COMPMID-1364: Add support for NHWC in NEDepthConcatenateLayer
Change-Id: I4f8e46d1c79afa9284f2c6dc00383c453a8e7bd5
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/140165
Reviewed-by: Giorgio Arena <giorgio.arena@arm.com>
Reviewed-by: Pablo Tello <pablo.tello@arm.com>
Tested-by: Jenkins <bsgcomp@arm.com>
diff --git a/src/runtime/NEON/functions/NEDepthConcatenateLayer.cpp b/src/runtime/NEON/functions/NEDepthConcatenateLayer.cpp
index cb01576..49db855 100644
--- a/src/runtime/NEON/functions/NEDepthConcatenateLayer.cpp
+++ b/src/runtime/NEON/functions/NEDepthConcatenateLayer.cpp
@@ -27,6 +27,7 @@
#include "arm_compute/core/Helpers.h"
#include "arm_compute/core/ITensor.h"
#include "arm_compute/core/PixelValue.h"
+#include "arm_compute/core/TensorInfo.h"
#include "arm_compute/core/Types.h"
#include "arm_compute/core/utils/misc/ShapeCalculator.h"
#include "arm_compute/runtime/NEON/NEScheduler.h"
@@ -42,10 +43,8 @@
{
}
-void NEDepthConcatenateLayer::configure(std::vector<ITensor *> inputs_vector, ITensor *output) // NOLINT
+void NEDepthConcatenateLayer::configure(const std::vector<ITensor *> &inputs_vector, ITensor *output) // NOLINT
{
- ARM_COMPUTE_ERROR_ON(inputs_vector.size() < 2);
-
_num_inputs = inputs_vector.size();
_concat_kernels_vector = arm_compute::support::cpp14::make_unique<NEDepthConcatenateLayerKernel[]>(_num_inputs);
_border_handlers_vector = arm_compute::support::cpp14::make_unique<NEFillBorderKernel[]>(_num_inputs);
@@ -59,6 +58,7 @@
// Output auto inizialitation if not yet initialized
auto_init_if_empty(*output->info(), output_shape, 1, inputs_vector[0]->info()->data_type());
+ ARM_COMPUTE_ERROR_THROW_ON(NEDepthConcatenateLayer::validate(inputs_vector_info, output->info()));
unsigned int depth_offset = 0;
for(unsigned int i = 0; i < _num_inputs; ++i)
@@ -73,6 +73,27 @@
output->info()->set_valid_region(ValidRegion(Coordinates(), output_shape));
}
+Status NEDepthConcatenateLayer::validate(const std::vector<ITensorInfo *> &inputs_vector, const ITensorInfo *output)
+{
+ ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(output);
+ ARM_COMPUTE_RETURN_ERROR_ON(inputs_vector.size() < 2);
+
+ // Output auto inizialitation if not yet initialized
+ TensorInfo tmp_output_info = *output->clone();
+ TensorShape output_shape = arm_compute::misc::shape_calculator::calculate_depth_concatenate_shape(inputs_vector);
+ auto_init_if_empty(tmp_output_info, output_shape, 1, inputs_vector[0]->data_type());
+
+ unsigned int depth_offset = 0;
+ for(const auto &input : inputs_vector)
+ {
+ ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input);
+ ARM_COMPUTE_RETURN_ON_ERROR(NEDepthConcatenateLayerKernel::validate(input, depth_offset, &tmp_output_info));
+ depth_offset += input->dimension(2);
+ }
+
+ return Status{};
+}
+
void NEDepthConcatenateLayer::run()
{
for(unsigned i = 0; i < _num_inputs; ++i)