diff options
author | Georgios Pinitas <georgios.pinitas@arm.com> | 2021-09-14 12:33:34 +0100 |
---|---|---|
committer | Freddie Liardet <frederick.liardet@arm.com> | 2021-10-07 10:59:05 +0000 |
commit | b6af482bc5d8e4f03f876e17909c561de198c4d3 (patch) | |
tree | f32c3a796cad01ffc27a4da2e8141cdf451ca453 /tests/framework | |
parent | 58e9e06102da7042bed34482ae89b3a6f8c77dca (diff) | |
download | ComputeLibrary-b6af482bc5d8e4f03f876e17909c561de198c4d3.tar.gz |
Per-operator build dependencies
Creates a list of operators their respective dependencies.
Alters the build system to walk-through them resolve the dependencies
and build Compute Library.
Removes the following unused kernels/functions:
-[NE|CL]MinMaxLayerKernel
-CLFillBorder
Resolves: COMPMID-4695,COMPMID-4696
Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com>
Change-Id: I35ebeef38dac25ec5459cfe9c5f7c9a708621124
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/c/VisualCompute/ComputeLibrary/+/357914
Tested-by: bsgcomp <bsgcomp@arm.com>
Reviewed-by: Michele DiGiorgio <michele.digiorgio@arm.com>
Comments-Addressed: bsgcomp <bsgcomp@arm.com>
Signed-off-by: Freddie Liardet <frederick.liardet@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/6295
Reviewed-by: Gunes Bayir <gunes.bayir@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Diffstat (limited to 'tests/framework')
-rw-r--r-- | tests/framework/instruments/OpenCLTimer.cpp | 45 | ||||
-rw-r--r-- | tests/framework/instruments/OpenCLTimer.h | 10 | ||||
-rw-r--r-- | tests/framework/instruments/SchedulerTimer.cpp | 34 | ||||
-rw-r--r-- | tests/framework/instruments/SchedulerTimer.h | 16 |
4 files changed, 66 insertions, 39 deletions
diff --git a/tests/framework/instruments/OpenCLTimer.cpp b/tests/framework/instruments/OpenCLTimer.cpp index 45eb4c5c60..e9f945bd95 100644 --- a/tests/framework/instruments/OpenCLTimer.cpp +++ b/tests/framework/instruments/OpenCLTimer.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2017-2019 Arm Limited. + * Copyright (c) 2017-2019, 2021 Arm Limited. * * SPDX-License-Identifier: MIT * @@ -54,7 +54,13 @@ std::string OpenCLClock<output_timestamps>::id() const template <bool output_timestamps> OpenCLClock<output_timestamps>::OpenCLClock(ScaleFactor scale_factor) - : _kernels(), _real_function(nullptr), _real_graph_function(nullptr), _prefix(), _timer_enabled(false) + : _kernels(), + _real_function(nullptr), +#ifdef ARM_COMPUTE_GRAPH_ENABLED + _real_graph_function(nullptr), +#endif /* ARM_COMPUTE_GRAPH_ENABLED */ + _prefix(), + _timer_enabled(false) { auto q = CLScheduler::get().queue(); cl_command_queue_properties props = q.getInfo<CL_QUEUE_PROPERTIES>(); @@ -91,19 +97,17 @@ void OpenCLClock<output_timestamps>::test_start() { // Start intercepting enqueues: ARM_COMPUTE_ERROR_ON(_real_function != nullptr); - ARM_COMPUTE_ERROR_ON(_real_graph_function != nullptr); - _real_function = CLSymbols::get().clEnqueueNDRangeKernel_ptr; - _real_graph_function = graph::TaskExecutor::get().execute_function; - auto interceptor = [this]( - cl_command_queue command_queue, - cl_kernel kernel, - cl_uint work_dim, - const size_t *gwo, - const size_t *gws, - const size_t *lws, - cl_uint num_events_in_wait_list, - const cl_event * event_wait_list, - cl_event * event) + _real_function = CLSymbols::get().clEnqueueNDRangeKernel_ptr; + auto interceptor = [this]( + cl_command_queue command_queue, + cl_kernel kernel, + cl_uint