I'd rather read the prompt (claytonwramsey.com)

271 points by claytonwramsey 3h ago 183 comments

Design for 3D-Printing (blog.rahix.de)

281 points by q3k 4h ago 75 comments

Graceful Shutdown in Go: Practical Patterns (victoriametrics.com)

17 points by mkl95 1h ago 2 comments

The New Control Society (thenewatlantis.com)

16 points by Michelangelo11 1h ago 7 comments

Dummy's Guide to Modern LLM Sampling (rentry.co)

125 points by nkko 5h ago 23 comments

The complicated business of electing a Doge (theballotboy.com)

40 points by dr_dshiv 3h ago 20 comments

Helmdar: 3D Scanning Brooklyn on Rollerblades (owentrueblood.com)

4 points by todsacerdoti 34m ago 0 comments

Thunderscope update: My take: Why open source is better (crowdsupply.com)

19 points by ChuckMcM 3d ago 1 comments

KaiPod Learning (YC S21) Is Hiring VP of Engineering (ycombinator.com)

1 points by amarkumar81 1h ago 0 comments

Typed Lisp, a Primer (alhassy.com)

34 points by todsacerdoti 5h ago 4 comments

An Alabama landline that keeps ringing (oxfordamerican.org)

239 points by bookofjoe 10h ago 81 comments

TScale – distributed training on consumer GPUs (github.com)

106 points by zX41ZdbW 8h ago 23 comments

Orders of Infinity (terrytao.wordpress.com)

46 points by matt_d 4h ago 5 comments

Minimal Linux Bootloader (2018) (raw.githubusercontent.com)

47 points by 1vuio0pswjnm7 5h ago 13 comments

LLMs as Unbiased Oracles (jazzberry.ai)

20 points by MarcoDewey 3h ago 10 comments

Critical Program Reading (1975) [video] (youtube.com)

21 points by blahaj 4h ago 3 comments

Ask HN: Hackathons feel fake now – anyone else noticing this?

21 points by sepidy 1h ago 12 comments

Load-Store Conflicts (zeux.io)

68 points by ashvardanian 5h ago 5 comments

In kids, EEG monitoring of consciousness safely reduces anesthetic use (news.mit.edu)

104 points by LorenDB 4d ago 59 comments

Evidence of controversial Planet 9 uncovered in sky surveys taken 23 years apart (space.com)

135 points by spchampion2 2d ago 106 comments

Oberon Pi (pascal.hansotten.com)

151 points by tosh 13h ago 25 comments

Cyborg cicadas play Pachelbel's Canon (arstechnica.com)

3 points by tomrod 1d ago 1 comments

Why Archers Didn't Volley Fire (acoup.blog)

66 points by StefanBatory 1d ago 13 comments

Technical analysis of TM SGNL, the unofficial Signal app Trump officials used (micahflee.com)

72 points by micahflee 1d ago 14 comments

AI code is legacy code? (text-incubation.com)

123 points by krrishd 4h ago 123 comments

Gorgeous-GRUB: collection of decent community-made GRUB themes (github.com)

402 points by todsacerdoti 23h ago 110 comments

Show HN: EZ-TRAK Satellite Hand Tracking Suite (github.com)

32 points by benbojangles 6h ago 5 comments

The Speed of VITs and CNNs (lucasb.eyer.be)

55 points by jxmorris12 2d ago 13 comments

Show HN: Driverless print server for legacy printers, profit goes to open-source (printserver.ink)

17 points by ValdikSS 4h ago 0 comments

The Merovingians: 'Do-Nothing Kings'? (historytoday.com)

8 points by Thevet 1d ago 2 comments

What went wrong with wireless USB (oldvcr.blogspot.com)

300 points by goldenskye 20h ago 174 comments

How Riot Games is fighting the war against video game hackers (techcrunch.com)

33 points by badmonster 3h ago 75 comments

Predictions from the METR AI scaling graph are based on a flawed premise (garymarcus.substack.com)

43 points by nsoonhui 15h ago 22 comments

A visual feast of galaxies, from infrared to X-ray (esa.int)

71 points by giuliomagnifico 2d ago 3 comments

A Texan who built an empire of ecstasy (texasmonthly.com)

83 points by wallflower 9h ago 81 comments

Why Flatpak apps use so much disk space on Linux (ostechnix.com)

56 points by dxs 7h ago 96 comments

Legendary Bose Magic Carpet Suspension Is Finally Going Global (thedrive.com)

255 points by PaulHoule 4d ago 161 comments

Reading Zanzibar (macwright.com)

16 points by surprisetalk 1d ago 2 comments

Feather: Feather: A web framework that skips Rust's async boilerplate and jus (github.com)

32 points by todsacerdoti 6h ago 38 comments

When flat rate movers won't answer your calls (aphyr.com)

353 points by kevincox 1d ago 291 comments

Metagenomics test saves woman's sight after mystery infection (bbc.co.uk)

190 points by neversaydie 4d ago 66 comments

Anyon_e: A highly integrated, high end, open source laptop (github.com)

9 points by rishikeshs 4h ago 1 comments

Pascal for Small Machines (pascal.hansotten.com)

106 points by ibobev 20h ago 20 comments

Show HN: I taught AI to commentate Pong in real time (github.com)

183 points by pncnmnp 2d ago 50 comments

Tippy Coco: A Free, Open-Source Game Inspired by Slime Volleyball (tippycoco.com)

55 points by stefankuehnel 10h ago 14 comments

Show HN: MP3 File Editor for Bulk Processing (cjmapp.net)

20 points by cutandjoin 22h ago 14 comments

Distributed Continuous GPU Profiling (zymtrace.com)

10 points by tdullien 2d ago 2 comments

Why I Am Not Going to Buy a Computer (1987) [pdf] (classes.matthewjbrown.net)

173 points by bookofjoe 1d ago 124 comments

Brian Eno's Theory of Democracy (programmablemutter.com)

157 points by akkartik 17h ago 164 comments

Lilith and Modula-2 (astrobe.com)

53 points by kristianp 10h ago 7 comments

Load-Store Conflicts

68 ashvardanian 5 5/4/2025, 5:18:42 PM zeux.io ↗

Comments (5)

Sesse__ · 3h ago

I find Clang generally a bit too eager to combine loads. This is especially bad when returning structs through the stack; you typically write them piecemeal in some function, return, and then the caller often wants to copy it from the stack into somewhere else, which it does with SIMD loads/stores.

This is a significant problem on AMD; Intel and Apple seems to be better.

cwzwarich · 3h ago

> This is a significant problem on AMD; Intel and Apple seems to be better.

When did this change? In my testing years ago (while I was writing Rosetta 2, so Icelake-era Intel), Intel only allowed a load to forward from a single store, and no partial forwarding (i.e. mixed cache/register) without a huge penalty, whereas AMD at least allowed partial forwarding (or had a considerably lower penalty than Intel).

Sesse__ · 3h ago

I don't know if AMD allows more or fewer _situations_, but empirically, I'm seeing a lot of total cycles lost to this on Zen 2 and 3, and much less on the Intel CPUs I've been testing (mostly Skylake derivatives and Alder Lake).

I haven't tested Zen 4 or 5, but I haven't heard anything that indicates they should be a lot better.

cwzwarich · 2h ago

Interesting! IIRC, the LLVM passes dedicated to dodging this issue were contributed by Intel engineers, so maybe there’s some bias.

haberman · 4h ago

A very interesting article that goes deeper into store-to-load forwarding than anything I’ve read before.