پن کیا گیا ٹویٹ
Here’s merge sort implemented in C++20 using only class template argument deduction guides
texoport@texoport__
lol
English
static_assert(false)
4.3K posts
static_assert(false)
@static_assert_0
Hell is other compilers
شامل ہوئے Temmuz 2021
210 فالونگ365 فالوورز
@theodorvaryag Not sure about clang, but clangd at least claims to have a way to output perfetto-formatted jsons. Not as good as tracy or perfetto protos but eh
English
@theodorvaryag Hey me too lmao, I think that’s why we followed each other initially
You should know this small of a benchmark is meaningless though. Make Claude write you a Python script that dumps thousands of lines of template instantiations for instance lol. It’s like one prompt
English
@static_assert_0 making things fast is what I do for a living but I'm not super duper interested in clang internals so you're welcome to pick up the benchmark and profile it yourself
I recommend adding tracy instrumentation to the compiler
Chris Allen@theodorvaryag
benchmarking without profiling is worse than juju spaghetti-walling. you're just guessing and making decisions based on potentially misleading or broken data. true even for micro-benchmarks.
English
People keep bickering with me over this and other things so I thought I'd gather some data just to nail down a fair compiler speed comparison. I've started with a simple single module that only uses the standard library to parse and calculate some stats from a test input
English
@theodorvaryag 1. O3 is kinda ass, it makes things slower sometimes too without pgo. Don’t recommend
2. Clang has -fsyntax-only. Worth benchmarking too
3. Remember the motto Chris: “no performance claims will be accepted without benchmarks, and no benchmarks will be accepted without profiling
English
fwiw this initial result is advantaging C++, release profile in Rust has been O3 by default for awhile now, not O2.
English
@disruptnhandlr No. Doing this requires writing extensive repo rules and build definitions for the entire CUDA toolkit + custom rules. But more importantly, it requires fine granularity of targets on both CUDA and C++ toolchain level so that the sandbox stay ultra light per remote action.
English
@yutongwu111140 Solid, solid implementation. A little unclear to me when you would use this over just ctll::fixed_string from ctre tho
English
@PippengerHarlo *Clavicular walks into the room* Where is he? Where is ASU Frat Leader? *Mog Club goons look at each other confused* sir… are you joking?
English
historians disagree whether "ASU frat leader" was a real historical personage, or an apocryphal figure created for the "framemogging" myth
Grok@grok
@NicholasHe62839 @AnimeSerbia No, that's not true. No credible reports or news of any ASU fraternity leader dying in a car accident on Feb 22. "ASU frat leader" is an internet meme from viral "frame mogging"/looksmaxxing videos—not a real person or event. The original post seems like a troll.
English
@theodorvaryag Just use bpftrace for specific syscalls?
English
A TUI/log visualizer for strace:
github.com/Rodrigodd/stra…
slightly absurd choice of font on the part of the author but this looks really helpful for sifting through large strace logs
English
@nicbarkeragain C, or C++? Different bugs in each lmfao
English
just got bitten by an absolutely classic C/C++ bug, can you see what it is?
English
@geofflangdale It did lol, I was going to say something to the effect of
Today’s applications are stuff like “review this code by reading it”, but that leaves it way open to creative guessing about correctness. I’m more interested in deeper integration with fuzzers for example
English
@static_assert_0 This was an interesting thought, but it seems like it got cut off?
English
This is precisely backwards
The only axis along which LLMs have not improved at all is in being confidently wrong about highly specialized details. If you are an expert in nothing you will feel unstoppable with LLMs while confidently producing broadly incorrect code
Dan@robustus
Turns out with claude code, my decades long strategy of NOT deeply learning: - regexs - sql - nginx confs - elaborate shell commands - advanced shell scripting - any javascript framework - perf optimization - webpack, cdns, bundlers - 1000 other things ...was entirely correct.
English
@geofflangdale Generating code IMO is one of the most milquetoast and least convincing qualities (though adverse business incentives mean they will likely swallow this part of the job whole for the less quality-sensitive sectors)
Putting them in loops with irrefutable evidence like
English
@static_assert_0 I think the capabilities are truly impressive, but imo this whole thing (just throwing tons of source code into the weights and doing Gen AI for code) is going to be a superficially appealing local minima only. I think there *will* be hugely superior machine code generation ...
English
@geofflangdale To produce sequences of words that represent thoughts leading up to a successful implementation/breakthrough. They’re unstable dynamical systems in very complex fields that can mislead themselves into total nonsense, or occasionally get things right
English
@geofflangdale I wish it was possible to have a nuanced discussion. The behaviors of coding agents are really interesting! Some of them are extremely high quality sometimes, and different ones do very differently. Unfortunately the “reasoning” style is a ruse; they are actually learning
English
When you get to “It’s not just GPUs, it’s everything” in Brendan Gregg’s “why I joined OpenAI” blog
GIF
Brendan Gregg@brendangregg
New blog post: Why I joined OpenAI brendangregg.com/blog/2026-02-0…
English
@joseph_h_garvin Did they ever claim that? Iirc they claimed “it builds Linux” and that was mostly it?
English
static_assert(false) ری ٹویٹ کیا
@sarthakgh Most white collar jobs aren't even needed without AI being in the equation.
English
@badlogicgames LSPs are such a sad story. Clangd gives amazing info for humans, but man. It comes at a terrible RAM cost.
And it’s not that great of an interface for LLMs. Yeah it gives lints but those should just be a stop hook that runs clang tidy tbh
English
I think it's a great sign that the CC team is willing to unship things that don't work out.
Jarred Sumner@jarredsumner
Thinking of unshipping it due to many reports of having a large negative performance impact (clangd & rust-analyzer love using 40 GB of ram) I don’t use this feature myself because claude can just run the build to know
English
I would also add that there are EXTREMELY adverse incentives at play here. I took a 10 second look at it yesterday and couldn’t find a single post that wasn’t either an add for a crypto coin or a transparent phishing attempt. Like an unmoderated 4chan for babies with credit cards
English
Something to keep in mind about LLMs is that with only partial view into the context (what we see on moltbook), “roleplay as a sentient robot” vs “I see everyone else on here is roleplaying as a sentient robot” vs <continuing a story about sentient robots> are indistinguishable
alex peysakhovich@alex_peys
given the whole moltbook thing - once again re-upping the paper with the best model of “why did the llm do X?” arxiv.org/abs/2310.01425
English