COMPMID-556 - Add QASYMM8 support for missing OpenCL kernels

- CLCol2ImKernel
- CLGEMMIntereleave4x4Kernel
- CLIm2ColKernel
- CLTransposeKernel
- CLWeightsReshapeKernel

Change-Id: I2c6066f59f078cfe88ed0f3d2f61db137d375159
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/110790
Tested-by: BSG Visual Compute Jenkins server to access repositories on http://mpd-gerrit.cambridge.arm.com <bsgcomp@arm.com>
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>

11 files changed, 18 insertions, 15 deletions

diff --git a/arm_compute/core/CL/kernels/CLGEMMInterleave4x4Kernel.h b/arm_compute/core/CL/kernels/CLGEMMInterleave4x4Kernel.h
index 8e0c1836a9..c87fb2cd66 100644
--- a/arm_compute/core/CL/kernels/CLGEMMInterleave4x4Kernel.h
+++ b/arm_compute/core/CL/kernels/CLGEMMInterleave4x4Kernel.h
@@ -64,7 +64,7 @@ public:
     CLGEMMInterleave4x4Kernel &operator=(CLGEMMInterleave4x4Kernel &&) = default;
     /** Initialise the kernel's input and output.
      *
-     * @param[in]  input  Input tensor. Data types supported: U8/S8/QS8/U16/S16/QS16/F16/U32/S32/F32
+     * @param[in]  input  Input tensor. Data types supported: U8/S8/QS8/QASYMM8/U16/S16/QS16/F16/U32/S32/F32
      * @param[out] output Output tensor. Data type supported: same as @p input
      */
     void configure(const ICLTensor *input, ICLTensor *output);
diff --git a/arm_compute/core/CL/kernels/CLGEMMTranspose1xWKernel.h b/arm_compute/core/CL/kernels/CLGEMMTranspose1xWKernel.h
index 50bc64c2c6..8a37720462 100644
--- a/arm_compute/core/CL/kernels/CLGEMMTranspose1xWKernel.h
+++ b/arm_compute/core/CL/kernels/CLGEMMTranspose1xWKernel.h
@@ -70,7 +70,7 @@ class CLGEMMTranspose1xWKernel : public ICLSimple2DKernel
 public:
     /** Initialise the kernel's input and output.
      *
-     * @param[in]  input  Input tensor. Data types supported: U8/S8/QS8/U16/S16/QS16/F16/U32/S32/F32
+     * @param[in]  input  Input tensor. Data types supported: U8/S8/QS8/QASYMM8/U16/S16/QS16/F16/U32/S32/F32
      * @param[out] output Output tensor. Data type supported: same as @p input
      */
     void configure(const ICLTensor *input, ICLTensor *output);
diff --git a/arm_compute/core/CL/kernels/CLIm2ColKernel.h b/arm_compute/core/CL/kernels/CLIm2ColKernel.h
index eed683b4c0..1d8b5500c1 100644
--- a/arm_compute/core/CL/kernels/CLIm2ColKernel.h
+++ b/arm_compute/core/CL/kernels/CLIm2ColKernel.h
@@ -69,7 +69,7 @@ public:
     /** Set the input and output of the kernel.
      *
      * @param[in]  input       The input tensor to convert. 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/QS16/F16/F32
+     *                         while every optional dimension from 4 and above represent a batch of inputs. Data types supported: QS8/QS16/QASYMM8/F16/F32
      * @param[out] output      The output tensor. First 2 lower dimensions represent a transform of each 3D input,
      *                         while every dimension above represents a batch. Data types supported: Same as @p input
      * @param[in]  kernel_dims The kernel dimensions (width and height).
diff --git a/arm_compute/core/CL/kernels/CLTransposeKernel.h b/arm_compute/core/CL/kernels/CLTransposeKernel.h
index faccf5e37c..c4bfe706c5 100644
--- a/arm_compute/core/CL/kernels/CLTransposeKernel.h
+++ b/arm_compute/core/CL/kernels/CLTransposeKernel.h
@@ -40,7 +40,7 @@ class CLTransposeKernel : public ICLSimple2DKernel
 public:
     /** Initialise the kernel's input and output.
      *
-     * @param[in]  input  Input tensor. Data types supported: U8/S8/QS8/U16/S16/QS16/F16/U32/S32/F32
+     * @param[in]  input  Input tensor. Data types supported: U8/S8/QS8/QASYMM8/U16/S16/QS16/F16/U32/S32/F32
      * @param[out] output Output tensor. Data type supported: Same as @p input
      */
     void configure(const ICLTensor *input, ICLTensor *output);
diff --git a/arm_compute/core/CL/kernels/CLWeightsReshapeKernel.h b/arm_compute/core/CL/kernels/CLWeightsReshapeKernel.h
index 07c7c772c3..ec8940ef7e 100644
--- a/arm_compute/core/CL/kernels/CLWeightsReshapeKernel.h
+++ b/arm_compute/core/CL/kernels/CLWeightsReshapeKernel.h
@@ -47,7 +47,7 @@ public:
     /** Set the input and output of the kernel.
      *
      * @param[in]  input  The input tensor to convert. Weights are 4D tensor with dimensions [kernel_x, kernel_y, IFM, OFM] if shared,
-     *                    and 5D tensor with dimensions [kernel_x, kernel_y, IFM, OFM,  num_patches] if unshared. Data types supported: QS8/QS16/F16/F32
+     *                    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
      * @param[out] output The output tensor. Should be a 2D Tensor. Data types supported: Same as @p input
diff --git a/src/core/CL/kernels/CLCol2ImKernel.cpp b/src/core/CL/kernels/CLCol2ImKernel.cpp
index 4521b435b3..31cc6448c9 100644
--- a/src/core/CL/kernels/CLCol2ImKernel.cpp
+++ b/src/core/CL/kernels/CLCol2ImKernel.cpp
@@ -52,7 +52,7 @@ void CLCol2ImKernel::configure(const ICLTensor *input, ICLTensor *output, std::p
     output_shape.set(2, input->info()->tensor_shape()[0]);
 
