Beating the compiler
In modern times, everyone knows that writing assembly is a fool's errand: compilers are the result of literal engineer-centuries of work, and they know the processor much better than you do. And yet – one hears rumors. Written in ancient tomes, muttered in quiet watering holes, scrawled on the walls of bygone temples, hinted at by mysterious texts; the rumors paint a specific picture: Compilers ar
Do Not Taunt Happy Fun Branch Predictor
I've been writing a lot of AArch64 assembly, for reasons. I recently came up with a "clever" idea to eliminate one jump from an inner loop, and was surprised to find that it slowed things down. Allow me to explain my terrible error, so that you don't fall victim in the future. A toy model of the relevant code looks something like this: float run(const float* data, size_t n) { float g = 0.0; while
The Solid-State Register Allocator
I recently found myself needing a very fast register allocator for very simple code. There's a huge amount of academic research on the subject, dating back decades, and just as many published implementations; my specific implementation – while not fundamentally novel – landed in an interesting-enough part of configuration space that it seemed worth writing up. The system has one big limitation: it
From Oscilloscope to Wireshark - A UDP Story
Physical Like many of you, I've got hardware on my desk that's sending UDP packets, and the time has come to take a closer look at them. Most "low-level" networking tutorials will bottom out somewhere at "use tcpdump to see raw packets". We'll be starting a bit lower in the stack; specifically, here: This is a high-speed active differential probe soldered to an Oxide Computer Company rack switch.
It Can Happen to You
A few days ago, a fascinating article about Grand Theft Auto Online made the rounds of the tech news ecosystem. I'd encourage you to read the whole thing, but in short, GTA Online had accidentally quadratic performance when parsing a large JSON blob (due to repeated calls to strlen); fixing this improved loading time by almost 70%. This sparked a great deal of discussion: Was this C's fault? Perha
Constraint Solver
Motivation In CAD software, we often want to express constraints, e.g. Point a is 10 cm from point b Line g is perpendicular to line h Line t is tangent to arc c Drag points in the systems below and see what constraints are obeyed: In general, constraints can be expressed as systems of equations. Given a system of equations, e.g. $$ (2x + 3y) \times (x - y) = 2 $$ $$ 3x + y = 5 $$ How would we go
2D Contouring
About The CAD systems that I design use functional representations to represent solid models. These f-reps need to be converted into meshes before they can be used in other programs (e.g. to be manufactured on a 3D printer). There are many meshing algorithms in the literature, but we want one that is: Watertight (without any cracks) Hierarchical (avoids many tiny triangles) Feature-preserving (cor
Dataflow Engine
overview A CAD program like Antimony has to keep track of many small pieces of data. For example, a 2D point has x and y coordinates: One of the major design decisions in Antimony is that every datum is a piece of code. This means we can put arbitrary expressions into the point's coordinates. Object properties are also free to refer to each other, with the graph acting as a local namespace. Given
1
リリース、障害情報などのサービスのお知らせ
最新の人気エントリーの配信
j次のブックマーク
k前のブックマーク
lあとで読む
eコメント一覧を開く
oページを開く
A Neat XOR Trick
Program synthesis with Z3
Pixel-Art Physics