work_dim, + const size_t *gwo, + const size_t *gws, + const size_t *lws, + cl_uint num_events_in_wait_list, + const cl_event * event_wait_list, + cl_event * event) { if(this->_timer_enabled) { @@ -138,7 +142,11 @@ void OpenCLClock<output_timestamps>::test_start() return this->_real_function(command_queue, kernel, work_dim, gwo, gws, lws, num_events_in_wait_list, event_wait_list, event); } }; + CLSymbols::get().clEnqueueNDRangeKernel_ptr = interceptor; +#ifdef ARM_COMPUTE_GRAPH_ENABLED + ARM_COMPUTE_ERROR_ON(_real_graph_function != nullptr); + _real_graph_function = graph::TaskExecutor::get().execute_function; // Start intercepting tasks: auto task_interceptor = [this](graph::ExecutionTask & task) { @@ -153,9 +161,8 @@ void OpenCLClock<output_timestamps>::test_start() this->_real_graph_function(task); this->_prefix = ""; }; - - CLSymbols::get().clEnqueueNDRangeKernel_ptr = interceptor; graph::TaskExecutor::get().execute_function = task_interceptor; +#endif /* ARM_COMPUTE_GRAPH_ENABLED */ } template <bool output_timestamps> @@ -175,9 +182,11 @@ void OpenCLClock<output_timestamps>::test_stop() { // Restore real function CLSymbols::get().clEnqueueNDRangeKernel_ptr = _real_function; + _real_function = nullptr; +#ifdef ARM_COMPUTE_GRAPH_ENABLED graph::TaskExecutor::get().execute_function = _real_graph_function; _real_graph_function = nullptr; - _real_function = nullptr; +#endif /* ARM_COMPUTE_GRAPH_ENABLED */ } template <bool output_timestamps> diff --git a/tests/framework/instruments/OpenCLTimer.h b/tests/framework/instruments/OpenCLTimer.h index 9904035c20..1812272435 100644 --- a/tests/framework/instruments/OpenCLTimer.h +++ b/tests/framework/instruments/OpenCLTimer.h @@ -1,5 +1,5 @@ /* - * Copyright (c) 2017-2018 Arm Limited. + * Copyright (c) 2017-2018, 2021 Arm Limited. * * SPDX-License-Identifier: MIT * @@ -67,9 +67,11 @@ private: }; std::list<kernel_info> _kernels; std::function<decltype(clEnqueueNDRangeKernel)> _real_function; - std::function<decltype(graph::execute_task)> _real_graph_function; - std::string _prefix; - bool _timer_enabled; +#ifdef ARM_COMPUTE_GRAPH_ENABLED + std::function<decltype(graph::execute_task)> _real_graph_function; +#endif /* ARM_COMPUTE_GRAPH_ENABLED */ + std::string _prefix; + bool _timer_enabled; #endif /* ARM_COMPUTE_CL */ private: diff --git a/tests/framework/instruments/SchedulerTimer.cpp b/tests/framework/instruments/SchedulerTimer.cpp index 35f960d368..b753485351 100644 --- a/tests/framework/instruments/SchedulerTimer.cpp +++ b/tests/framework/instruments/SchedulerTimer.cpp @@ -129,16 +129,24 @@ protected: private: std::list<struct SchedulerClock<output_timestamps>::kernel_info> &_kernels; - std::map<std::string, SchedulerTimer::LayerData> &_layer_data_map; - IScheduler &_real_scheduler; - WallClock<output_timestamps> _timer; - std::string _prefix; + std::map<std::string, SchedulerTimer::LayerData> &_layer_data_map; + IScheduler &_real_scheduler; + WallClock<output_timestamps> _timer; + std::string _prefix; }; template <bool output_timestamps> SchedulerClock<output_timestamps>::SchedulerClock(ScaleFactor scale_factor) - : _kernels(), _layer_data_map(), _real_scheduler(nullptr), _real_scheduler_type(), _real_graph_function(nullptr), - _scale_factor(scale_factor), _interceptor(nullptr), _scheduler_users() + : _kernels(), + _layer_data_map(), + _real_scheduler(nullptr), + _real_scheduler_type(), +#ifdef ARM_COMPUTE_GRAPH_ENABLED + _real_graph_function(nullptr), +#endif /* ARM_COMPUTE_GRAPH_ENABLED */ + _scale_factor(scale_factor), + _interceptor(nullptr), + _scheduler_users() { if(instruments_info != nullptr) { @@ -149,6 +157,7 @@ SchedulerClock<output_timestamps>::SchedulerClock(ScaleFactor scale_factor) template <bool output_timestamps> void SchedulerClock<output_timestamps>::test_start() { +#ifdef ARM_COMPUTE_GRAPH_ENABLED // Start intercepting tasks: ARM_COMPUTE_ERROR_ON(_real_graph_function != nullptr); _real_graph_function = graph::TaskExecutor::get().execute_function; @@ -182,6 +191,7 @@ void SchedulerClock<output_timestamps>::test_start() scheduler->set_prefix(""); } }; +#endif /* ARM_COMPUTE_GRAPH_ENABLED */ ARM_COMPUTE_ERROR_ON(_real_scheduler != nullptr); _real_scheduler_type = Scheduler::get_type(); @@ -191,7 +201,9 @@ void SchedulerClock<output_timestamps>::test_start() _real_scheduler = &Scheduler::get(); _interceptor = std::make_shared<Interceptor<output_timestamps>>(_kernels, _layer_data_map, *_real_scheduler, _scale_factor); Scheduler::set(std::static_pointer_cast<IScheduler>(_interceptor)); +#ifdef ARM_COMPUTE_GRAPH_ENABLED graph::TaskExecutor::get().execute_function = task_interceptor; +#endif /* ARM_COMPUTE_GRAPH_ENABLED */ // Create an interceptor for each scheduler // TODO(COMPID-2638) : Allow multiple schedulers, now it assumes the same scheduler is used. @@ -217,10 +229,12 @@ void SchedulerClock<output_timestamps>::test_stop() { // Restore real scheduler Scheduler::set(_real_scheduler_type); - _real_scheduler = nullptr; - _interceptor = nullptr; + _real_scheduler = nullptr; + _interceptor = nullptr; +#ifdef ARM_COMPUTE_GRAPH_ENABLED graph::TaskExecutor::get().execute_function = _real_graph_function; _real_graph_function = nullptr; +#endif /* ARM_COMPUTE_GRAPH_ENABLED */ // Restore schedulers std::for_each(std::begin(_scheduler_users), std::end(_scheduler_users), @@ -270,9 +284,9 @@ Instrument::MeasurementsMap SchedulerClock<output_timestamps>::measurements() co } template <bool output_timestamps> -std::string SchedulerClock<output_timestamps>::instrument_header() const +std::string SchedulerClock<output_timestamps>::instrument_header() const { - std::string output{""}; + std::string output{ "" }; output += R"("layer_data" : {)"; for(auto i_it = _layer_data_map.cbegin(), i_end = _layer_data_map.cend(); i_it != i_end; ++i_it) { diff --git a/tests/framework/instruments/SchedulerTimer.h b/tests/framework/instruments/SchedulerTimer.h index 9cc0381a9a..c437f2717c 100644 --- a/tests/framework/instruments/SchedulerTimer.h +++ b/tests/framework/instruments/SchedulerTimer.h @@ -97,14 +97,16 @@ public: }; private: - std::list<kernel_info> _kernels; - std::map<std::string, LayerData> _layer_data_map; - IScheduler *_real_scheduler; - Scheduler::Type _real_scheduler_type; + std::list<kernel_info> _kernels; + std::map<std::string, LayerData> _layer_data_map; + IScheduler *_real_scheduler; + Scheduler::Type _real_scheduler_type; +#ifdef ARM_COMPUTE_GRAPH_ENABLED std::function<decltype(graph::execute_task)> _real_graph_function; - ScaleFactor _scale_factor; - std::shared_ptr<IScheduler> _interceptor; - std::vector<ISchedulerUser *> _scheduler_users; +#endif /* ARM_COMPUTE_GRAPH_ENABLED */ + ScaleFactor _scale_factor; + std::shared_ptr<IScheduler> _interceptor; + std::vector<ISchedulerUser *> _scheduler_users; }; using SchedulerTimer = SchedulerClock<false>; |