Add output operator for dynamic fusion

* The output of the fused operator must be explicitly specified
  using GpuOutput operator.
* Any temporary tensors used to connect the output of an operator
  to the input of another operator will be marked as no-alloc
  and won't be allocated as a tensor in the memory.

Resolves: COMPMID-5771
Signed-off-by: Viet-Hoa Do <viet-hoa.do@arm.com>
Change-Id: I5ae8e800f8f737db23a055a92b01c4f1d78c3bb8
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/8794
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: SiCong Li <sicong.li@arm.com>
Reviewed-by: Gian Marco Iodice <gianmarco.iodice@arm.com>
Reviewed-by: Gunes Bayir <gunes.bayir@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>

1 files changed, 27 insertions, 0 deletions

diff --git a/src/dynamic_fusion/sketch/utils/DependencyGraph.h b/src/dynamic_fusion/sketch/utils/DependencyGraph.h
index 03678defae..633c5e4263 100644
--- a/src/dynamic_fusion/sketch/utils/DependencyGraph.h
+++ b/src/dynamic_fusion/sketch/utils/DependencyGraph.h
@@ -417,6 +417,33 @@ public:
         }
         return tensors;
     }
+    /** Get intermediate tensors of the whole graph.
+     *
+     * @return std::vector<TensorId>
+     */
+    std::vector<TensorId> intermediate_tensors() const
+    {
+        std::vector<TensorId> tensors;
+
+        // If a tensor is used to connect the input of an operator and the output of another operator,
+        // it is not allocated in the memory. The tensor exists as a temporary variable only.
+        for(auto src_tensor : _adj_src_ops)
+        {
+            if(!src_tensor.second.empty())
+            {
+                const auto dst_tensor = _adj_dst_ops.find(src_tensor.first);
+                if(dst_tensor != _adj_dst_ops.end())
+                {
+                    if(!dst_tensor->second.empty())
+                    {
+                        tensors.push_back(src_tensor.first);
+                    }
+                }
+            }
+        }
+
+        return tensors;
+    }
     /** Get all root ops. Root ops can also be referred to as "src ops" of the whole graph
      *
      * @return std::vector<OperatorId>