Fuse bn

• Fix Add + Mul recipies
• Add fuseBN recipies
• Fix parameters issue #8 (closed)
• Update tensor to get/set values
• Add approxEq Tensor utils method to check if two tensor are approximatly equals
• Update the way scheduler is updated
• Add unit test for:
  • fuse Add + Mul
  • remove padding
  • Add unit test for fuseBN recipies (TODO in aidge_backend_cpu because of dependancie to the backend for the test)

include/aidge/data/Tensor.hpp

@@ -580,13 +580,14 @@ class Tensor : public Data,
      * @param flatIdx 1D index of the value considering a flatten tensor.
      * @return std::vector<DimSize_t>
      */
-    std::vector<std::size_t> getCoord(std::size_t flatIdx){
-        std::vector<std::size_t> coordIdx = {};
+    std::vector<std::size_t> getCoord(std::size_t flatIdx) const {
+        std::vector<std::size_t> coordIdx = std::vector<std::size_t>(mDims.size());
         std::size_t idx = flatIdx;
-        for (std::size_t d: mDims){
-            coordIdx.push_back(idx % d);
-            idx/=d;
+        for (std::size_t i = mDims.size() - 1; i > 0; --i){
+            coordIdx[i] = (idx % mDims[i]);
+            idx/=mDims[i];
         }
+        coordIdx[0] = idx % mDims[0];
         return coordIdx;
     }
 
@@ -596,16 +597,17 @@ class Tensor : public Data,
      * @param coordIdx Coordinate to an element in the tensor
      * @return DimSize_t
      */
-    std::size_t getIdx(std::vector<std::size_t> coordIdx){
+    std::size_t getIdx(std::vector<std::size_t> coordIdx) const {
+        // std::size_t flatIdx = 0;
+        // std::size_t stride = 1;
         std::size_t flatIdx = 0;
-        std::size_t stride = 1;
         assert(coordIdx.size() == mDims.size() && "Coordinates does not match number of dimensions");
-        for(std::size_t i=0; i< mDims.size(); ++i){
+        std::size_t i = 0;
+        for(; i < mDims.size() - 1; ++i){
             assert(coordIdx[i] < mDims[i] && "Coordinates dimensions does not fit the dimensions of the tensor");
-            flatIdx += (coordIdx[i] * stride);
-            stride *= mDims[i];
+            flatIdx = (flatIdx + coordIdx[i]) * mDims[i + 1];
         }
-        return flatIdx;
+        return flatIdx + coordIdx[i];
     }
 
 private:

[laurent soulier]

I think that from tensor, either you iterate or you get from coordinates. getting from index is low level, and should be reserved to TensorImpl (possibly useless because TensorImpl gives you access to raw pointer anyway, and you can use [] on it).

[Cyril Moineau]

Getting from coordinate or index is the same.

[laurent soulier]

I think that only TensorImpl has the required informations to convert between coordinates and linear index.

[Maxence Naud]

Actually, only the Tensor class has got both "dimensions" and "size" attributes necessary for converting coordinates.

[laurent soulier]

Sorry I was speaking about ongoing refactoring. Yet, see issue #13 (closed) or #14 (closed), these data may effectively go back to Tensor

[laurent soulier]

conversion functions will be excessively expensive. As number of dimension is practically a constant for a given tensor, I propose to preallocate a vector of coordinates of the right size and pass it by reference to the coordinate computation function and not return it. The actual implementation makes reallocations and then a full vector copy. Even if the compiler is smart, is not that smart

[Cyril Moineau]

I do not understand your comment, you want to memoize the getIdx and getCoord method ?

[laurent soulier]

Actually I'm not sure how to do that properly in order to avoid threading issues, dangling references and so on...
I think that the best solution is to pass a reference to an user preallocated vector of coordinates to getCoord.
It should be documented as an unsafe function: if the input vector size is not the right one, the behavior is undefined.