m0th87
Joined 17 years ago
Karma
4,936
Submissions
771
Member since
Apr 21
Recent submissions
Comment · 1 year ago
https://www.intel.com/content/www/us/en/docs/intrinsics-guid...
Intel's docs are unfortunately spartan, but the guarantees around program order is a hint that this is what it does.
Comment · 1 year ago
I see, thanks. I had assumed incorrectly that NT writes operated the same as NT accesses, where there is no dedicated cache.
Comment · 1 year ago
There are no guarantees even if everything operates on the same core. Rust docs have some details: https://doc.rust-lang.org/stable/core/arch/x86_64/fn._mm_sfe...
Comment · 1 year ago
I had interpreted GP to mean that you don’t slap on NTs for correctness reasons, rather you do it for performance reasons.
Comment · 1 year ago
I work on optimizations like this at work, and yes this is largely correct. But do you have a source on this?
> or (more likely) go into just some special small subsection of it reserved for non-temporal writes only.
I hadn’t heard of this before. It looks like older x86 CPUs may have had a dedicated cache.
Comment · 1 year ago
That’s what I hope for, but everything that isn’t bananas expensive with unified memory has very low memory bandwidth. DGX (Digits), Framework Desktop, and non-Ultra Macs are all around 128 gb/s, and will produce single digits tokens per second for larger models: https://github.com/XiongjieDai/GPU-Benchmarks-on-LLM-Inferen...
So there’s a fundamental tradeoff between cost, inference speed, and hostable model size for the foreseeable future.
Comment · 2 years ago
This is almost perfect for my needs (communication and location sharing in areas without phone signal.) But it's missing GPS/GNSS. I'm not sure how niche that use case is, so more broadly I wish Lilygo offered something akin to Raspberry Pi hats -- i.e. some way to add extensions.
Comment · 3 years ago
I’ve heard Beaglebone Blacks are great (never used them), but have to say the experience with Blue was awful. I had a hardware defect that’s apparently commonplace. It gave me the feeling (maybe unjustified) that there’s a low bar for slapping the Beaglebone label on a product. In contrast, none of my several RPi’s have had any hardware issues, and I’m pretty brutal with them.
Comment · 3 years ago
I’m definitely interested! Was just looking over this code and wishing it were in rust or C.
Comment · 4 years ago
That's interesting, I didn't know they did that. But it could be better:
1) This alert doesn't show up at all on mobile.
2) IMO this should either affect the overall score or be displayed in search results as well, which is not currently the case: https://www.glassdoor.com/Search/results.htm?keyword=ZURU
Comment · 4 years ago
For what it’s worth, the person you’re replying to is one of the heavyweights of the rust community.
Comment · 4 years ago
Just whipped up a (naive) solve in rust: https://gist.github.com/ysimonson/01a1dee41b1b5990c30568fd25...
It usually solves this in under 6 guesses. The guessing could be improved; at the moment it's random, but it could select for words with non-repeating letters to narrow down the search space faster.
Comment · 5 years ago
R2 is marketed as "one less than S3!"
I used the exact same tagline for a project open sourced in 2019: https://github.com/pachyderm/s2/
So there you go VCs, I can market as well as cloudflare. Money please!
Comment · 5 years ago
I messed around with and contributed to this project a bit. The biggest challenge I saw in day to day usage was that line numbers aren't included in error messages: https://github.com/murarth/ketos/issues/11
Comment · 5 years ago
I wrote up some rust low-level bindings for guile [1]. The problem is that Guile liberally uses setjmp/longjmp, which breaks rust destructors. It might be possible to fix this by wrapping every call the way rlua does [2], but I'm not familiar enough with Guile to know.
Email (ROT13™): fvzbafba ng tznvy qbg pbz