input tensor
-   * @param[in]       dim_im_in_x  input tensor dimension x
-   * @param[in]       dim_im_in_y  input tensor dimension y
-   * @param[in]       ch_im_in     number of input tensor channels
-   * @param[in]       wt           pointer to kernel weights
-   * @param[in]       ch_im_out    number of filters, i.e., output tensor channels
-   * @param[in]       dim_kernel_x filter kernel size x
-   * @param[in]       dim_kernel_y filter kernel size y
-   * @param[in]       padding_x    padding size x
-   * @param[in]       padding_y    padding size y
-   * @param[in]       stride_x     convolution stride x
-   * @param[in]       stride_y     convolution stride y
-   * @param[in]       bias         pointer to bias
-   * @param[in]       bias_shift   amount of left-shift for bias
-   * @param[in]       out_shift    amount of right-shift for output
-   * @param[in,out]   Im_out       pointer to output tensor
-   * @param[in]       dim_im_out_x output tensor dimension x
-   * @param[in]       dim_im_out_y output tensor dimension y
-   * @param[in,out]   bufferA      pointer to buffer space for input
-   * @param[in,out]   bufferB      pointer to buffer space for output
-   * @return     The function returns either
-   * <code>ARM_MATH_SIZE_MISMATCH</code> or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking.
-   *
-   * @details
-   *
-   * <b>Buffer size:</b>
-   *
-   * bufferA size: 2*ch_im_in*dim_kernel*dim_kernel
-   *
-   * bufferB size: 0
-   *
-   * <b>Input dimension constraints:</b>
-   *
-   * ch_im_in is multiple of 2
-   *
-   * ch_im_out is multipe of 2
-   *
-   */
-
-    arm_status
-    arm_convolve_HWC_q15_fast_nonsquare(const q15_t * Im_in,
-                              const uint16_t dim_im_in_x,
-                              const uint16_t dim_im_in_y,
-                              const uint16_t ch_im_in,
-                              const q15_t * wt,
-                              const uint16_t ch_im_out,
-                              const uint16_t dim_kernel_x,
-                              const uint16_t dim_kernel_y,
-                              const uint16_t padding_x,
-                              const uint16_t padding_y,
-                              const uint16_t stride_x,
-                              const uint16_t stride_y,
-                              const q15_t * bias,
-                              const uint16_t bias_shift,
-                              const uint16_t out_shift,
-                              q15_t * Im_out,
-                              const uint16_t dim_im_out_x,
-                              const uint16_t dim_im_out_y,
-                              q15_t * bufferA,
-                              q7_t * bufferB);
-
-  /**
-   * @brief Q7 depthwise separable convolution function
-   * @param[in]       Im_in       pointer to input tensor
-   * @param[in]       dim_im_in   input tensor dimension
-   * @param[in]       ch_im_in    number of input tensor channels
-   * @param[in]       wt          pointer to kernel weights
-   * @param[in]       ch_im_out   number of filters, i.e., output tensor channels
-   * @param[in]       dim_kernel  filter kernel size
-   * @param[in]       padding     padding sizes
-   * @param[in]       stride      convolution stride
-   * @param[in]       bias        pointer to bias
-   * @param[in]       bias_shift  amount of left-shift for bias
-   * @param[in]       out_shift   amount of right-shift for output
-   * @param[in,out]   Im_out      pointer to output tensor
-   * @param[in]       dim_im_out  output tensor dimension
-   * @param[in,out]   bufferA     pointer to buffer space for input
-   * @param[in,out]   bufferB     pointer to buffer space for output
-   * @return     The function returns either
-   * <code>ARM_MATH_SIZE_MISMATCH</code> or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking.
-   *
-   * This function is the version with full list of optimization tricks, but with
-   * some contraints:
-   *   ch_im_in is multiple of 2
-   *   ch_im_out is multiple of 2
-   */
-
-    arm_status arm_depthwise_separable_conv_HWC_q7(const q7_t * Im_in,
-                                                   const uint16_t dim_im_in,
-                                                   const uint16_t ch_im_in,
-                                                   const q7_t * wt,
-                                                   const uint16_t ch_im_out,
-                                                   const uint16_t dim_kernel,
-                                                   const uint16_t padding,
-                                                   const uint16_t stride,
-                                                   const q7_t * bias,
-                                                   const uint16_t bias_shift,
-                                                   const uint16_t out_shift,
-                                                   q7_t * Im_out,
-                                                   const uint16_t dim_im_out,
-                                                   q15_t * bufferA,
-                                                   q7_t * bufferB);
-
-  /**
-   * @brief Q7 depthwise separable convolution function (non-square shape)
-   * @param[in]       Im_in         pointer to input tensor
-   * @param[in]       dim_im_in_x   input tensor dimension x
-   * @param[in]       dim_im_in_y   input tensor dimension y
-   * @param[in]       ch_im_in      number of input tensor channels
-   * @param[in]       wt            pointer to kernel weights
-   * @param[in]       ch_im_out     number of filters, i.e., output tensor channels
-   * @param[in]       dim_kernel_x  filter kernel size x
-   * @param[in]       dim_kernel_y  filter kernel size y
-   * @param[in]       padding_x     padding sizes x
-   * @param[in]       padding_y     padding sizes y
-   * @param[in]       stride_x      convolution stride x
-   * @param[in]       stride_y      convolution stride y
-   * @param[in]       bias          pointer to bias
-   * @param[in]       bias_shift    amount of left-shift for bias
-   * @param[in]       out_shift     amount of right-shift for output
-   * @param[in,out]   Im_out        pointer to output tensor
-   * @param[in]       dim_im_out_x  output tensor dimension x
-   * @param[in]       dim_im_out_y  output tensor dimension y
-   * @param[in,out]   bufferA       pointer to buffer space for input
-   * @param[in,out]   bufferB       pointer to buffer space for output
-   * @return     The function returns either
-   * <code>ARM_MATH_SIZE_MISMATCH</code> or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking.
-   *
-   * This function is the version with full list of optimization tricks, but with
-   * some contraints:
-   *   ch_im_in is multiple of 2
-   *   ch_im_out is multiple of 2
-   */
-    arm_status arm_depthwise_separable_conv_HWC_q7_nonsquare(const q7_t * Im_in,
-                                                             const uint16_t dim_im_in_x,
-                                                             const uint16_t dim_im_in_y,
-                                                             const uint16_t ch_im_in,
-                                                             const q7_t * wt,
-                                                             const uint16_t ch_im_out,
-                                                             const uint16_t dim_kernel_x,
-                                                             const uint16_t dim_kernel_y,
-                                                             const uint16_t padding_x,
-                                                             const uint16_t padding_y,
-                                                             const uint16_t stride_x,
-                                                             const uint16_t stride_y,
-                                                             const q7_t * bias,
-                                                             const uint16_t bias_shift,
-                                                             const uint16_t out_shift,
-                                                             q7_t * Im_out,
-                                                             const uint16_t dim_im_out_x,
-                                                             const uint16_t dim_im_out_y,
-                                                             q15_t * bufferA,
-                                                             q7_t * bufferB);
-
-   /**
-   * @brief Wrapper function to pick the right optimized s8 depthwise convolution function
-   *
-   * @param[in, out] ctx            Function context (e.g. temporary buffer). Check the function
-   *                                definition file to see if an additional buffer is required.
-   *                                Optional function {API}_get_buffer_size() provides the buffer
-   *                                size if required.
-   * @param[in]      dw_conv_params Depthwise convolution parameters (e.g. strides, dilations, pads,...)
-   *                                dw_conv_params->dilation is not used.
-   *                                Range of dw_conv_params->input_offset : [-127, 128]
-   *                                Range of dw_conv_params->output_offset : [-128, 127]
-   * @param[in]      quant_params   Per-channel quantization info.
-    *                               It contains the multiplier and shift values to be applied to each
-    *                               output channel
-   * @param[in]      input_dims     Input (activation) tensor dimensions. Format: [H, W, C_IN]
-   *                                Batch argument N is not used and assumed to be 1.
-   * @param[in]      input_data     Input (activation) data pointer. Data type: int8
-   * @param[in]      filter_dims    Filter tensor dimensions. Format: [1, H, W, C_OUT]
-   * @param[in]      filter_data    Filter data pointer. Data type: int8
-   * @param[in]      bias_dims      Bias tensor dimensions. Format: [C_OUT]
-   * @param[in]      bias_data      Bias data pointer. Data type: int32
-   * @param[in]      output_dims    Output tensor dimensions. Format: [1, H, W, C_OUT]
-   * @param[in, out] output_data    Output data pointer. Data type: int8
-   * @return     The function returns
-   *                <code>ARM_MATH_SUCCESS</code>   -  Successful completion.
-   *
-   * @details
-   *    - Supported framework: TensorFlow Lite
-   *    - Picks one of the the following functions
-   *        -# arm_depthwise_conv_s8()
-   *        -# arm_depthwise_conv_3x3_s8() - Cortex-M CPUs with DSP extension only
-   *        -# arm_depthwise_conv_s8_opt()
-   *    - q7 is used as data type eventhough it is s8 data. It is done so to be consistent with existing APIs.
-   *    - Check details of arm_depthwise_conv_s8_opt() for potential data that can be accessed outside of the boundary.
-   */
-   arm_status arm_depthwise_conv_wrapper_s8(const cmsis_nn_context *ctx,
-                                            const cmsis_nn_dw_conv_params *dw_conv_params,
-                                            const cmsis_nn_per_channel_quant_params *quant_params,
-                                            const cmsis_nn_dims *input_dims,
-                                            const q7_t *input_data,
-                                            const cmsis_nn_dims *filter_dims,
-                                            const q7_t *filter_data,
-                                            const cmsis_nn_dims *bias_dims,
-                                            const int32_t *bias_data,
-                                            const cmsis_nn_dims *output_dims,
-                                            q7_t *output_data);
-
-   /**
-   * @brief Get size of additional buffer required by arm_depthwise_conv_wrapper_s8()
-   *
-   * @param[in]      dw_conv_params Depthwise convolution parameters (e.g. strides, dilations, pads,...)
-   *                                dw_conv_params->dilation is not used.
-   *                                Range of dw_conv_params->input_offset : [-127, 128]
-   *                                Range of dw_conv_params->input_offset : [-128, 127]
-   * @param[in]      input_dims     Input (activation) tensor dimensions. Format: [H, W, C_IN]
-   *                                Batch argument N is not used and assumed to be 1.
-   * @param[in]      filter_dims    Filter tensor dimensions. Format: [1, H, W, C_OUT]
-   * @param[in]      output_dims    Output tensor dimensions. Format: [1, H, W, C_OUT]
-   * @return                        Size of additional memory required for optimizations in bytes.
-   *
-   */
-   int32_t arm_depthwise_conv_wrapper_s8_get_buffer_size(const cmsis_nn_dw_conv_params *dw_conv_params,
-                                                         const cmsis_nn_dims *input_dims,
-                                                         const cmsis_nn_dims *filter_dims,
-                                                         const cmsis_nn_dims *output_dims);
-
-   /**
-   * @brief Basic s8 depthwise convolution function that doesn't have any constraints on the input dimensions.
-   *
-   * @param[in, out] ctx            Function context (e.g. temporary buffer). Check the function
-   *                                definition file to see if an additional buffer is required.
-   *                                Optional function {API}_get_buffer_size() provides the buffer
-   *                                size if an additional buffer is required.
-   *                                exists if additional memory is.
-   * @param[in]      dw_conv_params Depthwise convolution parameters (e.g. strides, dilations, pads,...)
-   *                                dw_conv_params->dilation is not used.
-   *                                Range of dw_conv_params->input_offset : [-127, 128]
-   *                                Range of dw_conv_params->input_offset : [-128, 127]
-   * @param[in]      quant_params   Per-channel quantization info.
-    *                               It contains the multiplier and shift values to be applied to each
-    *                               output channel
-   * @param[in]      input_dims     Input (activation) tensor dimensions. Format: [1, H, W, C_IN]
-   *                                Batch argument N is not used.
-   * @param[in]      input_data     Input (activation) data pointer. Data type: int8
-   * @param[in]      filter_dims    Filter tensor dimensions. Format: [1, H, W, C_OUT]
-   * @param[in]      filter_data    Filter data pointer. Data type: int8
-   * @param[in]      bias_dims      Bias tensor dimensions. Format: [C_OUT]
-   * @param[in]      bias_data      Bias data pointer. Data type: int32
-   * @param[in]      output_dims    Output tensor dimensions. Format: [1, H, W, C_OUT]
-   * @param[in, out] output_data    Output data pointer. Data type: int8
-   * @return     The function returns <code>ARM_MATH_SUCCESS</code>
-   *
-   * @details
-   *    - Supported framework: TensorFlow Lite
-   *    - q7 is used as data type eventhough it is s8 data. It is done so to be consistent with existing APIs.
-   */
-   arm_status arm_depthwise_conv_s8(const cmsis_nn_context *ctx,
-                                    const cmsis_nn_dw_conv_params *dw_conv_params,
+/**
+ * @brief Fast s8 version for 1x1 convolution (non-square shape)
+ *
+ * @param[in, out] ctx            Function context that contains the additional buffer if required by the function.
+                                  arm_convolve_1x1_s8_fast_get_buffer_size will return the buffer_size if required
+ * @param[in]      conv_params    Convolution parameters (e.g. strides, dilations, pads,...).
+ *                                Range of conv_params->input_offset  : [-127, 128]
+ *                                Range of conv_params->output_offset : [-128, 127]
+ * @param[in]      quant_params   Per-channel quantization info.
+ *                                It contains the multiplier and shift values to be applied to each output channel
+ * @param[in]      input_dims     Input (activation) tensor dimensions. Format: [N, H, W, C_IN]
+ * @param[in]      input_data     Input (activation) data pointer. Data type: int8
+ * @param[in]      filter_dims    Filter tensor dimensions. Format: [C_OUT, 1, 1, C_IN]
+ * @param[in]      filter_data    Filter data pointer. Data type: int8
+ * @param[in]      bias_dims      Bias tensor dimensions. Format: [C_OUT]
+ * @param[in]      bias_data      Optional bias data pointer. Data type: int32
+ * @param[in]      output_dims    Output tensor dimensions. Format: [N, H, W, C_OUT]
+ * @param[out]     output_data    Output data pointer. Data type: int8
+ *
+ * @return     The function returns either
+ *                  <code>ARM_MATH_SIZE_MISMATCH</code> if argument constraints fail. or,
+ *                  <code>ARM_MATH_SUCCESS</code> on successful completion.
+ *
+ * @details
+ *   - Supported framework : TensorFlow Lite Micro
+ *   - The following constrains on the arguments apply
+ *      -# input_dims->c is a multiple of 4
+ *      -# conv_params->padding.w = conv_params->padding.h = 0
+ *      -# conv_params->stride.w = conv_params->stride.h = 1
+ *
+ */
+arm_status arm_convolve_1x1_s8_fast(const cmsis_nn_context *ctx,
+                                    const cmsis_nn_conv_params *conv_params,
                                     const cmsis_nn_per_channel_quant_params *quant_params,
                                     const cmsis_nn_dims *input_dims,
                                     const q7_t *input_data,
@@ -1029,305 +620,717 @@
                                     const cmsis_nn_dims *output_dims,
                                     q7_t *output_data);
 
