SethTro

  time ./goldbach -t8 -K 1024 -e 100''000''000''000
  real 0m44.133s
  user 3m32.193s
  sys 0m0.092s

  time ./goldbach -t12 -K 1500 -e 1''000''000''000''000
    267468893942 = 267468891139 + 2803 (407)
    926868341768 = 926868338701 + 3067 (437)
    599533546358 = 599533542901 + 3457 (481)
  real 2m3.909s
  user 16m0.729s
  sys 0m37.960s

I know of at least another 4x I could improve it but I've been nerd sniped enough for the night.

I wouldn't trust it 100% but it finds all terms in https://oeis.org/A025018 and https://oeis.org/A025019 and I can verify in https://sweet.ua.pt/tos/goldbach.html

Comment · 1 year ago

I don't see any benchmarks on your github.

How long does it take for you to test up to 10^9? 10^12?

Comment · 1 year ago

> Goldbach states: every even number ≥ 4 is a sum of two primes. The naive check for an even n tries many primes p and hopes that n − p is prime. Our idea is simpler: fix a small set of primes Q = {q1, . . . , qK} (the “gear”), and for each even n only test p = n − q with q ∈ Q

I don't see how your idea is different from the naive check. As far as I can tell you are basically saying do the naive check but only up to p > 250-300?

Comment · 1 year ago

That's the magic part, they aren't

Comment · 1 year ago

Totally agree with the title. I discovered Fenwick trees as part of solving a Project Euler, then from the problem forum I found out someone had invented this in the 90s and other people had imagined it much earlier.

Comment · 1 year ago

The board likely kept him till most of the bad PR had subsided so the new CEO can have a friendlier reception.

Comment · 2 years ago

This has 2 of the 3 features (float support, faster clock) + more POI that was keeping me on ESP32. For projects that need wifi, and can tolerate the random interrupts, I'll stick with ESP32.