diff options
| author | steniu01 <steven.niu@arm.com> | 2017-09-29 14:55:00 +0100 |
|---|---|---|
| committer | Anthony Barbier <anthony.barbier@arm.com> | 2018-11-02 16:35:24 +0000 |
| commit | 7ce53c620b50c718bac62017d28072cf61457233 (patch) | |
| tree | 80a4336ce63f7794986777e8434e9e440478f70d /src | |
| parent | b482ce1d601a777250f28ed118ac250943aca4eb (diff) | |
| download | ComputeLibrary-7ce53c620b50c718bac62017d28072cf61457233.tar.gz | |
COMPMID-546 Add auto config to depth concatenate
Change-Id: I7798a56677d541338a73e3888ed0a2cfe0375794
Reviewed-on: http://mpd-gerrit.cambridge.arm.com/89726
Tested-by: Kaizen <jeremy.johnson+kaizengerrit@arm.com>
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
Diffstat (limited to 'src')
| -rw-r--r-- | src/runtime/CL/functions/CLDepthConcatenate.cpp | 6 | ||||
| -rw-r--r-- | src/runtime/NEON/functions/NEDepthConcatenate.cpp | 6 |
2 files changed, 12 insertions, 0 deletions
diff --git a/src/runtime/CL/functions/CLDepthConcatenate.cpp b/src/runtime/CL/functions/CLDepthConcatenate.cpp index f42627f34c..89e44ca98e 100644 --- a/src/runtime/CL/functions/CLDepthConcatenate.cpp +++ b/src/runtime/CL/functions/CLDepthConcatenate.cpp @@ -25,6 +25,7 @@ #include "arm_compute/core/CL/ICLTensor.h" #include "arm_compute/core/Error.h" +#include "arm_compute/core/Helpers.h" #include "arm_compute/core/PixelValue.h" #include "arm_compute/core/Types.h" #include "arm_compute/runtime/CL/CLScheduler.h" @@ -51,6 +52,11 @@ void CLDepthConcatenate::configure(std::vector<ICLTensor *> inputs_vector, ICLTe _concat_kernels_vector = arm_compute::support::cpp14::make_unique<CLDepthConcatenateKernel[]>(_num_inputs); _border_handlers_vector = arm_compute::support::cpp14::make_unique<CLFillBorderKernel[]>(_num_inputs); + TensorShape output_shape = calculate_depth_concatenate_shape(inputs_vector); + + // Output auto inizialitation if not yet initialized + auto_init_if_empty(*output->info(), output_shape, 1, inputs_vector[0]->info()->data_type(), inputs_vector[0]->info()->fixed_point_position()); + for(unsigned int i = 0; i < _num_inputs; i++) { _concat_kernels_vector[i].configure(inputs_vector.at(i), depth_offset, output); diff --git a/src/runtime/NEON/functions/NEDepthConcatenate.cpp b/src/runtime/NEON/functions/NEDepthConcatenate.cpp index 90eee4f45f..f8ad2abe61 100644 --- a/src/runtime/NEON/functions/NEDepthConcatenate.cpp +++ b/src/runtime/NEON/functions/NEDepthConcatenate.cpp @@ -24,6 +24,7 @@ #include "arm_compute/runtime/NEON/functions/NEDepthConcatenate.h" #include "arm_compute/core/Error.h" +#include "arm_compute/core/Helpers.h" #include "arm_compute/core/ITensor.h" #include "arm_compute/core/PixelValue.h" #include "arm_compute/core/Types.h" @@ -48,6 +49,11 @@ void NEDepthConcatenate::configure(std::vector<ITensor *> inputs_vector, ITensor _concat_kernels_vector = arm_compute::support::cpp14::make_unique<NEDepthConcatenateKernel[]>(_num_inputs); _border_handlers_vector = arm_compute::support::cpp14::make_unique<NEFillBorderKernel[]>(_num_inputs); + TensorShape output_shape = calculate_depth_concatenate_shape(inputs_vector); + + // Output auto inizialitation if not yet initialized + auto_init_if_empty(*output->info(), output_shape, 1, inputs_vector[0]->info()->data_type(), inputs_vector[0]->info()->fixed_point_position()); + unsigned int depth_offset = 0; for(unsigned int i = 0; i < _num_inputs; ++i) { |