COMPMID-754: Add validation to kernels.

Adds validation method to:
- CLConvolutionLayer

Change-Id: I95516e20cfb71c1e603c60fc6491ac695883a856
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/117355
Reviewed-by: Anthony Barbier <anthony.barbier@arm.com>
Tested-by: Jenkins <bsgcomp@arm.com>

8 files changed, 114 insertions, 22 deletions

diff --git a/arm_compute/core/CL/kernels/CLCol2ImKernel.h b/arm_compute/core/CL/kernels/CLCol2ImKernel.h
index bd86da1b5e..24d0fdd914 100644
--- a/arm_compute/core/CL/kernels/CLCol2ImKernel.h
+++ b/arm_compute/core/CL/kernels/CLCol2ImKernel.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2017 ARM Limited.
+ * Copyright (c) 2017-2018 ARM Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -72,6 +72,16 @@ public:
      * @param[in]  convolved_dims Output convolved dimensions.
      */
     void configure(const ICLTensor *input, ICLTensor *output, std::pair<unsigned int, unsigned int> convolved_dims);
+    /** Static function to check if given info will lead to a valid configuration of @ref CLCol2ImKernel
+     *
+     * @param[in] input          The input tensor to convert. Data types supported: QS8/QS16/QASYMM8/F16/F32
+     * @param[in] output         The output tensor. 3 lower dimensions represent a single output [width, height, OFM],
+     *                           while the rest represent batch of outputs. Data types supported: Same as @p input
+     * @param[in] convolved_dims Output convolved dimensions.
+     *
+     * @return a status
+     */
+    static Status validate(const ITensorInfo *input, const ITensorInfo *output, std::pair<unsigned int, unsigned int> convolved_dims);
 
     // Inherited methods overridden:
     void run(const Window &window, cl::CommandQueue &queue) override;
diff --git a/arm_compute/core/CL/kernels/CLGEMMMatrixAdditionKernel.h b/arm_compute/core/CL/kernels/CLGEMMMatrixAdditionKernel.h
index 8f73d8c2c3..dc84a40ca8 100644
--- a/arm_compute/core/CL/kernels/CLGEMMMatrixAdditionKernel.h
+++ b/arm_compute/core/CL/kernels/CLGEMMMatrixAdditionKernel.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2017 ARM Limited.
+ * Copyright (c) 2017-2018 ARM Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -57,6 +57,15 @@ public:
      * @param[in]      beta   Weight of matrix C
      */
     void configure(const ICLTensor *input, ICLTensor *output, float beta);
+    /** Static function to check if given info will lead to a valid configuration of @ref CLGEMMMatrixAdditionKernel.
+     *
+     * @param[in] input  Input tensor (Matrix C). Data types supported: QS8/QS16/F16/F32
+     * @param[in] output Output tensor. If this kernel is used to finalize the GEMM result (alpha * AB + beta * C), output must contain the result obtained by @ref CLGEMMMatrixMultiplyKernel. Data type supported: same as @p input
+     * @param[in] beta   Weight of matrix C
+     *
+     * @return a status
+     */
+    static Status validate(const ITensorInfo *input, const ITensorInfo *output, const float beta);
 
     // Inherited methods overridden:
     void run(const Window &window, cl::CommandQueue &queue) override;
diff --git a/arm_compute/core/CL/kernels/CLIm2ColKernel.h b/arm_compute/core/CL/kernels/CLIm2ColKernel.h
index e38e7e8a49..1ad302eedb 100644
--- a/arm_compute/core/CL/kernels/CLIm2ColKernel.h
+++ b/arm_compute/core/CL/kernels/CLIm2ColKernel.h
@@ -77,9 +77,6 @@ public:
      * @param[in]  has_bias    In case biases are provided expands the matrix with 1.
      */
     void configure(const ICLTensor *input, ICLTensor *output, const Size2D &kernel_dims, const PadStrideInfo &conv_info, bool has_bias);
-
-    // Inherited methods overridden:
-    void run(const Window &window, cl::CommandQueue &queue) override;
     /** Static function to check if given info will lead to a valid configuration of @ref CLIm2ColKernel
      *
      * @param[in] input       The input tensor to convert. 3 lower dimensions represent a single input [width, height, IFM],
@@ -94,6 +91,9 @@ public:
      */
     static Status validate(const ITensorInfo *input, const ITensorInfo *output, const Size2D &kernel_dims, const PadStrideInfo &conv_info, bool has_bias);
 
