Explore Projects

<p>This project will explore extending a new feature for the Performance Co-Pilot (PCP). PCP is an open source framework at the system level that monitors, analyze and retrieve information to understand system performance. It has a distributed architecture of agents, called Performance Metric Domain Agents (PMDA), that extract performance values from the desired environments or systems.</p> <p>Every agent (PMDA) reports to a centralized daemon called Performance Metric Collection Daemon or PMCD per host. Then, accessing any metric is as easy as sending a request to PMCDs. PCP has several client tools to help the final user understand the metrics. Finally, PCP also offers a good API to create an agent for any software.</p> <p>Coming back to the interest of the proposal, PCP also supports instrumentation. This is a way to enable user applications like Firefox to send metrics in execution time to the PMDA. In this scenario, Memory Mapped Values (MMV) are used to offer an extremely lightweight mechanism for instrumenting applications using shared memory. However, there is no support for metadata labels added since the newest v4, and hence the goal of this project: add label support for instrumentation.</p>

<p>Goal of this project is to improve and build upon existing Grafana intergration for Performance Co-Pilot.</p> <ul> <li>Developing a <strong>reactive dashboard</strong> and using PCP metrics</li> <li>Implementing support for PCP <strong>derived metrics</strong> in the Vector data source, as described in the PCP REST API documentation</li> <li>Improving the <strong>metric search</strong> process by creating a new page for live, full-text search on PCP metric names, instances, instance domains and their descriptions, if any</li> </ul> <p>Based on performance metrics from PCP and a predefined rule set, the <strong>reactive dashboard</strong> should display only relevant metrics to the user. For example, if the CPU utilization is high, it should show an overview of CPU statistics and processes contributing load. The user should then be able to drill down into more specific areas of interest - e.g. present graphs about lock contention. For each recognized performance issue a link should be presented to the user on how to resolve the issue. In this way the dashboards guide users toward possible root cause of performance problems.</p>

Indigo Night is planned to be a beautiful Audio Visual journey, focused on storytelling. Based on pixel art, it will contain retro vibes and an old school style, which is immensely trending as of today. Salient features of this project include: Smooth Animations, Multiple Playable Characters, Beautiful environments as well as a musical delivery of several elements. Overall, I aim to create a groundbreaking modern experience in the Luna&Cat IDE app, by bringing together all those hours I invested into 2D games, indie as well as mainstream, and the game design experience I have, in Unity and RPG Maker.

<p>I propose the implementation of an idiomatic golang package that will allow applications written in golang to instrument performance simply by making the appropriate function calls exported by the package. The package’s implementation is to be based on both libpcp_mmv and parfait, and although built initially for PCP, it will be extensible with appropriate interfaces, so someone else can implement the said interfaces to implement instrumentation for another platform.</p>

<p>In some algorithms there is the need to compare two distances of two point pairs. Especially, computing distances on ellipsoid in Boost.Geometry used compare_distance() function, which need to calculate the actual distance of two geodesic segments on the ellipsoid. Boost.Geometry used Vincenty formula, the time complexity of this algorithm is hard to estimate and expensive.</p> <p>To reduce the time complexity, this project tried to use the following approaches:</p> <ul> <li><strong>Condition Division</strong>:Compute the Euclidean distances first, if the length of two segments are very close, defined not “far enough” means “very close”. then fall back to expensive geographic distance calculation otherwise return the result by comparing those numbers obtained by less expensive cartesian computation.</li> <li><strong>Calculation with Sectorial Area</strong>:Get cross-section through the center of ellipsoid and the geodesic segments, then calculate the sectorial area, which is easy to compare distances. </li> <li><strong>Calculation with Performing a local spheroid approximation and return the 2D distance by map projection</strong>:Perform a local spheroid approximation, then use methods of map projection to calculate the distances.</li> </ul>

At present, selecting a radius or bounding box to cover a desired region and then filtering the retrieved points is a viable option for narrowing down results by a specific region of interest. However, this method can become sluggish when dealing with expansive areas and entails handling location data separately in a database. Introducing a more efficient feature for intricate geo-filters would 1) speed up the process, 2) create opportunities for novel applications, 3) open the door for further opportunities of optimization. Such a feature would be particularly beneficial for Qdrant's users dealing with vast data sets or intricate geometries

The goal of this project is to add the JSON_* functions support in Select Handler mode. After the project, we can use json function with Columnstore in Select Handler mode. We can design and implement JSON functions. My implementation idea is as follows(take the JSON_VALID function as example): ● Related files and classes: define the relevant files required by the JSON_VALID function according to the specification of the SQL function definition in Columnstore 1. functor_json.cpp: According to the return value of the function, create new class Func_json_valid derived from class Func_bool. 2. func_json_valid.cpp: the JSON_VALID function is implemented here. Implement any other methods that behave differently from the default. 3. funcexp.{h,cpp}: in FuncExp::FuncExp(), add a new entry fFuncMap["json_valid"] = new Func_json_valid(). 4. CMakeLists.txt: Add any source files to build. ● Borrow and translate: 1. JSON_* functions are already implemented in MariaDB leveraging its own JSON format parser 2. The parser itself is couple files only library. Here is the main file. 3. So the next step is to read/understand this implementation and translate the code into Columnstore using the mentioned JSON format parser.

