Boost C++ Libraries

Mentors: Student: Pan Yue

<p>GIL is restricting Python code performance. There have been a long time that the Pythondevelopers are struggling to solve the performance issue brought byGIL. In PEP554, peoplecame up with the idea of multi-interpreter in the same program. Which can start multiple in-terpreter concurrently to solve the problem. But due to some reasons, it wasn’t be implementedin Python 3.9 Thanks to boost::asio::iocontext::strand, it’s possible for us to use multiple Python interpretersin the same program in theactor modelfashion. Hereby I propose this project to implementPython eventloop usingboost::asio::iocontext::strand Those developers who wants to use numerous Python libraries and share rich Python ecosystemwill benefit from this project.</p>

Mentors: Student: Suraj Kumar-3

<p>Boost.Real aims at providing a data-type that can represent all types of computable real numbers and provides the flexibility of performing arbitrary precision arithmetic, the user can always decide and control the amount of precision to be used for a calculation thus controlling the computational power used by the machine. I propose to benchmark the library and compare it against other similar libraries. There is a scope of optimization for Trigonometric functions and performance can be improved by a significant amount. I also plan for several other small optimizations</p>

Mentors: Student: Eduardo Quintana Miranda

<p>We propose the implementation of Fast Fourier Transform (FFT) capabilities inside the Boost.Math library. This project answers the need for a C++ library of this sort, and the lack of FFT within Boost.Math. Our main goal is to design a modern C++ template interface for FFT that supports multiple types that satisfy the Ring Axioms and specialized routines for complex numbers, which constitute the most used cases.</p>

Mentors: Student: Prathamesh Tagore

<ul> <li>A layer of separation between two dimensional convolution and correlation was added.</li> <li>Support for many different boundary options along with optimisation was developed.</li> <li>2D convolution implementation was completely aligned with existing implementation of 1D convolution(APIs of both implementations are consistent and provide suitable functionalities). </li> <li>All changes were complemented with tests and documentation.</li> </ul>

Mentors: Student: Syed Fahad

<p>A new <strong>quad-double</strong> type would be implemented, initially as a wrapper around the QD package's <code>qd_real</code> class authored by <em>David Bailey</em>. Tests for correctness and benchmarks will be subsequently written along with the documentation. If time permits and if the mentors find it feasible, a new <strong>multiple-double</strong> type based on the same algorithms as QD package will be written with no dependency on external libraries.</p>