tavert

12 ms·

61.
by tavert 9y ago
A subset of non-performance-critical functions, and things like argument validation layers, are written in m code that you can read. Most of the actual work happens in obfuscated p code or calls into compiled libraries. Many of those librar
62.
by tavert 9y ago
Fedora (and other rpm-based distros, OpenSUSE in particular) are easier to build/modify packages for than deb-based distros and have easier to use build services with better automation, IMO.
63.
by tavert 9y ago
JuMP, in Julia: https://github.com/JuliaOpt/JuMP.jl/blob/master/examples/bas... It's rapidly displacing proprietary DSL's like AMPL and GAMS in operations research. Neither of those happen
64.
by tavert 9y ago
Numpy is good for vectorized calculations on dense multidimensional arrays of homogenous 64 or 32 bit floating point numbers, or integers. If that describes your problem and you like Python, by all means use numpy. If that doesn't desc
65.
by tavert 9y ago
How many of them don't support MSVC due to requiring features that MSVC can't handle (like inline assembly or Fortran), vs not wanting to deal with build system contortions?
66.
by tavert 9y ago
> That issue is much more complicated than what most FOSS projects deal with. R or Julia would have the exact same problems if they cared at all about MSVC. > I agree with you, building a lot of non-Windows stuff on Windows sucks. It&
67.
by tavert 9y ago
Easy how? Otherwise the scipy maintainers would be doing it.
68.
by tavert 9y ago
Try compiling SciPy from source using Microsoft dev tools. Or libgmp, the standard library for arbitrary precision integer arithmetic.
69.
by tavert 9y ago
They're sort of trying to work on this, see https://github.com/Microsoft/vcpkg . If things are strictly C++ and build using cmake, then MSVC is okay. Anything autotools or relying on inline assembly or overly compl
70.
by tavert 9y ago
Although unequally inconvenient when working with user-defined types. Possible now (wasn't always) but not seamless. Are there examples of libraries that depend on Numba?
71.
by tavert 10y ago
There are mac vagrant boxes available that work on a linux host. You may need to put vb.customize ["modifyvm", :id, "--usbehci", "off"] in the vagrantfile.
72.
by tavert 10y ago
You can build xterm with sixel support https://github.com/saitoha/libsixel
73.
by tavert 10y ago
Lots more info at https://byuu.org/emulation/preservation/lost-package/
74.
by tavert 10y ago
GPL is a major issue for a number of commercial users. It prevents us from shipping MKL as an option, for example. A lot of people can get by just fine without FFTW or SuiteSparse (they aren't dependencies of everything written in the
75.
by tavert 10y ago
Julia isn't GPL, it's MIT. Currently includes a few GPL libraries in the default distribution which makes the overall thing GPL, but that won't be the case for much longer. You can leave those libraries out for a no-GPL build
76.
by tavert 10y ago
Looks like that was one of the several third-party Octave GUI projects that predated the incorporation of an official GUI upstream in version 3.8 or 4.0 of Octave.
77.
by tavert 10y ago
Two libraries used by Octave, FFTW and SuiteSparse, both do exactly this - GPL for open source usage, or commercial paid licenses for inclusion in proprietary software (like Matlab).
78.
by tavert 10y ago
You mean like http://www.numfocus.org ? That's not quite how their fiscal sponsorship agreements work, but close enough.
79.
by tavert 10y ago
They conceal it more than they should, but you can shut Cortana off and create a local-only account.
80.
by tavert 10y ago
> a company that is built around Octave How many of those are there? I give talks pretty frequently where I start by asking people in the audience what tools they regularly use for technical computing, and I'm legitimately surprised
81.
by tavert 10y ago
Octave has a GUI nowadays which is pretty good, but a bit late to make much difference.
82.
by tavert 10y ago
All I want is to hide the toolbar by default. Wastes several clicks and a few seconds of setup time every time I give a talk from a notebook.
83.
by tavert 10y ago
> is just plain dog shit slow for certain things compared to SciPy Do you have real examples here or is this just FUD? SciPy is not known for using absolute state of the art algorithms or perfectly optimized implementations. It can be pr
84.
by tavert 10y ago
> Imagine if you did dev on Julia - how do you hire devs? Julia doesn't take long to learn if you've ever written any Python, or Matlab, or R. It's very good at calling into existing code written in other languages (the ab
85.
by tavert 10y ago
The installation, update, and packaging story of VS Code is so much more polished on Windows than it is with Atom. I gave Atom a shot for a few months (my employer has products based on it), but it had a tendency to overwrite my settings or
86.
by tavert 10y ago
Of course you don't build the reference BLAS if you care about performance. But OpenBLAS is quite competitive, and open source. A non-free compiler is a big impediment to build reproducibility and accessibility for new contributors for
87.
by tavert 10y ago
Or Intel will sell you their Fortran compiler. I guess you can get academic or open source license deals too, but it ends up feeling like you need a law degree to determine whether or not you can redistribute the results. Kind of annoying t
88.
by tavert 10y ago
The mingwpy project is slowly working towards making mingw capable of linking to the new ucrt, but it isn't ready yet - http://mingwpy.github.io/issues.html#the-vs-14-2015-runtime I'm sure they'd appreciate h
89.
by tavert 10y ago
If everything is using cmake, and you don't need any inline assembly or fortran for scientific code, then maybe. But you end up in a parallel universe where everything that touches MSVC is just needlessly different than every other pla
90.
by tavert 10y ago
Travis-CI's mac builders run on real Mac Mini hardware hosted on MacStadium.

More ›