← Feed

Posts

• View post

  first person to suggest writing that replacement in rust gets the meat hook

• View post

  fontconfig 2.18 is broken and looking at the commit log makes me want to throw up i was right to start my own replacement but i’m gonna need stims for it to get anywhere

• View post

  AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA

• View post

  turns out state of the art AI models, the smallest of which have trillions of parameters and need 400 GB VRAM, are still useless at the translation tasks i do for a living. this is like the one job that LLMs had, and they still can’t do it. they’re not even CLOSE. oh i’m sure they can find bugs in software and speedrun video games but like, that’s basically pure pattern recognition and brute force. those are problems that can be fully understood and even trivially modeled as equations. that does...

• View post

• View post

  sometimes GCC is pretty clever. it recognized that i’m calling atan2f element-wise on a SIMD vector type and uses the libmvec function instead of plain libm accordingly, reducing a lot of expensive scalar extraction and function call instructions to just this one clang is not that clever. it has a __builtin_elementwise_atan2 but that just uses the non-vectorized version. it also has no clean way to do masked shuffles because its shuffle builtin only takes scalar indices

• View post

  ported most of the oklab/okhsl code to GNU vector extensions laptop can now render a full-screen okhsl color wheel at 40 fps instead of 4. SIMD is magic

• View post

  clang go home you’re drunk

• View post

• View post

  squishy birb

• View post

  🐝

• View post

  rare

• View post

  justice is a spook your honor

• View post

  c++ people be like i love header-only libraries i love measuring compile times in hours

• View post

• View post

  this is where i post from

• View post

  GCC is very good at auto-vectorizing code if only you allow it and the benefits are very real (distros: please consider using -march=x86_64-v3 by default)

• View post

  @vriska@lizards.live man, the bullshit we’re subjected to here “separate cap and carton for the environment” “cap stays on — for the environment”

• View post

  glad to see i’m not the only one who considered building heat-seeking missiles by strapping wiimotes to them

• View post

  @snacks@netzsphaere.xyz a couple years ago krita got it right in a bunch of places but then messed up the rotations. i hope that’s fixed now

• View post

  studying digital color science and figuring out how incredibly wrong a lot of image processing software and even widespread standards like CSS are about color, much stuff we could be doing better but aren’t (like the 1931 color primaries and observers i mentioned before; you can calibrate displays to what measures as identical in that system but looks very different to human eyes) fucking CSS specifying effects to be applied in non-linear sRGB, SVG renderers basically doing all their alpha blen...

• View post

• View post

• View post

  @tknarr@mstdn.social @navi@social.vlhl.dev @ska@social.treehouse.systems this is just normal for corpo code. they do not stop to think about their choices.

• View post

  @snacks@netzsphaere.xyz all of them are

• View post

  @snacks@netzsphaere.xyz oh that seems likely

• View post

  @snacks@netzsphaere.xyz ESD?

• View post

  github.com/V4bel/dirtyfrag it happened again

• View post

  @cell@shitposter.world we barely have a good enough encoder for av1! there’s rav1e which has excellent results at speed <= 4, far better than the others, but that is dogshit slow and development has stalled. then there’s svt-av1 which is fast but optimized for PSNR and has bad rate control no matter what you do (including the PSYEX fork), so it sucks at preserving detail. also doesn’t support 4​:4:​4 sampling. and there’s the reference libaom which is has absolute garbage rate control and is...

• View post

  ​:paimon_nomming:​