+    // Inherited methods overridden:
+    void run(const Window &window, cl::CommandQueue &queue) override;
+
 private:
     /** Run the reshape kernel optimised for the special case (stride is 1, padding is 0 and kernel's low 3 dimensions are same as input)
      *
diff --git a/arm_compute/core/CL/kernels/CLWeightsReshapeKernel.h b/arm_compute/core/CL/kernels/CLWeightsReshapeKernel.h
index 6c84ded49e..b9ede12e3d 100644
--- a/arm_compute/core/CL/kernels/CLWeightsReshapeKernel.h
+++ b/arm_compute/core/CL/kernels/CLWeightsReshapeKernel.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2017 ARM Limited.
+ * Copyright (c) 2017-2018 ARM Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -43,7 +43,6 @@ public:
     CLWeightsReshapeKernel &operator=(CLWeightsReshapeKernel &&) = default;
     /** Default destructor */
     ~CLWeightsReshapeKernel() = default;
-
     /** Set the input and output of the kernel.
      *
      * @param[in]  input  The input tensor to convert. Weights are 4D tensor with dimensions [kernel_x, kernel_y, IFM, OFM] if shared,
@@ -54,6 +53,18 @@ public:
      * @param[out] output The output tensor. Should be a 2D Tensor. Data types supported: Same as @p input
      */
     void configure(const ICLTensor *input, const ICLTensor *biases, ICLTensor *output);
+    /** Static function to check if given info will lead to a valid configuration of @ref CLWeightsReshapeKernel
+     *
+     * @param[in] input  The input tensor to convert. Weights are 4D tensor with dimensions [kernel_x, kernel_y, IFM, OFM] if shared,
+     *                   and 5D tensor with dimensions [kernel_x, kernel_y, IFM, OFM,  num_patches] if unshared. Data types supported: QS8/QS16/QASYMM8/F16/F32
+     * @param[in] biases The shared biases tensor to append.  Bias is 1D tensor with dimensions [OFM] if shared and 2D tensor with
+     *                   dimensions [OFM, num_patches] if unshared. Data types supported: Same as @p input
+     *                   @warning Appending biases to weights reshaped matrix is not supported for quantized asymmetric types.
+     * @param[in] output The output tensor. Should be a 2D Tensor. Data types supported: Same as @p input
+     *
+     * @return a status
+     */
+    static Status validate(const ITensorInfo *input, const ITensorInfo *biases, const ITensorInfo *output);
 
     // Inherited methods overridden:
     void run(const Window &window, cl::CommandQueue &queue) override;
diff --git a/arm_compute/core/utils/misc/ShapeCalculator.h b/arm_compute/core/utils/misc/ShapeCalculator.h
index e51c6bbe98..c53ac4c71f 100644
--- a/arm_compute/core/utils/misc/ShapeCalculator.h
+++ b/arm_compute/core/utils/misc/ShapeCalculator.h
@@ -40,6 +40,17 @@ inline TensorShape compute_permutation_output_shape(const ITensorInfo &input, co
     permute(output_shape, perm);
     return output_shape;
 }
+inline TensorShape compute_weights_reshaped_shape(const ITensorInfo &weights, bool has_bias = false)
+{
+    // Calculate output shape
+    TensorShape weights_reshaped{ weights.tensor_shape() };
+    weights_reshaped.collapse(3);
+    const size_t tmp_dim = weights_reshaped[0];
+    weights_reshaped.set(0, weights_reshaped[1]);
+    weights_reshaped.set(1, tmp_dim + (has_bias ? 1 : 0));
+
+    return weights_reshaped;
+}
 inline TensorShape compute_interleaved_shape(const ITensorInfo &a, int mult_interleave4x4_height = 1)
 {
     // The interleaved output matrix will have the following shape: [ a_height * W, ceil(a_width / W) ] where W = 4 * mult_interleave4x4_height
@@ -101,6 +112,15 @@ inline TensorShape compute_im2col_shape(const ITensorInfo &input)
 
