From 173ba9bbb19ea83f951318d9989e440768b4de8f Mon Sep 17 00:00:00 2001 From: Michalis Spyrou Date: Tue, 23 Jun 2020 17:25:43 +0100 Subject: COMPMID-3373: Async support to NEArithmetic* kernels/functions (Pt. 1) Added support on NEArithmeticAddition and NEArithmeticSubtraction Signed-off-by: Michalis Spyrou Change-Id: Ifa805f8455ef6eff1ee627752dc1c7fe9740ec47 Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/3451 Tested-by: Arm Jenkins Reviewed-by: Georgios Pinitas --- src/runtime/NEON/functions/NERNNLayer.cpp | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) (limited to 'src/runtime/NEON/functions/NERNNLayer.cpp') diff --git a/src/runtime/NEON/functions/NERNNLayer.cpp b/src/runtime/NEON/functions/NERNNLayer.cpp index 4a15777be9..19b84e7fb8 100644 --- a/src/runtime/NEON/functions/NERNNLayer.cpp +++ b/src/runtime/NEON/functions/NERNNLayer.cpp @@ -34,7 +34,7 @@ namespace arm_compute { NERNNLayer::NERNNLayer(std::shared_ptr memory_manager) - : _memory_group(std::move(memory_manager)), _gemm_state_f(), _add_kernel(), _activation(), _fully_connected(memory_manager), _copy_kernel(), _fully_connected_out(), _gemm_output(), _add_output(), + : _memory_group(std::move(memory_manager)), _gemm_state_f(), _add_f(), _activation(), _fully_connected(memory_manager), _copy_kernel(), _fully_connected_out(), _gemm_output(), _add_output(), _is_prepared(false) { } @@ -59,7 +59,7 @@ Status NERNNLayer::validate(const ITensorInfo *input, const ITensorInfo *weights auto shape_info = TensorInfo(misc::shape_calculator::compute_rnn_shape(recurrent_weights, hidden_state->dimension(idx_height)), 1, input->data_type()); ARM_COMPUTE_RETURN_ON_ERROR(NEFullyConnectedLayer::validate(input, weights, bias, &shape_info)); - ARM_COMPUTE_RETURN_ON_ERROR(NEArithmeticAdditionKernel::validate(&shape_info, &shape_info, &shape_info, ConvertPolicy::SATURATE)); + ARM_COMPUTE_RETURN_ON_ERROR(NEArithmeticAddition::validate(&shape_info, &shape_info, &shape_info, ConvertPolicy::SATURATE)); ARM_COMPUTE_RETURN_ON_ERROR(NEActivationLayer::validate(&shape_info, &shape_info, info)); return Status{}; @@ -90,7 +90,7 @@ void NERNNLayer::configure(const ITensor *input, const ITensor *weights, const I _add_output.allocator()->init(TensorInfo(shape, 1, input->info()->data_type())); _memory_group.manage(&_add_output); - _add_kernel.configure(&_fully_connected_out, &_gemm_output, &_add_output, ConvertPolicy::SATURATE); + _add_f.configure(&_fully_connected_out, &_gemm_output, &_add_output, ConvertPolicy::SATURATE); _fully_connected_out.allocator()->allocate(); _gemm_output.allocator()->allocate(); @@ -111,7 +111,7 @@ void NERNNLayer::run() _gemm_state_f.run(); - NEScheduler::get().schedule(&_add_kernel, Window::DimY); + _add_f.run(); _activation.run(); // copy hidden out to output -- cgit v1.2.1