stabbles

19 ms·

301.
by stabbles 3y ago
Rust's borrow checker could probably be used for alias analysis, and avoid emitting redundant load instructions. That would give it a competitive advantage to C++. Does something like this exist yet?
302.
by stabbles 3y ago
That's equally problematic... '--progress=yes please' with whitespace runs into problems. In Bash you can use arrays though: args+=('--progress=yes please')
303.
by stabbles 3y ago
read one more sentence of the article
304.
by stabbles 3y ago
> Many people have asked for research into this for I don't know 20 years Sometimes I wonder if people in machine learning ever look at literature. Basic iteration schemes like the secant method (ok, 1-dimensional) have been known w
305.
by stabbles 3y ago
If you solve f(x) = 0 where f: ℝ^k -> ℝ^k maps vectors to vectors, most schemes are based on Taylor expansion around x_n f(x) = f(x_n) + Df(x_n)(x - x_n) + O(||x - x_n||^2) Drop the quadratic term, equate to 0, and you get an ap
306.
by stabbles 3y ago
> Python arguments are evaluated when the function definition is encountered This is a giant pain. Easy to miss. Sometimes forces you to deal with Optional[Something] instead of just Something. Compare with Julia where default arguments
307.
by stabbles 3y ago
Another fun `ldd` fact: some executables can be run, even when ldd says its libraries cannot be located. This is thanks to $ORIGIN rpaths. 1. Create an executable in `./bin/exe` with say `$ORIGIN/../lib` rpath, and link
308.
by stabbles 3y ago
(author of libtree): I believe that downloading, compiling and running libtree is faster than the startup overhead of pax-utils lddtree ;) maybe C is the better scripting language? $ alias latest_libtree='curl -Lfs https:/&#
309.
by stabbles 3y ago
I think that sums it up nicely. To add to that: Spack's solver effectively generates these 'concrete' packages for you. In the example, suppose you're packaging `X`, it depends on already existing packages `A` and `B`, a
310.
by stabbles 3y ago
Maybe I'm misunderstanding, but my impression is that Spack is more composable: In Spack pkg A can say `depends_on("curl tls=gnutls")`, pkg B can say `depends_on("curl +nghttp2")`, and if you then request a package
311.
by stabbles 3y ago
The main difference is that Spack has a single definition of a package, and in Spack language `curlMinimal`, `curlWithGnuTls`, and `curlHTTP3` are all "concretizations" of curl. That has the advantage that dependents can say `depe
312.
by stabbles 3y ago
It's interesting how Nix's selling point is that its non-standard file system structure allows multiple versions/flavors of the same package to be installed at the same time, yet if I understand correctly, it often has one ve
313.
by stabbles 3y ago
The lesson here is different: package developers must put upperbounds on major versions of dependencies that follow semver.
314.
by stabbles 3y ago
My impression is that the Python ecosystem rarely specifies upperbounds on dependencies even if they follow semver. In the Julia ecosystem it's the default, and you basically have to release a new patch version for updated compat bound
315.
by stabbles 3y ago
In Spack [1] we can express all these constraints for the dependency solver, and we also try to always re-cythonize sources [2]. The latter is because bundled cythonized files are sometimes forward incompatible with Python, so it's bet
316.
by stabbles 3y ago
You're linking to an article about brotli. I don't think gzip is affected by it. If anything, its decompression should be faster given fewer bytes to process. The usual reason to stick to gzip level 4 is that compression gets ex
317.
by stabbles 3y ago
GNU parallel devs really want you to cite their work... it's just a tool folks, it's like citing the manual of a hammer.
318.
by stabbles 3y ago
I think you have to remove escaping: `\*` should be `*`
319.
by stabbles 3y ago
No official download of sources/binaries available yet
320.
by stabbles 3y ago
Maybe hackernews would benefit from an LLM that refuses appeal to authority type of arguments
321.
by stabbles 3y ago
CSS has a feature creep that makes it harder and harder to write a new browser from scratch
322.
by stabbles 3y ago
In a way SO is the perfect platform for human feedback in reinforcement learning
323.
by stabbles 3y ago
This is a very common theme, same with Python 2. It's EOL, but in practice it's really not, because distros have an independent EOL date and cannot make breaking changes.
324.
by stabbles 3y ago
Even if true it's unnecessary. I just can't read anything written by Wolfram without rolling my eyes
325.
by stabbles 3y ago
Also works with zlib-ng
326.
by stabbles 3y ago
Is there any difference between "-ffunction-sections -fdata-sections -Wl,--gc-sections" and gcc's -fwhole-program?
327.
by stabbles 3y ago
LD_PRELOAD is too global to be useful, it's hard to scope it to one process (and not child processes). macOS is better in the sense that it clears DYLD_* variables when the dynamic linker has done its work and the process starts. (Alth
328.
by stabbles 3y ago
I agree that you shouldn't dismiss an airplane by the fact that it doesn't flap its wings, but we also don't call them artificial birds. The debate is around the word 'intelligence', which is misleading.
329.
by stabbles 4y ago
The only thing that's missing is a custom repeated background image like it's myspace
330.
by stabbles 4y ago
In theory of cache oblivious algorithms this is modeled; for optimality the complexity should also be independent of block / cache size

More ›