     return shape_im2col;
 }
+inline TensorShape compute_col2im_shape(const ITensorInfo &input, std::pair<unsigned int, unsigned int> convolved_dims)
+{
+    TensorShape col2im_shape{ input.tensor_shape() };
+    col2im_shape.set(0, convolved_dims.first);
+    col2im_shape.set(1, convolved_dims.second);
+    col2im_shape.set(2, input.tensor_shape()[0]);
+
+    return col2im_shape;
+}
 inline TensorShape compute_transposed_shape(const ITensorInfo &input)
 {
     TensorShape shape_transposed{ input.tensor_shape() };
diff --git a/arm_compute/runtime/CL/functions/CLGEMM.h b/arm_compute/runtime/CL/functions/CLGEMM.h
index 0f144915d7..2e82457ee2 100644
--- a/arm_compute/runtime/CL/functions/CLGEMM.h
+++ b/arm_compute/runtime/CL/functions/CLGEMM.h
@@ -72,6 +72,20 @@ public:
      *                       in case matrix A and matrix B have been already transformed.
      */
     void configure(const ICLTensor *a, const ICLTensor *b, const ICLTensor *c, ICLTensor *output, float alpha, float beta, const GEMMInfo &gemm_info = GEMMInfo());
+    /** Static function to check if given info will lead to a valid configuration of @ref CLGEMM.
+     *
+     * @param[in]  a         First input tensor  (Matrix or Vector A). Data types supported: QS8/QS16/F16/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 ICLTensor *c, const ITensorInfo *output, const float alpha, const float beta, const GEMMInfo &gemm_info = GEMMInfo());
 
     // Inherited methods overridden:
     void run() override;
diff --git a/arm_compute/runtime/CL/functions/CLGEMMConvolutionLayer.h b/arm_compute/runtime/CL/functions/CLGEMMConvolutionLayer.h
index 7126688f8b..24029509b8 100644
--- a/arm_compute/runtime/CL/functions/CLGEMMConvolutionLayer.h
+++ b/arm_compute/runtime/CL/functions/CLGEMMConvolutionLayer.h
@@ -58,14 +58,22 @@ public:
     CLConvolutionLayerReshapeWeights(std::shared_ptr<IMemoryManager> memory_manager = nullptr);
     /** Set the input and output tensors.
      *
-     * @param[in]  weights      Weights tensor. Weights are 4D tensor with dimensions [kernel_x, kernel_y, IFM, OFM].
-     *                          Data type supported: QS8/QASYMM8/QS16/F16/F32.
-     * @param[in]  biases       Biases tensor. Shared biases supported. Biases are 1D tensor with dimensions [OFM]. Data type supported: Same as @p weights.
-     * @param[out] output       Destination tensor. Data types supported: Same as @p weights.
-     * @param[in]  transpose1xW True if the weights are to undergo a 1xW transposition after reshaping (in case of GEMM operation), false otherwise.
-     *                          Data types supported: Same as @p weights.
+     * @param[in]  weights Weights tensor. Weights are 4D tensor with dimensions [kernel_x, kernel_y, IFM, OFM].
+     *                     Data type supported: QS8/QASYMM8/QS16/F16/F32.
+     * @param[in]  biases  Biases tensor. Shared biases supported. Biases are 1D tensor with dimensions [OFM]. Data type supported: Same as @p weights.
+     * @param[out] output  Destination tensor. Data types supported: Same as @p weights.
      */
-    void configure(const ICLTensor *weights, const ICLTensor *biases, ICLTensor *output, bool transpose1xW);
+    void configure(const ICLTensor *weights, const ICLTensor *biases, ICLTensor *output);
+    /** Static function to check if given info will lead to a valid configuration of @ref CLConvolutionLayerReshapeWeights
+     *
+     * @param[in] weights Weights tensor. Weights are 4D tensor with dimensions [kernel_x, kernel_y, IFM, OFM].
+     *                    Data type supported: QS8/QASYMM8/QS16/F16/F32.
+     * @param[in] biases  Biases tensor. Shared biases supported. Biases are 1D tensor with dimensions [OFM]. Data type supported: Same as @p weights.
+     * @param[in] output  Destination tensor. Data types supported: Same as @p weights.
+     *
+     * @return a status
+     */
+    static Status validate(const ITensorInfo *weights, const ITensorInfo *biases, const ITensorInfo *output);
     // Inherited methods overridden:
     void run() override;
 
@@ -74,7 +82,6 @@ private:
     CLWeightsReshapeKernel   _weights_reshape_kernel;
     CLGEMMTranspose1xWKernel _weights_transposed_kernel;
     CLTensor                 _weights_reshaped;
-    bool                     _transpose1xW;
 };
 