<p>Metal Renegades is meant to feel like an immersive Old West-style world. Right now, it’s just a desert, with some mountains made out of sand. I propose to expand that with more interesting terrain typical of the Old West: mesas where rock around a plateau has eroded away, canyons where rivers have cut deep into the rock, mountains made of rock instead of sand, and more. Additionally, I plan to create a realistic distribution of flora and fauna.</p>

As MariaDB continues to evolve and improve, MySQL users may consider it as a better alternative. The absence of some spatial functions could be a drawback for some of them and potentially make the migration process difficult. My goal is to implement the following spatial functions: ST_IsValid, ST_Validate, - MBRCoveredBy, ST_Collect, ST_GeoHash, ST_Simplify, ST_LatFromGeoHash, ST_LatFromGeoHash and ST_PointFromGeoHash. For each new function to implement, I will start by importing the corresponding tests from the MySQL test suite. These tests serve as an optimal starting point. Given that the primary goal is to ensure compatibility between MySQL and MariaDB, passing the same set of tests on both codebases ensures uniform behaviour. Additionally, these tests serve as a valuable supplement to the documentation, offering detailed insights into the expected behaviour of each function while also documenting edge cases. Moreover, as new tests are written when fixing a bug, we can mitigate the risk of encountering previously resolved issues by the MySQL developers. The spatial functions already implemented, are defined in the sql/item_geofunc.h file. The new functions will also be placed in the same file and will be implemented by extending the interfaces that are already defined. For instance, ST_IsValid will extend Item_bool_func_args_geometry which is the class that defines the interface for all the MariaDB native functions returning boolean measurements of a single GEOMETRY argument. In MySQL, the functions ST_GeoHash(longitude, latitude, max_length) and ST_GeoHash(point, max_length) do not return a special datatype but a string. I want to make a more in-depth analysis regarding this function and evaluate if defining a custom datatype would be a viable option and if it would grant meaningful performance improvements.

This project aims to implement a Vector Tracking Loop (VTL) architecture in GNSS-SDR, a widely used open-source software-defined GNSS receiver. Traditional scalar tracking architecture tracks satellite signals independently, limiting robustness in dynamic or degraded signal environments. In contrast, VTL integrates tracking and navigation tasks into a single algorithm, where all channels collaborate to enhance signal tracking performance, particularly under challenging conditions such as interference or temporary outages. The proposed work will design and integrate a modular, configurable VTL architecture supporting multi-constellation and multi-frequency operation, aligning with the existing capabilities of GNSS-SDR. Deliverables include: VTL algorithm implementation, real-time feedback to tracking channels, multi-constellation and multi-frequency support, benchmarking, and documentation. This contribution will enable researchers and developers to explore advanced GNSS techniques related to the VTL architecture within a real-time, open-source SDR.

<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>

<p>This project involves developing a new library which tries to cover the major concepts of a complex and diverse field of Astronomy. This library will provide functionalities to mitigate the complexities of astronomical calculations using C++.</p> <p><strong>1) Celestial Coordinate System:</strong><br> For any Astronomical observation, it is essential to know the location of any object on the celestial sphere. These classes will provide functionalities (e.g.: find sun, find the moon, the angular separation between two points in the sky, find the constellation of coordinate, conversion from cartesian coordinates to spherical polar coordinates and vice versa, etc.) for using different coordinate systems, process them for different purposes and convert them into each other.</p> <p><strong>2) Handling FITS (Flexible Image Transport System) File:</strong><br> This library will provide functionality for reading, manipulating and verifying file which can further be used for other processes like making light curves or spectroscopy. It will deal with data unit consisting of 8 bit, 16 bit and 32 bit also. It will also provide functionality for converting the normal image to FITS and applying Astrometry to find positions of objects in the data unit</p>

<p>Currently Paintroid only has the basic brush and the spray can. The goal is to add additional brush options such as a watercolor brush, rubbing around existing color. Also, when drawing a little bit faster, the drawing line should be slightly smoothed. When drawing slowly, it should however be possible to draw without smoothing of the lines. Also, add an option in the brush settings that allows to turn off (and on again) antialiasing, independently of the width of the brush. Currently, antialiasing is only turned off if the width is set to 1.</p>

This project aims to transform MariaDB's INTERVAL keyword into a first-class data type, enabling its use in table columns, stored routines, and hybrid functions.