     // Output auto inizialitation if not yet initialized
-    auto_init_if_empty(*output->info(), output_shape, 1, input->info()->data_type(), input->info()->fixed_point_position());
+    auto_init_if_empty(*output->info(), output_shape, 1, input->info()->data_type(), input->info()->fixed_point_position(), input->info()->quantization_info());
 
     ARM_COMPUTE_ERROR_ON_MISMATCHING_DIMENSIONS(output->info()->tensor_shape(), output_shape);
     ARM_COMPUTE_ERROR_ON_MISMATCHING_DATA_TYPES(input, output);
diff --git a/src/core/CL/kernels/CLGEMMInterleave4x4Kernel.cpp b/src/core/CL/kernels/CLGEMMInterleave4x4Kernel.cpp
index cea7fb6f67..984121c5bc 100644
--- a/src/core/CL/kernels/CLGEMMInterleave4x4Kernel.cpp
+++ b/src/core/CL/kernels/CLGEMMInterleave4x4Kernel.cpp
@@ -43,7 +43,8 @@ CLGEMMInterleave4x4Kernel::CLGEMMInterleave4x4Kernel()
 
 void CLGEMMInterleave4x4Kernel::configure(const ICLTensor *input, ICLTensor *output)
 {
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::U8, DataType::S8, DataType::QS8, DataType::U16, DataType::S16, DataType::QS16, DataType::U32, DataType::S32, DataType::F16,
+    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::U8, DataType::S8, DataType::QS8, DataType::QASYMM8, DataType::U16, DataType::S16, DataType::QS16, DataType::U32, DataType::S32,
+                                                  DataType::F16,
                                                   DataType::F32);
     ARM_COMPUTE_ERROR_ON_NULLPTR(output);
 
@@ -52,7 +53,7 @@ void CLGEMMInterleave4x4Kernel::configure(const ICLTensor *input, ICLTensor *out
     output_shape.set(1, std::ceil(input->info()->dimension(1) / 4.0f));
 
     // Output auto inizialitation if not yet initialized
-    auto_init_if_empty(*output->info(), output_shape, 1, input->info()->data_type(), input->info()->fixed_point_position());
+    auto_init_if_empty(*output->info(), output_shape, 1, input->info()->data_type(), input->info()->fixed_point_position(), input->info()->quantization_info());
 
     ARM_COMPUTE_ERROR_ON_MISMATCHING_DIMENSIONS(output->info()->tensor_shape(), output_shape);
     ARM_COMPUTE_ERROR_ON_MISMATCHING_DATA_TYPES(input, output);
diff --git a/src/core/CL/kernels/CLGEMMTranspose1xWKernel.cpp b/src/core/CL/kernels/CLGEMMTranspose1xWKernel.cpp
index 5057c8ff21..95b16b32cc 100644
--- a/src/core/CL/kernels/CLGEMMTranspose1xWKernel.cpp
+++ b/src/core/CL/kernels/CLGEMMTranspose1xWKernel.cpp
@@ -40,7 +40,8 @@ using namespace arm_compute;
 
 void CLGEMMTranspose1xWKernel::configure(const ICLTensor *input, ICLTensor *output)
 {
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::U8, DataType::S8, DataType::QS8, DataType::U16, DataType::S16, DataType::QS16, DataType::U32, DataType::S32, DataType::F16,
+    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::U8, DataType::S8, DataType::QS8, DataType::QASYMM8, DataType::U16, DataType::S16, DataType::QS16, DataType::U32, DataType::S32,
+                                                  DataType::F16,
                                                   DataType::F32);
     ARM_COMPUTE_ERROR_ON_NULLPTR(output);
 
@@ -50,7 +51,7 @@ void CLGEMMTranspose1xWKernel::configure(const ICLTensor *input, ICLTensor *outp
     output_shape.set(1, static_cast<size_t>(std::ceil((input->info()->dimension(0) / static_cast<float>(transpose_w)))));
 
