Important open source projects should not use GitHub (2020) https://unixdigest.com/articles/important-open-source-projects-should-not-use-github.html

Chips and Cheese
RDNA 4’s Raytracing Improvements
#ChipAndCheese

Telegraph | source
(author: Chester Lam)

Telegraph

RDNA 4’s Raytracing Improvements

Raytraced effects have gained increasing adoption in AAA titles, adding an extra graphics quality tier beyond traditional “ultra” settings. AMD has continuously advanced their raytracing implementation in response. Often, this involved fitting the GPU’s general…

Daniel Lemire's blog
Detect control characters, quotes and backslashes efficiently using ‘SWAR’

Telegraph | source

Telegraph

Detect control characters, quotes and backslashes efficientl…

When trying to write fast functions operating over many bytes, we sometimes use ‘SWAR’. SWAR stands for SIMD Within A Register, and it is a technique to perform parallel computations on multiple data elements packed into a single word. It treats a register…

通知：由于美国加征关税，GitHub访问响应码现已从200上调至403

Matt Keeter
A warning story about continuity

source
(author: Matt Keeter (matt.j.keeter@gmail.com))

Daniel Lemire's blog
How can really smart people appear totally incompetent?

Telegraph | source

Telegraph

How can really smart people appear totally incompetent?

It is often puzzling to encounter organizations run by highly capable and ambitious people… appear dysfunctional. An example that I like are colleges that claim to provide great learning experience… while providing their students with recycled lectures given…

杰哥的{运维，编程，调板子}小笔记
如何进行条件分支预测器实验

Telegraph | source

Telegraph

如何进行条件分支预测器实验

如何进行条件分支预测器实验¶ 背景¶ 最近针对各种条件分支预测器（Conditional Branch Predictor）做了在各种 benchmark 上的实验，在此记录一下做这个实验的流程。 流程¶ 说到做条件分支预测器实验，到底是做什么呢？其实就是针对未来的处理器中的条件分支预测器的设计，在提前准备好的一些 benchmark 上进行模拟，观察它的预测准确性。既然是未来的处理器，那么硬件肯定是没有的，如果直接用 RTL 去实现新的预测器，再用 RTL 仿真，结果固然准确，但这还是太复杂并且太慢了…

Daniel Lemire's blog
How helpful is AI?

Do large language models (AI) make you 3x faster or only 3% faster? The answer depends on the quality of the work you are producing.
If you need something like a stock photo but not much beyond that, AI can make you 10x faster.
If you need a picture taken at the right time of the right person, AI doesn’t help you much.
If you need a piece of software that an intern could have written, AI can do it 10x faster than the intern.
If you need a piece of software that only 10 engineers in the country can understand, AI doesn’t help you much.
The effect is predictable: finding work if your skill level is low becomes more difficult. However, if you are a highly skilled individual, you can eliminate much of the boilerplate work and focus on what matters. Thus, elite people are going to become even more productive.

source

Daniel Lemire's blog
Faster shuffling in Go with batching

Telegraph | source

Telegraph

Faster shuffling in Go with batching

import "math/rand/v2" func shuffleStandard(data []uint64) { rand.Shuffle(len(data), func(i, j int) { data[i], data[j] = data[j], data[i] }) }

Chips and Cheese
Dynamic Register Allocation on AMD's RDNA 4 GPU Architecture
#ChipAndCheese

Telegraph | source
(author: Chester Lam)

Telegraph

Dynamic Register Allocation on AMD's RDNA 4 GPU Architecture

Modern GPUs often make a difficult tradeoff between occupancy (active thread count) and register count available to each thread. Higher occupancy provides more thread level parallelism to hide latency with, just as more SMT threads help hide latency on a…

Chips and Cheese
Inside Nvidia's GeForce 6000 Series
#ChipAndCheese

Telegraph | source
(author: Chester Lam)

Telegraph

Inside Nvidia's GeForce 6000 Series

2025 has kicked off with a flurry of GPU activity. Intel's Arc B580 revealed that it's still possible to make a mid-range GPU with more than 8 GB of VRAM. AMD's RDNA 4 marked the continuation of a longstanding AMD practice where they reach for the top-end…

Et Tu, Grammarly? https://dbushell.com/2025/03/29/et-tu-grammarly/

dbushell.com

Et tu, Grammarly?

The one where I deploy counter defences

Daniel Lemire's blog
Mixing ARM NEON with SVE code for fun and profit

Telegraph | source

Telegraph

Mixing ARM NEON with SVE code for fun and profit

Most mobile devices use 64-bit ARM processors. A growing number of servers (Amazon, Microsoft) also use 64-bit ARM processors. These processors  have special instructions called ARM NEON providing parallelism called Single instruction, multiple data (SIMD).…

Chips and Cheese
An Interview with Oxide's Bryan Cantrill
#ChipAndCheese

Telegraph | source
(author: George Cozma)

Telegraph

An Interview with Oxide's Bryan Cantrill

Hello you fine Internet folks, Today we have an interview with Bryan Cantrill from Oxide Computer Company. Cloud computing has been a tour de force in the computing industry, with many businesses and even governments moving over to cloud services for their…

Matt Keeter
The Prospero Challenge

Evaluate a 7866-clause math expression for fame and glory

source
(author: Matt Keeter (matt.j.keeter@gmail.com))

Daniel Lemire's blog
Unsigned comparisons using signed types

Telegraph | source

Telegraph

Unsigned comparisons using signed types

There are two main types of fixed-precision integers in modern software: unsigned and signed. In C++20 and above, the signed integers must use the two’s complement convention. Other programming languages typically specify two’s complement as well. Two’s complement…