diff --git a/src/runtime/CL/CLSubTensor.cpp b/src/runtime/CL/CLSubTensor.cpp
index 5f58024..d0e7d76 100644
--- a/src/runtime/CL/CLSubTensor.cpp
+++ b/src/runtime/CL/CLSubTensor.cpp
@@ -29,6 +29,11 @@
 
 using namespace arm_compute;
 
+CLSubTensor::CLSubTensor()
+    : _parent(nullptr), _info()
+{
+}
+
 CLSubTensor::CLSubTensor(ICLTensor *parent, const TensorShape &tensor_shape, const Coordinates &coords, bool extend_parent)
     : _parent(nullptr), _info()
 {
diff --git a/src/runtime/CL/functions/CLGEMM.cpp b/src/runtime/CL/functions/CLGEMM.cpp
index e6f8f26..c1926a7 100644
--- a/src/runtime/CL/functions/CLGEMM.cpp
+++ b/src/runtime/CL/functions/CLGEMM.cpp
@@ -138,8 +138,10 @@
 
         // Manage intermediate buffers
         _memory_group.manage(&_tmp_a);
-        _memory_group.manage(&_tmp_b);
-
+        if(_reshape_b_only_on_first_run)
+        {
+            _memory_group.manage(&_tmp_b);
+        }
         // _tmp_a and _tmp_b will be auto configured in _interleave_kernel and in _transpose_kernel
 
         // Configure interleave kernel
diff --git a/src/runtime/CL/functions/CLGEMMConvolutionLayer.cpp b/src/runtime/CL/functions/CLGEMMConvolutionLayer.cpp
index 3cba98c..bc339f1 100644
--- a/src/runtime/CL/functions/CLGEMMConvolutionLayer.cpp
+++ b/src/runtime/CL/functions/CLGEMMConvolutionLayer.cpp
@@ -38,8 +38,8 @@
 using namespace arm_compute;
 using namespace arm_compute::misc::shape_calculator;
 
-CLConvolutionLayerReshapeWeights::CLConvolutionLayerReshapeWeights(std::shared_ptr<IMemoryManager> memory_manager)
-    : _memory_group(std::move(memory_manager)), _weights_reshape_kernel(), _weights_transposed_kernel(), _weights_reshaped()
+CLConvolutionLayerReshapeWeights::CLConvolutionLayerReshapeWeights()
+    : _weights_reshape_kernel()
 {
 }
 
@@ -86,16 +86,12 @@
 
 void CLConvolutionLayerReshapeWeights::run()
 {
-    _memory_group.acquire();
-
     CLScheduler::get().enqueue(_weights_reshape_kernel);
-
-    _memory_group.release();
 }
 
 CLGEMMConvolutionLayer::CLGEMMConvolutionLayer(std::shared_ptr<IMemoryManager> memory_manager)
     : _memory_group(memory_manager), _reshape_weights(), _im2col_kernel(), _mm_gemm(memory_manager), _mm_gemmlowp(memory_manager), _gemmlowp_output_stage(), _col2im_kernel(), _im2col_output(),
-      _interleave_output(), _weights_reshaped(), _weights_transposed(), _gemm_output(), _tmp_output(), _is_quantized(false), _is_first_run(true)
+      _weights_reshaped(), _gemm_output(), _tmp_output(), _is_quantized(false), _is_first_run(true)
 {
 }
 
diff --git a/src/runtime/NEON/functions/NEConvolutionLayer.cpp b/src/runtime/NEON/functions/NEConvolutionLayer.cpp
index 0a49158..c16ce9b 100644
--- a/src/runtime/NEON/functions/NEConvolutionLayer.cpp
+++ b/src/runtime/NEON/functions/NEConvolutionLayer.cpp
@@ -30,6 +30,7 @@
 
 #include <cmath>
 #include <tuple>
+#include <utility>
 
 namespace arm_compute
 {
diff --git a/src/runtime/SubTensor.cpp b/src/runtime/SubTensor.cpp
index c5b8f33..b010a32 100644
--- a/src/runtime/SubTensor.cpp
+++ b/src/runtime/SubTensor.cpp
@@ -27,6 +27,11 @@
 
 using namespace arm_compute;
 
+SubTensor::SubTensor()
+    : _parent(nullptr), _info()
+{
+}
+
 SubTensor::SubTensor(ITensor *parent, const TensorShape &tensor_shape, const Coordinates &coords, bool extend_parent)
     : _parent(nullptr), _info()
 {