Internals refers to the inner workings of compilers or interpreters, detailing how they are structured and operate, documenting their internal structures such as the low-level runtime library, intermediate representations, control graph analysis and optimization, machine descriptions, and more.
The 1979 HP-41 series programmable calculator, produced until 1990, was famously used in the Space Shuttle missions. It featured an LCD alphanumeric display and was powered by a processor initially called "Coconut" (or 1LE3 CPU), later more broadly known as the Nut CPU. The calculator included RAM, ROM, and I/O capabilities, utilizing the interpreted programming language FOCAL ("Forty One Calculator Language"). Both the operating system and the interpreter were stored in ROM, programmed using the Coconut processor's assembly language, also called MCODE (or "M-Code"). A later upgrade, the NEWT Microprocessor, further enhanced its capabilities.
The site is a repository dedicated to the preservation and sharing of software and documentation for classical computer systems, including a wide variety of classical computer systems and their software, such as the CP/M operating system. The site hosts software and documentation of all types for classical computer systems, providing a valuable resource for enthusiasts and researchers interested in retrocomputing. Maintained by classiccmp.org community.
Very nice WebGL application to visualize the loss landscape for some common ANN. Currently features the models Resnet-20 (short/no-short), Resnet-56 (short/no-short), Vgg 16 and DenseNet 121.
Good post by Matthew Stewart's "Neural Network Optimization" from June 27, 2019.
https://towardsdatascience.com/neural-network-optimization-7ca72d4db3e0
Last edited: 2024-05-28
My personal notes about the seminar Using Physics-informed Neural Networks for Inverse Problems by João Pereira - IMPA at National Scientific Computing Laboratory (LNCC) on 2024-05-13.
The National Scientific Computing Laboratory (LNCC) and Eviden/Atos signed a new contract worth us$ 19.4 million, which will allow the machine to go from the current 5.1 Petaflop/s to 17 Petaflop/s of capacity. The technology will be based on the BullSequana XH3000 architecture, and with the expansion it will be the most powerful supercomputer in Latin America dedicated to academic studies.
https://agenciabrasil.ebc.com.br/geral/noticia/2024-04/supercomputador-mais-potente-do-pais-tera-capacidade-aumentada (in Portuguese)
LNCC SDumont website (in Portuguese): https://sdumont.lncc.br
Very nice Wikibook based on Parallel Spectral Numerical Methods by Chen et al. (2012) from University of Michigan. Discusses how to solve ordinary differential equations (ODE) and partial differential equations (PDE) using separation of variables. Next, it introduces numerical time-stepping schemes that can be used to solve ODEs and PDEs. This is followed by an introduction to pseudo spectral methods through an overview of the discrete Fourier Transform (DFT) and the Fast Fourier Transform (FFT) algorithm that is used to quickly calculate the DFT. Finally it will combine all of these to solve a couple of different PDEs first in a serial setting and then in a parallel setting. The programs will use Matlab and Fortran. A Python implementation of some of the Matlab programs is also provided.
https://en.wikibooks.org/wiki/Parallel_Spectral_Numerical_Methods
Some related links I found interesting (mostly in Portuguese):
SDumont support manual: https://sdumont.lncc.br/support_manual.php
LNCC Youtube channel: https://www.youtube.com/@LNCCbr
2015 Simple presentation including some photos of the physical construction of the facilities: https://www.slideshare.net/RobsondaCosta2/supercomputador-sdumont-top500org-146-2015
2016 Presentation. First SDumont architecture when it was first deployed: https://sdumont.lncc.br/media/01_General_overview_of_SANTOS_DUMONT_architecture.pdf
More info:
Zenodo is a general-purpose open repository developed under the European OpenAIRE program and operated by the European Organization for Nuclear Research (CERN) and allows the deposit of research articles, datasets, research software, reports and any other related digital artifacts to research. For each submission, a persistent digital object identifier (DOI) is created, which makes stored items easily citable. It also allows the deposit of data sets of up to 50 GB. One supported source is GitHub repositories. It is hosted on CERN's high-performance computing infrastructure that is operated primarily for the needs of high-energy physics.
Lecture describing the use (in Portuguese): https://youtu.be/6tum4jq4mNs