commit 002e6530f6218b00a28aef9be8b21efb08cf3602
author Ramy Elgammal <ramelg01@e129512.arm.com> Wed Jan 11 18:48:04 2023 +0000
committer Ramy Elgammal <ramy.elgammal@arm.com> Wed Jan 25 10:24:23 2023 +0000
tree f3e2f9d064b985ffe283512825b34cdd59f29f50
parent cc2877368d5e15d9ea89d31c84ec651fc0fffd13

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>

tests/validation/fixtures/dynamic_fusion/gpu/cl/Pool2dFixture.h

diff --git a/tests/validation/fixtures/dynamic_fusion/gpu/cl/Pool2dFixture.h b/tests/validation/fixtures/dynamic_fusion/gpu/cl/Pool2dFixture.h
index efb67f8..efb5cf1 100644
--- a/tests/validation/fixtures/dynamic_fusion/gpu/cl/Pool2dFixture.h
+++ b/tests/validation/fixtures/dynamic_fusion/gpu/cl/Pool2dFixture.h
@@ -109,9 +109,10 @@
         // (Important) Allocate auxiliary tensor memory if there are any
         for(auto &data : runtime.get_auxiliary_tensors())
         {
-            auto       tensor      = data.first;
-            const auto 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
         }
         // Construct user tensors
@@ -142,8 +143,8 @@
         return reference::pooling_layer<T>(src, pool_info, QuantizationInfo(), nullptr, DataLayout::NCHW);
     }
 
-    TensorType       _target{};
-    SimpleTensor<T>  _reference{};
+    TensorType      _target{};
+    SimpleTensor<T> _reference{};
 };
 
 template <typename TensorType, typename AccessorType, typename FunctionType, typename T>