Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
the8472
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
51 ms
·
841.
▲
by
the8472
4y ago
Is there a way to run podman as non-root but with a privileged helper (or some caps) for proper bridged network setup?
842.
▲
by
the8472
4y ago
Note that that kind of research was only possible on the (smaller) open models. When the new models are proprietary ones then outsiders can't do that kind of analysis. If they had regenerated the tokenization dictionary to match the tr
843.
▲
by
the8472
4y ago
> At that point, does it keep scaling or is there an S curve where 100x more data and compute only leads to a 2x improvement? Careful with the scales. 2x improvement could be interpreted from 80% human performance to 160% human performan
844.
▲
by
the8472
4y ago
Some isotopes do glow in the dark if sufficiently enriched, due to their decay heat. https://en.wikipedia.org/wiki/File:Plutonium_pellet.jpg
845.
▲
by
the8472
4y ago
That PR is about buggy or niche systems. VMs lying about their clocks, CPUs with faulty firmware revisions, kernel bugs. Rust is just going the extra mile to uphold its guarantee even when the OS fails to do its job. It would have been legi
846.
▲
by
the8472
4y ago
Few things need atime at all, even fewer things need synchronous durability guarantees for atime (I'm not aware of any). So writing it out in the background (either when evicting disk cache or after 24h) while keeping the in-memory rep
847.
▲
by
the8472
4y ago
Primary source (or at least one of several published versions): https://cse-robotics.engr.tamu.edu/dshell/cs625/Thompson96Ev...
848.
▲
by
the8472
4y ago
> It's not only about performance, but also about all the guarantees that you get "for free" Note that some filesystems have mount options that let you remove some of those guarantees.
849.
▲
by
the8472
4y ago
Basic atomic filesystem operations support consensus numbers[0] 1 or 2 (via rename and RENAME_EXCHANGE, respectively), but for some concurrent algorithms one needs higher ones. File locks are the only way to get those (if we ignore IPC), an
850.
▲
by
the8472
4y ago
That advice is mostly obsolete, lazytime has been a thing for years.
851.
▲
by
the8472
4y ago
Windows is slow due to all the filter drivers used by antivirus, custom FS drivers and other stuff. And the lack of a dentry cache. Basically all the reasons why WSL1 is slow. On linux you'll get slowdowns too if you have some online a
852.
▲
by
the8472
4y ago
A recent change[0] on nightly rustc might help with incremental builds. And for repeated clean + full build cycles there's sccache[1]. [0] https://github.com/rust-lang/rust/pull/84762 [1] https:/&#
853.
▲
by
the8472
4y ago
If one uses an offline password manager then you want the stored passwords to be approximately as secure as memorized passwords. So how do you deal with emergency family access to memorized information in the deceased person's brain? S
854.
▲
by
the8472
4y ago
If numpy uses runtime detection of available SIMD instructions while rust is only compiled with the x86-64 baseline (which only includes SSE2) then compiling the module with `RUSTFLAGS=-Ctarget-cpu=native` might provide some additional perf
855.
▲
by
the8472
4y ago
Argon2 can be used as KDF, not just as a password hash in an authentication database.
856.
▲
by
the8472
4y ago
Use an offline password manager, avoid SaaS. Make sure it's secured with Argon2 and burns a a few seconds of compute to unlock. Use hardware MFA tokens for your most critical accounts. You could also tie individual copies of the databa
857.
▲
by
the8472
4y ago
Wasm doesn't necessarily help if it's performance-critical because that comes with its own overhead. And the kernel component of ksmbd interacts with other parts of the kernel such as the vfs, sockets/rdma. If you had marshal
858.
▲
by
the8472
4y ago
I didn't write "I think" or "I believe". If you want to know my epistemic status: I know about windows based on observable surface behavior and bugs related to accelerated cursors, but I haven't looked at the s
859.
▲
by
the8472
4y ago
JSON cannot be efficient for binary data or even utf8 strings because it needs escaping which means you can't have a zero-copy implementation. CBOR is better in that regard, yeah.
860.
▲
by
the8472
4y ago
Having C functions isn't all that much better. You have replaced a crufty text format with crufty data structures full of paddings, unions, bitfields, VLAs, unaligned nested structs and other crazy stuff. Look at ioctls or cmsg. With C
861.
▲
by
the8472
4y ago
This is fairly basic functionality. Windows and wlroots-based compositors have it at least. I'm reasonably certain that other major compositors have it too.
862.
▲
by
the8472
4y ago
Why stop at a containment building ? It's economical to build underground mines to extract coal and minerals. Might as well build mines to house nuclear reactors in stable geologic structures. If it goes wrong seal it and return a hun
863.
▲
by
the8472
4y ago
Things that are technically necessary to facilitate functions requested by the user like shopping carts or login tokens are exempt under clause 22 of the ePrivacy directive. Don't set any other cookies and you won't need to ask fo
864.
▲
by
the8472
4y ago
$employer procures workstations equipped with 64C Threadripper Pro + A6000 48GB GPUs for CG artists. They haven't said what they're actually paying but list prices of those machines are north of 15k. Devs get a similar setup excep
865.
▲
by
the8472
4y ago
Many data-/number-crunching C libraries already do dynamic-link-time or runtime cpu feature detection and select appropriate implementations. E.g. glibc already has bunch of specialized assembly impls for many of the string operations
866.
▲
by
the8472
4y ago
Many string search impls are adaptive and use different algorithms depending on needle and haystack lengths and possibly also contents of the needle. E.g. https://github.com/BurntSushi/memchr#user-content-algorithms...
867.
▲
by
the8472
4y ago
The upside is that babies get to interact with the environment they're training on. Image models can't move the camera a few cm to the right if they're interested in the perspective of a particular scene.
868.
▲
by
the8472
4y ago
It's not like you shove a blank wafer into the machine and finished chips pop out. ASML's machines only do the exposure step.
869.
▲
by
the8472
4y ago
Some of the AVX512 instructions are nice additions though, even if you only use it on YMM or XMM registers. All the masked instructions make it much simpler to vectorize loops over variable amounts of data without special treatment for th
870.
▲
by
the8472
4y ago
That's pretty much the background plot of La Forge de Vulcain https://www.goodreads.com/book/show/1768022.La_forge_de_Vulc...
More ›