-   /**
-   * @brief Optimized s8 depthwise convolution function for 3x3 kernel size with some constraints on
-   *        the input arguments(documented below). Refer arm_depthwise_conv_s8() for function
-   *        argument details.
-   *
-   * @return     The function returns one of the following
-   *                <code>ARM_MATH_SIZE_MISMATCH</code> - Unsupported dimension of tensors
-   *                <code>ARM_MATH_ARGUMENT_ERROR</code> - Unsupported pad size along the x axis
-   *                <code>ARM_MATH_SUCCESS</code> - Successful operation
-   *
-   * @details
-   *   - Supported framework : TensorFlow Lite Micro
-   *   - The following constrains on the arguments apply
-   *      -# Number of input channel equals number of output channels
-   *      -# Filter height and width equals 3
-   *      -# Padding along x is either 0 or 1.
-   *
-   */
-   arm_status arm_depthwise_conv_3x3_s8(const cmsis_nn_context *ctx,
-                                        const cmsis_nn_dw_conv_params *dw_conv_params,
-                                        const cmsis_nn_per_channel_quant_params *quant_params,
-                                        const cmsis_nn_dims *input_dims,
-                                        const q7_t *input_data,
-                                        const cmsis_nn_dims *filter_dims,
-                                        const q7_t *filter_data,
-                                        const cmsis_nn_dims *bias_dims,
-                                        const int32_t *bias_data,
-                                        const cmsis_nn_dims *output_dims,
-                                        q7_t *output_data);
-
-   /**
-   * @brief Optimized s8 depthwise convolution function with constraint that in_channel equals out_channel.
-   *        Refer arm_depthwise_conv_s8() for function argument details.
-   *
-   * @return     The function returns one of the following
-   *                <code>ARM_MATH_SIZE_MISMATCH</code> - input channel != output channel or
-   *                                                      ch_mult != 1
-   *                <code>ARM_MATH_SUCCESS</code> - Successful operation
-   *
-   * @note       If number of channels is not a multiple of 4, upto 3 elements outside the boundary will be read out
-   *             for the following if MVE optimizations(Arm Helium Technology) are used.
-   *               - Output shift
-   *               - Output multiplier
-   *               - Output bias
-   *               - kernel
-   * @details
-   *    - Supported framework: TensorFlow Lite
-   *    - The following constrains on the arguments apply
-   *        -# Number of input channel equals number of output channels or ch_mult equals 1
-   *    - q7 is used as data type eventhough it is s8 data. It is done so to be consistent with existing APIs.
-   *    - Reccomended when number of channels is 4 or greater.
-   *
-   */
-   arm_status arm_depthwise_conv_s8_opt(const cmsis_nn_context *ctx,
-                                        const cmsis_nn_dw_conv_params *dw_conv_params,
-                                        const cmsis_nn_per_channel_quant_params *quant_params,
-                                        const cmsis_nn_dims *input_dims,
-                                        const q7_t *input_data,
-                                        const cmsis_nn_dims *filter_dims,
-                                        const q7_t *filter_data,
-                                        const cmsis_nn_dims *bias_dims,
-                                        const int32_t *bias_data,
-                                        const cmsis_nn_dims *output_dims,
-                                        q7_t *output_data);
-
-   /**
-   * @brief Get the required buffer size for optimized s8 depthwise convolution
-   * function with constraint that in_channel equals out_channel.
-   * @param[in]       input_dims     Input (activation) tensor dimensions. Format: [1, H, W, C_IN]
-   *                                 Batch argument N is not used.
-   * @param[in]       filter_dims    Filter tensor dimensions. Format: [1, H, W, C_OUT]
-   * @return          The function returns  required buffer size in bytes
-   *
-   */
-   int32_t arm_depthwise_conv_s8_opt_get_buffer_size(const cmsis_nn_dims* input_dims,
-                                                     const cmsis_nn_dims* filter_dims);
-
- /**
- * @defgroup FC Fully-connected Layer Functions
+/**
+ * @brief Get the required buffer size for arm_convolve_1x1_s8_fast
  *
- * Collection of fully-connected and matrix multiplication functions.
+ * @param[in]       input_dims            Input (activation) dimensions
+ * @return          The function returns the required buffer size in bytes
  *
- * Fully-connected layer is basically a matrix-vector multiplication
- * with bias. The matrix is the weights and the input/output vectors
- * are the activation values. Supported {weight, activation} precisions
- * include {8-bit, 8-bit}, {16-bit, 16-bit}, and {8-bit, 16-bit}.
+ */
+int32_t arm_convolve_1x1_s8_fast_get_buffer_size(const cmsis_nn_dims *input_dims);
+
+/**
+ * @brief 1xn convolution
  *
- * Here we have two types of kernel functions. The basic function
- * implements the function using regular GEMV approach. The opt functions
- * operates with weights in interleaved formats.
+ * @param[in, out] ctx            Function context that contains the additional buffer if required by the function.
+                                  arm_convolve_1_x_n_s8_get_buffer_size will return the buffer_size if required
+ * @param[in]      conv_params    Convolution parameters (e.g. strides, dilations, pads,...).
+ *                                Range of conv_params->input_offset  : [-127, 128]
+ *                                Range of conv_params->output_offset : [-128, 127]
+ * @param[in]      quant_params   Per-channel quantization info.
+ *                                It contains the multiplier and shift values to be applied to each output channel
+ * @param[in]      input_dims     Input (activation) tensor dimensions. Format: [N, H, W, C_IN]
+ * @param[in]      input_data     Input (activation) data pointer. Data type: int8
+ * @param[in]      filter_dims    Filter tensor dimensions. Format: [C_OUT, 1, WK, C_IN] where WK is the horizontal
+ *                                spatial filter dimension
+ * @param[in]      filter_data    Filter data pointer. Data type: int8
+ * @param[in]      bias_dims      Bias tensor dimensions. Format: [C_OUT]
+ * @param[in]      bias_data      Optional bias data pointer. Data type: int32
+ * @param[in]      output_dims    Output tensor dimensions. Format: [N, H, W, C_OUT]
+ * @param[out]     output_data    Output data pointer. Data type: int8
+ *
+ * @return     The function returns either
+ *                  <code>ARM_MATH_SIZE_MISMATCH</code> if argument constraints fail. or,
+ *                  <code>ARM_MATH_SUCCESS</code> on successful completion.
+ *
+ * @details
+ *   - Supported framework : TensorFlow Lite Micro
+ *   - The following constrains on the arguments apply
+ *      -# input_dims->n equals 1
+ *      -# ouput_dims->w is a multiple of 4
+ *      -# Explicit constraints(since it is for 1xN convolution)
+ *      -## input_dims->h equals 1
+ *      -## output_dims->h equals 1
+ *      -## filter_dims->h equals 1
+ *@todo  Remove constraint on output_dims->w to make the function generic.
+ *
+ */
+arm_status arm_convolve_1_x_n_s8(const cmsis_nn_context *ctx,
+                                 const cmsis_nn_conv_params *conv_params,
+                                 const cmsis_nn_per_channel_quant_params *quant_params,
+                                 const cmsis_nn_dims *input_dims,
+                                 const q7_t *input_data,
+                                 const cmsis_nn_dims *filter_dims,
+                                 const q7_t *filter_data,
+                                 const cmsis_nn_dims *bias_dims,
+                                 const int32_t *bias_data,
+                                 const cmsis_nn_dims *output_dims,
+                                 q7_t *output_data);
+
+/**
+ * @brief Get the required additional buffer size for 1xn convolution
+ *
+ * @param[in]       input_dims            Input (activation) tensor dimensions. Format: [N, H, W, C_IN]
+ * @param[in]       filter_dims           Filter tensor dimensions. Format: [C_OUT, 1, WK, C_IN] where WK is the
+ *                                        horizontal spatial filter dimension
+ * @return          The function returns  required buffer size(bytes)
+ *
+ */
+int32_t arm_convolve_1_x_n_s8_get_buffer_size(const cmsis_nn_dims *input_dims, const cmsis_nn_dims *filter_dims);
+
+/**
+ * @brief Q7 version of convolution for RGB image
+ * @param[in]       Im_in       pointer to input tensor
+ * @param[in]       dim_im_in   input tensor dimension
+ * @param[in]       ch_im_in    number of input tensor channels
+ * @param[in]       wt          pointer to kernel weights
+ * @param[in]       ch_im_out   number of filters, i.e., output tensor channels
+ * @param[in]       dim_kernel  filter kernel size
+ * @param[in]       padding     padding sizes
+ * @param[in]       stride      convolution stride
+ * @param[in]       bias        pointer to bias
+ * @param[in]       bias_shift  amount of left-shift for bias
+ * @param[in]       out_shift   amount of right-shift for output
+ * @param[in,out]   Im_out      pointer to output tensor
+ * @param[in]       dim_im_out  output tensor dimension
+ * @param[in,out]   bufferA     pointer to buffer space for input
+ * @param[in,out]   bufferB     pointer to buffer space for output
+ * @return     The function returns either
+ * <code>ARM_MATH_SIZE_MISMATCH</code> or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking.
+ *
+ * This kernel is written exclusively for convolution with ch_im_in
+ * equals 3. This applies on the first layer of CNNs which has input
+ * image with RGB format.
+ */
+
+arm_status arm_convolve_HWC_q7_RGB(const q7_t *Im_in,
+                                   const uint16_t dim_im_in,
+                                   const uint16_t ch_im_in,
+                                   const q7_t *wt,
+                                   const uint16_t ch_im_out,
+                                   const uint16_t dim_kernel,
+                                   const uint16_t padding,
+                                   const uint16_t stride,
+                                   const q7_t *bias,
+                                   const uint16_t bias_shift,
+                                   const uint16_t out_shift,
+                                   q7_t *Im_out,
+                                   const uint16_t dim_im_out,
+                                   q15_t *bufferA,
+                                   q7_t *bufferB);
+
+/**
+ * @brief Fast Q15 convolution function
+ * @param[in]       Im_in       pointer to input tensor
+ * @param[in]       dim_im_in   input tensor dimension
+ * @param[in]       ch_im_in    number of input tensor channels
+ * @param[in]       wt          pointer to kernel weights
+ * @param[in]       ch_im_out   number of filters, i.e., output tensor channels
+ * @param[in]       dim_kernel  filter kernel size
+ * @param[in]       padding     padding sizes
+ * @param[in]       stride      convolution stride
+ * @param[in]       bias        pointer to bias
+ * @param[in]       bias_shift  amount of left-shift for bias
+ * @param[in]       out_shift   amount of right-shift for output
+ * @param[in,out]   Im_out      pointer to output tensor
+ * @param[in]       dim_im_out  output tensor dimension
+ * @param[in,out]   bufferA     pointer to buffer space for input
+ * @param[in,out]   bufferB     pointer to buffer space for output
+ * @return     The function returns either
+ * <code>ARM_MATH_SIZE_MISMATCH</code> or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking.
+ *
+ * This function is the version with full list of optimization tricks, but with
+ * some contraints:
+ *   ch_im_in is multiple of 2
+ *   ch_im_out is multiple of 2
+ */
+
+arm_status arm_convolve_HWC_q15_fast(const q15_t *Im_in,
+                                     const uint16_t dim_im_in,
+                                     const uint16_t ch_im_in,
+                                     const q15_t *wt,
+                                     const uint16_t ch_im_out,
+                                     const uint16_t dim_kernel,
+                                     const uint16_t padding,
+                                     const uint16_t stride,
+                                     const q15_t *bias,
+                                     const uint16_t bias_shift,
+                                     const uint16_t out_shift,
+                                     q15_t *Im_out,
+                                     const uint16_t dim_im_out,
+                                     q15_t *bufferA,
+                                     q7_t *bufferB);
+
+/**
+ * @brief Fast Q15 convolution function (non-sqaure shape)
+ * @param[in]       Im_in        pointer to input tensor
+ * @param[in]       dim_im_in_x  input tensor dimension x
+ * @param[in]       dim_im_in_y  input tensor dimension y
+ * @param[in]       ch_im_in     number of input tensor channels
+ * @param[in]       wt           pointer to kernel weights
+ * @param[in]       ch_im_out    number of filters, i.e., output tensor channels
+ * @param[in]       dim_kernel_x filter kernel size x
+ * @param[in]       dim_kernel_y filter kernel size y
+ * @param[in]       padding_x    padding size x
+ * @param[in]       padding_y    padding size y
+ * @param[in]       stride_x     convolution stride x
+ * @param[in]       stride_y     convolution stride y
+ * @param[in]       bias         pointer to bias
+ * @param[in]       bias_shift   amount of left-shift for bias
+ * @param[in]       out_shift    amount of right-shift for output
+ * @param[in,out]   Im_out       pointer to output tensor
+ * @param[in]       dim_im_out_x output tensor dimension x
+ * @param[in]       dim_im_out_y output tensor dimension y
+ * @param[in,out]   bufferA      pointer to buffer space for input
+ * @param[in,out]   bufferB      pointer to buffer space for output
+ * @return     The function returns either
+ * <code>ARM_MATH_SIZE_MISMATCH</code> or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking.
+ *
+ * @details
+ *
+ * <b>Buffer size:</b>
+ *
+ * bufferA size: 2*ch_im_in*dim_kernel*dim_kernel
+ *
+ * bufferB size: 0
+ *
+ * <b>Input dimension constraints:</b>
+ *
+ * ch_im_in is multiple of 2
+ *
+ * ch_im_out is multipe of 2
  *
  */
 
-    /**
-   * @brief Q7 basic fully-connected layer function
-   * @param[in]       pV          pointer to input vector
-   * @param[in]       pM          pointer to matrix weights
-   * @param[in]       dim_vec     length of the vector
-   * @param[in]       num_of_rows number of rows in weight matrix
-   * @param[in]       bias_shift  amount of left-shift for bias
-   * @param[in]       out_shift   amount of right-shift for output
-   * @param[in]       bias        pointer to bias
-   * @param[in,out]   pOut        pointer to output vector
-   * @param[in,out]   vec_buffer  pointer to buffer space for input
-   * @return     The function returns <code>ARM_MATH_SUCCESS</code>
-   *
-   */
+arm_status arm_convolve_HWC_q15_fast_nonsquare(const q15_t *Im_in,
+                                               const uint16_t dim_im_in_x,
+                                               const uint16_t dim_im_in_y,
+                                               const uint16_t ch_im_in,
+                                               const q15_t *wt,
+                                               const uint16_t ch_im_out,
+                                               const uint16_t dim_kernel_x,
+                                               const uint16_t dim_kernel_y,
+                                               const uint16_t padding_x,
+                                               const uint16_t padding_y,
+                                               const uint16_t stride_x,
+                                               const uint16_t stride_y,
+                                               const q15_t *bias,
+                                               const uint16_t bias_shift,
+                                               const uint16_t out_shift,
+                                               q15_t *Im_out,
+                                               const uint16_t dim_im_out_x,
+                                               const uint16_t dim_im_out_y,
+                                               q15_t *bufferA,
+                                               q7_t *bufferB);
 