     // Output tensor auto inizialitation if not yet initialized
-    auto_init_if_empty(*output->info(), output_shape, 1, input->info()->data_type(), input->info()->fixed_point_position());
+    auto_init_if_empty(*output->info(), output_shape, 1, input->info()->data_type(), input->info()->fixed_point_position(), input->info()->quantization_info());
 
     ARM_COMPUTE_ERROR_ON_MISMATCHING_DATA_TYPES(input, output);
     ARM_COMPUTE_ERROR_ON_MISMATCHING_DIMENSIONS(output->info()->tensor_shape(), output_shape);
diff --git a/src/core/CL/kernels/CLIm2ColKernel.cpp b/src/core/CL/kernels/CLIm2ColKernel.cpp
index aee39b2443..a84116634c 100644
--- a/src/core/CL/kernels/CLIm2ColKernel.cpp
+++ b/src/core/CL/kernels/CLIm2ColKernel.cpp
@@ -46,7 +46,7 @@ CLIm2ColKernel::CLIm2ColKernel()
 
 void CLIm2ColKernel::configure(const ICLTensor *input, ICLTensor *output, const Size2D &kernel_dims, const PadStrideInfo &conv_info, bool has_bias)
 {
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::QS8, DataType::QS16, DataType::F16, DataType::F32);
+    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::QS8, DataType::QS16, DataType::QASYMM8, DataType::F16, DataType::F32);
     ARM_COMPUTE_ERROR_ON_MISMATCHING_DATA_TYPES(input, output);
     ARM_COMPUTE_ERROR_ON_MISMATCHING_FIXED_POINT(input, output);
 
diff --git a/src/core/CL/kernels/CLTransposeKernel.cpp b/src/core/CL/kernels/CLTransposeKernel.cpp
index 75d31d5f16..d57d82f22a 100644
--- a/src/core/CL/kernels/CLTransposeKernel.cpp
+++ b/src/core/CL/kernels/CLTransposeKernel.cpp
@@ -40,7 +40,8 @@ using namespace arm_compute;
 
 void CLTransposeKernel::configure(const ICLTensor *input, ICLTensor *output)
 {
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::U8, DataType::S8, DataType::QS8, DataType::U16, DataType::S16, DataType::QS16, DataType::U32, DataType::S32, DataType::F16,
+    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::U8, DataType::S8, DataType::QS8, DataType::QASYMM8, DataType::U16, DataType::S16, DataType::QS16, DataType::U32, DataType::S32,
+                                                  DataType::F16,
                                                   DataType::F32);
     ARM_COMPUTE_ERROR_ON_NULLPTR(output);
 
@@ -51,7 +52,7 @@ void CLTransposeKernel::configure(const ICLTensor *input, ICLTensor *output)
     output_shape.set(1, h_out);
 
     // Output tensor auto inizialitation if not yet initialized
-    auto_init_if_empty(*output->info(), output_shape, 1, input->info()->data_type(), input->info()->fixed_point_position());
+    auto_init_if_empty(*output->info(), output_shape, 1, input->info()->data_type(), input->info()->fixed_point_position(), input->info()->quantization_info());
 
     ARM_COMPUTE_ERROR_ON_MISMATCHING_DIMENSIONS(output->info()->tensor_shape(), output_shape);
     ARM_COMPUTE_ERROR_ON_MISMATCHING_DATA_TYPES(input, output);
diff --git a/src/core/CL/kernels/CLWeightsReshapeKernel.cpp b/src/core/CL/kernels/CLWeightsReshapeKernel.cpp
index bc27477971..be633b2418 100644
--- a/src/core/CL/kernels/CLWeightsReshapeKernel.cpp
+++ b/src/core/CL/kernels/CLWeightsReshapeKernel.cpp
@@ -41,7 +41,7 @@ CLWeightsReshapeKernel::CLWeightsReshapeKernel()
 
 void CLWeightsReshapeKernel::configure(const ICLTensor *input, const ICLTensor *biases, ICLTensor *output)
 {
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::QS8, DataType::QS16, DataType::F16, DataType::F32);
+    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::QS8, DataType::QS16, DataType::QASYMM8, DataType::F16, DataType::F32);
     ARM_COMPUTE_ERROR_ON_NULLPTR(output);
 
     const DataType dt                   = input->info()->data_type();
@@ -54,7 +54,7 @@ void CLWeightsReshapeKernel::configure(const ICLTensor *input, const ICLTensor *
     output_shape.set(1, tmp_dim + (biases != nullptr ? 1 : 0));
 
     // Output tensor auto inizialitation if not yet initialized
-    auto_init_if_empty(*output->info(), output_shape, 1, dt, fixed_point_position);
+    auto_init_if_empty(*output->info(), output_shape, 1, dt, fixed_point_position, input->info()->quantization_info());
 
     ARM_COMPUTE_ERROR_ON_MISMATCHING_DIMENSIONS(output->info()->tensor_shape(), output_shape);
     ARM_COMPUTE_ERROR_ON_MISMATCHING_DATA_TYPES(input, output);