From 4c6bd514a8d424a29b776754f1b3426fa3a8c339 Mon Sep 17 00:00:00 2001 From: Manuel Bottini Date: Wed, 8 Apr 2020 10:15:51 +0100 Subject: COMPMID-3280: Make all ML primitives for CL use the new interface - Part 1 - Only CLKernels have been updated Change-Id: Ife55b847c2e39e712a186eb6ca452503d5b66937 Signed-off-by: Manuel Bottini Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/3001 Reviewed-by: Michele Di Giorgio Reviewed-by: Michalis Spyrou Comments-Addressed: Arm Jenkins Tested-by: Arm Jenkins --- src/core/CL/kernels/CLBoundingBoxTransformKernel.cpp | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-) (limited to 'src/core/CL/kernels/CLBoundingBoxTransformKernel.cpp') diff --git a/src/core/CL/kernels/CLBoundingBoxTransformKernel.cpp b/src/core/CL/kernels/CLBoundingBoxTransformKernel.cpp index bca23c7259..5ed5523632 100644 --- a/src/core/CL/kernels/CLBoundingBoxTransformKernel.cpp +++ b/src/core/CL/kernels/CLBoundingBoxTransformKernel.cpp @@ -89,6 +89,11 @@ CLBoundingBoxTransformKernel::CLBoundingBoxTransformKernel() } void CLBoundingBoxTransformKernel::configure(const ICLTensor *boxes, ICLTensor *pred_boxes, const ICLTensor *deltas, const BoundingBoxTransformInfo &info) +{ + configure(CLKernelLibrary::get().get_compile_context(), boxes, pred_boxes, deltas, info); +} + +void CLBoundingBoxTransformKernel::configure(CLCompileContext &compile_context, const ICLTensor *boxes, ICLTensor *pred_boxes, const ICLTensor *deltas, const BoundingBoxTransformInfo &info) { ARM_COMPUTE_ERROR_ON_NULLPTR(boxes, pred_boxes, deltas); auto_init_if_empty(*pred_boxes->info(), deltas->info()->clone()->set_data_type(boxes->info()->data_type()).set_quantization_info(boxes->info()->quantization_info())); @@ -137,7 +142,7 @@ void CLBoundingBoxTransformKernel::configure(const ICLTensor *boxes, ICLTensor * // Create kernel const std::string kernel_name = (is_quantized) ? "bounding_box_transform_quantized" : "bounding_box_transform"; - _kernel = static_cast(CLKernelLibrary::get().create_kernel(kernel_name, build_opts.options())); + _kernel = create_kernel(compile_context, kernel_name, build_opts.options()); // Since the number of columns is a multiple of 4 by definition, we don't need to pad the tensor const unsigned int num_elems_processed_per_iteration = 4; -- cgit v1.2.1