-    arm_status arm_fully_connected_q7(const q7_t * pV,
-                                      const q7_t * pM,
+/**
+ * @brief Q7 depthwise separable convolution function
+ * @param[in]       Im_in       pointer to input tensor
+ * @param[in]       dim_im_in   input tensor dimension
+ * @param[in]       ch_im_in    number of input tensor channels
+ * @param[in]       wt          pointer to kernel weights
+ * @param[in]       ch_im_out   number of filters, i.e., output tensor channels
+ * @param[in]       dim_kernel  filter kernel size
+ * @param[in]       padding     padding sizes
+ * @param[in]       stride      convolution stride
+ * @param[in]       bias        pointer to bias
+ * @param[in]       bias_shift  amount of left-shift for bias
+ * @param[in]       out_shift   amount of right-shift for output
+ * @param[in,out]   Im_out      pointer to output tensor
+ * @param[in]       dim_im_out  output tensor dimension
+ * @param[in,out]   bufferA     pointer to buffer space for input
+ * @param[in,out]   bufferB     pointer to buffer space for output
+ * @return     The function returns either
+ * <code>ARM_MATH_SIZE_MISMATCH</code> or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking.
+ *
+ * This function is the version with full list of optimization tricks, but with
+ * some contraints:
+ *   ch_im_in is multiple of 2
+ *   ch_im_out is multiple of 2
+ */
+
+arm_status arm_depthwise_separable_conv_HWC_q7(const q7_t *Im_in,
+                                               const uint16_t dim_im_in,
+                                               const uint16_t ch_im_in,
+                                               const q7_t *wt,
+                                               const uint16_t ch_im_out,
+                                               const uint16_t dim_kernel,
+                                               const uint16_t padding,
+                                               const uint16_t stride,
+                                               const q7_t *bias,
+                                               const uint16_t bias_shift,
+                                               const uint16_t out_shift,
+                                               q7_t *Im_out,
+                                               const uint16_t dim_im_out,
+                                               q15_t *bufferA,
+                                               q7_t *bufferB);
+
+/**
+ * @brief Q7 depthwise separable convolution function (non-square shape)
+ * @param[in]       Im_in         pointer to input tensor
+ * @param[in]       dim_im_in_x   input tensor dimension x
+ * @param[in]       dim_im_in_y   input tensor dimension y
+ * @param[in]       ch_im_in      number of input tensor channels
+ * @param[in]       wt            pointer to kernel weights
+ * @param[in]       ch_im_out     number of filters, i.e., output tensor channels
+ * @param[in]       dim_kernel_x  filter kernel size x
+ * @param[in]       dim_kernel_y  filter kernel size y
+ * @param[in]       padding_x     padding sizes x
+ * @param[in]       padding_y     padding sizes y
+ * @param[in]       stride_x      convolution stride x
+ * @param[in]       stride_y      convolution stride y
+ * @param[in]       bias          pointer to bias
+ * @param[in]       bias_shift    amount of left-shift for bias
+ * @param[in]       out_shift     amount of right-shift for output
+ * @param[in,out]   Im_out        pointer to output tensor
+ * @param[in]       dim_im_out_x  output tensor dimension x
+ * @param[in]       dim_im_out_y  output tensor dimension y
+ * @param[in,out]   bufferA       pointer to buffer space for input
+ * @param[in,out]   bufferB       pointer to buffer space for output
+ * @return     The function returns either
+ * <code>ARM_MATH_SIZE_MISMATCH</code> or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking.
+ *
+ * This function is the version with full list of optimization tricks, but with
+ * some contraints:
+ *   ch_im_in is multiple of 2
+ *   ch_im_out is multiple of 2
+ */
+arm_status arm_depthwise_separable_conv_HWC_q7_nonsquare(const q7_t *Im_in,
+                                                         const uint16_t dim_im_in_x,
+                                                         const uint16_t dim_im_in_y,
+                                                         const uint16_t ch_im_in,
+                                                         const q7_t *wt,
+                                                         const uint16_t ch_im_out,
+                                                         const uint16_t dim_kernel_x,
+                                                         const uint16_t dim_kernel_y,
+                                                         const uint16_t padding_x,
+                                                         const uint16_t padding_y,
+                                                         const uint16_t stride_x,
+                                                         const uint16_t stride_y,
+                                                         const q7_t *bias,
+                                                         const uint16_t bias_shift,
+                                                         const uint16_t out_shift,
+                                                         q7_t *Im_out,
+                                                         const uint16_t dim_im_out_x,
+                                                         const uint16_t dim_im_out_y,
+                                                         q15_t *bufferA,
+                                                         q7_t *bufferB);
+
+/**
+* @brief Wrapper function to pick the right optimized s8 depthwise convolution function
+*
+* @param[in, out] ctx            Function context (e.g. temporary buffer). Check the function
+*                                definition file to see if an additional buffer is required.
+*                                Optional function {API}_get_buffer_size() provides the buffer
+*                                size if required.
+* @param[in]      dw_conv_params Depthwise convolution parameters (e.g. strides, dilations, pads,...)
+*                                dw_conv_params->dilation is not used.
+*                                Range of dw_conv_params->input_offset : [-127, 128]
+*                                Range of dw_conv_params->output_offset : [-128, 127]
+* @param[in]      quant_params   Per-channel quantization info.
+ *                               It contains the multiplier and shift values to be applied to each
+ *                               output channel
+* @param[in]      input_dims     Input (activation) tensor dimensions. Format: [H, W, C_IN]
+*                                Batch argument N is not used and assumed to be 1.
+* @param[in]      input_data     Input (activation) data pointer. Data type: int8
+* @param[in]      filter_dims    Filter tensor dimensions. Format: [1, H, W, C_OUT]
+* @param[in]      filter_data    Filter data pointer. Data type: int8
+* @param[in]      bias_dims      Bias tensor dimensions. Format: [C_OUT]
+* @param[in]      bias_data      Bias data pointer. Data type: int32
+* @param[in]      output_dims    Output tensor dimensions. Format: [1, H, W, C_OUT]
+* @param[in, out] output_data    Output data pointer. Data type: int8
+* @return     The function returns
+*                <code>ARM_MATH_SUCCESS</code>   -  Successful completion.
+*
+* @details
+*    - Supported framework: TensorFlow Lite
+*    - Picks one of the the following functions
+*        -# arm_depthwise_conv_s8()
+*        -# arm_depthwise_conv_3x3_s8() - Cortex-M CPUs with DSP extension only
+*        -# arm_depthwise_conv_s8_opt()
+*    - q7 is used as data type eventhough it is s8 data. It is done so to be consistent with existing APIs.
+*    - Check details of arm_depthwise_conv_s8_opt() for potential data that can be accessed outside of the boundary.
+*/
+arm_status arm_depthwise_conv_wrapper_s8(const cmsis_nn_context *ctx,
+                                         const cmsis_nn_dw_conv_params *dw_conv_params,
+                                         const cmsis_nn_per_channel_quant_params *quant_params,
+                                         const cmsis_nn_dims *input_dims,
+                                         const q7_t *input_data,
+                                         const cmsis_nn_dims *filter_dims,
+                                         const q7_t *filter_data,
+                                         const cmsis_nn_dims *bias_dims,
+                                         const int32_t *bias_data,
+                                         const cmsis_nn_dims *output_dims,
+                                         q7_t *output_data);
+
+/**
+* @brief Get size of additional buffer required by arm_depthwise_conv_wrapper_s8()
+*
+* @param[in]      dw_conv_params Depthwise convolution parameters (e.g. strides, dilations, pads,...)
+*                                dw_conv_params->dilation is not used.
+*                                Range of dw_conv_params->input_offset : [-127, 128]
+*                                Range of dw_conv_params->input_offset : [-128, 127]
+* @param[in]      input_dims     Input (activation) tensor dimensions. Format: [H, W, C_IN]
+*                                Batch argument N is not used and assumed to be 1.
+* @param[in]      filter_dims    Filter tensor dimensions. Format: [1, H, W, C_OUT]
+* @param[in]      output_dims    Output tensor dimensions. Format: [1, H, W, C_OUT]
+* @return                        Size of additional memory required for optimizations in bytes.
+*
+*/
+int32_t arm_depthwise_conv_wrapper_s8_get_buffer_size(const cmsis_nn_dw_conv_params *dw_conv_params,
+                                                      const cmsis_nn_dims *input_dims,
+                                                      const cmsis_nn_dims *filter_dims,
+                                                      const cmsis_nn_dims *output_dims);
+
+/**
+* @brief Basic s8 depthwise convolution function that doesn't have any constraints on the input dimensions.
+*
+* @param[in, out] ctx            Function context (e.g. temporary buffer). Check the function
+*                                definition file to see if an additional buffer is required.
+*                                Optional function {API}_get_buffer_size() provides the buffer
+*                                size if an additional buffer is required.
+*                                exists if additional memory is.
+* @param[in]      dw_conv_params Depthwise convolution parameters (e.g. strides, dilations, pads,...)
+*                                dw_conv_params->dilation is not used.
+*                                Range of dw_conv_params->input_offset : [-127, 128]
+*                                Range of dw_conv_params->input_offset : [-128, 127]
+* @param[in]      quant_params   Per-channel quantization info.
+ *                               It contains the multiplier and shift values to be applied to each
+ *                               output channel
+* @param[in]      input_dims     Input (activation) tensor dimensions. Format: [1, H, W, C_IN]
+*                                Batch argument N is not used.
+* @param[in]      input_data     Input (activation) data pointer. Data type: int8
+* @param[in]      filter_dims    Filter tensor dimensions. Format: [1, H, W, C_OUT]
+* @param[in]      filter_data    Filter data pointer. Data type: int8
+* @param[in]      bias_dims      Bias tensor dimensions. Format: [C_OUT]
+* @param[in]      bias_data      Bias data pointer. Data type: int32
+* @param[in]      output_dims    Output tensor dimensions. Format: [1, H, W, C_OUT]
+* @param[in, out] output_data    Output data pointer. Data type: int8
+* @return     The function returns <code>ARM_MATH_SUCCESS</code>
+*
+* @details
+*    - Supported framework: TensorFlow Lite
+*    - q7 is used as data type eventhough it is s8 data. It is done so to be consistent with existing APIs.
+*/
+arm_status arm_depthwise_conv_s8(const cmsis_nn_context *ctx,
+                                 const cmsis_nn_dw_conv_params *dw_conv_params,
+                                 const cmsis_nn_per_channel_quant_params *quant_params,
+                                 const cmsis_nn_dims *input_dims,
+                                 const q7_t *input_data,
+                                 const cmsis_nn_dims *filter_dims,
+                                 const q7_t *filter_data,
+                                 const cmsis_nn_dims *bias_dims,
+                                 const int32_t *bias_data,
+                                 const cmsis_nn_dims *output_dims,
+                                 q7_t *output_data);
+
+/**
+* @brief Optimized s8 depthwise convolution function for 3x3 kernel size with some constraints on
+*        the input arguments(documented below). Refer arm_depthwise_conv_s8() for function
+*        argument details.
+*
+* @return     The function returns one of the following
+*                <code>ARM_MATH_SIZE_MISMATCH</code> - Unsupported dimension of tensors
+*                <code>ARM_MATH_ARGUMENT_ERROR</code> - Unsupported pad size along the x axis
+*                <code>ARM_MATH_SUCCESS</code> - Successful operation
+*
+* @details
+*   - Supported framework : TensorFlow Lite Micro
+*   - The following constrains on the arguments apply
+*      -# Number of input channel equals number of output channels
+*      -# Filter height and width equals 3
+*      -# Padding along x is either 0 or 1.
+*
+*/
+arm_status arm_depthwise_conv_3x3_s8(const cmsis_nn_context *ctx,
+                                     const cmsis_nn_dw_conv_params *dw_conv_params,
+                                     const cmsis_nn_per_channel_quant_params *quant_params,
+                                     const cmsis_nn_dims *input_dims,
+                                     const q7_t *input_data,
+                                     const cmsis_nn_dims *filter_dims,
+                                     const q7_t *filter_data,
+                                     const cmsis_nn_dims *bias_dims,
+                                     const int32_t *bias_data,
+                                     const cmsis_nn_dims *output_dims,
+                                     q7_t *output_data);
+
+/**
+* @brief Optimized s8 depthwise convolution function with constraint that in_channel equals out_channel.
+*        Refer arm_depthwise_conv_s8() for function argument details.
+*
+* @return     The function returns one of the following
+*                <code>ARM_MATH_SIZE_MISMATCH</code> - input channel != output channel or
+*                                                      ch_mult != 1
+*                <code>ARM_MATH_SUCCESS</code> - Successful operation
+*
+* @note       If number of channels is not a multiple of 4, upto 3 elements outside the boundary will be read out
+*             for the following if MVE optimizations(Arm Helium Technology) are used.
+*               - Output shift
+*               - Output multiplier
+*               - Output bias
+*               - kernel
+* @details
+*    - Supported framework: TensorFlow Lite
+*    - The following constrains on the arguments apply
+*        -# Number of input channel equals number of output channels or ch_mult equals 1
+*    - q7 is used as data type eventhough it is s8 data. It is done so to be consistent with existing APIs.
+*    - Reccomended when number of channels is 4 or greater.
+*
+*/
+arm_status arm_depthwise_conv_s8_opt(const cmsis_nn_context *ctx,
+                                     const cmsis_nn_dw_conv_params *dw_conv_params,
+                                     const cmsis_nn_per_channel_quant_params *quant_params,
+                                     const cmsis_nn_dims *input_dims,
+                                     const q7_t *input_data,
+                                     const cmsis_nn_dims *filter_dims,
+                                     const q7_t *filter_data,
+                                     const cmsis_nn_dims *bias_dims,
+                                     const int32_t *bias_data,
+                                     const cmsis_nn_dims *output_dims,
+                                     q7_t *output_data);
+
+/**
+* @brief Get the required buffer size for optimized s8 depthwise convolution
+* function with constraint that in_channel equals out_channel.
+* @param[in]       input_dims     Input (activation) tensor dimensions. Format: [1, H, W, C_IN]
+*                                 Batch argument N is not used.
+* @param[in]       filter_dims    Filter tensor dimensions. Format: [1, H, W, C_OUT]
+* @return          The function returns  required buffer size in bytes
+*
+*/
+int32_t arm_depthwise_conv_s8_opt_get_buffer_size(const cmsis_nn_dims *input_dims, const cmsis_nn_dims *filter_dims);
+
+/**
+* @defgroup FC Fully-connected Layer Functions
+*
+* Collection of fully-connected and matrix multiplication functions.
+*
+* Fully-connected layer is basically a matrix-vector multiplication
+* with bias. The matrix is the weights and the input/output vectors
+* are the activation values. Supported {weight, activation} precisions
+* include {8-bit, 8-bit}, {16-bit, 16-bit}, and {8-bit, 16-bit}.
+*
+* Here we have two types of kernel functions. The basic function
+* implements the function using regular GEMV approach. The opt functions
+* operates with weights in interleaved formats.
+*
+*/
+
+/**
+*@brief Q7 basic fully-connected layer function
+*@param[in]       pV          pointer to input vector
+*@param[in]       pM          pointer to matrix weights
+*@param[in]       dim_vec     length of the vector
+*@param[in]       num_of_rows number of rows in weight matrix
+*@param[in]       bias_shift  amount of left-shift for bias
+*@param[in]       out_shift   amount of right-shift for output
+*@param[in]       bias        pointer to bias
+*@param[in,out]   pOut        pointer to output vector
+*@param[in,out]   vec_buffer  pointer to buffer space for input
+*@return     The function returns <code>ARM_MATH_SUCCESS</code>
+*
+*/
+
+arm_status arm_fully_connected_q7(const q7_t *pV,
+                                  const q7_t *pM,
+                                  const uint16_t dim_vec,
+                                  const uint16_t num_of_rows,
+                                  const uint16_t bias_shift,
+                                  const uint16_t out_shift,
+                                  const q7_t *bias,
+                                  q7_t *pOut,
+                                  q15_t *vec_buffer);
+
+/**
+* @brief Basic s8 Fully Connected function.
+*
+* @param[in, out] ctx            Function context (e.g. temporary buffer). Check the function
+*                                definition file to see if an additional buffer is required.
+*                                Optional function {API}_get_buffer_size() provides the buffer
+*                                size if an additional buffer is required.
+* @param[in]      fc_params      Fully Connected layer parameters (e.g. strides, dilations, pads,...)
+*                                Range of fc_params->input_offset  : [-127, 128]
+*                                Range of fc_params->filter_offset : [-127, 128]
+*                                Range of fc_params->output_offset : [-128, 127]
+* @param[in]      quant_params   Per-tensor quantization info.
+*                                It contains the multiplier and shift values to be applied to the output tensor.
+* @param[in]      input_dims     Input (activation) tensor dimensions. Format: [N, H, W, C_IN]
+*                                Input dimension is taken as Nx(H * W * C_IN)
+* @param[in]      input_data     Input (activation) data pointer. Data type: int8
+* @param[in]      filter_dims    Two dimensional filter dimensions. Format: [N, C]
+*                                N : accumulation depth and equals (H * W * C_IN) from input_dims
+*                                C : output depth and equals C_OUT in output_dims
+*                                H & W : Not used
+* @param[in]      filter_data    Filter data pointer. Data type: int8
+* @param[in]      bias_dims      Bias tensor dimensions. Format: [C_OUT]
+*                                N, H, W : Not used
+* @param[in]      bias_data      Bias data pointer. Data type: int32
+* @param[in]      output_dims    Output tensor dimensions. Format: [N, C_OUT]
+*                                N : Batches
+*                                C_OUT : Output depth
+*                                H & W : Not used.
+* @param[in, out] output_data    Output data pointer. Data type: int8
+* @return     The function returns <code>ARM_MATH_SUCCESS</code>
+*
+* @details
+*    - Supported framework: TensorFlow Lite
+*    - q7 is used as data type eventhough it is s8 data. It is done so to be consistent with existing APIs.
+*/
+arm_status arm_fully_connected_s8(const cmsis_nn_context *ctx,
+                                  const cmsis_nn_fc_params *fc_params,
+                                  const cmsis_nn_per_tensor_quant_params *quant_params,
+                                  const cmsis_nn_dims *input_dims,
+                                  const q7_t *input_data,
+                                  const cmsis_nn_dims *filter_dims,
+                                  const q7_t *filter_data,
+                                  const cmsis_nn_dims *bias_dims,
+                                  const int32_t *bias_data,
+                                  const cmsis_nn_dims *output_dims,
+                                  q7_t *output_data);
+
+/**
+ * @brief Get the required buffer size for S8 basic fully-connected and
+ * matrix multiplication layer function for TF Lite
+ * @param[in]      filter_dims             dimension of filter
+ * @return         The function returns    required buffer size in bytes
+ *
+ */
+int32_t arm_fully_connected_s8_get_buffer_size(const cmsis_nn_dims *filter_dims);
+
+/**
+ * @brief Q7 opt fully-connected layer function
+ * @param[in]       pV          pointer to input vector
+ * @param[in]       pM          pointer to matrix weights
+ * @param[in]       dim_vec     length of the vector
+ * @param[in]       num_of_rows number of rows in weight matrix
+ * @param[in]       bias_shift  amount of left-shift for bias
+ * @param[in]       out_shift   amount of right-shift for output
+ * @param[in]       bias        pointer to bias
+ * @param[in,out]   pOut        pointer to output vector
+ * @param[in,out]   vec_buffer  pointer to buffer space for input
+ * @return     The function returns <code>ARM_MATH_SUCCESS</code>
+ *
+ */
+
+arm_status arm_fully_connected_q7_opt(const q7_t *pV,
+                                      const q7_t *pM,
                                       const uint16_t dim_vec,
                                       const uint16_t num_of_rows,
                                       const uint16_t bias_shift,
                                       const uint16_t out_shift,
-                                      const q7_t * bias,
-                                      q7_t * pOut,
-                                      q15_t * vec_buffer);
+                                      const q7_t *bias,
+                                      q7_t *pOut,
+                                      q15_t *vec_buffer);
 
-   /**
-   * @brief Basic s8 Fully Connected function.
-   *
-   * @param[in, out] ctx            Function context (e.g. temporary buffer). Check the function
-   *                                definition file to see if an additional buffer is required.
-   *                                Optional function {API}_get_buffer_size() provides the buffer
-   *                                size if an additional buffer is required.
-   * @param[in]      fc_params      Fully Connected layer parameters (e.g. strides, dilations, pads,...)
-   *                                Range of fc_params->input_offset  : [-127, 128]
-   *                                Range of fc_params->filter_offset : [-127, 128]
-   *                                Range of fc_params->output_offset : [-128, 127]
-   * @param[in]      quant_params   Per-tensor quantization info.
-   *                                It contains the multiplier and shift values to be applied to the output tensor.
-   * @param[in]      input_dims     Input (activation) tensor dimensions. Format: [N, H, W, C_IN]
-   *                                Input dimension is taken as Nx(H * W * C_IN)
-   * @param[in]      input_data     Input (activation) data pointer. Data type: int8
-   * @param[in]      filter_dims    Two dimensional filter dimensions. Format: [N, C]
-   *                                N : accumulation depth and equals (H * W * C_IN) from input_dims
-   *                                C : output depth and equals C_OUT in output_dims
-   *                                H & W : Not used
-   * @param[in]      filter_data    Filter data pointer. Data type: int8
-   * @param[in]      bias_dims      Bias tensor dimensions. Format: [C_OUT]
-   *                                N, H, W : Not used
-   * @param[in]      bias_data      Bias data pointer. Data type: int32
-   * @param[in]      output_dims    Output tensor dimensions. Format: [N, C_OUT]
-   *                                N : Batches
-   *                                C_OUT : Output depth
-   *                                H & W : Not used.
-   * @param[in, out] output_data    Output data pointer. Data type: int8
-   * @return     The function returns <code>ARM_MATH_SUCCESS</code>
-   *
-   * @details
-   *    - Supported framework: TensorFlow Lite
-   *    - q7 is used as data type eventhough it is s8 data. It is done so to be consistent with existing APIs.
-   */
-    arm_status
-    arm_fully_connected_s8(const cmsis_nn_context *ctx,
-                           const cmsis_nn_fc_params *fc_params,
-                           const cmsis_nn_per_tensor_quant_params *quant_params,
-                           const cmsis_nn_dims *input_dims,
-                           const q7_t *input_data,
-                           const cmsis_nn_dims *filter_dims,
-                           const q7_t *filter_data,
-                           const cmsis_nn_dims *bias_dims,
-                           const int32_t *bias_data,
-                           const cmsis_nn_dims *output_dims,
-                           q7_t *output_data);
+/**
+ * @brief Q15 basic fully-connected layer function
+ * @param[in]       pV          pointer to input vector
+ * @param[in]       pM          pointer to matrix weights
+ * @param[in]       dim_vec     length of the vector
+ * @param[in]       num_of_rows number of rows in weight matrix
+ * @param[in]       bias_shift  amount of left-shift for bias
+ * @param[in]       out_shift   amount of right-shift for output
+ * @param[in]       bias        pointer to bias
+ * @param[in,out]   pOut        pointer to output vector
+ * @param[in,out]   vec_buffer  pointer to buffer space for input
+ * @return     The function returns <code>ARM_MATH_SUCCESS</code>
+ *
+ */
 
-  /**
-   * @brief Get the required buffer size for S8 basic fully-connected and
-   * matrix multiplication layer function for TF Lite
-   * @param[in]      filter_dims             dimension of filter
-   * @return         The function returns    required buffer size in bytes
-   *
-   */
-    int32_t arm_fully_connected_s8_get_buffer_size(const cmsis_nn_dims *filter_dims);
+arm_status arm_fully_connected_q15(const q15_t *pV,
+                                   const q15_t *pM,
+                                   const uint16_t dim_vec,
+                                   const uint16_t num_of_rows,
+                                   const uint16_t bias_shift,
+                                   const uint16_t out_shift,
+                                   const q15_t *bias,
+                                   q15_t *pOut,
+                                   q15_t *vec_buffer);
 
