calcyman wrote:Thanks! I presume you're referring to the recent B3/S23 speed improvements which take it beyond v3.28? I still have a few ideas about how to make B3/S23 even faster by reducing the instruction count in the iterator.
On the subject of speed improvements, 4.4 is still slower than 3.28 for me, or so it seems --- the (admittedly extremely unscientific) experiment of letting 4.4 run for a while, concurrently with 3.28, and compare the speeds, yielded about 1270 soups per (second times HT core) for 4.4, versus about 1370 for 3.28.
This was in C4_4, in case that may have made a difference (I doubt it should have though). I think I also accidently knocked off a zero from the 4.4 haul size, but even at 1 million soups per haul rather than 10 I expected the numbers to be closer to 3.28's...
FWIW my computer doesn't support the newer instruction sets (AVX1 is the most it'll do).
I'll happily test upcoming, new-and-improved!!! versions of 4.x though. (Alternatively I'll be happy to help diagnose what the reason for 4.x not having caught up on my computer might be.)
calcyman wrote:
I'm also
currently working on the cryptocurrency implementation, where blocks are mined by finding rare objects* in b3s23/C1 soups. I've made a few other departures from Bitcoin, such as using lattice-based digital signatures instead of elliptic curves, which should future-proof this against quantum computers and number theory breakthroughs.
* of types xp, xq, or yl.
Once completed and launched, be prepared for natural loafers and p120 oscillators and suchlike.
Excellent, looking forward to that being launched.
(Any hope of trading existing Catagolue contributions for CAcoins? No? Dang.)