Mul, Sub and Add Layer with broadcasting support

[Semyon1104]

I will also add the functionality of element-by-element subtraction and element-by-element addition layers - sub and add

[Copilot]

Pull Request Overview

This PR implements a BinaryOpLayer class that supports element-wise multiplication, addition, and subtraction operations with broadcasting support for tensors. It extends existing layer functionality to handle binary operations between tensors of different but compatible shapes.

Key changes:

• New BinaryOpLayer class with support for multiplication, addition, and subtraction operations
• Broadcasting logic to handle tensors with compatible but different shapes
• Scalar tensor operations for element-wise operations with single values

Reviewed Changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 3 comments.

File	Description
include/layers/BinaryOpLayer.hpp	Header file defining the BinaryOpLayer class interface and operation enums
src/layers/BinaryOpLayer.cpp	Implementation of binary operations with broadcasting and scalar support
include/layers/Shape.hpp	Added equality operators for Shape class to support testing
test/single_layer/test_binaryoplayer.cpp	Comprehensive test suite covering various broadcasting scenarios and operations

Comments suppressed due to low confidence (2)

test/single_layer/test_binaryoplayer.cpp:180

• Test is named 'BroadcastingTestAdd' but is part of the 'BinaryOpLayerMulTests' test fixture. Consider creating a separate test fixture for Add operations or renaming to reflect the actual operation being tested.

TEST_F(BinaryOpLayerMulTests, BroadcastingTestAdd) {

test/single_layer/test_binaryoplayer.cpp:204

• Test is named 'BroadcastingTestSubGooglNet' but is part of the 'BinaryOpLayerMulTests' test fixture. Consider creating a separate test fixture for Sub operations or renaming to reflect the actual operation being tested.

TEST_F(BinaryOpLayerMulTests, BroadcastingTestSubGooglNet) {

[allnes]

It's part of elementwise operation

[Semyon1104]

I thought it would be better to put the logic of interaction between two tensors with a fairly comprehensive broadcast logic in an another layer to separate the situations where the input is a single tensor, EWLayer, and where there are two input arrays and 1 output

[Semyon1104]

Or is it better to combine these layers into one?

[allnes]

Ok we can choose compromise, please create operation - Logic and unite all logic operations

[Semyon1104]

Do you mean to add the kDiv operation and name the enumerated type logic?

[allnes]

The current implementation recalculates indices for each element. Consider caching stride calculations for better performance with large tensors.

[allnes]

While Mul, Add, and Sub are implemented, division would complete the basic arithmetic set. Consider adding kDiv to the BinaryOpType enum.

[allnes]

@aobolensk