Rename functions/classes for elementwise unary operations

* Create CpuElementwiseUnary operator
* Rename kernel classes
* Make the kernels stateless

Partially implements: COMPMID-4003

Change-Id: Ie0440cd01d4924847d6991b4df7ccaf311439297
Signed-off-by: Sang-Hoon Park <sang-hoon.park@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/4912
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com>
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>

5 files changed, 57 insertions, 58 deletions

diff --git a/src/core/NEON/NEKernels.h b/src/core/NEON/NEKernels.h
index c009a6d3af..87eec38605 100644
--- a/src/core/NEON/NEKernels.h
+++ b/src/core/NEON/NEKernels.h
@@ -54,7 +54,6 @@
 #include "src/core/NEON/kernels/NEDilateKernel.h"
 #include "src/core/NEON/kernels/NEDirectConvolutionLayerKernel.h"
 #include "src/core/NEON/kernels/NEDirectConvolutionLayerOutputStageKernel.h"
-#include "src/core/NEON/kernels/NEElementwiseUnaryKernel.h"
 #include "src/core/NEON/kernels/NEErodeKernel.h"
 #include "src/core/NEON/kernels/NEFFTDigitReverseKernel.h"
 #include "src/core/NEON/kernels/NEFFTRadixStageKernel.h"
diff --git a/src/core/NEON/kernels/NEElementwiseUnaryKernel.cpp b/src/core/cpu/kernels/CpuElementwiseUnaryKernel.cpp
index fdd2aabf60..d2681bb060 100644
--- a/src/core/NEON/kernels/NEElementwiseUnaryKernel.cpp
+++ b/src/core/cpu/kernels/CpuElementwiseUnaryKernel.cpp
@@ -21,31 +21,35 @@
  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
  */
-#include "src/core/NEON/kernels/NEElementwiseUnaryKernel.h"
+#include "src/core/cpu/kernels/CpuElementwiseUnaryKernel.h"
 
 #include "arm_compute/core/Error.h"
 #include "arm_compute/core/Helpers.h"
 #include "arm_compute/core/ITensor.h"
 #include "arm_compute/core/Validate.h"
 #include "src/core/CPP/Validate.h"
-#include "src/core/NEON/kernels/elementwise/impl/elementwise_unary_list.h"
-#include "src/core/SVE/kernels/elementwise/impl/elementwise_unary_list.h"
 #include "src/core/common/Registrars.h"
+#include "src/core/cpu/kernels/elementwise/neon/elementwise_unary_list.h"
+#include "src/core/cpu/kernels/elementwise/sve/elementwise_unary_list.h"
 #include "src/core/helpers/AutoConfiguration.h"
 #include "src/core/helpers/WindowHelpers.h"
 #include "support/ToolchainSupport.h"
 
 namespace arm_compute
 {
+namespace cpu
+{
+namespace kernels
+{
 namespace
 {
 using ElementwiseUnarySelector = std::add_pointer<bool(DataType)>::type;
 
 struct ElementwiseUnaryKernel
 {
-    const char                                          *name;
-    const ElementwiseUnarySelector                       is_selected;
-    NEElementwiseUnaryKernel::ElementwiseUnaryUkernelPtr ukernel;
+    const char                                           *name;
+    const ElementwiseUnarySelector                        is_selected;
+    CpuElementwiseUnaryKernel::ElementwiseUnaryUkernelPtr ukernel;
 };
 
 static const ElementwiseUnaryKernel available_kernels[] =
@@ -99,41 +103,35 @@ const ElementwiseUnaryKernel *get_implementation(DataType dt)
 }
 } // namespace
 
-NEElementwiseUnaryKernel::NEElementwiseUnaryKernel()
-    : _func(nullptr), _input(nullptr), _output(nullptr), _op()
+CpuElementwiseUnaryKernel::CpuElementwiseUnaryKernel()
+    : _op()
 {
 }
 
-void NEElementwiseUnaryKernel::configure(ElementWiseUnary op, const ITensor *input, ITensor *output)
+void CpuElementwiseUnaryKernel::configure(ElementWiseUnary op, const ITensorInfo &input, ITensorInfo &output)
 {
-    ARM_COMPUTE_ERROR_THROW_ON(validate(op, input->info(), output->info()));
-    ARM_COMPUTE_ERROR_ON_NULLPTR(input, output);
+    ARM_COMPUTE_ERROR_THROW_ON(validate(op, input, output));
 
