@Aella_Girl You're just in time for the fourth installment of the roughly-every-9-months floorplan poll! How am I the only person that's noticed this, I need answers
https://t.co/dvmQyckXG0
@Aella_Girl This is round 3, in case someone wants to make a graph.
Different survey options make it a bit hard to compare though
https://t.co/w43Pck2618
@mrbillstunes drive -> +60 dB โ
Seriously though, what difference does this make besides making the master meter not red? I would think >0dB would be hard clipped by necessity when exporting
@marc_b_reynolds I didn't know that. Kinda ridiculous that they did that when it's only one extra instruction to get the proper clz semantics, and anyone using clz (as opposed to a bsr intrinsic) probably wants those proper semantics
@johnregehr @MDTom A variation on that idea: use an array instead, and use some hash bits to index it, so you'd only have to store what isn't used in the index.
As a bonus, you could store the maximum seen hash in each slot, so you'd sometimes get a definite "no" answer (if that matters at all)
@danluu This doesn't feel like a fair reading of the comment, IMO. There was no judgment given about the number of bugs itself, only the open development processes that might contribute to it (in both finding and creating bugs), and that judgment was rather reserved.
@HaroldAptroot @marc_b_reynolds @vegard_no @slembcke Yep, from playing around a bit looks like it's:
number of bytes from start of line to end of literal << 24 + 2 * line number where literal starts + 34
@pervognsen But each continuation byte position isn't unique...? E.g. a 4 byte sequence has three bytes of the form 10xxxxxx that can be shuffled around to form different code points that would still sort to the same value.