-  /**
-   * @brief Q7 opt fully-connected layer function
-   * @param[in]       pV          pointer to input vector
-   * @param[in]       pM          pointer to matrix weights
-   * @param[in]       dim_vec     length of the vector
-   * @param[in]       num_of_rows number of rows in weight matrix
-   * @param[in]       bias_shift  amount of left-shift for bias
-   * @param[in]       out_shift   amount of right-shift for output
-   * @param[in]       bias        pointer to bias
-   * @param[in,out]   pOut        pointer to output vector
-   * @param[in,out]   vec_buffer  pointer to buffer space for input
-   * @return     The function returns <code>ARM_MATH_SUCCESS</code>
-   *
-   */
+/**
+ * @brief Q15 opt fully-connected layer function
+ * @param[in]       pV          pointer to input vector
+ * @param[in]       pM          pointer to matrix weights
+ * @param[in]       dim_vec     length of the vector
+ * @param[in]       num_of_rows number of rows in weight matrix
+ * @param[in]       bias_shift  amount of left-shift for bias
+ * @param[in]       out_shift   amount of right-shift for output
+ * @param[in]       bias        pointer to bias
+ * @param[in,out]   pOut        pointer to output vector
+ * @param[in,out]   vec_buffer  pointer to buffer space for input
+ * @return     The function returns <code>ARM_MATH_SUCCESS</code>
+ *
+ */
 
-    arm_status arm_fully_connected_q7_opt(const q7_t * pV,
-                                          const q7_t * pM,
-                                          const uint16_t dim_vec,
-                                          const uint16_t num_of_rows,
-                                          const uint16_t bias_shift,
-                                          const uint16_t out_shift,
-                                          const q7_t * bias,
-                                          q7_t * pOut,
-                                          q15_t * vec_buffer);
-
-  /**
-   * @brief Q15 basic fully-connected layer function
-   * @param[in]       pV          pointer to input vector
-   * @param[in]       pM          pointer to matrix weights
-   * @param[in]       dim_vec     length of the vector
-   * @param[in]       num_of_rows number of rows in weight matrix
-   * @param[in]       bias_shift  amount of left-shift for bias
-   * @param[in]       out_shift   amount of right-shift for output
-   * @param[in]       bias        pointer to bias
-   * @param[in,out]   pOut        pointer to output vector
-   * @param[in,out]   vec_buffer  pointer to buffer space for input
-   * @return     The function returns <code>ARM_MATH_SUCCESS</code>
-   *
-   */
-
-    arm_status arm_fully_connected_q15(const q15_t * pV,
-                                       const q15_t * pM,
+arm_status arm_fully_connected_q15_opt(const q15_t *pV,
+                                       const q15_t *pM,
                                        const uint16_t dim_vec,
                                        const uint16_t num_of_rows,
                                        const uint16_t bias_shift,
                                        const uint16_t out_shift,
-                                       const q15_t * bias,
-                                       q15_t * pOut,
-                                       q15_t * vec_buffer);
+                                       const q15_t *bias,
+                                       q15_t *pOut,
+                                       q15_t *vec_buffer);
 
-  /**
-   * @brief Q15 opt fully-connected layer function
-   * @param[in]       pV          pointer to input vector
-   * @param[in]       pM          pointer to matrix weights
-   * @param[in]       dim_vec     length of the vector
-   * @param[in]       num_of_rows number of rows in weight matrix
-   * @param[in]       bias_shift  amount of left-shift for bias
-   * @param[in]       out_shift   amount of right-shift for output
-   * @param[in]       bias        pointer to bias
-   * @param[in,out]   pOut        pointer to output vector
-   * @param[in,out]   vec_buffer  pointer to buffer space for input
-   * @return     The function returns <code>ARM_MATH_SUCCESS</code>
-   *
-   */
+/**
+ * @brief Mixed Q15-Q7 fully-connected layer function
+ * @param[in]       pV          pointer to input vector
+ * @param[in]       pM          pointer to matrix weights
+ * @param[in]       dim_vec     length of the vector
+ * @param[in]       num_of_rows number of rows in weight matrix
+ * @param[in]       bias_shift  amount of left-shift for bias
+ * @param[in]       out_shift   amount of right-shift for output
+ * @param[in]       bias        pointer to bias
+ * @param[in,out]   pOut        pointer to output vector
+ * @param[in,out]   vec_buffer  pointer to buffer space for input
+ * @return     The function returns <code>ARM_MATH_SUCCESS</code>
+ *
+ */
 
-    arm_status arm_fully_connected_q15_opt(const q15_t * pV,
-                                           const q15_t * pM,
-                                           const uint16_t dim_vec,
-                                           const uint16_t num_of_rows,
-                                           const uint16_t bias_shift,
-                                           const uint16_t out_shift,
-                                           const q15_t * bias,
-                                           q15_t * pOut,
-                                           q15_t * vec_buffer);
+arm_status arm_fully_connected_mat_q7_vec_q15(const q15_t *pV,
+                                              const q7_t *pM,
+                                              const uint16_t dim_vec,
+                                              const uint16_t num_of_rows,
+                                              const uint16_t bias_shift,
+                                              const uint16_t out_shift,
+                                              const q7_t *bias,
+                                              q15_t *pOut,
+                                              q15_t *vec_buffer);
 
-  /**
-   * @brief Mixed Q15-Q7 fully-connected layer function
-   * @param[in]       pV          pointer to input vector
-   * @param[in]       pM          pointer to matrix weights
-   * @param[in]       dim_vec     length of the vector
-   * @param[in]       num_of_rows number of rows in weight matrix
-   * @param[in]       bias_shift  amount of left-shift for bias
-   * @param[in]       out_shift   amount of right-shift for output
-   * @param[in]       bias        pointer to bias
-   * @param[in,out]   pOut        pointer to output vector
-   * @param[in,out]   vec_buffer  pointer to buffer space for input
-   * @return     The function returns <code>ARM_MATH_SUCCESS</code>
-   *
-   */
+/**
+ * @brief Mixed Q15-Q7 opt fully-connected layer function
+ * @param[in]       pV          pointer to input vector
+ * @param[in]       pM          pointer to matrix weights
+ * @param[in]       dim_vec     length of the vector
+ * @param[in]       num_of_rows number of rows in weight matrix
+ * @param[in]       bias_shift  amount of left-shift for bias
+ * @param[in]       out_shift   amount of right-shift for output
+ * @param[in]       bias        pointer to bias
+ * @param[in,out]   pOut        pointer to output vector
+ * @param[in,out]   vec_buffer  pointer to buffer space for input
+ * @return     The function returns <code>ARM_MATH_SUCCESS</code>
+ *
+ */
 
-    arm_status arm_fully_connected_mat_q7_vec_q15(const q15_t * pV,
-                                                  const q7_t * pM,
+arm_status arm_fully_connected_mat_q7_vec_q15_opt(const q15_t *pV,
+                                                  const q7_t *pM,
                                                   const uint16_t dim_vec,
                                                   const uint16_t num_of_rows,
                                                   const uint16_t bias_shift,
                                                   const uint16_t out_shift,
-                                                  const q7_t * bias,
-                                                  q15_t * pOut,
-                                                  q15_t * vec_buffer);
-
-  /**
-   * @brief Mixed Q15-Q7 opt fully-connected layer function
-   * @param[in]       pV          pointer to input vector
-   * @param[in]       pM          pointer to matrix weights
-   * @param[in]       dim_vec     length of the vector
-   * @param[in]       num_of_rows number of rows in weight matrix
-   * @param[in]       bias_shift  amount of left-shift for bias
-   * @param[in]       out_shift   amount of right-shift for output
-   * @param[in]       bias        pointer to bias
-   * @param[in,out]   pOut        pointer to output vector
-   * @param[in,out]   vec_buffer  pointer to buffer space for input
-   * @return     The function returns <code>ARM_MATH_SUCCESS</code>
-   *
-   */
-
-    arm_status arm_fully_connected_mat_q7_vec_q15_opt(const q15_t * pV,
-                                                      const q7_t * pM,
-                                                      const uint16_t dim_vec,
-                                                      const uint16_t num_of_rows,
-                                                      const uint16_t bias_shift,
-                                                      const uint16_t out_shift,
-                                                      const q7_t * bias,
-                                                      q15_t * pOut,
-                                                      q15_t * vec_buffer);
+                                                  const q7_t *bias,
+                                                  q15_t *pOut,
+                                                  q15_t *vec_buffer);
 
 /**
  * @brief Matrix-Multiplication Kernels for Convolution
@@ -1341,108 +1344,108 @@
  *
  */
 
-   /**
-   * @brief Matrix-multiplication function for convolution
-   * @param[in]       pA          pointer to operand A
-   * @param[in]       pInBuffer   pointer to operand B, always conssists of 2 vectors
-   * @param[in]       ch_im_out   numRow of A
-   * @param[in]       numCol_A    numCol of A
-   * @param[in]       bias_shift  amount of left-shift for bias
-   * @param[in]       out_shift   amount of right-shift for output
-   * @param[in]       bias        the bias
-   * @param[in,out]   pOut        pointer to output
-   * @return     The function returns the incremented output pointer
-   */
+/**
+* @brief Matrix-multiplication function for convolution
+* @param[in]       pA          pointer to operand A
+* @param[in]       pInBuffer   pointer to operand B, always conssists of 2 vectors
+* @param[in]       ch_im_out   numRow of A
+* @param[in]       numCol_A    numCol of A
+* @param[in]       bias_shift  amount of left-shift for bias
+* @param[in]       out_shift   amount of right-shift for output
+* @param[in]       bias        the bias
+* @param[in,out]   pOut        pointer to output
+* @return     The function returns the incremented output pointer
+*/
 
-    q7_t     *arm_nn_mat_mult_kernel_q7_q15(const q7_t * pA,
-                                            const q15_t * pInBuffer,
-                                            const uint16_t ch_im_out,
-                                            const uint16_t numCol_A,
-                                            const uint16_t bias_shift,
-                                            const uint16_t out_shift,
-                                            const q7_t * bias,
-                                            q7_t * pOut);
-   /**
-   * @brief Matrix-multiplication function for convolution with per-channel requantization.
-   * @param[in]       input_a     pointer to operand A
-   * @param[in]       input_b     pointer to operand B, always consists of 2 vectors.
-   * @param[in]       output_ch   number of rows of A
-   * @param[in]       out_shift  pointer to per output channel requantization shift parameter.
-   * @param[in]       out_mult   pointer to per output channel requantization multiplier parameter.
-   * @param[in]       out_offset      output tensor offset.
-   * @param[in]       activation_min   minimum value to clamp the output to. Range : int8
-   * @param[in]       activation_max   maximum value to clamp the output to. Range : int8
-   * @param[in]       num_col_a   number of columns of A
-   * @param[in]       output_bias per output channel bias. Range : int32
-   * @param[in,out]   out_0       pointer to output
-   * @return     The function returns one of the two
-   *              1. The incremented output pointer for a successful operation or
-   *              2. NULL if implementation is not available.
-   *
-   * @details   This function does the matrix multiplication of weight matrix for all output channels
-   *            with 2 columns from im2col and produces two elements/output_channel. The outputs are
-   *            clamped in the range provided by activation min and max.
-   *            Supported framework: TensorFlow Lite micro.
-   */
-    q7_t *arm_nn_mat_mult_kernel_s8_s16(const q7_t *input_a,
-                                        const q15_t *input_b,
-                                        const uint16_t output_ch,
-                                        const int32_t *out_shift,
-                                        const int32_t *out_mult,
-                                        const int32_t out_offset,
-                                        const int16_t activation_min,
-                                        const int16_t activation_max,
-                                        const uint16_t num_col_a,
-                                        const int32_t *const output_bias,
-                                        q7_t *out_0);
+q7_t *arm_nn_mat_mult_kernel_q7_q15(const q7_t *pA,
+                                    const q15_t *pInBuffer,
+                                    const uint16_t ch_im_out,
+                                    const uint16_t numCol_A,
+                                    const uint16_t bias_shift,
+                                    const uint16_t out_shift,
+                                    const q7_t *bias,
+                                    q7_t *pOut);
+/**
+* @brief Matrix-multiplication function for convolution with per-channel requantization.
+* @param[in]       input_a     pointer to operand A
+* @param[in]       input_b     pointer to operand B, always consists of 2 vectors.
+* @param[in]       output_ch   number of rows of A
+* @param[in]       out_shift  pointer to per output channel requantization shift parameter.
+* @param[in]       out_mult   pointer to per output channel requantization multiplier parameter.
+* @param[in]       out_offset      output tensor offset.
+* @param[in]       activation_min   minimum value to clamp the output to. Range : int8
+* @param[in]       activation_max   maximum value to clamp the output to. Range : int8
+* @param[in]       num_col_a   number of columns of A
+* @param[in]       output_bias per output channel bias. Range : int32
+* @param[in,out]   out_0       pointer to output
+* @return     The function returns one of the two
+*              1. The incremented output pointer for a successful operation or
+*              2. NULL if implementation is not available.
+*
+* @details   This function does the matrix multiplication of weight matrix for all output channels
+*            with 2 columns from im2col and produces two elements/output_channel. The outputs are
+*            clamped in the range provided by activation min and max.
+*            Supported framework: TensorFlow Lite micro.
+*/
+q7_t *arm_nn_mat_mult_kernel_s8_s16(const q7_t *input_a,
+                                    const q15_t *input_b,
+                                    const uint16_t output_ch,
+                                    const int32_t *out_shift,
+                                    const int32_t *out_mult,
+                                    const int32_t out_offset,
+                                    const int16_t activation_min,
+                                    const int16_t activation_max,
+                                    const uint16_t num_col_a,
+                                    const int32_t *const output_bias,
+                                    q7_t *out_0);
 
-   /**
-   * @brief Matrix-multiplication of re-ordered input B with A.
-   *
-   * @details  For arguments, refer arm_nn_mat_mult_kernel_s8_s16. The re-ordering is a consequence
-   *           of sign extension done by the SXTB16 command on input_b. The outputs are clamped in the range
-   *           provided by activation min and max.
-   *   * @details
-   *   - Supported framework : TensorFlow Lite Micro
-   *   - The following constrains on the arguments apply
-   *      -# num_col_a is a multiple of 4
-   *      -# output_ch is a multiple of 2
-   *
-   */
-    q7_t *arm_nn_mat_mult_kernel_s8_s16_reordered(const q7_t *input_a,
-                                                  const q15_t *input_b,
-                                                  const uint16_t output_ch,
-                                                  const int32_t *out_shift,
-                                                  const int32_t *out_mult,
-                                                  const int32_t out_offset,
-                                                  const int16_t activation_min,
-                                                  const int16_t activation_max,
-                                                  const uint16_t num_col_a,
-                                                  const int32_t *const output_bias,
-                                                  q7_t *out_0);
+/**
+* @brief Matrix-multiplication of re-ordered input B with A.
+*
+* @details  For arguments, refer arm_nn_mat_mult_kernel_s8_s16. The re-ordering is a consequence
+*           of sign extension done by the SXTB16 command on input_b. The outputs are clamped in the range
+*           provided by activation min and max.
+*   * @details
+*   - Supported framework : TensorFlow Lite Micro
+*   - The following constrains on the arguments apply
+*      -# num_col_a is a multiple of 4
+*      -# output_ch is a multiple of 2
+*
+*/
+q7_t *arm_nn_mat_mult_kernel_s8_s16_reordered(const q7_t *input_a,
+                                              const q15_t *input_b,
+                                              const uint16_t output_ch,
+                                              const int32_t *out_shift,
+                                              const int32_t *out_mult,
+                                              const int32_t out_offset,
+                                              const int16_t activation_min,
+                                              const int16_t activation_max,
+                                              const uint16_t num_col_a,
+                                              const int32_t *const output_bias,
+                                              q7_t *out_0);
 
-    /**
-   * @brief Matrix-multiplication function for convolution with reordered columns
-   * @param[in]       pA          pointer to operand A
-   * @param[in]       pInBuffer   pointer to operand B, always conssists of 2 vectors
-   * @param[in]       ch_im_out   numRow of A
-   * @param[in]       numCol_A    numCol of A
-   * @param[in]       bias_shift  amount of left-shift for bias
-   * @param[in]       out_shift   amount of right-shift for output
-   * @param[in]       bias        the bias
-   * @param[in,out]   pOut        pointer to output
-   * @return     The function returns the incremented output pointer
-   *
-   * @details  This function assumes that data in pInBuffer are reordered
-   */
-    q7_t     *arm_nn_mat_mult_kernel_q7_q15_reordered(const q7_t * pA,
-                                                      const q15_t * pInBuffer,
-                                                      const uint16_t ch_im_out,
-                                                      const uint16_t numCol_A,
-                                                      const uint16_t bias_shift,
-                                                      const uint16_t out_shift,
-                                                      const q7_t * bias,
-                                                      q7_t * pOut);
+/**
+*@brief Matrix-multiplication function for convolution with reordered columns
+*@param[in]       pA          pointer to operand A
+*@param[in]       pInBuffer   pointer to operand B, always conssists of 2 vectors
+*@param[in]       ch_im_out   numRow of A
+*@param[in]       numCol_A    numCol of A
+*@param[in]       bias_shift  amount of left-shift for bias
+*@param[in]       out_shift   amount of right-shift for output
+*@param[in]       bias        the bias
+*@param[in,out]   pOut        pointer to output
+*@return     The function returns the incremented output pointer
+*
+*@details  This function assumes that data in pInBuffer are reordered
+*/
+q7_t *arm_nn_mat_mult_kernel_q7_q15_reordered(const q7_t *pA,
+                                              const q15_t *pInBuffer,
+                                              const uint16_t ch_im_out,
+                                              const uint16_t numCol_A,
+                                              const uint16_t bias_shift,
+                                              const uint16_t out_shift,
+                                              const q7_t *bias,
+                                              q7_t *pOut);
 
