Implement dynamic fusion softmax operator

 - Return aux tensorInfo by get_aux_tensors() at runtime to init the aux
   tensor with the right size.
 - Keep softmax unfusable for this commit
 - Hence, added Tensor3D to template writer arguments declaration, for sake of
   keeping dynamic fusion softmax componenets' kernels matching their cl
   counterparts.

Resolves: COMPMID-5523
Change-Id: I667f39545db925f667036ef448302c79a0330373
Signed-off-by: Ramy Elgammal <ramy.elgammal@arm.com>
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/c/VisualCompute/ComputeLibrary/+/483924
Tested-by: bsgcomp <bsgcomp@arm.com>
Reviewed-by: Gunes Bayir <gunes.bayir@arm.com>
Comments-Addressed: bsgcomp <bsgcomp@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/8986
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Jakub Sujak <jakub.sujak@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>

1 files changed, 12 insertions, 9 deletions

diff --git a/tests/validation/dynamic_fusion/gpu/Integration.cpp b/tests/validation/dynamic_fusion/gpu/Integration.cpp
index 7f2d439183..6a283f8082 100644
--- a/tests/validation/dynamic_fusion/gpu/Integration.cpp
+++ b/tests/validation/dynamic_fusion/gpu/Integration.cpp
@@ -90,9 +90,10 @@ TEST_CASE(Conv2d, framework::DatasetMode::ALL)
     // Instead of using ACL allocated memory, the user can choose to import memory into the tensors
     for(auto &data : runtime.get_auxiliary_tensors())
     {
-        CLTensor     *tensor      = data.first;
-        AuxMemoryInfo aux_mem_req = data.second;
-        tensor->allocator()->init(*data.first->info(), aux_mem_req.alignment);
+        CLTensor     *tensor      = std::get<0>(data);
+        TensorInfo    info        = std::get<1>(data);
+        AuxMemoryInfo aux_mem_req = std::get<2>(data);
+        tensor->allocator()->init(info, aux_mem_req.alignment);
         tensor->allocator()->allocate(); // Use ACL allocated memory
         // auto buf = cl::Buffer();
         // tensor->allocator()->import_memory(buf);  // Or, import external memory
@@ -178,9 +179,10 @@ TEST_CASE(Add_Output_Add_Output, framework::DatasetMode::ALL)
     // Instead of using ACL allocated memory, the user can choose to import memory into the tensors
     for(auto &data : runtime.get_auxiliary_tensors())
     {
-        CLTensor     *tensor      = data.first;
-        AuxMemoryInfo aux_mem_req = data.second;
-        tensor->allocator()->init(*data.first->info(), aux_mem_req.alignment);
+        CLTensor     *tensor      = std::get<0>(data);
+        TensorInfo    info        = std::get<1>(data);
+        AuxMemoryInfo aux_mem_req = std::get<2>(data);
+        tensor->allocator()->init(info, aux_mem_req.alignment);
         tensor->allocator()->allocate(); // Use ACL allocated memory
         // auto buf = cl::Buffer();
         // tensor->allocator()->import_memory(buf);  // Or, import external memory
@@ -282,9 +284,10 @@ TEST_CASE(Add_Output_Add_Cast_Cast_Output, framework::DatasetMode::ALL)
     // Instead of using ACL allocated memory, the user can choose to import memory into the tensors
     for(auto &data : runtime.get_auxiliary_tensors())
     {
-        CLTensor     *tensor      = data.first;
-        AuxMemoryInfo aux_mem_req = data.second;
-        tensor->allocator()->init(*data.first->info(), aux_mem_req.alignment);
+        CLTensor     *tensor      = std::get<0>(data);
+        TensorInfo    info        = std::get<1>(data);
+        AuxMemoryInfo aux_mem_req = std::get<2>(data);
+        tensor->allocator()->init(info, aux_mem_req.alignment);
         tensor->allocator()->allocate(); // Use ACL allocated memory
         // auto buf = cl::Buffer();
         // tensor->allocator()->import_memory(buf);  // Or, import external memory