eapriv

577 karma · joined July 27, 2012

eapriv··on No elephants: Breakthroughs in image generation

It’s always fun to read posts like that: they say “look at this amazing thing it drew”, and the image is utter garbage.

Caution: this seems to be an ad for "quadratic", which promises "The spreadsheet with AI". I'm sure it will turn out much better than Excel, a spreadsheet without "AI".

eapriv··on The order of files in /etc/ssh/sshd_config.d/ matters

Of course the order matters, that’s why the file names have numbers in them.

eapriv··on A 10x Faster TypeScript

Yet people wouldn’t write a Fortran compiler in Fortran, or a MATLAB compiler in MATLAB.

eapriv··on A 10x Faster TypeScript

“Nice” doesn’t mean “well suitable for writing a compiler in”. It’s strange to think that all languages should be equally good for writing all kinds of things, and choosing a web language for a non-web task is doubly strange.

eapriv··on A 10x Faster TypeScript

It’s not obvious from the text, but the compiler was previously written in TypeScript (which was kind of a strange choice for the language to write a compiler in).

eapriv··on The Meter, Golden Ratio, Pyramids, and Cubits, Oh My

Or there are no “coinciding measurements”, and someone intentionally selected some random dates and numbers based on the sizes of the existing kiosk?

eapriv··on The Meter, Golden Ratio, Pyramids, and Cubits, Oh My

Lots of meaningless numerology, so what’s the point?

eapriv··on Discord client that works on Win95*, Win98 and above

I’m pretty sure they didn’t mean “exponentially more” in the asymptotic complexity sense. They could be both O(N log(N)), but with dramatically different constants.

eapriv··on Optimizing uint64_t Digit Counting: A Method that Beats Lemire's by up to 143%

Isn’t “the ratio of logs” going to be slower?

eapriv··on CUDA Moat Still Alive

Schönhage-Strassen is not about matrix multiplication.

eapriv··on CUDA Moat Still Alive

For matrix multiplication? How?

eapriv··on Emulating the FMAdd Instruction, Part 1: 32-bit Floats

It doesn’t even compile (the last code snippet uses some undefined variables).

eapriv··on CUDA Moat Still Alive

> FFT based matrix multiplication, which is O((n^2)log(n))

What?

eapriv··on Against Best Practices

The name “best practices” kind of implies that they actually are practiced somewhere. So it’s different from theoretical abstract ideas “how we should write software”, which maybe nobody follows.

eapriv··on Against Best Practices

Software around us is quite terrible, so it’s kind of obvious that one shouldn’t follow “best practices” that led to this state of things.

eapriv··on Memory64

Why would anyone do that?

eapriv··on Wild Ball

It’s not “optimized for desktop”, it only works on desktop. I think it’s hilarious when something plainly doesn’t work and people say “oh, it’s not optimized for this case”. That’s now what “optimization” means.

eapriv··on Battleships Logic Puzzle

There’s a bug: if a puzzle has multiple solutions, only one of them is considered “correct”.

eapriv··on Calculating the norm of a complex number

No need to apologize, 0 == -0 even if we are talking about floats.

eapriv··on John Carmack on inlined code (2014)

“Comment threat” is a nice one.

eapriv··on AVX Bitwise ternary logic instruction busted

It is actually supported since .NET 8.

eapriv··on AVX Bitwise ternary logic instruction busted

It’s “ternary (logic instruction)”, not “(ternary logic) instruction”.

eapriv··on Dissecting the gzip format (2011)

Yes. https://caseymuratori.com/blog_0015