C++Now 2023 - BLADE: CUDA-Accelerated DSP for Allen Telescope Array⌗
May, 2023 - Aspen, CO - USA — The Allen Telescope Array, comprising 42 antennas, utilizes the BLADE (Breakthrough Listen Accelerated DSP Engine) - a C++20 GPU-based software. This talk delves into the capabilities of a software-defined telescope with a GPU backend, emphasizing the real-time processing of a ~1 Tbps data stream. Highlighting the challenges of hardware bottlenecks and the intricacies of building a distributed pipeline, the discussion also sheds light on BLADE’s modular interface that seamlessly integrates optimization techniques such as CUDA Graphs, JIT compilation of CUDA kernels, and smart memory management. For more information about the project, visit BLADE repository.
GNU Radio Conference 2022 - CyberEther Heterogeneous GUI⌗
September, 2022 - Washington, DC - USA — CyberEther offers a GPU accelerated frequency sink interface (2D Waterfall, Lineplot, Spectrogram). It works natively using modern graphical APIs like Metal, Vulkan, and WebGPU. The internal Digital Signal Processing (DSP) is accelerated whenever possible using parallel computing APIs like CUDA, Metal, and Vulkan. By being modular, CyberEther can adapt to the target device and run with the best-supported combination possible. For more information about the project, visit CyberEther repository.
SciPy Conference 2022 - GPU Accelerated FM Demodulation⌗
July, 2022 - Austin, Texas - USA — Talk about the process of demodulating real-world FM stations in parallel with elemental Numpy and SciPy Signal functions. Focusing on techniques, like vectorization, buffering, and frequency-domain filters, to make the computation more computationally efficient to achieve real-time operation. We also discuss how to apply GPU acceleration with CuPy and cuSignal. For more information about the project, visit radio-core repository.
SciPy Conference 2022 - GPU Accelerated TurboSETI Poster⌗
July, 2022 - Austin, Texas - USA — A common technique adopted by the SETI community is monitoring electromagnetic radiation for signs of extraterrestrial technosignatures using ground-based radio observatories. The data stream generated by a telescope can easily reach the rate of terabits per second. The analysis is made using a Python-based software called TurboSETI to detect narrowband drifting signals inside the recordings that can mean a technosignature (E.T.). We discuss how we ported a CPU-only program to leverage the parallel capabilities of a GPU using CuPy, Numba, and custom CUDA kernels. For more information about the project, visit TurboSETI repository.
Equinix Interview - Metal Meets: Radio⌗
March, 2020 - Livestream — Me, Jeremy, and Ed talk about Software Defined Radio and ARM Single Board Computers and PiSDR.
Cyberspectrum #23 - WirelessVillage DEFCON⌗
August, 2018 - Caesars Palace, Las Vegas - USA — Talk about the development of the Open Satellite Project software to demodulate, decode and parse data sent from the new generation of weather satellites from NASA and NOAA (NPOESS/NPP) via the High Data Rate X-Band Datalink.
AMSAT-BR & LABRE - Satellite Amateur Workshop⌗
November, 2017 - Pardinho, São Paulo - Brazil — Introductory talk about Open Satellite Project software. From geostationary satellites to polar-orbiting X-Band satellites.
Campus Party Brasil 11
February, 2018 - Anhembi Convention Center, Brazil — Talk with Lucas Teske about the software stack provided by the Open Satellite Project to decode, demodulate and parse weather satellite data from sun-synchronous or geostationary orbits.