Port ClGemmLowp to new API

Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com>
Change-Id: Icef9ca564e61a00a3f4fd4ae7f465a711ff8c51d
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5939
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>

1 files changed, 4 insertions, 56 deletions

diff --git a/arm_compute/runtime/CL/functions/CLGEMMLowpMatrixMultiplyCore.h b/arm_compute/runtime/CL/functions/CLGEMMLowpMatrixMultiplyCore.h
index e62db8e644..a8ee9e5b56 100644
--- a/arm_compute/runtime/CL/functions/CLGEMMLowpMatrixMultiplyCore.h
+++ b/arm_compute/runtime/CL/functions/CLGEMMLowpMatrixMultiplyCore.h
@@ -28,33 +28,14 @@
 #include "arm_compute/runtime/IFunction.h"
 #include "arm_compute/runtime/MemoryGroup.h"
 
+#include <memory>
+
 namespace arm_compute
 {
 class CLCompileContext;
 class IMemoryManager;
 class ICLTensor;
 class ITensorInfo;
-namespace opencl
-{
-namespace kernels
-{
-class ClGemmReshapeRhsMatrixKernel;
-class ClGemmLowpMatrixMultiplyNativeKernel;
-class ClGemmLowpMatrixMultiplyReshapedOnlyRhsKernel;
-class ClGemmLowpOffsetContributionKernel;
-class ClGemmLowpOffsetContributionOutputStageKernel;
-class ClGemmLowpMatrixAReductionKernel;
-class ClGemmLowpMatrixBReductionKernel;
-} // namespace kernels
-} // namespace opencl
-
-namespace opencl
-{
-namespace kernels
-{
-class ClCastKernel;
-} // namespace kernels
-} // namespace opencl
 
 /** Basic function to execute GEMMLowpMatrixMultiplyCore on OpenCL. */
 class CLGEMMLowpMatrixMultiplyCore : public IFunction
@@ -147,41 +128,8 @@ public:
     void prepare() override;
 
 private:
-    MemoryGroup _memory_group;
-
-    // Kernels used
-    std::unique_ptr<opencl::kernels::ClCastKernel>                                  _weights_to_qasymm8;
-    std::unique_ptr<opencl::kernels::ClGemmLowpMatrixMultiplyNativeKernel>          _mm_native_kernel;
-    std::unique_ptr<opencl::kernels::ClGemmLowpMatrixMultiplyReshapedOnlyRhsKernel> _mm_reshaped_only_rhs_kernel;
-    std::unique_ptr<opencl::kernels::ClGemmReshapeRhsMatrixKernel>                  _mtx_b_reshape_kernel;
-    std::unique_ptr<opencl::kernels::ClGemmLowpMatrixAReductionKernel>              _mtx_a_reduction_kernel;
-    std::unique_ptr<opencl::kernels::ClGemmLowpMatrixBReductionKernel>              _mtx_b_reduction_kernel;
-    std::unique_ptr<opencl::kernels::ClGemmLowpOffsetContributionKernel>            _offset_contribution_kernel;
-    std::unique_ptr<opencl::kernels::ClGemmLowpOffsetContributionOutputStageKernel> _offset_contribution_output_stage_kernel;
-
-    // Temporary tensors
-    CLTensor _qasymm8_weights;
-    CLTensor _vector_sum_col;
-    CLTensor _vector_sum_row;
-    CLTensor _tmp_b;
-    CLTensor _mm_result_s32;
-    CLTensor _gemm_output_stage_multipliers;
-    CLTensor _gemm_output_stage_shifts;
-
-    // Tensor pointers
-    const ICLTensor *_matrix_a;
-    const ICLTensor *_original_b;
-    const ICLTensor *_c;
-    ICLTensor       *_output;
-
-    int32_t _a_offset;
-    int32_t _b_offset;
-    bool    _is_gemm_reshaped;
-    bool    _reshape_b_only_on_first_run;
-    bool    _is_prepared;
-    bool    _run_output_stage;
-    bool    _convert_to_qasymm8;
-    bool    _run_offset_contribution;
+    struct Impl;
+    std::unique_ptr<Impl> _impl;
 };
 } // namespace arm_compute
 #endif /*ARM_COMPUTE_CLGEMMLOWPMATRIXMULTIPLYCORE_H */
\ No newline at end of file