@ihodes@packyM Color the board as a chessboard, b&w. It has 6 w and 6 b squares. All the pieces except C cover 2w+2b squares. C covers either 3b+1w, or 1b+3w. If you use C, you need to cover the remaining odd b and odd w squares with 2x 2w, 2b pieces. Impossible, so C can't be used, QED.
@valigo Ticking sound in headphones in xfce while no audio is playing. Unplugging auto-selects an alternative device, then re-plugging keeps the alternative device selected. Crap UX all around.
@rfleury@DemetriSpanos OTOH, even complex tasks can be decomposed to subtasks, that then are reasonably constrained. Since the process is highly iterative, constraints are juggled by tooling at each step, tweaking context per step. Output is also iteratively refined to match concrete and general reqs.
@tsoding You overestimate programmers. Most software can be written by a Pavlov trained child. Shallow world needs shallow problems solved and 99,9% of programmers do exactly this. And an LLM "knows" a long term code base better than the people that've written it 10 years ago.
@thierrifilms@Kinefinity Yeah, will be testing and adding tweaks if needed once user samples start to appear. I don't think there are official raw samples from any of the 8.0 updated cameras?
@tigregalis @LubaRaphael@Jonathan_Blow On the other hand, the more experienced, capable & prepared you are, the less you profit from Rust's safety machinery preventing you from shooting yourself in the leg, since you rarely shoot your legs in the 1st place. So by reductio ad absurdum we've rejected your conclusion. ;)
@Jonathan_Blow@rflaherty71 A common base set of intrinsics + ability to fallback to platform specific + easy way to use external asm would cover almost all use cases. It is possible to decently translate sse to neon almost in its entirety (see e.g. sse2neon) - a very practical if not future proof approach.
@Jonathan_Blow@rflaherty71 I do lots of asm/intrinsics. I like the convenience of inline asm, but huge perf gains do come from SIMD, which is almost universally now coded as (platform specific) intrinsics in C/C++ land, due to MSVC limitations and AT&T ugliness.
@tjcampanella9@Jonathan_Blow Not in a game dev environment. Unit tests are next to useless and logging is more than nothing, but will get you very little knowledge about the interaction of the multiple systems that constitute a game.
@tsoding Actually, books have been written defending the thesis that simpler systems are superior to more complex systems in general, not only in software.
Since problem solutions are closed systems, there is a minimum entropy "best" solution. Any additional entropy would be just noise.
@MichaelKruglos@cmuratori E.g., in Go for any non-trivial type it is worse than virtual calls in C++. Interface2interface resolve hits the global ifaces hash table, possibly with locks, and type2interface conversion involves mallocgc+memmove. Cheap it is not. [2/2]
@MichaelKruglos@cmuratori Worth pointing out -- since the discussion originated with performance -- that while composition may have its own conceptual allures, it too doesn't escape the performance overhead of encapsulation and polymorphic dispatch. [1/2]
#slimRAW 1.8.3: slimRAW'ed Sigma fp CinemaDNG raw video can now be imported into Premiere CC, even though the original uncompressed raw video can't. https://t.co/apMGmNBwz6
#slimRAW 1.8.3 released with full support for @Sigma_Photo's Sigma fp CinemaDNG video. Raw video from the Sigma fp is beautiful and compresses well.
Update here: https://t.co/rHeKWhA6No