COMPMID-718 : Winograd: add validate method and tests

Validate methods added to Winograd kernels and function.
Renamed validation test suit

Change-Id: I0a88df436aff0bbaf4fd82213eeda089b87ac5bf
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/127781
Tested-by: Jenkins <bsgcomp@arm.com>
Reviewed-by: Anthony Barbier <anthony.barbier@arm.com>
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>

3 files changed, 66 insertions, 4 deletions

diff --git a/arm_compute/core/Error.h b/arm_compute/core/Error.h
index 590da9b58e..3635e93244 100644
--- a/arm_compute/core/Error.h
+++ b/arm_compute/core/Error.h
@@ -175,6 +175,16 @@ Status create_error(ErrorCode error_code, const char *function, const char *file
  */
 #define ARM_COMPUTE_CREATE_ERROR_LOC(error_code, func, file, line, ...) ::arm_compute::create_error(error_code, func, file, line, __VA_ARGS__) // NOLINT
 
+/** An error is returned with the given description.
+ *
+ * @param[in] ... Error description message.
+ */
+#define ARM_COMPUTE_RETURN_ERROR_MSG(...)                                                    \
+    do                                                                                       \
+    {                                                                                        \
+        return ARM_COMPUTE_CREATE_ERROR(arm_compute::ErrorCode::RUNTIME_ERROR, __VA_ARGS__); \
+    } while(false)
+
 /** Checks if a status contains an error and returns it
  *
  * @param[in] status Status value to check
diff --git a/arm_compute/core/NEON/kernels/NEWinogradLayerKernel.h b/arm_compute/core/NEON/kernels/NEWinogradLayerKernel.h
index 2f44d19b4f..69d8cc6851 100644
--- a/arm_compute/core/NEON/kernels/NEWinogradLayerKernel.h
+++ b/arm_compute/core/NEON/kernels/NEWinogradLayerKernel.h
@@ -124,7 +124,7 @@ public:
 
     /** Configure the output transform kernel.
      *
-     * @param[in]  input         Input tensor data
+     * @param[in]  input         Input tensor data. Data types supported: F32.
      * @param[in]  n_batches     Number of batches in input tensor.
      * @param[in]  n_rows        Number of rows in input tensor.
      * @param[in]  n_cols        Number of columns in input tensor.
@@ -152,6 +152,17 @@ public:
     /** Winograd convolution kernel */
     using WinogradConv = typename WinogradBase::template Convolution<T, T>;
 
+    /** Static function to check if given info will lead to a valid configuration of @ref NEWinogradLayerTransformInputKernel
+     *
+     * @param[in] input       First tensor input info. Data types supported: F32.
+     * @param[in] output      Output tensor info. Data types supported: same as @p input.
+     * @param[in] conv_info   Contains padding and stride information described in @ref PadStrideInfo. Currently only unit strides are supported.
+     * @param[in] kernel_dims Kernel dimensions. Currently only 3x3 and 5x5 kernels are supported
+     *
+     * @return a status
+     */
+    static Status validate(const ITensorInfo *input, const ITensorInfo *output, const PadStrideInfo &conv_info, const Size2D &kernel_dims);
+
 private:
     using InputTransform = typename WinogradBase::template InputTransform<T>;
     std::unique_ptr<InputTransform> _transform;
@@ -301,6 +312,19 @@ public:
     void run(const Window &window, const ThreadInfo &info) override;
     bool is_parallelisable() const override;
 
+    /** Static function to check if given info will lead to a valid configuration of @ref NEWinogradLayerTransformOutputKernel
+     *
+     * @param[in]  input                 Source tensor with shape [C, N, 16, batches] or [C, N, 36, batches]. Data types supported: F32.
+     * @param[in]  bias                  Biases tensor. Shared biases supported. Biases are 1D tensor with dimensions [OFM]. It can be a nullptr. Data type supported: as @p input
+     * @param[out] output                Destination tensor with shape [output_convolved_dims.width, output_convolved_dims.height, C, batches]. Data type supported: same as @p input
+     * @param[in]  kernel_dims           Kernel dimensions (Width and height). Currently only supported 3x3 and 5x5 kernels
+     * @param[in]  output_convolved_dims Output dimensions after the convolution (Width and height)
+     * @param[in]  num_tiles             Number of tiles of size 2x2 or 4x4 in the output tensor along the X and Y direction
+     *
+     * @return a status
+     */
+    static Status validate(const ITensorInfo *input, const ITensorInfo *bias, const ITensorInfo *output, const Size2D &kernel_dims, const Size2D &output_convolved_dims, const Size2D &num_tiles);
+
 private:
     using WinogradBase    = winograd::WinogradGEMM<OutputTileRows, OutputTileCols, KernelRows, KernelCols>;
     using WinogradConv    = typename WinogradBase::template Convolution<T, T>;
@@ -366,6 +390,17 @@ public:
         return "NEWinogradLayerTransformWeightsKernel";
     }
 
