Fixed reshape

aidge_export_cpp/kernels/reshape.hpp

@@ -9,18 +9,18 @@ template<int M,
          typename Input_T, typename Output_T>
 __attribute__((always_inline)) inline
 void reshape_forward (
-    const Input_T* __restrict, // First input is useless as it only dictate the resulting layout of the reshape
-    const Input_T* __restrict inputs2,
+    const Input_T* __restrict inputs,
+    const Input_T* __restrict /*shape*/,
     Output_T* __restrict outputs)
 {
     // If inputs and outputs pointers are the same, the memory manager has already optimized this function so it is a no-op !
-    if (inputs2 == outputs)
+    if (inputs == outputs)
         return;
 
     // A reshape in c++ world should equal to a Noop
     // We only need to copy the input buffer to the output
     for (int m = 0; m < M; ++m) {
-        outputs[m] = inputs2[m];
+        outputs[m] = inputs[m];
     }
 }
 

aidge_export_cpp/templates/configuration/reshape_config.jinja

@@ -6,3 +6,5 @@
 {% include "./_meminfo.jinja" %}
 {# For layer configuration -#}
 #define {{ name|upper }}_NB_ELTS {{ in_dims[0]|join('*') }}
+
+#endif /* {{ name|upper }}_LAYER_H */

aidge_export_cpp/unit_tests/test_export.py

@@ -427,12 +427,31 @@ class test_operator_export(unittest.TestCase):
     def test_transpose(self):
         print("Transpose")
         model = aidge_core.sequential([
-            aidge_core.Producer([1, 2, 7, 8], name="producer"),
             aidge_core.Transpose(output_dims_order=[0, 2, 3, 1], name="transpose")
         ])
 
         self.unit_test_export(model, "Transpose", [[1, 7, 8, 2]])
 
+    def test_reshape(self):
+        print("Reshape")
+        model = aidge_core.sequential([
+            aidge_core.Reshape(name="reshape")
+        ])
+        shape = aidge_core.Producer(aidge_core.Tensor(np.array([7, 2, 1, 8], dtype=np.int64)), name="shape")
+        shape.add_child(model.get_node("reshape"), 0, 1)
+        model.add(shape)
+
+        self.unit_test_export(model, "Reshape", [[1, 7, 8, 2]], random_weights=False)
+
+    def test_matmul(self):
+        print("MatMul")
+        model = aidge_core.sequential([
+            aidge_core.Producer([4, 8], name="producer"),
+            aidge_core.MatMul(name="matmul")
+        ])
+
+        self.unit_test_export(model, "MatMul", [[8, 4]])
+
     def test_concat_axis_2(self):
         print("ConcatAxis2")
         model = aidge_core.sequential([