 #ifdef __cplusplus
 }
@@ -1455,8 +1458,7 @@
  */
 
 #ifdef __cplusplus
-extern    "C"
-{
+extern "C" {
 #endif
 
 /**
@@ -1486,22 +1488,22 @@
    * @param[in]       block_size              number of samples
    * @return          The function returns    ARM_MATH_SUCCESS
    */
-    arm_status arm_elementwise_add_s8(const int8_t *input_1_vect,
-                                      const int8_t *input_2_vect,
-                                      const int32_t input_1_offset,
-                                      const int32_t input_1_mult,
-                                      const int32_t input_1_shift,
-                                      const int32_t input_2_offset,
-                                      const int32_t input_2_mult,
-                                      const int32_t input_2_shift,
-                                      const int32_t left_shift,
-                                      int8_t *output,
-                                      const int32_t out_offset,
-                                      const int32_t out_mult,
-                                      const int32_t out_shift,
-                                      const int32_t out_activation_min,
-                                      const int32_t out_activation_max,
-                                      const uint32_t block_size);
+arm_status arm_elementwise_add_s8(const int8_t *input_1_vect,
+                                  const int8_t *input_2_vect,
+                                  const int32_t input_1_offset,
+                                  const int32_t input_1_mult,
+                                  const int32_t input_1_shift,
+                                  const int32_t input_2_offset,
+                                  const int32_t input_2_mult,
+                                  const int32_t input_2_shift,
+                                  const int32_t left_shift,
+                                  int8_t *output,
+                                  const int32_t out_offset,
+                                  const int32_t out_mult,
+                                  const int32_t out_shift,
+                                  const int32_t out_activation_min,
+                                  const int32_t out_activation_max,
+                                  const uint32_t block_size);
 
 /**
    * @brief s8 element wise multiplication
@@ -1520,17 +1522,17 @@
    *
    * @details   Supported framework: TensorFlow Lite micro
    */
-  arm_status arm_elementwise_mul_s8(const int8_t *input_1_vect,
-                                    const int8_t *input_2_vect,
-                                    const int32_t input_1_offset,
-                                    const int32_t input_2_offset,
-                                    int8_t *output,
-                                    const int32_t out_offset,
-                                    const int32_t out_mult,
-                                    const int32_t out_shift,
-                                    const int32_t out_activation_min,
-                                    const int32_t out_activation_max,
-                                    const uint32_t block_size);
+arm_status arm_elementwise_mul_s8(const int8_t *input_1_vect,
+                                  const int8_t *input_2_vect,
+                                  const int32_t input_1_offset,
+                                  const int32_t input_2_offset,
+                                  int8_t *output,
+                                  const int32_t out_offset,
+                                  const int32_t out_mult,
+                                  const int32_t out_shift,
+                                  const int32_t out_activation_min,
+                                  const int32_t out_activation_max,
+                                  const uint32_t block_size);
 /**
  * @defgroup Acti Activation Functions
  *
@@ -1539,63 +1541,61 @@
  *
  */
 
-  /**
-   * @brief Q7 RELU function
-   * @param[in,out]   data        pointer to input
-   * @param[in]       size        number of elements
-   * @return none.
-   */
+/**
+ * @brief Q7 RELU function
+ * @param[in,out]   data        pointer to input
+ * @param[in]       size        number of elements
+ * @return none.
+ */
 
-    void      arm_relu_q7(q7_t *data, uint16_t size);
+void arm_relu_q7(q7_t *data, uint16_t size);
 
-  /**
-   * @brief s8 ReLU6 function
-   * @param[in,out]   data        pointer to input
-   * @param[in]       size        number of elements
-   */
+/**
+ * @brief s8 ReLU6 function
+ * @param[in,out]   data        pointer to input
+ * @param[in]       size        number of elements
+ */
 
-    void      arm_relu6_s8(q7_t *data, uint16_t size);
+void arm_relu6_s8(q7_t *data, uint16_t size);
 
-  /**
-   * @brief Q15 RELU function
-   * @param[in,out]   data        pointer to input
-   * @param[in]       size        number of elements
-   * @return none.
-   */
+/**
+ * @brief Q15 RELU function
+ * @param[in,out]   data        pointer to input
+ * @param[in]       size        number of elements
+ * @return none.
+ */
 
-    void      arm_relu_q15(q15_t *data, uint16_t size);
+void arm_relu_q15(q15_t *data, uint16_t size);
 
-  /**
-   * @brief Q7 neural network activation function using direct table look-up
-   * @param[in,out]   data        pointer to input
-   * @param[in]       size        number of elements
-   * @param[in]       int_width   bit-width of the integer part, assume to be smaller than 3
-   * @param[in]       type        type of activation functions
-   * @return none.
-   */
+/**
+ * @brief Q7 neural network activation function using direct table look-up
+ * @param[in,out]   data        pointer to input
+ * @param[in]       size        number of elements
+ * @param[in]       int_width   bit-width of the integer part, assume to be smaller than 3
+ * @param[in]       type        type of activation functions
+ * @return none.
+ */
 
-    void      arm_nn_activations_direct_q7(q7_t * data, uint16_t size, uint16_t int_width,
-                                           arm_nn_activation_type type);
+void arm_nn_activations_direct_q7(q7_t *data, uint16_t size, uint16_t int_width, arm_nn_activation_type type);
 
-  /**
-   * @brief Q15 neural network activation function using direct table look-up
-   * @param[in,out]   data        pointer to input
-   * @param[in]       size        number of elements
-   * @param[in]       int_width   bit-width of the integer part, assume to be smaller than 3
-   * @param[in]       type        type of activation functions
-   * @return none.
-   *
-   * @details
-   *
-   * This is the direct table look-up approach.
-   *
-   * Assume here the integer part of the fixed-point is <= 3.
-   * More than 3 just not making much sense, makes no difference with
-   * saturation followed by any of these activation functions.
-   */
+/**
+ * @brief Q15 neural network activation function using direct table look-up
+ * @param[in,out]   data        pointer to input
+ * @param[in]       size        number of elements
+ * @param[in]       int_width   bit-width of the integer part, assume to be smaller than 3
+ * @param[in]       type        type of activation functions
+ * @return none.
+ *
+ * @details
+ *
+ * This is the direct table look-up approach.
+ *
+ * Assume here the integer part of the fixed-point is <= 3.
+ * More than 3 just not making much sense, makes no difference with
+ * saturation followed by any of these activation functions.
+ */
 
-    void      arm_nn_activations_direct_q15(q15_t * data, uint16_t size, uint16_t int_width,
-                                            arm_nn_activation_type type);
+void arm_nn_activations_direct_q15(q15_t *data, uint16_t size, uint16_t int_width, arm_nn_activation_type type);
 
 /**
  * @defgroup Pooling Pooling Functions
@@ -1604,129 +1604,128 @@
  *
  */
 
-  /**
-   * @brief Q7 max pooling function
-   * @param[in]       Im_in       pointer to input tensor
-   * @param[in]       dim_im_in   input tensor dimension
-   * @param[in]       ch_im_in    number of input tensor channels
-   * @param[in]       dim_kernel  filter kernel size
-   * @param[in]       padding     padding sizes
-   * @param[in]       stride      convolution stride
-   * @param[in]       dim_im_out  output tensor dimension
-   * @param[in,out]   bufferA     pointer to buffer space for input
-   * @param[in,out]   Im_out      pointer to output tensor
-   * @return none.
-   *
-   */
+/**
+ * @brief Q7 max pooling function
+ * @param[in]       Im_in       pointer to input tensor
+ * @param[in]       dim_im_in   input tensor dimension
+ * @param[in]       ch_im_in    number of input tensor channels
+ * @param[in]       dim_kernel  filter kernel size
+ * @param[in]       padding     padding sizes
+ * @param[in]       stride      convolution stride
+ * @param[in]       dim_im_out  output tensor dimension
+ * @param[in,out]   bufferA     pointer to buffer space for input
+ * @param[in,out]   Im_out      pointer to output tensor
+ * @return none.
+ *
+ */
 
-    void      arm_maxpool_q7_HWC(q7_t * Im_in,
-                                 const uint16_t dim_im_in,
-                                 const uint16_t ch_im_in,
-                                 const uint16_t dim_kernel,
-                                 const uint16_t padding,
-                                 const uint16_t stride,
-                                 const uint16_t dim_im_out,
-                                 q7_t * bufferA,
-                                 q7_t * Im_out);
+void arm_maxpool_q7_HWC(q7_t *Im_in,
+                        const uint16_t dim_im_in,
+                        const uint16_t ch_im_in,
+                        const uint16_t dim_kernel,
+                        const uint16_t padding,
+                        const uint16_t stride,
+                        const uint16_t dim_im_out,
+                        q7_t *bufferA,
+                        q7_t *Im_out);
 
-  /**
-   * @brief Q7 average pooling function
-   * @param[in]       Im_in       pointer to input tensor
-   * @param[in]       dim_im_in   input tensor dimension
-   * @param[in]       ch_im_in    number of input tensor channels
-   * @param[in]       dim_kernel  filter kernel size
-   * @param[in]       padding     padding sizes
-   * @param[in]       stride      convolution stride
-   * @param[in]       dim_im_out  output tensor dimension
-   * @param[in,out]   bufferA     pointer to buffer space for input
-   * @param[in,out]   Im_out      pointer to output tensor
-   * @return none.
-   *
-   */
+/**
+ * @brief Q7 average pooling function
+ * @param[in]       Im_in       pointer to input tensor
+ * @param[in]       dim_im_in   input tensor dimension
+ * @param[in]       ch_im_in    number of input tensor channels
+ * @param[in]       dim_kernel  filter kernel size
+ * @param[in]       padding     padding sizes
+ * @param[in]       stride      convolution stride
+ * @param[in]       dim_im_out  output tensor dimension
+ * @param[in,out]   bufferA     pointer to buffer space for input
+ * @param[in,out]   Im_out      pointer to output tensor
+ * @return none.
+ *
+ */
 
-    void      arm_avepool_q7_HWC(q7_t * Im_in,
-                                 const uint16_t dim_im_in,
-                                 const uint16_t ch_im_in,
-                                 const uint16_t dim_kernel,
-                                 const uint16_t padding,
-                                 const uint16_t stride,
-                                 const uint16_t dim_im_out,
-                                 q7_t * bufferA,
-                                 q7_t * Im_out);
+void arm_avepool_q7_HWC(q7_t *Im_in,
+                        const uint16_t dim_im_in,
+                        const uint16_t ch_im_in,
+                        const uint16_t dim_kernel,
+                        const uint16_t padding,
+                        const uint16_t stride,
+                        const uint16_t dim_im_out,
+                        q7_t *bufferA,
+                        q7_t *Im_out);
 
-   /**
-   * @brief s8 average pooling function.
-   *
-   * @param[in, out] ctx            Function context (e.g. temporary buffer). Check the function
-   *                                definition file to see if an additional buffer is required.
-   *                                Optional function {API}_get_buffer_size() provides the buffer
-   *                                size if an additional buffer is required.
-   * @param[in]      pool_params    Pooling parameters
-   * @param[in]      input_dims     Input (activation) tensor dimensions. Format: [H, W, C_IN]
-   *                                Argument 'N' is not used.
-   * @param[in]      input_data     Input (activation) data pointer. Data type: int8
-   * @param[in]      filter_dims    Filter tensor dimensions. Format: [H, W]
-   *                                Argument N and C are not used.
-   * @param[in]      output_dims    Output tensor dimensions. Format: [H, W, C_OUT]
-   *                                Argument N is not used.
-   *                                C_OUT equals C_IN.
-   * @param[in, out] output_data    Output data pointer. Data type: int8
-   * @return                        The function returns
-   *                                    <code>ARM_MATH_SUCCESS</code> - Successful operation
-   *
-   * @details
-   *    - Supported Framework: TensorFlow Lite
-   *
-   */
-   arm_status arm_avgpool_s8(const cmsis_nn_context *ctx,
-                             const cmsis_nn_pool_params *pool_params,
-                             const cmsis_nn_dims *input_dims,
-                             const q7_t *input_data,
-                             const cmsis_nn_dims *filter_dims,
-                             const cmsis_nn_dims *output_dims,
-                             q7_t *output_data);
+/**
+* @brief s8 average pooling function.
+*
+* @param[in, out] ctx            Function context (e.g. temporary buffer). Check the function
+*                                definition file to see if an additional buffer is required.
+*                                Optional function {API}_get_buffer_size() provides the buffer
+*                                size if an additional buffer is required.
+* @param[in]      pool_params    Pooling parameters
+* @param[in]      input_dims     Input (activation) tensor dimensions. Format: [H, W, C_IN]
+*                                Argument 'N' is not used.
+* @param[in]      input_data     Input (activation) data pointer. Data type: int8
+* @param[in]      filter_dims    Filter tensor dimensions. Format: [H, W]
+*                                Argument N and C are not used.
+* @param[in]      output_dims    Output tensor dimensions. Format: [H, W, C_OUT]
+*                                Argument N is not used.
+*                                C_OUT equals C_IN.
+* @param[in, out] output_data    Output data pointer. Data type: int8
+* @return                        The function returns
+*                                    <code>ARM_MATH_SUCCESS</code> - Successful operation
+*
+* @details
+*    - Supported Framework: TensorFlow Lite
+*
+*/
+arm_status arm_avgpool_s8(const cmsis_nn_context *ctx,
+                          const cmsis_nn_pool_params *pool_params,
+                          const cmsis_nn_dims *input_dims,
+                          const q7_t *input_data,
+                          const cmsis_nn_dims *filter_dims,
+                          const cmsis_nn_dims *output_dims,
+                          q7_t *output_data);
 
-  /**
-   * @brief Get the required buffer size for S8 average pooling function
-   * @param[in]       dim_dst_width         output tensor dimension
-   * @param[in]       ch_src                number of input tensor channels
-   * @return          The function returns  required buffer size in bytes
-   *
-   */
-    int32_t arm_avgpool_s8_get_buffer_size(const int dim_dst_width,
-                                           const int ch_src);
+/**
+ * @brief Get the required buffer size for S8 average pooling function
+ * @param[in]       dim_dst_width         output tensor dimension
+ * @param[in]       ch_src                number of input tensor channels
+ * @return          The function returns  required buffer size in bytes
+ *
+ */
+int32_t arm_avgpool_s8_get_buffer_size(const int dim_dst_width, const int ch_src);
 
-   /**
-   * @brief s8 max pooling function.
-   *
-   * @param[in, out] ctx            Function context (e.g. temporary buffer). Check the function
-   *                                definition file to see if an additional buffer is required.
-   *                                Optional function {API}_get_buffer_size() provides the buffer
-   *                                size if an additional buffer is required.
-   * @param[in]      pool_params    Pooling parameters
-   * @param[in]      input_dims     Input (activation) tensor dimensions. Format: [H, W, C_IN]
-   *                                Argument 'N' is not used.
-   * @param[in]      input_data     Input (activation) data pointer. Data type: int8
-   * @param[in]      filter_dims    Filter tensor dimensions. Format: [H, W]
-   *                                Argument N and C are not used.
-   * @param[in]      output_dims    Output tensor dimensions. Format: [H, W, C_OUT]
-   *                                Argument N is not used.
-   *                                C_OUT equals C_IN.
-   * @param[in, out] output_data    Output data pointer. Data type: int8
-   * @return                        The function returns
-   *                                    <code>ARM_MATH_SUCCESS</code> - Successful operation
-   *
-   * @details
-   *    - Supported Framework: TensorFlow Lite
-   *
-   */
-    arm_status arm_max_pool_s8(const cmsis_nn_context *ctx,
-                               const cmsis_nn_pool_params *pool_params,
-                               const cmsis_nn_dims *input_dims,
-                               const q7_t *input_data,
-                               const cmsis_nn_dims *filter_dims,
-                               const cmsis_nn_dims *output_dims,
-                               q7_t *output_data);
+/**
+* @brief s8 max pooling function.
+*
+* @param[in, out] ctx            Function context (e.g. temporary buffer). Check the function
+*                                definition file to see if an additional buffer is required.
+*                                Optional function {API}_get_buffer_size() provides the buffer
+*                                size if an additional buffer is required.
+* @param[in]      pool_params    Pooling parameters
+* @param[in]      input_dims     Input (activation) tensor dimensions. Format: [H, W, C_IN]
+*                                Argument 'N' is not used.
+* @param[in]      input_data     Input (activation) data pointer. Data type: int8
+* @param[in]      filter_dims    Filter tensor dimensions. Format: [H, W]
+*                                Argument N and C are not used.
+* @param[in]      output_dims    Output tensor dimensions. Format: [H, W, C_OUT]
+*                                Argument N is not used.
+*                                C_OUT equals C_IN.
+* @param[in, out] output_data    Output data pointer. Data type: int8
+* @return                        The function returns
+*                                    <code>ARM_MATH_SUCCESS</code> - Successful operation
+*
+* @details
+*    - Supported Framework: TensorFlow Lite
+*
+*/
+arm_status arm_max_pool_s8(const cmsis_nn_context *ctx,
+                           const cmsis_nn_pool_params *pool_params,
+                           const cmsis_nn_dims *input_dims,
+                           const q7_t *input_data,
+                           const cmsis_nn_dims *filter_dims,
+                           const cmsis_nn_dims *output_dims,
+                           q7_t *output_data);
 /**
  * @defgroup Softmax Softmax Functions
  *
@@ -1734,61 +1733,61 @@
  *
  */
 
-  /**
-   * @brief Q7 softmax function
-   * @param[in]       vec_in      pointer to input vector
-   * @param[in]       dim_vec     input vector dimension
-   * @param[out]      p_out       pointer to output vector
-   *
-   * @note This function is an optimized version which is not bit-accurate with
-   *       TensorFlow Lite's kernel
-   *
-   */
+/**
+ * @brief Q7 softmax function
+ * @param[in]       vec_in      pointer to input vector
+ * @param[in]       dim_vec     input vector dimension
+ * @param[out]      p_out       pointer to output vector
+ *
+ * @note This function is an optimized version which is not bit-accurate with
+ *       TensorFlow Lite's kernel
+ *
+ */
 
-void arm_softmax_q7(const q7_t * vec_in, const uint16_t dim_vec, q7_t * p_out);
+void arm_softmax_q7(const q7_t *vec_in, const uint16_t dim_vec, q7_t *p_out);
 
