pervognsen

wingo, 12 days ago to random

they have been playing us for fools &c &c &c

joew, 12 days ago to random

Turns out intermittently restarting a server with stochastic service times operating on a queue can improve service times and fairness. I love a good, simple and counter intuitive solution.

“To implement resetting, imagine this server being stopped intermittently and then restarted – thus, jobs whose service has been reset are now assigned fresh service times.”

https://arxiv.org/abs/2404.08961

tobinbaker, 11 days ago to random

Just noticed that Github code search doesn't work anymore if you're not logged in. Enshittification proceeds apace.

pervognsen, 12 days ago to random

Zen 5 announced: https://www.anandtech.com/show/21415/amd-unveils-ryzen-9000-cpus-for-desktop-zen-5-takes-center-stage-at-computex-2024

chandlerc, 12 days ago to random

Playing with "overflow byte" approach to metadata+simd augmented hashtable design.

I've read that this is part of the F14 design, but I found the explanations for Boost's new unordered_flag_map to be what got the point across.

I implemented the core of this in my own way on top of my hashtable -- rather than stealing a byte from the metadata the way Boost does, I just tack on an array of 1-byte-per-group "probe markers" (a term that makes more sense to me than "overflow byte".

pervognsen, 6 days ago to random

@dneto Am I reading this correctly as WebGPU not defining a fill rule? Hopefully just an oversight. "Fragments are associated with pixel centers. That is, all the points with coordinates C, where fract(C) = vector2(0.5, 0.5) in the framebuffer space, enclosed into the polygon, are included. If a pixel center is on the edge of the polygon, whether or not it’s included is not defined. NOTE: this becomes a subject of precision for the rasterizer."

pkhuong, 11 days ago to random

After ~4 years, I actually handled poisoned locks for the first time! I have a sequence number + 2 copies thing for lock-free snapshots, but writers must serialise. The write mutex doesn't really protect any invariant, except that we mustn't interleave write sequences. Interruption is fine, as long as the interrupted writer never resumes.

pervognsen, 7 days ago (edited 2 days ago) to random

The top-left fill rule in rasterization is lexicographic:

is_left = dy > 0
is_top = dy == 0 && dx > 0
is_top_left = is_left || is_top

This is equivalent to (dy, dx) > (0, 0) with lexicographic ordering.

pervognsen, 7 days ago (edited 1 day ago) to random

While I'm commenting on rasterization, you can do perspective-correct interpolation with non-normalized barycentric coordinates (a, b, c) by interpolating (a/w, b/w, c/w) and then normalizing the barycentric coordinates per pixel by dividing by their sum. The conventional approach is to interpolate (a/w, b/w, 1/w) and then you can solve for c in terms of a and b if needed. The conventional approach is slightly more efficient. You get a reference scale: either a + b + c or 1.

pervognsen, 8 days ago to random

Precompiling shaders...

pervognsen, 9 days ago (edited 6 days ago) to random

This is anecdotal but I don't think I've run a high-end Windows game yet on Linux/Proton that didn't run better than it did on Windows. It's kind of silly.

pervognsen, 6 days ago (edited 5 days ago) to random

Left-to-right top-down parsers prefer separators over terminators. Consider:

// (',' item)*
while token is comma {
parse_comma()
parse_item()
}

// (item ';')*
while token in first(item) {
parse_item()
parse_semicolon()
}

The terminator does not play an active role in predictive parsing. It only exists to prevent first/follow conflicts for item. (It's also useful for Wirth-style error recovery.)

pervognsen, 6 days ago to random

Pointers to pointers are good for you. Here's how to build a DFS tree:

void push(void) {
*path++ = node;
*link = node = end++;
link = &node->down;
}

void pop(void) {
*link = 0;
link = &node->next;
node = *--path;
}

pervognsen, 6 days ago (edited 3 days ago) to random

I was randomly reminded of this by a stray mention of Sebastian Thrun. Back when Udacity started one of the first courses was Peter Norvig's Design of Computer Programs. It was great. Basically a whole course of short, well-produced video walkthroughs of problems and solutions in the vein of his classic https://norvig.com/spell-correct.html combined with programming exercises. Highly recommended: https://www.udacity.com/course/design-of-computer-programs--cs212

svat, 12 days ago to random

Wrote up a quick post mentioning Don Knuth's latest toy program he put up on his website: https://shreevatsa.net/post/dek-partition-square/ (Mostly just some context and explaining what the problem is; for anyone who'd like to just read the actual program directly, I have a typeset version at: https://shreevatsa.github.io/knuth-literate-programs/programs/perfect-partition-square.pdf )

pervognsen, 12 days ago to random

No need to fear, the optimizing compiler is here:

mov r8d, 1
test r8, r8
jne .LBB1_2

pervognsen, 14 days ago to random

I switched from Windows to Linux not directly because of the most recent shit show with Recall (though it's part of the general trend). But it sure does make me feel even better about making that choice when I'm struggling with some minor annoyance on Linux.

pervognsen, 9 days ago (edited 6 days ago) to random

The fact that plugging a PS4 controller into my computer changes the default sound output device to the tiny speaker in the controller is very confusing for a few seconds whenever it happens.