jheriko

16 ms·

121.
by jheriko 4y ago
reminds me of C++ templated types being used for similar... except in this case there is no performance advantage from removing run-time logic by force. this kind of stuff is often confusing when working with teams. using simple dumb stuff
122.
by jheriko 4y ago
sounds believable, i'm not sure its helpful. i think that also sums up John Romero tbh...
123.
by jheriko 4y ago
its interesting to see how this differs from my own experience, from a time without game engines, 3d graphics libraries, the internet or even books for reference. the point about actually finishing things is a great takeaway. the first game
124.
by jheriko 4y ago
* dishonest and ethically corrupt contractors now take home less thanks to IR35. for the rest of us, it was a non-event.
125.
by jheriko 4y ago
the richer insight here is into the bias present in arrangements of binary classifiers imo. this article is dripping with it.
126.
by jheriko 4y ago
i never expect anyone in silly valley to do due diligence... it was a pretty obvious problem from the outset. if you are making a product, then using copyrighted material is far from fair use, at least outside of the US.
127.
by jheriko 4y ago
Almost as bad as the child porn and flagrant criminality... Almost
128.
by jheriko 4y ago
removing the "how" from the title totally changes my thoughts from "this might be interesting" to "huh? an obvious statement? what could this article be?" (p.s. it is slightly interesting. and very popular scie
129.
by jheriko 4y ago
this was always obvious to me from day one. i didn't need a leak to notice this. it was flagrant and blatant at the time. everyone was so busy being a fanboy it drove me a bit mad... especially operating without licenses and looking as
130.
by jheriko 4y ago
this is not revolutionary, all of this has been discussed before. the stuff about inspiration from the brain confuses me too... it reads a bit like neither ai nor neuroscience is particularly deeply understood. its very basic stuff.
131.
by jheriko 4y ago
he used a gigantic and elaborate framework/engine filled with cruft... so i doubt that was the thinking here.
132.
by jheriko 4y ago
i remember this from my time at school, some 25 years ago now, and it was extremely unhelpful to education as well... a lot of time spent explaining things like "how do i do something where i would normally right click?" or that t
133.
by jheriko 5y ago
its nice to see this. the idea that speed, safety and usability are mutually exclusive has always been wrong. trivially so imo... but clearly not given how long the myth has persisted.
134.
by jheriko 5y ago
a compiler generator working from descriptions of the language producing a clang quality compiler (with richer errors actually...) with LLVM backend. it actually already works, with a few short cuts for generating itself... and ive used bit
135.
by jheriko 5y ago
"we don't need error handling because we output the thing" i'm wondering why the hell its not a faster format? json is awful for storing ... anything, but a good compromise given that "everyone knows what it is&quot
136.
by jheriko 5y ago
the algorithm ensures that the point is approximately in the triangle. using the midpoint with one of the vertices, excludes the triangle in the middle...
137.
by jheriko 5y ago
how does it know the value remains constant? this is annoyingly non-trivial, and exactly the sort of thing that gets culled for JIT performance. the problem isn't generating code properly targeted at the hardware, or unrolling late-bin
138.
by jheriko 5y ago
as with everything written about division there are a host of mistakes. :) i'm still wondering what the "out by 2 case" is for algorithm D... most notably for 'word-size' remainders the compiler most certainly does
139.
by jheriko 5y ago
i got a similar impression, although i thought of it rather than being "anti-old-school" as less flattering "naive oss/linux mentality".... and a bunch of thoughts cursing how much this hurts progress. the mention o
140.
by jheriko 5y ago
what about the payload bits?
141.
by jheriko 5y ago
*tragedy
142.
by jheriko 5y ago
the attitude in a lot of the comments there is bizarre... in most of the world i'd hope an accident caused by using these features is the drivers fault for switching them on and not 'driving with due care and attention' all o
143.
by jheriko 5y ago
how about an actual fan to improve performance? :)
144.
by jheriko 5y ago
the blue channel in this image is not as filled as the green, the levels are lower. the conclusions here are a stretch.
145.
by jheriko 5y ago
i think this is most compilers.
146.
by jheriko 5y ago
what?!?! static doesn't imply the value can be known, only in some cases where its not modified in that compilation unit (which is trivial to detect with SSA) const is something else.
147.
by jheriko 5y ago
const_cast and linkage make const weaker than static on a variable that doesn't change.
148.
by jheriko 5y ago
this is totally wrong in my experience, although some of the understanding is right. by using static to limit to the compilation unit the compiler can workout the value is constant. const doesn't do this thanks to const_cast etc. maybe
149.
by jheriko 5y ago
the biggest win here is informing the compiler sufficiently to swap out div for and. the use of static is just a tool to inform the compiler that the value is a constant (which const /might/)
150.
by jheriko 5y ago
really? people mentioned const? you can tell how much low-level optimisation they have done if they think its gonna change codegen reliably, or at ll.

More ›