 /** Basic function to compute the convolution layer. This function calls the following OpenCL kernels/functions:
@@ -112,6 +119,22 @@ public:
      *                          tensor has also been transposed with CLGEMMTranspose1xWKernel. Data type supported: Same as @p input.
      */
     void configure(const ICLTensor *input, const ICLTensor *weights, const ICLTensor *biases, ICLTensor *output, const PadStrideInfo &conv_info, const WeightsInfo &weights_info = WeightsInfo());
+    /** Static function to check if given info will lead to a valid configuration of @ref CLGEMMConvolutionLayer.
+     *
+     * @param[in]  input        Source tensor. 3 lower dimensions represent a single input [width, height, IFM],
+     *                          while every optional dimension from 4 and above represent a batch of inputs.
+     *                          Data types supported: QS8/QASYMM8/QS16/F16/F32.
+     * @param[in]  weights      Weights tensor. Weights are 4D tensor with dimensions [kernel_x, kernel_y, IFM, OFM]. Data type supported: Same as @p input.
+     * @param[in]  biases       Biases tensor. Shared biases supported. Biases are 1D tensor with dimensions [OFM].
+     *                          Data type supported: Should match @p input data type, except for input of QASYMM8 type where biases should be of S32 type.
+     * @param[out] output       Destination tensor. 3 lower dimensions represent a single output [width, height, OFM], while the rest represent batch of outputs.
+     *                          Data types supported: Same as @p input.
+     * @param[in]  conv_info    Contains padding and stride information described in @ref PadStrideInfo.
+     * @param[in]  weights_info Specifies if the weights tensor has been reshaped with CLWeightsReshapeKernel. If this is not part of the fully connected layer the weights
+     *                          tensor has also been transposed with CLGEMMTranspose1xWKernel. Data type supported: Same as @p input.
+     */
+    static Status validate(const ITensorInfo *input, const ITensorInfo *weights, const ITensorInfo *biases, const ITensorInfo *output, const PadStrideInfo &conv_info,
+                           const WeightsInfo &weights_info = WeightsInfo());
 
     // Inherited methods overridden:
     void run() override;
@@ -123,16 +146,23 @@ private:
      * @param weights                   Weights tensor. Data type supported: Same as @p input.
      * @param output                    Output tensor. Data types supported: Same as @p input,
      *                                                 except for input of QASYMM8 type where output should be of S32 type.
-     * @param is_interleaved_transposed Flag that signals if matrix is interleaved transposed
      */
-    void configure_mm(const ICLTensor *input, const ICLTensor *weights, ICLTensor *output, bool is_interleaved_transposed, bool are_weights_reshaped);
+    void configure_mm(const ICLTensor *input, const ICLTensor *weights, ICLTensor *output);
+    /** Static function to check if given info will lead to a valid configuration of @ref CLGEMMConvolutionLayer matrix multiply routines
+     *
+     * @param[in] input   Input tensor. Data types supported: QS8/QASYMM8/QS16/F16/F32.
+     * @param[in] weights Weights tensor. Data type supported: Same as @p input.
+     * @param[in] output  Output tensor. Data types supported: Same as @p input,
+     *                                      except for input of QASYMM8 type where output should be of S32 type.
+     *
+     * @return a status
+     */
+    static Status validate_mm(const ITensorInfo *input, const ITensorInfo *weights, const ITensorInfo *output);
 
 private:
     CLMemoryGroup                                       _memory_group;
     CLConvolutionLayerReshapeWeights                    _reshape_weights;
     CLIm2ColKernel                                      _im2col_kernel;
-    CLGEMMInterleave4x4Kernel                           _interleave_kernel;
-    CLGEMMMatrixMultiplyKernel                          _mm_kernel;
     CLGEMM                                              _mm_gemm;
     CLGEMMLowpMatrixMultiplyCore                        _mm_gemmlowp;
     CLGEMMLowpQuantizeDownInt32ToUint8ScaleByFixedPoint _gemmlowp_output_stage;
@@ -145,9 +175,7 @@ private:
     CLTensor _gemm_output;
     CLTensor _tmp_output;
 
-    bool _are_weights_reshaped;
     bool _is_quantized;
-    bool _is_interleaved_transposed;
 };
 }
 #endif /* __ARM_COMPUTE_CLGEMMCONVOLUTIONLAYER_H__ */
diff --git a/arm_compute/runtime/NEON/functions/NEGEMM.h b/arm_compute/runtime/NEON/functions/NEGEMM.h
index 4b0614badc..f2b6ef77bd 100644
--- a/arm_compute/runtime/NEON/functions/NEGEMM.h
+++ b/arm_compute/runtime/NEON/functions/NEGEMM.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2017 ARM Limited.
+ * Copyright (c) 2017-2018 ARM Limited.
  *
  * SPDX-License-Identifier: MIT
  *