Remove dynamic fusion prototype with tests and examples

Public headers of the new experimental dynamic fusion can be found in arm_compute/dynamic_fusion/
New examples on how to use the interface can be found in tests/validation/dynamic_fusion/gpu/Integration.cpp

Resolves COMPMID-5683

Change-Id: I7ccb902a227fb487562df15fc3c30118d1d95bbd
Signed-off-by: SiCong Li <sicong.li@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/8671
Reviewed-by: Jakub Sujak <jakub.sujak@arm.com>
Reviewed-by: Gunes Bayir <gunes.bayir@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>

2 files changed, 0 insertions, 73 deletions

diff --git a/src/runtime/CL/CLScheduler.cpp b/src/runtime/CL/CLScheduler.cpp
index 8d30c05361..49fb724cdb 100644
--- a/src/runtime/CL/CLScheduler.cpp
+++ b/src/runtime/CL/CLScheduler.cpp
@@ -27,10 +27,6 @@
 #include "arm_compute/runtime/CL/CLTuner.h"
 #include "src/core/CL/ICLKernel.h"
 
-#if defined(ENABLE_EXPERIMENTAL_DYNAMIC_FUSION)
-#include "src/gpu/cl/kernels/experimental/dynamic_fusion/ClCompositeKernel.h"
-#endif // defined(ENABLE_EXPERIMENTAL_DYNAMIC_FUSION)
-
 namespace arm_compute
 {
 cl::Context &CLScheduler::context()
@@ -190,34 +186,6 @@ void CLScheduler::enqueue_common(ICLKernel &kernel, ITensorPack &tensors, bool f
     flush_queue(flush);
 }
 
-#if defined(ENABLE_EXPERIMENTAL_DYNAMIC_FUSION)
-
-void CLScheduler::enqueue_common(ICLKernel &kernel, ITensorPack &tensors, const experimental::dynamic_fusion::ClExecutionDescriptor &exec_desc, bool flush)
-{
-    ARM_COMPUTE_ERROR_ON_MSG(!_is_initialised,
-                             "The CLScheduler is not initialised yet! Please call the CLScheduler::get().default_init(), \
-                             or CLScheduler::get()::init() and CLKernelLibrary::get()::init() function before running functions!");
-
-    // ClCompositeKernel is stateless thus alway requires memory injection
-
-    // Tune the kernel if the CLTuner has been provided
-    if(_cl_tuner != nullptr)
-    {
-        _cl_tuner->tune_kernel_dynamic(kernel, tensors, exec_desc);
-    }
-
-    // Run kernel
-    kernel.run_composite_op(tensors, kernel.window(), _queue, exec_desc);
-    if(_job_chaining_enabled)
-    {
-        ++_job_chaining_count;
-    }
-
-    flush_queue(flush);
-}
-
-#endif // defined(ENABLE_EXPERIMENTAL_DYNAMIC_FUSION)
-
 void CLScheduler::flush_queue(bool flush)
 {
     if(_job_chaining_enabled)
@@ -245,15 +213,6 @@ void CLScheduler::enqueue_op(ICLKernel &kernel, ITensorPack &tensors, bool flush
     enqueue_common(kernel, tensors, flush);
 }
 
-#if defined(ENABLE_EXPERIMENTAL_DYNAMIC_FUSION)
-
-void CLScheduler::enqueue_op(ICLKernel &kernel, ITensorPack &tensors, const experimental::dynamic_fusion::ClExecutionDescriptor &exec_desc, bool flush)
-{
-    enqueue_common(kernel, tensors, exec_desc, flush);
-}
-
-#endif // defined(ENABLE_EXPERIMENTAL_DYNAMIC_FUSION)
-
 void CLScheduler::enable_job_chaining(int job_chaining_size)
 {
     _job_chaining_enabled = true;
diff --git a/src/runtime/CL/CLTuner.cpp b/src/runtime/CL/CLTuner.cpp
index 8ce5177847..1cc20f0c1e 100644
--- a/src/runtime/CL/CLTuner.cpp
+++ b/src/runtime/CL/CLTuner.cpp
@@ -28,9 +28,6 @@
 #include "arm_compute/runtime/CL/CLScheduler.h"
 #include "src/core/CL/ICLKernel.h"
 #include "support/StringSupport.h"
-#if defined(ENABLE_EXPERIMENTAL_DYNAMIC_FUSION)
-#include "src/gpu/cl/kernels/experimental/dynamic_fusion/ClCompositeKernel.h"
-#endif // defined(ENABLE_EXPERIMENTAL_DYNAMIC_FUSION)
 
 #include <cerrno>
 #include <fstream>
@@ -65,26 +62,6 @@ private:
     ITensorPack &_tensors;
 };
 
-#if defined(ENABLE_EXPERIMENTAL_DYNAMIC_FUSION)
-struct CompositeKernelData : public CLTuner::IKernelData
-{
-    CompositeKernelData(ITensorPack &tensors, const experimental::dynamic_fusion::ClExecutionDescriptor &exec_desc)
-        : _tensors{ tensors }, _exec_desc{ exec_desc }
-    {
-    }
-    ~CompositeKernelData() override = default;
-    void do_run(ICLKernel &kernel, cl::CommandQueue &queue) override
-    {
-        // ClCompositeKernel is purely stateless, and thus always requires memory injection
-        kernel.run_composite_op(_tensors, kernel.window(), queue, _exec_desc);
-    }
-
-private:
-    ITensorPack                                               &_tensors;
-    const experimental::dynamic_fusion::ClExecutionDescriptor &_exec_desc;
-};
-#endif // defined(ENABLE_EXPERIMENTAL_DYNAMIC_FUSION)
-
 bool CLTuner::kernel_event_is_set() const
 {
     return _kernel_event() != nullptr;
@@ -165,15 +142,6 @@ void CLTuner::tune_kernel_dynamic(ICLKernel &kernel, ITensorPack &tensors)
     do_tune_kernel_dynamic(kernel, &data);
 }
 
-#if defined(ENABLE_EXPERIMENTAL_DYNAMIC_FUSION)
-void CLTuner::tune_kernel_dynamic(ICLKernel &kernel, ITensorPack &tensors, const experimental::dynamic_fusion::ClExecutionDescriptor &exec_desc)
-{
-    CompositeKernelData data{ tensors, exec_desc };
-
-    do_tune_kernel_dynamic(kernel, &data);
-}
-#endif // defined(ENABLE_EXPERIMENTAL_DYNAMIC_FUSION)
-
 void CLTuner::add_tuning_params(const std::string &kernel_id, CLTuningParams optimal_tuning_params)
 {
     _tuning_params_table.emplace(kernel_id, optimal_tuning_params);