EASENA 3.1.1

Performance enhancement on CPU. This should bring a significant improvement for programs using a lot of mutations and individuals.

30% performance improvement reported on 4 cores for a program optimizing quantum circuits using 100'000 individuals.

Performance

• Mutation is now done in parallel for all templates based on CPopulation.

Patches

3.1.1

• Fix first immigrant not being cleared from the queue and being read instead of newer ones.
• An island can no longer send an individual to itself.
• Fix formatting when an individual was received.

EASENA 3.0.2

Major improvements in portability, performances and usability. This version also fix a lot of bugs.

Portability

• Creation of a CI pipeline to ensure Easea and all examples compile and run on Windows (latest msvc), Linux (latest gcc on Ubuntu) and MacOS (latest AppleClang version on MacOS).
• Rewritten CMakeLists.txt to install in standard paths while maintaining backward compatibility.
• Easena now generates a portable CMakeLists.txt along with the old deprecated Makefile.
• New Install chapter in the README to simplify installation for all users.
• New UDP layer based on Boost::asio.

Usability

• C++ Templates can now be used inside the genome.
• All types can be used inside the genome as long as the user provides a way to serialize them.

Performance

• New random generator using modern C++.
• Removed lot of slow deep copies in multi-objective and selection operators.

Fix

• Lot of wrong type casts have been fixed.
• There was a few memory leaks inside the compiler.
• Some includes were missing.
• At least 3 buffer overflow
• At least 2 null pointer referencing.
• Floating point selection pressure was not working anymore.

Patches

1. Fixed CUDA's version of genetic programming not compiling anymore.
2. Fixed statistics header in command line output.

EASEA V2.0.0

v2.0.0

Update README.md

Stable

Stable version of EASEA

Please refers to the differents installations guides:

• For Linux/OSX (English/French)
• For Windows (French)