-  /**
-   * @brief Q7 softmax function with batch parameter
-   * @param[in]       vec_in      pointer to input vector
-   * @param[in]       nb_batches  number of batches
-   * @param[in]       dim_vec     input vector dimension
-   * @param[out]      p_out       pointer to output vector
-   * @return none.
-   *
-   * @note This function is an optimized version which is not bit-accurate with
-   *       TensorFlow Lite's kernel
-   *
-   */
+/**
+ * @brief Q7 softmax function with batch parameter
+ * @param[in]       vec_in      pointer to input vector
+ * @param[in]       nb_batches  number of batches
+ * @param[in]       dim_vec     input vector dimension
+ * @param[out]      p_out       pointer to output vector
+ * @return none.
+ *
+ * @note This function is an optimized version which is not bit-accurate with
+ *       TensorFlow Lite's kernel
+ *
+ */
 
-void arm_softmax_with_batch_q7(const q7_t * vec_in, const uint16_t nb_batches,const uint16_t dim_vec, q7_t * p_out );
-  /**
-   * @brief Q15 softmax function
-   * @param[in]       vec_in      pointer to input vector
-   * @param[in]       dim_vec     input vector dimension
-   * @param[out]      p_out       pointer to output vector
-   * @return none.
-   *
-   * @note This function is an optimized version which is not bit-accurate with
-   *       TensorFlow Lite's kernel
-   *
-   */
+void arm_softmax_with_batch_q7(const q7_t *vec_in, const uint16_t nb_batches, const uint16_t dim_vec, q7_t *p_out);
+/**
+ * @brief Q15 softmax function
+ * @param[in]       vec_in      pointer to input vector
+ * @param[in]       dim_vec     input vector dimension
+ * @param[out]      p_out       pointer to output vector
+ * @return none.
+ *
+ * @note This function is an optimized version which is not bit-accurate with
+ *       TensorFlow Lite's kernel
+ *
+ */
 
-void arm_softmax_q15(const q15_t * vec_in, const uint16_t dim_vec, q15_t * p_out);
+void arm_softmax_q15(const q15_t *vec_in, const uint16_t dim_vec, q15_t *p_out);
 
-  /**
-   * @brief S8 softmax function
-   * @param[in]  input     Pointer to the input tensor
-   * @param[in]  num_rows  Number of rows in the input tensor
-   * @param[in]  row_size  Number of elements in each input row
-   * @param[in]  mult      Input quantization multiplier
-   * @param[in]  shift     Input quantization shift within the range [0, 31]
-   * @param[in]  diff_min  Minimum difference with max in row. Used to check if
-   *                       the quantized exponential operation can be performed
-   * @param[out] output    Pointer to the output tensor
-   *
-   * @note Supported framework: TensorFlow Lite micro (bit-accurate)
-   *
-   */
+/**
+ * @brief S8 softmax function
+ * @param[in]  input     Pointer to the input tensor
+ * @param[in]  num_rows  Number of rows in the input tensor
+ * @param[in]  row_size  Number of elements in each input row
+ * @param[in]  mult      Input quantization multiplier
+ * @param[in]  shift     Input quantization shift within the range [0, 31]
+ * @param[in]  diff_min  Minimum difference with max in row. Used to check if
+ *                       the quantized exponential operation can be performed
+ * @param[out] output    Pointer to the output tensor
+ *
+ * @note Supported framework: TensorFlow Lite micro (bit-accurate)
+ *
+ */
 
 void arm_softmax_s8(const int8_t *input,
                     const int32_t num_rows,
@@ -1798,20 +1797,20 @@
                     const int32_t diff_min,
                     int8_t *output);
 
-  /**
-   * @brief U8 softmax function
-   * @param[in]  input     Pointer to the input tensor
-   * @param[in]  num_rows  Number of rows in the input tensor
-   * @param[in]  row_size  Number of elements in each input row
-   * @param[in]  mult      Input quantization multiplier
-   * @param[in]  shift     Input quantization shift within the range [0, 31]
-   * @param[in]  diff_min  Minimum difference with max in row. Used to check if
-   *                       the quantized exponential operation can be performed
-   * @param[out] output    Pointer to the output tensor
-   *
-   * @note Supported framework: TensorFlow Lite micro (bit-accurate)
-   *
-   */
+/**
+ * @brief U8 softmax function
+ * @param[in]  input     Pointer to the input tensor
+ * @param[in]  num_rows  Number of rows in the input tensor
+ * @param[in]  row_size  Number of elements in each input row
+ * @param[in]  mult      Input quantization multiplier
+ * @param[in]  shift     Input quantization shift within the range [0, 31]
+ * @param[in]  diff_min  Minimum difference with max in row. Used to check if
+ *                       the quantized exponential operation can be performed
+ * @param[out] output    Pointer to the output tensor
+ *
+ * @note Supported framework: TensorFlow Lite micro (bit-accurate)
+ *
+ */
 
 void arm_softmax_u8(const uint8_t *input,
                     const int32_t num_rows,
@@ -1821,313 +1820,313 @@
                     const int32_t diff_min,
                     uint8_t *output);
 
-  /**
-   * @brief uint8 depthwise convolution function with asymmetric quantization
-   *        Unless specified otherwise, arguments are mandatory.
-   *
-   * @param[in]     input     Pointer to input tensor
-   * @param[in]     input_x   Width of input tensor
-   * @param[in]     input_y   Height of input tensor
-   * @param[in]     input_ch  Channels in input tensor
-   * @param[in]     kernel    Pointer to kernel weights
-   * @param[in]     kernel_x  Width of kernel
-   * @param[in]     kernel_y  Height of kernel
-   * @param[in]     ch_mult   Number of channel multiplier
-   * @param[in]     pad_x     Padding sizes x
-   * @param[in]     pad_y     Padding sizes y
-   * @param[in]     stride_x  stride along the width
-   * @param[in]     stride_y  stride along the height
-   * @param[in]     dilation_x Dilation along width. Not used and intended for future enhancement.
-   * @param[in]     dilation_y Dilation along height. Not used and intended for future enhancement.
-   * @param[in]     bias       Pointer to optional bias values. If no bias is
-   *                           availble, NULL is expected
-   * @param[in]     input_offset  Input tensor zero offset
-   * @param[in]     filter_offset Kernel tensor zero offset
-   * @param[in]     output_offset Output tensor zero offset
-   * @param[in,out] output        Pointer to output tensor
-   * @param[in]     output_x  Width of output tensor
-   * @param[in]     output_y  Height of output tensor
-   * @param[in]     output_activation_min   Minimum value to clamp the output to. Range : {0, 255}
-   * @param[in]     output_activation_max   Minimum value to clamp the output to. Range : {0, 255}
-   * @param[in]     out_shift  Amount of right-shift for output
-   * @param[in]     out_mult   Output multiplier for requantization
-   * @return        The function returns the following
-   *                <code>ARM_MATH_SUCCESS</code> - Successful operation
-   *
-   */
-    arm_status arm_depthwise_conv_u8_basic_ver1(const uint8_t *input,
-                                                const uint16_t input_x,
-                                                const uint16_t input_y,
-                                                const uint16_t input_ch,
-                                                const uint8_t *kernel,
-                                                const uint16_t kernel_x,
-                                                const uint16_t kernel_y,
-                                                const int16_t ch_mult,
-                                                const int16_t pad_x,
-                                                const int16_t pad_y,
-                                                const int16_t stride_x,
-                                                const int16_t stride_y,
-                                                const int16_t dilation_x,
-                                                const int16_t dilation_y,
-                                                const int32_t *bias,
-                                                const int32_t input_offset,
-                                                const int32_t filter_offset,
-                                                const int32_t output_offset,
-                                                uint8_t *output,
-                                                const uint16_t output_x,
-                                                const uint16_t output_y,
-                                                const int32_t output_activation_min,
-                                                const int32_t output_activation_max,
-                                                const int32_t out_shift,
-                                                const int32_t out_mult);
+/**
+ * @brief uint8 depthwise convolution function with asymmetric quantization
+ *        Unless specified otherwise, arguments are mandatory.
+ *
+ * @param[in]     input     Pointer to input tensor
+ * @param[in]     input_x   Width of input tensor
+ * @param[in]     input_y   Height of input tensor
+ * @param[in]     input_ch  Channels in input tensor
+ * @param[in]     kernel    Pointer to kernel weights
+ * @param[in]     kernel_x  Width of kernel
+ * @param[in]     kernel_y  Height of kernel
+ * @param[in]     ch_mult   Number of channel multiplier
+ * @param[in]     pad_x     Padding sizes x
+ * @param[in]     pad_y     Padding sizes y
+ * @param[in]     stride_x  stride along the width
+ * @param[in]     stride_y  stride along the height
+ * @param[in]     dilation_x Dilation along width. Not used and intended for future enhancement.
+ * @param[in]     dilation_y Dilation along height. Not used and intended for future enhancement.
+ * @param[in]     bias       Pointer to optional bias values. If no bias is
+ *                           availble, NULL is expected
+ * @param[in]     input_offset  Input tensor zero offset
+ * @param[in]     filter_offset Kernel tensor zero offset
+ * @param[in]     output_offset Output tensor zero offset
+ * @param[in,out] output        Pointer to output tensor
+ * @param[in]     output_x  Width of output tensor
+ * @param[in]     output_y  Height of output tensor
+ * @param[in]     output_activation_min   Minimum value to clamp the output to. Range : {0, 255}
+ * @param[in]     output_activation_max   Minimum value to clamp the output to. Range : {0, 255}
+ * @param[in]     out_shift  Amount of right-shift for output
+ * @param[in]     out_mult   Output multiplier for requantization
+ * @return        The function returns the following
+ *                <code>ARM_MATH_SUCCESS</code> - Successful operation
+ *
+ */
+arm_status arm_depthwise_conv_u8_basic_ver1(const uint8_t *input,
+                                            const uint16_t input_x,
+                                            const uint16_t input_y,
+                                            const uint16_t input_ch,
+                                            const uint8_t *kernel,
+                                            const uint16_t kernel_x,
+                                            const uint16_t kernel_y,
+                                            const int16_t ch_mult,
+                                            const int16_t pad_x,
+                                            const int16_t pad_y,
+                                            const int16_t stride_x,
+                                            const int16_t stride_y,
+                                            const int16_t dilation_x,
+                                            const int16_t dilation_y,
+                                            const int32_t *bias,
+                                            const int32_t input_offset,
+                                            const int32_t filter_offset,
+                                            const int32_t output_offset,
+                                            uint8_t *output,
+                                            const uint16_t output_x,
+                                            const uint16_t output_y,
+                                            const int32_t output_activation_min,
+                                            const int32_t output_activation_max,
+                                            const int32_t out_shift,
+                                            const int32_t out_mult);
 
 /**
  * @defgroup Reshape Reshape Functions
  *
  */
 
-   /**
-    * @brief Reshape a s8 vector into another with different shape
-    * @param[in]  input      points to the s8 input vector
-    * @param[out] output     points to the s8 output vector
-    * @param[in]  total_size total size of the input and output vectors in bytes
-    *
-    * @note The output is expected to be in a memory area that does not overlap with the input's
-    *
-    */
-    void arm_reshape_s8(const int8_t *input,
-                        int8_t *output,
-                        const uint32_t total_size);
+/**
+ * @brief Reshape a s8 vector into another with different shape
+ * @param[in]  input      points to the s8 input vector
+ * @param[out] output     points to the s8 output vector
+ * @param[in]  total_size total size of the input and output vectors in bytes
+ *
+ * @note The output is expected to be in a memory area that does not overlap with the input's
+ *
+ */
+void arm_reshape_s8(const int8_t *input, int8_t *output, const uint32_t total_size);
 
 /**
  * @defgroup Concatenation Concatenation Functions
  *
  */
 
-  /**
-   * @brief int8/uint8 concatenation function to be used for concatenating N-tensors along the X axis
-   *        This function should be called for each input tensor to concatenate. The argument offset_x
-   *        will be used to store the input tensor in the correct position in the output tensor
-   *
-   *        i.e.    offset_x = 0
-   *                for(i = 0 i < num_input_tensors; ++i)
-   *                {
-   *                    arm_concatenation_s8_x(&input[i], ..., &output, ..., ..., offset_x)
-   *                    offset_x += input_x[i]
-   *                }
-   *
-   *        This function assumes that the output tensor has:
-   *        -# The same height of the input tensor
-   *        -# The same number of channels of the input tensor
-   *        -# The same batch size of the input tensor
-   *
-   *        Unless specified otherwise, arguments are mandatory.
-   *
-   * @note This function, data layout independent, can be used to concatenate either int8 or uint8 tensors because does not involve any arithmetic operation
-   *
-   * @param[in]  input    Pointer to input tensor
-   * @param[in]  input_x  Width of input tensor
-   * @param[in]  input_y  Height of input tensor
-   * @param[in]  input_z  Channels in input tensor
-   * @param[in]  input_w  Batch size in input tensor
-   * @param[out] output   Pointer to output tensor
-   * @param[in]  output_x Width of output tensor
-   * @param[in]  offset_x The offset (in number of elements) on the X axis to start concatenating the input tensor
-   *                      It is user responsibility to provide the correct value
-   *
-   * <b> Input constraints</b>
-   * offset_x is less than output_x
-   *
-   */
-    void arm_concatenation_s8_x(const int8_t *input,
-                                const uint16_t input_x,
-                                const uint16_t input_y,
-                                const uint16_t input_z,
-                                const uint16_t input_w,
-                                int8_t *output,
-                                const uint16_t output_x,
-                                const uint32_t offset_x);
+/**
+ * @brief int8/uint8 concatenation function to be used for concatenating N-tensors along the X axis
+ *        This function should be called for each input tensor to concatenate. The argument offset_x
+ *        will be used to store the input tensor in the correct position in the output tensor
+ *
+ *        i.e.    offset_x = 0
+ *                for(i = 0 i < num_input_tensors; ++i)
+ *                {
+ *                    arm_concatenation_s8_x(&input[i], ..., &output, ..., ..., offset_x)
+ *                    offset_x += input_x[i]
+ *                }
+ *
+ *        This function assumes that the output tensor has:
+ *        -# The same height of the input tensor
+ *        -# The same number of channels of the input tensor
+ *        -# The same batch size of the input tensor
+ *
+ *        Unless specified otherwise, arguments are mandatory.
+ *
+ * @note This function, data layout independent, can be used to concatenate either int8 or uint8 tensors because it
+ *      does not involve any arithmetic operation
+ *
+ * @param[in]  input    Pointer to input tensor
+ * @param[in]  input_x  Width of input tensor
+ * @param[in]  input_y  Height of input tensor
+ * @param[in]  input_z  Channels in input tensor
+ * @param[in]  input_w  Batch size in input tensor
+ * @param[out] output   Pointer to output tensor
+ * @param[in]  output_x Width of output tensor
+ * @param[in]  offset_x The offset (in number of elements) on the X axis to start concatenating the input tensor
+ *                      It is user responsibility to provide the correct value
+ *
+ * <b> Input constraints</b>
+ * offset_x is less than output_x
+ *
+ */
+void arm_concatenation_s8_x(const int8_t *input,
+                            const uint16_t input_x,
+                            const uint16_t input_y,
+                            const uint16_t input_z,
+                            const uint16_t input_w,
+                            int8_t *output,
+                            const uint16_t output_x,
+                            const uint32_t offset_x);
 
-  /**
-   * @brief int8/uint8 concatenation function to be used for concatenating N-tensors along the Y axis
-   *        This function should be called for each input tensor to concatenate. The argument offset_y
-   *        will be used to store the input tensor in the correct position in the output tensor
-   *
-   *        i.e.    offset_y = 0
-   *                for(i = 0 i < num_input_tensors; ++i)
-   *                {
-   *                    arm_concatenation_s8_y(&input[i], ..., &output, ..., ..., offset_y)
-   *                    offset_y += input_y[i]
-   *                }
-   *
-   *        This function assumes that the output tensor has:
-   *        -# The same width of the input tensor
-   *        -# The same number of channels of the input tensor
-   *        -# The same batch size of the input tensor
-   *
-   *        Unless specified otherwise, arguments are mandatory.
-   *
-   * @note This function, data layout independent, can be used to concatenate either int8 or uint8 tensors because does not involve any arithmetic operation
-   *
-   * @param[in]  input    Pointer to input tensor
-   * @param[in]  input_x  Width of input tensor
-   * @param[in]  input_y  Height of input tensor
-   * @param[in]  input_z  Channels in input tensor
-   * @param[in]  input_w  Batch size in input tensor
-   * @param[out] output   Pointer to output tensor
-   * @param[in]  output_y Height of output tensor
-   * @param[in]  offset_y The offset on the Y axis to start concatenating the input tensor
-   *                      It is user responsibility to provide the correct value
-   *
-   * <b> Input constraints</b>
-   * offset_y is less than output_y
-   *
-   */
-    void arm_concatenation_s8_y(const int8_t *input,
-                                const uint16_t input_x,
-                                const uint16_t input_y,
-                                const uint16_t input_z,
-                                const uint16_t input_w,
-                                int8_t *output,
-                                const uint16_t output_y,
-                                const uint32_t offset_y);
+/**
+ * @brief int8/uint8 concatenation function to be used for concatenating N-tensors along the Y axis
+ *        This function should be called for each input tensor to concatenate. The argument offset_y
+ *        will be used to store the input tensor in the correct position in the output tensor
+ *
+ *        i.e.    offset_y = 0
+ *                for(i = 0 i < num_input_tensors; ++i)
+ *                {
+ *                    arm_concatenation_s8_y(&input[i], ..., &output, ..., ..., offset_y)
+ *                    offset_y += input_y[i]
+ *                }
+ *
+ *        This function assumes that the output tensor has:
+ *        -# The same width of the input tensor
+ *        -# The same number of channels of the input tensor
+ *        -# The same batch size of the input tensor
+ *
+ *        Unless specified otherwise, arguments are mandatory.
+ *
+ * @note This function, data layout independent, can be used to concatenate either int8 or uint8 tensors because it
+ *       does not involve any arithmetic operation
+ *
+ * @param[in]  input    Pointer to input tensor
+ * @param[in]  input_x  Width of input tensor
+ * @param[in]  input_y  Height of input tensor
+ * @param[in]  input_z  Channels in input tensor
+ * @param[in]  input_w  Batch size in input tensor
+ * @param[out] output   Pointer to output tensor
+ * @param[in]  output_y Height of output tensor
+ * @param[in]  offset_y The offset on the Y axis to start concatenating the input tensor
+ *                      It is user responsibility to provide the correct value
+ *
+ * <b> Input constraints</b>
+ * offset_y is less than output_y
+ *
+ */
+void arm_concatenation_s8_y(const int8_t *input,
+                            const uint16_t input_x,
+                            const uint16_t input_y,
+                            const uint16_t input_z,
+                            const uint16_t input_w,
+                            int8_t *output,
+                            const uint16_t output_y,
+                            const uint32_t offset_y);
 
