All pull requests

Pull requests

• Store the shifted initial seed unweighted, and always forward-weight at the start of a test

  buggithub_actions
  #389·MarkRose opened Nov 1, 2026Member·

  ·

  · Fix crashes
  15

• Mersenne PRP: carry-normalize the seed before the first squaring

  bug
  #388·MarkRose opened Nov 1, 2026Member·

  ·

  · Correctness
  3

• Mlucas: reject a value-taking command-line flag given as the last argument

  bug
  #386·MarkRose opened Nov 1, 2026Member·

  ·

  · Fix crashes
  3

• PRP-CF: restarting after the PRP base was changed in worktodo.txt aborts in the known-factor check

  bug
  #385·MarkRose opened Oct 28, 2026Member·

  ·

  · Correctness
  16

• PRP-CF: spurious "Full-residue check fails!" abort when a known factor >= 2^64 is listed before a shorter one

  bug
  #384·MarkRose opened Oct 27, 2026Member·

  ·

  · Fix crashes
  1

• Mersenne PRP with a large base: abort after the full test, wrong worktype, and wrong residues once the abort is fixed

  bug
  #383·MarkRose opened Oct 27, 2026Member·

  ·

  · Correctness
  14

• PRP-CF: a bad primary savefile throws away the good secondary (restart from iteration 0, or a hang at the end of the run)

  bug
  #382·MarkRose opened Oct 27, 2026Member·

  ·

  · Fix crashes
  3

• get_nextlarger_fft_length: return a whole-K length below 8K (fixes an endless FFT-length switch loop)

  bug
  #381·MarkRose opened Oct 27, 2026Member·

  ·

  · Fix crashes
  3

• mi64_scalar_modpow_lr: fix the "Unxpected borrow!" abort on a zero-padded modulus, and a one-limb overrun of c[]

  bug
  #380·MarkRose opened Oct 27, 2026Member·

  ·

  · Fix crashes
  9

• p-1: propagate the borrow when forming the stage 1 residue - 1

  bug
  #378·MarkRose opened Oct 27, 2026Member·

  ·

  · Correctness
  3

• Stop with a clear error on a roundoff error at the largest FFT length, instead of switching to "0K"

  bug
  #377·MarkRose opened Oct 27, 2026Member·

  ·

  · Fix crashes
  3

• Accept two-radix sets when decoding a faster larger-length cfg entry (1K runs abort on a 2K 32 32 row)

  #376·MarkRose opened Oct 27, 2026Member·

  ·

  · Fix crashes
  1

• Mfactor: don't route the exponent PRP test of p = 2^n+1 into the Fermat-factor path (-m 65537 aborted)

  bug
  #375·MarkRose opened Oct 26, 2026Member·

  ·

  · Fix crashes
  3

• test_fac: check the 128-bit multiply macros against unsigned __int128 (stacked on #173 and #198)

  bug
  #374·MarkRose opened Oct 26, 2026Member·

  ·
  3

• mi64_is_div_by_scalar64_x4: make every asm stream test all of x[], and use the asm arm again

  bug
  #373·MarkRose opened Oct 26, 2026Member·

  ·

  · Fix latent issues
  3

• Mfactor: allocate the modulus-dependent vectors for at least lenP+1 words (heap overflow in the multiword builds)

  bug
  #372·MarkRose opened Oct 25, 2026Member·

  ·

  · Fix latent issues
  9

• Mfactor: make QUIT_WHEN_FACTOR_FOUND stop the run (it deadlocked threaded builds and lost candidates)

  bug
  #371·MarkRose opened Oct 23, 2026Member·

  ·

  · Fix latent issues
  9

• Restart from scratch when the first Gerbicz check fails, instead of dropping the assignment

  bug
  #370·MarkRose opened Oct 22, 2026Member·

  ·

  · Fix crashes
  8

• radix60: fix the Fermat-mod AVX/AVX2 carry reading past its subchain tables (SIGSEGV in short/hiacc chain mode)

  bug
  #369·MarkRose opened Oct 21, 2026Member·

  ·

  · Correctness
  3

• AVX-512 carry: use exact DWT weights in HIACC mode for leading radices 36-60 (wrong PRP self-test residues)

  #364·MarkRose opened Oct 17, 2026Member·

  ·

  · Correctness
  1

• Split Linux and ASan CI into build and test jobs

  github_actionsperformance
  #363·MarkRose opened Oct 17, 2026Member·

  ·
  34

• Skip scratch sources in CI and CodeRabbit review

  #362·MarkRose opened Oct 17, 2026Member·

  ·
  3

• Mfactor: don't compare a multiword p's low word against the sieving primes (AVX-512 sieve)

  #349·MarkRose opened Sep 25, 2026Member·

  ·

  · Fix crashes
  3

• Mfactor: on resume, rebuild the lists of the prime and composite factors found so far (depends on #338, #340, #344, #345, #346, #347)

  bug
  #348·MarkRose opened Sep 25, 2026Member·

  ·

  · Fix latent issues
  7

• Mfactor: close the window in which a kill repeats a factor report on resume (depends on #338, #340, #344)

  bug
  #347·MarkRose opened Sep 24, 2026Member·

  ·

  · Fix latent issues
  19