     // Configure kernel window
-    const std::pair<TensorShape, ValidRegion> broadcast_pair = ITensorInfo::broadcast_shape_and_valid_region(*input->info());
+    const std::pair<TensorShape, ValidRegion> broadcast_pair = ITensorInfo::broadcast_shape_and_valid_region(input);
     const TensorShape &out_shape    = broadcast_pair.first;
     const ValidRegion &valid_region = broadcast_pair.second;
 
     // Auto initialize output if not initialized
-    auto_init_if_empty(*output->info(), out_shape, 1, input->info()->data_type());
+    auto_init_if_empty(output, out_shape, 1, input.data_type());
 
     Window win = calculate_max_window(valid_region);
 
-    _input  = input;
-    _output = output;
-    _op     = op;
+    _op = op;
 
-    INEKernel::configure(win);
-
-    _func = get_implementation(input->info()->data_type())->ukernel;
+    ICpuKernel::configure(win);
 }
 
-Status NEElementwiseUnaryKernel::validate(ElementWiseUnary op, const ITensorInfo *input, const ITensorInfo *output)
+Status CpuElementwiseUnaryKernel::validate(ElementWiseUnary op, const ITensorInfo &input, const ITensorInfo &output)
 {
-    ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input, output);
-    ARM_COMPUTE_RETURN_ERROR_ON_CPU_F16_UNSUPPORTED(input);
+    ARM_COMPUTE_RETURN_ERROR_ON_CPU_F16_UNSUPPORTED(&input);
 
-    const auto *uk = get_implementation(input->data_type());
+    const auto *uk = get_implementation(input.data_type());
     ARM_COMPUTE_RETURN_ERROR_ON(uk == nullptr || uk->ukernel == nullptr);
 
     switch(op)
@@ -143,30 +141,36 @@ Status NEElementwiseUnaryKernel::validate(ElementWiseUnary op, const ITensorInfo
         case ElementWiseUnary::LOG:
         case ElementWiseUnary::ROUND:
         case ElementWiseUnary::SIN:
-            ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::F16, DataType::F32);
+            ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(&input, 1, DataType::F16, DataType::F32);
             break;
         case ElementWiseUnary::NEG:
         case ElementWiseUnary::ABS:
-            ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::F16, DataType::F32, DataType::S32);
+            ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(&input, 1, DataType::F16, DataType::F32, DataType::S32);
             break;
         default:
             ARM_COMPUTE_ERROR("ElementWiseUnary operation not supported");
     }
     // Validate in case of configured output
-    if(output->total_size() > 0)
+    if(output.total_size() > 0)
     {
-        ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input, output);
+        ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(&input, &output);
     }
 
     return Status{};
 }
 
-void NEElementwiseUnaryKernel::run(const Window &window, const ThreadInfo &info)
+void CpuElementwiseUnaryKernel::run_op(ITensorPack &tensors, const Window &window, const ThreadInfo &info)
 {
     ARM_COMPUTE_UNUSED(info);
     ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
-    ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(INEKernel::window(), window);
-    ARM_COMPUTE_ERROR_ON(_func == nullptr);
-    (*_func)(_input, _output, window, _op);
+    ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(ICpuKernel::window(), window);
+
+    auto src  = tensors.get_const_tensor(TensorType::ACL_SRC);
+    auto dst  = tensors.get_tensor(TensorType::ACL_DST);
+    auto func = get_implementation(src->info()->data_type())->ukernel;
+    ARM_COMPUTE_ERROR_ON(func == nullptr);
+    func(src, dst, window, _op);
 }
+} // namespace kernels
+} // namespace cpu
 } // namespace arm_compute
diff --git a/src/core/NEON/kernels/NEElementwiseUnaryKernel.h b/src/core/cpu/kernels/CpuElementwiseUnaryKernel.h
index b248e821c3..193f6f1e4f 100644
--- a/src/core/NEON/kernels/NEElementwiseUnaryKernel.h
+++ b/src/core/cpu/kernels/CpuElementwiseUnaryKernel.h
@@ -21,51 +21,48 @@
  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
  */
-#ifndef ARM_COMPUTE_NEELEMENTWISEUNARYKERNEL_H
-#define ARM_COMPUTE_NEELEMENTWISEUNARYKERNEL_H
+#ifndef ARM_COMPUTE_CPU_ELEMENTWISE_UNARY_KERNEL_H
+#define ARM_COMPUTE_CPU_ELEMENTWISE_UNARY_KERNEL_H
 