-  /**
-   * @brief int8/uint8 concatenation function to be used for concatenating N-tensors along the Z axis
-   *        This function should be called for each input tensor to concatenate. The argument offset_z
-   *        will be used to store the input tensor in the correct position in the output tensor
-   *
-   *        i.e.    offset_z = 0
-   *                for(i = 0 i < num_input_tensors; ++i)
-   *                {
-   *                    arm_concatenation_s8_z(&input[i], ..., &output, ..., ..., offset_z)
-   *                    offset_z += input_z[i]
-   *                }
-   *
-   *        This function assumes that the output tensor has:
-   *        -# The same width of the input tensor
-   *        -# The same height of the input tensor
-   *        -# The same batch size of the input tensor
-   *
-   *        Unless specified otherwise, arguments are mandatory.
-   *
-   * @note This function, data layout independent, can be used to concatenate either int8 or uint8 tensors because does not involve any arithmetic operation
-   *
-   * @param[in]  input    Pointer to input tensor
-   * @param[in]  input_x  Width of input tensor
-   * @param[in]  input_y  Height of input tensor
-   * @param[in]  input_z  Channels in input tensor
-   * @param[in]  input_w  Batch size in input tensor
-   * @param[out] output   Pointer to output tensor
-   * @param[in]  output_z Channels in output tensor
-   * @param[in]  offset_z The offset on the Z axis to start concatenating the input tensor
-   *                      It is user responsibility to provide the correct value
-   *
-   * <b> Input constraints</b>
-   * offset_z is less than output_z
-   *
-   */
-    void arm_concatenation_s8_z(const int8_t *input,
-                                const uint16_t input_x,
-                                const uint16_t input_y,
-                                const uint16_t input_z,
-                                const uint16_t input_w,
-                                int8_t *output,
-                                const uint16_t output_z,
-                                const uint32_t offset_z);
+/**
+ * @brief int8/uint8 concatenation function to be used for concatenating N-tensors along the Z axis
+ *        This function should be called for each input tensor to concatenate. The argument offset_z
+ *        will be used to store the input tensor in the correct position in the output tensor
+ *
+ *        i.e.    offset_z = 0
+ *                for(i = 0 i < num_input_tensors; ++i)
+ *                {
+ *                    arm_concatenation_s8_z(&input[i], ..., &output, ..., ..., offset_z)
+ *                    offset_z += input_z[i]
+ *                }
+ *
+ *        This function assumes that the output tensor has:
+ *        -# The same width of the input tensor
+ *        -# The same height of the input tensor
+ *        -# The same batch size of the input tensor
+ *
+ *        Unless specified otherwise, arguments are mandatory.
+ *
+ * @note This function, data layout independent, can be used to concatenate either int8 or uint8 tensors because it
+ *       does not involve any arithmetic operation
+ *
+ * @param[in]  input    Pointer to input tensor
+ * @param[in]  input_x  Width of input tensor
+ * @param[in]  input_y  Height of input tensor
+ * @param[in]  input_z  Channels in input tensor
+ * @param[in]  input_w  Batch size in input tensor
+ * @param[out] output   Pointer to output tensor
+ * @param[in]  output_z Channels in output tensor
+ * @param[in]  offset_z The offset on the Z axis to start concatenating the input tensor
+ *                      It is user responsibility to provide the correct value
+ *
+ * <b> Input constraints</b>
+ * offset_z is less than output_z
+ *
+ */
+void arm_concatenation_s8_z(const int8_t *input,
+                            const uint16_t input_x,
+                            const uint16_t input_y,
+                            const uint16_t input_z,
+                            const uint16_t input_w,
+                            int8_t *output,
+                            const uint16_t output_z,
+                            const uint32_t offset_z);
 
-  /**
-   * @brief int8/uint8 concatenation function to be used for concatenating N-tensors along the W axis (Batch size)
-   *        This function should be called for each input tensor to concatenate. The argument offset_w
-   *        will be used to store the input tensor in the correct position in the output tensor
-   *
-   *        i.e.    offset_w = 0
-   *                for(i = 0 i < num_input_tensors; ++i)
-   *                {
-   *                    arm_concatenation_s8_w(&input[i], ..., &output, ..., ..., offset_w)
-   *                    offset_w += input_w[i]
-   *                }
-   *
-   *        This function assumes that the output tensor has:
-   *        -# The same width of the input tensor
-   *        -# The same height of the input tensor
-   *        -# The same number o channels of the input tensor
-   *
-   *        Unless specified otherwise, arguments are mandatory.
-   *
-   * @note This function, data layout independent, can be used to concatenate either int8 or uint8 tensors because does not involve any arithmetic operation
-   *
-   * @param[in]  input    Pointer to input tensor
-   * @param[in]  input_x  Width of input tensor
-   * @param[in]  input_y  Height of input tensor
-   * @param[in]  input_z  Channels in input tensor
-   * @param[in]  input_w  Batch size in input tensor
-   * @param[out] output   Pointer to output tensor
-   * @param[in]  offset_w The offset on the W axis to start concatenating the input tensor
-   *                      It is user responsibility to provide the correct value
-   *
-   */
-    void arm_concatenation_s8_w(const int8_t *input,
-                                const uint16_t input_x,
-                                const uint16_t input_y,
-                                const uint16_t input_z,
-                                const uint16_t input_w,
-                                int8_t *output,
-                                const uint32_t offset_w);
+/**
+ * @brief int8/uint8 concatenation function to be used for concatenating N-tensors along the W axis (Batch size)
+ *        This function should be called for each input tensor to concatenate. The argument offset_w
+ *        will be used to store the input tensor in the correct position in the output tensor
+ *
+ *        i.e.    offset_w = 0
+ *                for(i = 0 i < num_input_tensors; ++i)
+ *                {
+ *                    arm_concatenation_s8_w(&input[i], ..., &output, ..., ..., offset_w)
+ *                    offset_w += input_w[i]
+ *                }
+ *
+ *        This function assumes that the output tensor has:
+ *        -# The same width of the input tensor
+ *        -# The same height of the input tensor
+ *        -# The same number o channels of the input tensor
+ *
+ *        Unless specified otherwise, arguments are mandatory.
+ *
+ * @note This function, data layout independent, can be used to concatenate either int8 or uint8 tensors because it
+ *       does not involve any arithmetic operation
+ *
+ * @param[in]  input    Pointer to input tensor
+ * @param[in]  input_x  Width of input tensor
+ * @param[in]  input_y  Height of input tensor
+ * @param[in]  input_z  Channels in input tensor
+ * @param[in]  input_w  Batch size in input tensor
+ * @param[out] output   Pointer to output tensor
+ * @param[in]  offset_w The offset on the W axis to start concatenating the input tensor
+ *                      It is user responsibility to provide the correct value
+ *
+ */
+void arm_concatenation_s8_w(const int8_t *input,
+                            const uint16_t input_x,
+                            const uint16_t input_y,
+                            const uint16_t input_z,
+                            const uint16_t input_w,
+                            int8_t *output,
+                            const uint32_t offset_w);
 /**
  * @defgroup SVDF SVDF Layer Functions
  *
  */
 
-    /**
-     * @brief s8 SVDF function
-     *
-     * @param[in]   input_ctx Temporary scratch buffer
-     * @param[in]   output_ctx Temporary output scratch buffer
-     * @param[in]   svdf_params SVDF Parameters
-     *              Range of svdf_params->input_offset  : [-128, 127]
-     *              Range of svdf_params->output_offset  : [-128, 127]
-     * @param[in]   input_quant_params Input quantization parameters
-     * @param[in]   output_quant_params Output quantization parameters
-     * @param[in]   input_dims Input tensor dimensions
-     * @param[in]   input_data Pointer to input tensor
-     * @param[in]   state_dims State tensor dimensions
-     * @param[in]   state_data Pointer to state tensor
-     * @param[in]   weights_feature_dims Weights (feature) tensor dimensions
-     * @param[in]   weights_feature_data Pointer to the weights (feature) tensor
-     * @param[in]   weights_time_dims Weights (time) tensor dimensions
-     * @param[in]   weights_time_data Pointer to the weights (time) tensor
-     * @param[in]   bias_dims Bias tensor dimensions
-     * @param[in]   bias_data Pointer to bias tensor
-     * @param[in]   output_dims Output tensor dimensions
-     * @param[out]  output_data Pointer to the output tensor
-     *
-     * @return     The function returns <code>ARM_MATH_SUCCESS</code>
-     *
-     * @details
-     *    1. Supported framework: TensorFlow Lite micro
-     *    2. q7 is used as data type eventhough it is s8 data. It is done so to be consistent with existing APIs.
-     *
-     */
-    arm_status
-    arm_svdf_s8(const cmsis_nn_context *input_ctx,
-                const cmsis_nn_context *output_ctx,
-                const cmsis_nn_svdf_params *svdf_params,
-                const cmsis_nn_per_tensor_quant_params *input_quant_params,
-                const cmsis_nn_per_tensor_quant_params *output_quant_params,
-                const cmsis_nn_dims *input_dims,
-                const q7_t *input_data,
-                const cmsis_nn_dims *state_dims,
-                q15_t *state_data,
-                const cmsis_nn_dims *weights_feature_dims,
-                const q7_t *weights_feature_data,
-                const cmsis_nn_dims *weights_time_dims,
-                const q15_t *weights_time_data,
-                const cmsis_nn_dims *bias_dims,
-                const q31_t *bias_data,
-                const cmsis_nn_dims *output_dims,
-                q7_t *output_data);
-
+/**
+ * @brief s8 SVDF function
+ *
+ * @param[in]   input_ctx Temporary scratch buffer
+ * @param[in]   output_ctx Temporary output scratch buffer
+ * @param[in]   svdf_params SVDF Parameters
+ *              Range of svdf_params->input_offset  : [-128, 127]
+ *              Range of svdf_params->output_offset  : [-128, 127]
+ * @param[in]   input_quant_params Input quantization parameters
+ * @param[in]   output_quant_params Output quantization parameters
+ * @param[in]   input_dims Input tensor dimensions
+ * @param[in]   input_data Pointer to input tensor
+ * @param[in]   state_dims State tensor dimensions
+ * @param[in]   state_data Pointer to state tensor
+ * @param[in]   weights_feature_dims Weights (feature) tensor dimensions
+ * @param[in]   weights_feature_data Pointer to the weights (feature) tensor
+ * @param[in]   weights_time_dims Weights (time) tensor dimensions
+ * @param[in]   weights_time_data Pointer to the weights (time) tensor
+ * @param[in]   bias_dims Bias tensor dimensions
+ * @param[in]   bias_data Pointer to bias tensor
+ * @param[in]   output_dims Output tensor dimensions
+ * @param[out]  output_data Pointer to the output tensor
+ *
+ * @return     The function returns <code>ARM_MATH_SUCCESS</code>
+ *
+ * @details
+ *    1. Supported framework: TensorFlow Lite micro
+ *    2. q7 is used as data type eventhough it is s8 data. It is done so to be consistent with existing APIs.
+ *
+ */
+arm_status arm_svdf_s8(const cmsis_nn_context *input_ctx,
+                       const cmsis_nn_context *output_ctx,
+                       const cmsis_nn_svdf_params *svdf_params,
+                       const cmsis_nn_per_tensor_quant_params *input_quant_params,
+                       const cmsis_nn_per_tensor_quant_params *output_quant_params,
+                       const cmsis_nn_dims *input_dims,
+                       const q7_t *input_data,
+                       const cmsis_nn_dims *state_dims,
+                       q15_t *state_data,
+                       const cmsis_nn_dims *weights_feature_dims,
+                       const q7_t *weights_feature_data,
+                       const cmsis_nn_dims *weights_time_dims,
+                       const q15_t *weights_time_data,
+                       const cmsis_nn_dims *bias_dims,
+                       const q31_t *bias_data,
+                       const cmsis_nn_dims *output_dims,
+                       q7_t *output_data);
 
 #ifdef __cplusplus
 }
diff --git a/CMSIS/NN/Include/arm_nnsupportfunctions.h b/CMSIS/NN/Include/arm_nnsupportfunctions.h
index 44a677b..0e8fbd3 100644
--- a/CMSIS/NN/Include/arm_nnsupportfunctions.h
+++ b/CMSIS/NN/Include/arm_nnsupportfunctions.h
@@ -30,35 +30,33 @@
 #ifndef _ARM_NNSUPPORTFUNCTIONS_H_
 #define _ARM_NNSUPPORTFUNCTIONS_H_
 
-#include "arm_math_types.h"
 #include "arm_common_tables.h"
+#include "arm_math_types.h"
 
 #ifdef __cplusplus
-extern    "C"
-{
+extern "C" {
 #endif
 
-#define LEFT_SHIFT(_shift)  (_shift > 0 ? _shift : 0)
+#define LEFT_SHIFT(_shift) (_shift > 0 ? _shift : 0)
 #define RIGHT_SHIFT(_shift) (_shift > 0 ? 0 : -_shift)
-#define MASK_IF_ZERO(x)     (x) == 0 ? ~0 : 0
+#define MASK_IF_ZERO(x) (x) == 0 ? ~0 : 0
 #define MASK_IF_NON_ZERO(x) (x) != 0 ? ~0 : 0
 #define SELECT_USING_MASK(mask, a, b) ((mask) & (a)) ^ (~(mask) & (b))
 
-#define MAX(A,B) ((A) > (B) ? (A) : (B))
-#define MIN(A,B) ((A) < (B) ? (A) : (B))
+#define MAX(A, B) ((A) > (B) ? (A) : (B))
+#define MIN(A, B) ((A) < (B) ? (A) : (B))
 #define CLAMP(x, h, l) MAX(MIN((x), (h)), (l))
 
 /**
  * @brief Union for SIMD access of q31/q15/q7 types
  */
-union arm_nnword
-{
-    q31_t     word;
-               /**< q31 type */
-    q15_t     half_words[2];
-               /**< q15 type */
-    q7_t      bytes[4];
-               /**< q7 type */
+union arm_nnword {
+    q31_t word;
+    /**< q31 type */
+    q15_t half_words[2];
+    /**< q15 type */
+    q7_t bytes[4];
+    /**< q7 type */
 };
 
 /**
@@ -66,14 +64,13 @@
  */
 struct arm_nn_double
 {
-  uint32_t low;
-  int32_t high;
+    uint32_t low;
+    int32_t high;
 };
 
-union arm_nn_long_long
-{
-  int64_t long_long;
-  struct arm_nn_double word;
+union arm_nn_long_long {
+    int64_t long_long;
+    struct arm_nn_double word;
 };
 
 /**
@@ -118,7 +115,7 @@
  * @return none.
  *
  */
-void arm_q7_to_q15_reordered_no_shift(const q7_t * pSrc, q15_t * pDst, uint32_t blockSize);
+void arm_q7_to_q15_reordered_no_shift(const q7_t *pSrc, q15_t *pDst, uint32_t blockSize);
 
 /**
  * @brief Converts the elements from a q7 vector to a q15 vector with an added offset
@@ -300,11 +297,13 @@
 *
 * @param[in]  lhs                Pointer to the LHS input matrix
 * @param[in]  rhs                Pointer to the RHS input matrix
-* @param[in]  bias               Pointer to the bias vector. The length of this vector is equal to the number of output columns (or RHS input rows)
+* @param[in]  bias               Pointer to the bias vector. The length of this vector is equal to the number of output
+*                                columns (or RHS input rows)
 * @param[out] dst                Pointer to the output matrix with "m" rows and "n" columns
-* @param[in]  dst_multipliers    Pointer to the multipliers vector needed for the per-channel requantization. The length of this vector is equal to
-*                                the number of output columns (or RHS input rows)
-* @param[in]  dst_shifts         Pointer to the shifts vector needed for the per-channel requantization. The length of this vector is equal to
+* @param[in]  dst_multipliers    Pointer to the multipliers vector needed for the per-channel requantization.
+*                                The length of this vector is equal to the number of output columns (or RHS input rows)
+* @param[in]  dst_shifts         Pointer to the shifts vector needed for the per-channel requantization. The length of
+*                                this vector is equal to
 *                                the number of output columns (or RHS input rows)
 * @param[in]  lhs_rows           Number of LHS input rows
 * @param[in]  rhs_rows           Number of RHS input rows
@@ -338,8 +337,10 @@
  * @param[in]      rhs             Input right-hand side matrix (transposed)
  * @param[in]      bias            Input bias
  * @param[out]     dst             Output vector
- * @param[in]      lhs_offset      Offset to be added to the input values of the left-hand side vector. Range: -127 to 128
- * @param[in]      rhs_offset      Offset to be added to the input values of the right-hand side matrix. Range: -127 to 128
+ * @param[in]      lhs_offset      Offset to be added to the input values of the left-hand side vector.
+ *                                 Range: -127 to 128
+ * @param[in]      rhs_offset      Offset to be added to the input values of the right-hand side matrix.
+ *                                 Range: -127 to 128
  * @param[in]      dst_offset      Offset to be added to the output values. Range: -127 to 128
  * @param[in]      dst_multiplier  Output multiplier
  * @param[in]      dst_shift       Output shift
@@ -454,12 +455,12 @@
  */
 __STATIC_FORCEINLINE q31_t arm_nn_read_q15x2_ia(const q15_t **in_q15)
 {
-  q31_t val;
+    q31_t val;
 
-  memcpy(&val, *in_q15, 4);
-  *in_q15 += 2;
+    memcpy(&val, *in_q15, 4);
+    *in_q15 += 2;
 