+    /** Static function to check if given info will lead to a valid configuration of @ref NEWinogradLayerTransformWeightsKernel
+     *
+     * @param[in] input       Source tensor info. The input is a 4D tensor with dimensions [kernel_x, kernel_y, IFM, OFM] (NCHW data layout).
+     *                        kernel_x must be 3 and equal to kernel_y. Data types supported: F32.
+     * @param[in] output      Destination tensor info. The output is a 3D tensor with dimensions [OFM, IFM, 16] or [OFM, IFM, 36]. Data type supported: same as @p input
+     * @param[in] output_tile Output tile. Currently only 2x2 and 4x4 tiles are supported.
+     *
+     * @return a status
+     */
+    static Status validate(const ITensorInfo *input, const ITensorInfo *output, const Size2D &output_tile);
+
     // Inherited methods overridden:
     void configure(const ITensor *weights_hwio, T *const output, const int matrix_stride, const int n_output_channels, const int n_input_channels) override;
     unsigned int get_weight_storage_size(int n_output_channels, int n_input_channels) const override;
@@ -496,6 +531,21 @@ public:
 
     void run(const Window &window, const ThreadInfo &info) override;
 
+    /** Static function to check if given info will lead to a valid configuration of @ref NEWinogradLayerBatchedGEMMKernel.
+     *
+     * @param[in]  a         First input tensor  (Matrix or Vector A). Data types supported: F32
+     * @param[in]  b         Second input tensor (Matrix B). Data type supported: same as @p a.
+     * @param[in]  c         Third input tensor  (Matrix C). It can be a nullptr if just the multiplication between @p a and @p b is needed. Data type supported: same as @p a.
+     * @param[out] output    Output tensor. Data type supported: same as @p a
+     * @param[in]  alpha     Weight of the matrix product
+     * @param[in]  beta      Weight of matrix C
+     * @param[in]  gemm_info (Optional) Specifies if the matrix A and/or matrix B have been reshaped and
+     *                       if the reshape of matrix B should happen only for the first run
+     *
+     * @return a status
+     */
+    static Status validate(const ITensorInfo *a, const ITensorInfo *b, const ITensor *c, const ITensorInfo *output, const float alpha, const float beta, const GEMMInfo &gemm_info = GEMMInfo());
+
 private:
     static const int           _output_tile_rows = OutputTileRows;
     static const int           _output_tile_cols = OutputTileCols;
diff --git a/arm_compute/runtime/NEON/functions/NEWinogradLayer.h b/arm_compute/runtime/NEON/functions/NEWinogradLayer.h
index 61a4caae3a..27b1e84201 100644
--- a/arm_compute/runtime/NEON/functions/NEWinogradLayer.h
+++ b/arm_compute/runtime/NEON/functions/NEWinogradLayer.h
@@ -57,7 +57,7 @@ public:
      *                       while every optional dimension from 4 and above represent a batch of inputs.
      *                       Data types supported: F32.
      * @param[in]  weights   Weights tensor. Weights are 4D tensor with dimensions [kernel_x, kernel_y, IFM, OFM]. Data type supported: Same as @p input.
-     *                       Currently only 3x3 kernels are supported.
+     *                       Currently only 3x3 and 5x5 kernels are supported.
      * @param[in]  biases    Biases tensor. Shared biases supported. Biases are 1D tensor with dimensions [OFM]. Data type supported: Same as @p weights.
      * @param[out] output    Destination tensor. 3 lower dimensions represent a single output [width, height, OFM], while the rest represent batch of outputs.
      *                       Data types supported: Same as @p input.
@@ -75,15 +75,17 @@ public:
      *                      while every optional dimension from 4 and above represent a batch of inputs.
      *                      Data types supported: F32.
      * @param[in] weights   Weights tensor. Weights are 4D tensor with dimensions [kernel_x, kernel_y, IFM, OFM]. Data type supported:Same as @p input.
-     *                      Currently only 3x3 kernels are supported.
+     *                      Currently only 3x3 and 5x5 kernels are supported.
      * @param[in] biases    Biases tensor. Shared biases supported. Biases are 1D tensor with dimensions [OFM]. Data type supported: Same as @p weights.
      * @param[in] output    Destination tensor. 3 lower dimensions represent a single output [width, height, OFM], while the rest represent batch of outputs.
      *                      Data types supported: Same as @p input.
      * @param[in] conv_info Contains padding and stride information described in @ref PadStrideInfo. Currently only unit strides are supported.
+     * @param[in] act_info  (Optional) Activation layer information in case of a fused activation.
      *
      * @return a status
      */
-    static Status validate(const ITensorInfo *input, const ITensorInfo *weights, const ITensorInfo *biases, const ITensorInfo *output, const PadStrideInfo &conv_info);
+    static Status validate(const ITensorInfo *input, const ITensorInfo *weights, const ITensorInfo *biases, const ITensorInfo *output, const PadStrideInfo &conv_info,
+                           const ActivationLayerInfo &act_info = ActivationLayerInfo());
 
     /** Prevent instances of this class from being copied (As this class contains pointers) */
     NEWinogradLayer(const NEWinogradLayer &) = delete;