 #include "arm_compute/core/Types.h"
-#include "src/core/NEON/INEKernel.h"
+#include "src/core/common/Macros.h"
+#include "src/core/cpu/ICpuKernel.h"
 
 namespace arm_compute
 {
 class ITensor;
-
+namespace cpu
+{
+namespace kernels
+{
 /** Interface for an element-wise unary operation kernel
  *
  * Element-wise operation is computed by:
  * @f[ output(x) = OP(input(x))@f]
  *
  */
-class NEElementwiseUnaryKernel : public INEKernel
+class CpuElementwiseUnaryKernel : public ICpuKernel
 {
 public:
     const char *name() const override
     {
-        return "NEElementwiseUnaryKernel";
+        return "CpuElementwiseUnaryKernel";
     }
     /** Default constructor */
-    NEElementwiseUnaryKernel();
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    NEElementwiseUnaryKernel(const NEElementwiseUnaryKernel &) = delete;
-    /** Prevent instances of this class from being copied (As this class contains pointers) */
-    NEElementwiseUnaryKernel &operator=(const NEElementwiseUnaryKernel &) = delete;
-    /** Allow instances of this class to be moved */
-    NEElementwiseUnaryKernel(NEElementwiseUnaryKernel &&) = default;
-    /** Allow instances of this class to be moved */
-    NEElementwiseUnaryKernel &operator=(NEElementwiseUnaryKernel &&) = default;
+    CpuElementwiseUnaryKernel();
     /** Default destructor */
-    ~NEElementwiseUnaryKernel() = default;
+    ~CpuElementwiseUnaryKernel() = default;
+    ARM_COMPUTE_DISALLOW_COPY_ALLOW_MOVE(CpuElementwiseUnaryKernel);
 
-    /** Function to configure the @ref NEElementwiseUnaryKernel
+    /** Function to configure the @ref CpuElementwiseUnaryKernel
      *
      * @param[in]  op     Arithmetic operation to be executed.
      * @param[in]  input  First tensor input. Data types supported: F16/F32, F16/F32/S32 for NEG/ABS operations.
      * @param[out] output Output tensor. Data types supported: Same as @p input.
      */
-    void configure(ElementWiseUnary op, const ITensor *input, ITensor *output);
+    void configure(ElementWiseUnary op, const ITensorInfo &input, ITensorInfo &output);
 
-    /** Static function to check if given info will lead to a valid configuration of @ref NEElementwiseUnaryKernel
+    /** Static function to check if given info will lead to a valid configuration of @ref CpuElementwiseUnaryKernel
      *
      * @param[in] op     Arithmetic operation to be executed.
      * @param[in] input  First tensor input info. Data types supported: F16/F32, F16/F32/S32 for NEG/ABS operations.
@@ -73,10 +70,10 @@ public:
      *
      * @return a Status
      */
-    static Status validate(ElementWiseUnary op, const ITensorInfo *input, const ITensorInfo *output);
+    static Status validate(ElementWiseUnary op, const ITensorInfo &input, const ITensorInfo &output);
 
     // Inherited methods overridden:
-    void run(const Window &window, const ThreadInfo &info) override;
+    void run_op(ITensorPack &tensors, const Window &window, const ThreadInfo &info) override;
 
     /** Common signature for all the specialised elementwise unary micro-kernels
      *
@@ -85,10 +82,9 @@ public:
     using ElementwiseUnaryUkernelPtr = std::add_pointer<void(const ITensor *, ITensor *, const Window &, ElementWiseUnary)>::type;
 
 private:
-    ElementwiseUnaryUkernelPtr _func;
-    const ITensor             *_input;
-    ITensor                   *_output;
-    ElementWiseUnary           _op;
+    ElementWiseUnary _op;
 };
+} // namespace kernels
+} // namespace cpu
 } // namespace arm_compute
-#endif /* ARM_COMPUTE_NEELEMENTWISEUNARYKERNEL_H */
+#endif /* ARM_COMPUTE_CPU_ELEMENTWISE_UNARY_KERNEL_H */
diff --git a/src/core/NEON/kernels/elementwise/impl/elementwise_unary_list.h b/src/core/cpu/kernels/elementwise/neon/elementwise_unary_list.h
index 307e95fae9..307e95fae9 100644
--- a/src/core/NEON/kernels/elementwise/impl/elementwise_unary_list.h
+++ b/src/core/cpu/kernels/elementwise/neon/elementwise_unary_list.h
diff --git a/src/core/SVE/kernels/elementwise/impl/elementwise_unary_list.h b/src/core/cpu/kernels/elementwise/sve/elementwise_unary_list.h
index 23502c71e5..23502c71e5 100644
--- a/src/core/SVE/kernels/elementwise/impl/elementwise_unary_list.h
+++ b/src/core/cpu/kernels/elementwise/sve/elementwise_unary_list.h