-  return (val);
+    return (val);
 }
 
 /**
@@ -469,11 +470,11 @@
  */
 __STATIC_FORCEINLINE q31_t arm_nn_read_q7x4_ia(const q7_t **in_q7)
 {
-  q31_t val;
-  memcpy(&val, *in_q7, 4);
-  *in_q7 += 4;
+    q31_t val;
+    memcpy(&val, *in_q7, 4);
+    *in_q7 += 4;
 
-  return (val);
+    return (val);
 }
 
 /**
@@ -483,10 +484,10 @@
  */
 __STATIC_FORCEINLINE q31_t arm_nn_read_q15x2(const q15_t *in_q15)
 {
-  q31_t val;
-  memcpy(&val, in_q15, 4);
+    q31_t val;
+    memcpy(&val, in_q15, 4);
 
-  return (val);
+    return (val);
 }
 
 /**
@@ -496,10 +497,10 @@
  */
 __STATIC_FORCEINLINE q31_t arm_nn_read_q7x4(const q7_t *in_q7)
 {
-  q31_t val;
-  memcpy(&val, in_q7, 4);
+    q31_t val;
+    memcpy(&val, in_q7, 4);
 
-  return (val);
+    return (val);
 }
 
 /**
@@ -509,91 +510,87 @@
  * @param[in]       block_size  Number of bytes to copy.
  *
  */
-__STATIC_FORCEINLINE void arm_memset_q7(q7_t *dst,
-                                        const q7_t val,
-                                        uint32_t block_size)
+__STATIC_FORCEINLINE void arm_memset_q7(q7_t *dst, const q7_t val, uint32_t block_size)
 {
 #if defined(ARM_MATH_MVEI)
-     __asm volatile (
-        "   vdup.8                  q0, %[set_val]             \n"
-        "   wlstp.8                 lr, %[cnt], 1f             \n"
-        "2:                                                    \n"
-        "   vstrb.8                 q0, [%[in]], 16            \n"
-        "   letp                    lr, 2b                     \n"
-        "1:                                                    \n"
-        :[in] "+r"(dst)
-        :[cnt] "r"(block_size), [set_val] "r"(val)
-        :"q0", "memory", "r14");
+    __asm volatile("   vdup.8                  q0, %[set_val]             \n"
+                   "   wlstp.8                 lr, %[cnt], 1f             \n"
+                   "2:                                                    \n"
+                   "   vstrb.8                 q0, [%[in]], 16            \n"
+                   "   letp                    lr, 2b                     \n"
+                   "1:                                                    \n"
+                   : [in] "+r"(dst)
+                   : [cnt] "r"(block_size), [set_val] "r"(val)
+                   : "q0", "memory", "r14");
 #else
     memset(dst, val, block_size);
 #endif
 }
 
-#if defined (ARM_MATH_DSP)
+#if defined(ARM_MATH_DSP)
 
 /**
  * @brief read and expand one q7 word into two q15 words
  */
 
-__STATIC_FORCEINLINE const q7_t *read_and_pad(const q7_t *source, q31_t * out1, q31_t * out2)
+__STATIC_FORCEINLINE const q7_t *read_and_pad(const q7_t *source, q31_t *out1, q31_t *out2)
 {
-        q31_t     inA = arm_nn_read_q7x4_ia(&source);
-        q31_t     inAbuf1 = __SXTB16(__ROR((uint32_t)inA, 8));
-        q31_t     inAbuf2 = __SXTB16(inA);
+    q31_t inA = arm_nn_read_q7x4_ia(&source);
+    q31_t inAbuf1 = __SXTB16(__ROR((uint32_t)inA, 8));
+    q31_t inAbuf2 = __SXTB16(inA);
 
 #ifndef ARM_MATH_BIG_ENDIAN
-  *out2 = (int32_t) (__PKHTB (inAbuf1, inAbuf2, 16));
-  *out1 = (int32_t) (__PKHBT (inAbuf2, inAbuf1, 16));
+    *out2 = (int32_t)(__PKHTB(inAbuf1, inAbuf2, 16));
+    *out1 = (int32_t)(__PKHBT(inAbuf2, inAbuf1, 16));
 #else
-  *out1 = (int32_t) (__PKHTB(inAbuf1, inAbuf2, 16));
-  *out2 = (int32_t) (__PKHBT(inAbuf2, inAbuf1, 16));
+    *out1 = (int32_t)(__PKHTB(inAbuf1, inAbuf2, 16));
+    *out2 = (int32_t)(__PKHBT(inAbuf2, inAbuf1, 16));
 #endif
 
-        return source;
+    return source;
 }
 
 /**
  * @brief read and expand one q7 word into two q15 words with reordering
  */
 
-__STATIC_FORCEINLINE const q7_t *read_and_pad_reordered(const q7_t *source, q31_t * out1, q31_t * out2)
+__STATIC_FORCEINLINE const q7_t *read_and_pad_reordered(const q7_t *source, q31_t *out1, q31_t *out2)
 {
-        q31_t     inA = arm_nn_read_q7x4_ia(&source);
+    q31_t inA = arm_nn_read_q7x4_ia(&source);
 #ifndef ARM_MATH_BIG_ENDIAN
-        *out2 = __SXTB16(__ROR((uint32_t)inA, 8));
-        *out1 = __SXTB16(inA);
+    *out2 = __SXTB16(__ROR((uint32_t)inA, 8));
+    *out1 = __SXTB16(inA);
 #else
-        *out1 = __SXTB16(__ROR((uint32_t)inA, 8));
-        *out2 = __SXTB16(inA);
+    *out1 = __SXTB16(__ROR((uint32_t)inA, 8));
+    *out2 = __SXTB16(inA);
 #endif
 
-        return source;
+    return source;
 }
 
 /**
  * @brief read and expand one q7 word into two q15 words with reordering and add an offset
  */
-__STATIC_FORCEINLINE const q7_t *read_and_pad_reordered_with_offset(const q7_t *source, q31_t * out1, q31_t * out2, q31_t offset)
+__STATIC_FORCEINLINE const q7_t *
+read_and_pad_reordered_with_offset(const q7_t *source, q31_t *out1, q31_t *out2, q31_t offset)
 {
-        q31_t     inA = arm_nn_read_q7x4_ia(&source);
+    q31_t inA = arm_nn_read_q7x4_ia(&source);
 
 #ifndef ARM_MATH_BIG_ENDIAN
-        *out2 = __SXTB16(__ROR((uint32_t)inA, 8));
-        *out1 = __SXTB16(inA);
+    *out2 = __SXTB16(__ROR((uint32_t)inA, 8));
+    *out1 = __SXTB16(inA);
 #else
-        *out1 = __SXTB16(__ROR((uint32_t)inA, 8));
-        *out2 = __SXTB16(inA);
+    *out1 = __SXTB16(__ROR((uint32_t)inA, 8));
+    *out2 = __SXTB16(inA);
 #endif
-        *out1 = __QADD16(*out1,offset);
-        *out2 = __QADD16(*out2,offset);
+    *out1 = __QADD16(*out1, offset);
+    *out2 = __QADD16(*out2, offset);
 
-        return source;
+    return source;
 }
 
 #endif
 
-
-
 /**
  * @defgroup NNBasicMath Basic Math Functions for Neural Network Computation
  *
@@ -616,12 +613,7 @@
  * Results outside of the allowable q15 range [0x8000 0x7FFF] will be saturated.
  */
 
-void arm_nn_mult_q15(
-  q15_t * pSrcA,
-  q15_t * pSrcB,
-  q15_t * pDst,
-  const uint16_t out_shift,
-  uint32_t blockSize);
+void arm_nn_mult_q15(q15_t *pSrcA, q15_t *pSrcB, q15_t *pDst, const uint16_t out_shift, uint32_t blockSize);
 
 /**
  * @brief           q7 vector multiplication with variable output shifts
@@ -638,34 +630,27 @@
  * Results outside of the allowable q7 range [0x80 0x7F] will be saturated.
  */
 
-void arm_nn_mult_q7(
-  q7_t * pSrcA,
-  q7_t * pSrcB,
-  q7_t * pDst,
-  const uint16_t out_shift,
-  uint32_t blockSize);
+void arm_nn_mult_q7(q7_t *pSrcA, q7_t *pSrcB, q7_t *pDst, const uint16_t out_shift, uint32_t blockSize);
 
 /**
  * @brief macro for adding rounding offset
  */
 #ifndef ARM_NN_TRUNCATE
-    #define NN_ROUND(out_shift) ( (0x1u << out_shift) >> 1 )
+#define NN_ROUND(out_shift) ((0x1u << out_shift) >> 1)
 #else
-    #define NN_ROUND(out_shift) 0
+#define NN_ROUND(out_shift) 0
 #endif
 
 // Macros for shortening quantization functions' names and avoid long lines
-#define MUL_SAT(a, b)  arm_nn_doubling_high_mult((a), (b))
+#define MUL_SAT(a, b) arm_nn_doubling_high_mult((a), (b))
 #define MUL_SAT_MVE(a, b) arm_doubling_high_mult_mve_32x4((a), (b))
 #define MUL_POW2(a, b) arm_nn_mult_by_power_of_two((a), (b))
 
-
 #define DIV_POW2(a, b) arm_nn_divide_by_power_of_two((a), (b))
 #define DIV_POW2_MVE(a, b) arm_divide_by_power_of_two_mve((a), (b))
 
-
-#define EXP_ON_NEG(x)  arm_nn_exp_on_negative_values((x))
-#define ONE_OVER1(x)   arm_nn_one_over_one_plus_x_for_x_in_0_1((x))
+#define EXP_ON_NEG(x) arm_nn_exp_on_negative_values((x))
+#define ONE_OVER1(x) arm_nn_one_over_one_plus_x_for_x_in_0_1((x))
 
 /**
  * @brief           Saturating doubling high multiply. Result matches
@@ -690,7 +675,7 @@
 
     // Utilize all of the upper 32 bits. This is the doubling step
     // as well.
-    result = (int32_t) (mult / (1ll << 31));
+    result = (int32_t)(mult / (1ll << 31));
 
     if ((m1 == m2) && (m1 == (int32_t)Q31_MIN))
     {
@@ -774,9 +759,8 @@
  */
 __STATIC_FORCEINLINE q31_t arm_nn_requantize(const q31_t val, const q31_t multiplier, const q31_t shift)
 {
-  return arm_nn_divide_by_power_of_two(
-      arm_nn_doubling_high_mult_no_sat(val * (1 << LEFT_SHIFT(shift)), multiplier),
-      RIGHT_SHIFT(shift));
+    return arm_nn_divide_by_power_of_two(arm_nn_doubling_high_mult_no_sat(val * (1 << LEFT_SHIFT(shift)), multiplier),
+                                         RIGHT_SHIFT(shift));
 }
 
 /**
@@ -786,22 +770,18 @@
  * @param[in]       block_size  Number of bytes to copy.
  *
  */
-__STATIC_FORCEINLINE void arm_memcpy_q7(q7_t *__RESTRICT dst,
-                                        const q7_t *__RESTRICT src,
-                                        uint32_t block_size)
+__STATIC_FORCEINLINE void arm_memcpy_q7(q7_t *__RESTRICT dst, const q7_t *__RESTRICT src, uint32_t block_size)
 {
 #if defined(ARM_MATH_MVEI)
-     __asm volatile (
-        "   wlstp.8                 lr, %[cnt], 1f             \n"
-        "2:                                                    \n"
-        "   vldrb.8                 q0, [%[in]], 16            \n"
-        "   vstrb.8                 q0, [%[out]], 16           \n"
-        "   letp                    lr, 2b                     \n"
-        "1:                                                    \n"
-        :[in] "+r"(src)
-        ,[out] "+r"(dst)
-        :[cnt] "r"(block_size)
-        :"q0", "memory", "r14");
+    __asm volatile("   wlstp.8                 lr, %[cnt], 1f             \n"
+                   "2:                                                    \n"
+                   "   vldrb.8                 q0, [%[in]], 16            \n"
+                   "   vstrb.8                 q0, [%[out]], 16           \n"
+                   "   letp                    lr, 2b                     \n"
+                   "1:                                                    \n"
+                   : [in] "+r"(src), [out] "+r"(dst)
+                   : [cnt] "r"(block_size)
+                   : "q0", "memory", "r14");
 #else
     memcpy(dst, src, block_size);
 #endif
@@ -830,10 +810,10 @@
  */
 __STATIC_FORCEINLINE int32x4_t arm_divide_by_power_of_two_mve(const int32x4_t dividend, const q31_t exponent)
 {
-  const int32x4_t shift = vdupq_n_s32(-exponent);
-  const int32x4_t fixup = vshrq_n_s32(vandq_s32(dividend, shift), 31);
-  const int32x4_t fixed_up_dividend = vqaddq_s32(dividend, fixup);
-  return vrshlq_s32(fixed_up_dividend, shift);
+    const int32x4_t shift = vdupq_n_s32(-exponent);
+    const int32x4_t fixup = vshrq_n_s32(vandq_s32(dividend, shift), 31);
+    const int32x4_t fixed_up_dividend = vqaddq_s32(dividend, fixup);
+    return vrshlq_s32(fixed_up_dividend, shift);
 }
 
 /**
@@ -847,33 +827,35 @@
  */
 __STATIC_FORCEINLINE int32x4_t arm_requantize_mve(const int32x4_t val, const q31_t multiplier, const q31_t shift)
 {
-  return arm_divide_by_power_of_two_mve(
-          arm_doubling_high_mult_mve(vshlq_s32(val, vdupq_n_s32(LEFT_SHIFT(shift))), multiplier),
-          RIGHT_SHIFT(shift));
+    return arm_divide_by_power_of_two_mve(
+        arm_doubling_high_mult_mve(vshlq_s32(val, vdupq_n_s32(LEFT_SHIFT(shift))), multiplier), RIGHT_SHIFT(shift));
 }
 
 __STATIC_FORCEINLINE int32x4_t arm_doubling_high_mult_mve_32x4(const int32x4_t m1, const int32x4_t m2)
 {
-  return vqrdmulhq_s32(m1, m2);
+    return vqrdmulhq_s32(m1, m2);
 }
 
 __STATIC_FORCEINLINE int32x4_t arm_divide_by_power_of_two_mve_32x4(const int32x4_t dividend, const int32x4_t exponent)
 {
-  const int32x4_t shift = -exponent;
-  const int32x4_t fixup = vshrq_n_s32(vandq_s32(dividend, shift), 31);
-  const int32x4_t fixed_up_dividend = vqaddq_s32(dividend, fixup);
-  return vrshlq_s32(fixed_up_dividend, shift);
+    const int32x4_t shift = -exponent;
+    const int32x4_t fixup = vshrq_n_s32(vandq_s32(dividend, shift), 31);
+    const int32x4_t fixed_up_dividend = vqaddq_s32(dividend, fixup);
+    return vrshlq_s32(fixed_up_dividend, shift);
 }
 
-__STATIC_FORCEINLINE int32x4_t arm_requantize_mve_32x4(const int32x4_t val, const int32x4_t multiplier, const int32x4_t shift)
+__STATIC_FORCEINLINE int32x4_t arm_requantize_mve_32x4(const int32x4_t val,
+                                                       const int32x4_t multiplier,
+                                                       const int32x4_t shift)
 {
-  const int32x4_t zz = vdupq_n_s32(0);
-  const mve_pred16_t p = vcmpgtq_n_s32(shift, 0);
+    const int32x4_t zz = vdupq_n_s32(0);
+    const mve_pred16_t p = vcmpgtq_n_s32(shift, 0);
 
-  const int32x4_t left_shift = vpselq_s32(shift, zz, p);
-  const int32x4_t right_shift = -vpselq_s32(zz, shift, p);
+    const int32x4_t left_shift = vpselq_s32(shift, zz, p);
+    const int32x4_t right_shift = -vpselq_s32(zz, shift, p);
 
-  return arm_divide_by_power_of_two_mve_32x4(arm_doubling_high_mult_mve_32x4(vshlq_s32(val, left_shift), multiplier), right_shift);
+    return arm_divide_by_power_of_two_mve_32x4(arm_doubling_high_mult_mve_32x4(vshlq_s32(val, left_shift), multiplier),
+                                               right_shift);
 }
 #endif
 
@@ -881,22 +863,22 @@
 
 __STATIC_FORCEINLINE int32_t arm_nn_exp_on_negative_values(int32_t val)
 {
-    int32_t mask  = 0;
+    int32_t mask = 0;
     int32_t shift = 24;
 
     const int32_t val_mod_minus_quarter = (val & ((1 << shift) - 1)) - (1 << shift);
-    const int32_t remainder             = val_mod_minus_quarter - val;
-    const int32_t x                     = (val_mod_minus_quarter << 5) + (1 << 28);
-    const int32_t x2                    = MUL_SAT(x, x);
+    const int32_t remainder = val_mod_minus_quarter - val;
+    const int32_t x = (val_mod_minus_quarter << 5) + (1 << 28);
+    const int32_t x2 = MUL_SAT(x, x);
 
-    int32_t result = 1895147668 + MUL_SAT(1895147668, x +
-        DIV_POW2(MUL_SAT(DIV_POW2(MUL_SAT(x2, x2), 2) + MUL_SAT(x2, x), 715827883) + x2, 1));
+    int32_t result = 1895147668 +
+        MUL_SAT(1895147668, x + DIV_POW2(MUL_SAT(DIV_POW2(MUL_SAT(x2, x2), 2) + MUL_SAT(x2, x), 715827883) + x2, 1));
 
-#define SELECT_IF_NON_ZERO(x)                                     \
-{                                                                 \
-    mask   = MASK_IF_NON_ZERO(remainder & (1 << shift++));        \
-    result = SELECT_USING_MASK(mask, MUL_SAT(result, x), result); \
-}
+#define SELECT_IF_NON_ZERO(x)                                                                                          \
+    {                                                                                                                  \
+        mask = MASK_IF_NON_ZERO(remainder & (1 << shift++));                                                           \
+        result = SELECT_USING_MASK(mask, MUL_SAT(result, x), result);                                                  \
+    }
 
     SELECT_IF_NON_ZERO(1672461947)
     SELECT_IF_NON_ZERO(1302514674)