<p>Complete support for <a href="https://github.com/openshift/origin" target="_blank">OpenShift</a> and <a href="https://github.com/kubernetes/kubernetes" target="_blank">Kubernetes</a> as a backend in <a href="https://github.com/fedora-modularity/conu" target="_blank">conu</a></p>

<p>This project aims to improve radare2 debugger with program tracing and support remote debugging on more platforms. A program trace contains all states of a running process in time so that users can analyze the program execution at any point in time quickly. Besides, a new timeless debugging view is implemented in Cutter to provide more in-depth analysis with information from the trace. In addition, this project implements network support for WinDbg/KD debugger and remote debugging support for qira and kgdb.</p>

ReZero is a rogue-lite, Dead Cells-inspired action-platformer, allowing you to explore a sprawling, ever-changing castle… assuming you’re able to fight your way past its keepers. To beat the game, you’ll have to master 2D "souls-lite combat" with the ever-present threat of permadeath looming. Kill, die, learn, repeat. The gameplay will be very similar to Dead Cells with cinematic cutscenes, high-paced action, and magic items. This project aims to showcase the true potential of Pocketcode by creating a good quality high-level game. The game will be playable on a phone as well as on smart TV.

XRefer is a binary analysis tool built by Mandiant FLARE that uses cross-reference correlation, capa integration, and LLM-powered cluster analysis to help malware analysts navigate complex binaries faster. Today it requires IDA Pro, Binary Ninja, or a configured Ghidra environment to run, locking out students and researchers who lack commercial licenses or heavyweight setups. This project implements a fully functional Radare2 backend for XRefer via the r2pipe Python bindings, paired with a standalone CLI that auto-detects the best available disassembler and falls back to Radare2 when none are found. Radare2 is free, open-source, cross-platform, and pip-installable, making it the right foundation for democratizing XRefer access. The backend maps every method of the BackEnd abstract base class to verified r2pipe commands, handles known r2 edge cases including the aflj NaN-value bug and pdfj integer overflow, and includes targeted refinements to the base abstraction layer. Deliverables are: a complete Radare2 backend module passing integration tests against the Ghidra backend on a three-binary reference corpus; a standalone CLI installable with pip; a test suite with pre-recorded JSON fixtures enabling CI without a live r2 installation; and full documentation covering setup, architecture, and guidance for adding future backends.

<p>During last decades Global Navigation Satellite Systems (GNSS) have increasingly become a fundamental tool for high accuracy positioning purposes. For this reason, it is crucial to guarantee a certain high quality of service (QoS). This QoS may be degraded by the presence of spurious undesirable signals coming from jammers that could degrade partially (or totally) the bandwidth of the desired GNSS signal. Basically, two kinds of Radio Frequency Interferences (RFI) can disturb a GNSS signal. Each one has its own characteristic spectral feature and a classification based on that is described as following: • Continuous wave interferences (CWI): a perturbation of a relative small part of the spectrum and can be mitigated by means of a notch filter. • Short pulsed interferences: with an instantaneous bandwidth that usually covers the entire GNSS band and can be partially removed by means of a blanking filter. Two interference detection and mitigation algorithms will be evaluated in order to implement a new Signal conditioner block into GNSS-SDR. These algorithms will be coded in C++</p>

<p>Matrix multiplication is one of the most important operations in linear algebra. The project "Optimising multiplication in uBLAS" is concerned with optimising the multiplications under matrices. Optimising multiplication in uBLAS can be done in done following two lines: 1) By finding patterns to optimise in expression templates using template programming. 2) By applying optimising algorithms (written with templates). I would be optimising the matrices multiplication operation following both the above mentioned ways.</p>

This project's goal is to make GRC as a stand alone application, separated from GNU Radio code base and make the workflow pluggable. The main goal of this project is broken down into several milestones that is, separation of GRC from GNU Radio code base, modularized option block, modularized templates, and jinja template support. The project intends to let people define their own workflow so that people can generate code from the flow graph that is created in GRC into any programming language and can make use of different GUI frameworks. Another main goal of this project is to support jinja template for code generation besides mako template.

<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>

<p>The current implementation of the GNSS-SDR is a hard correlation scheme with a known scheme. The goal of this project is to implement and validate an optimal detector, based on Massey’s frame synchronization metric.</p>