* src/factor.c (factor_using_pollard_rho)
(factor_using_pollard_rho2, mp_factor_using_pollard_rho):
Use int_fast64_t for internal counters rather than int, as int
could overflow on some somewhat-practical examples. Problem
discovered on a hypothetical platform where W_TYPE_SIZE is neither
32 nor 64, when factoring 0x7ffffffffffffeab7fffffffffff7369 ==
170141183460469225450570946617781744489, causing k to overflow in
mp_factor_using_pollard_rho. Presumably a similar problem exists
in the previous stable coreutils 9.7, too, on 32-bit platforms
with somewhat-larger test cases, though I haven’t take the
somewhat-extensive CPU time to discover it.
* src/factor.c (mp_factor_using_pollard_rho):
Fix recently-introduced aliasing bug by computing q
before g gets updated in place. Problem discovered
on a hypothetical platform where W_TYPE_SIZE
is neither 32 nor 64.
* src/factor.c: Do not include c-ctype.h.
(strtouuint): Don’t bother generating a number on
error; just return a strtol_error value other than LONGINT_OK.
Speed up overflow checking.
* src/factor.c (primes_ptab): New table of primes, replacing
primes_diff and primes_diff8. All uses changed. This is simpler
and should improve performance slightly. Although this limits the
table’s primes to 2**15 instead of to 668221, the limit can easily
grow to 2**32 by changing the type of ‘prime’, without hurting
performance significantly compared to the primes_diff and
primes_diff8 approach.
* src/make-prime-list.c (output_primes):
For each prime p, output p instead of two differences.
* src/factor.c (factor_using_pollard_rho)
(factor_using_pollard_rho2, mp_factor_using_pollard_rho):
Use int, not unsigned long int, for counters that won’t go above
2**31 on practical platforms. This yields a significant speedup
on GCC 15 x86-64, and using signed values allows for automatic
checks for overflow when using gcc -fsanitize=undefined.
* src/factor.c (mp_factor_insert_ui): Rename fom
mp_factor_insert_ui, and change arg type from unsigned long int to
mp_limb_t. All uses changed. This avoids creating and freeing a
small mpz_t.
(mp_factor_using_division): Add a static assert requiring that
that mp_limb_t be wide enough, which it should be (and is in
standard GMP).
* src/factor.c (factor_using_pollard_rho)
(factor_using_pollard_rho2): Use mp_limb_t, not unsigned long int,
for a parameter. This avoids some casts, and avoids a theoretical
bug where converting to mp_limb_t loses info.
lt2 a bit more natural, given the current implementation.
* src/factor.c (lt2): New function.
(ge2): Rewrite in terms of lt2.
(gt2): Remove. All callers changed to use lt2.
* src/factor.c (umul_ppmm): When !USE_LONGLONG_H so we need to
define this, speed things up if there is an unsigned type uuroom_t
wide enough to hold two words. Do not make a similar change for
udiv_qrnnd, as it is not performance critical and anyway on GCC 15
x86-64 that would mean a subroutine call.
* src/factor.c (gcd_odd, gcd2_odd): Speed up, given that the
second argument is always odd.
(gcd_odd): Avoid recomputing a temporary.
(gcd2_odd): Test for zero only if a multiple of B.
This saves an ‘assume’.
* src/factor.c (submod2): Use ckd_sub to subtract by hand rather
than using sub_ddmmss plus a compare. This speeds things up a
bit, on x86-64 with GCC 15 anyway.
* src/make-prime-list.c (output_primes): Output
SQUARE_OF_FIRST_OMITTED_PRIME, not FIRST_OMITTED_PRIME. All uses
changed. This way, the uses don’t need to worry about casts to
avoid overflow.
* src/factor.c (make_uuint2): New function.
(powm2, millerrabin2): Pass two-word args as uuints,
not as mp_limb_t const [2] pointers. All uses changed.
(prime2_p): Rework to use the new API, fixing a FIXME.
* src/factor.c (factor_using_division, mp_finish_up_in_single)
(mp_finish_in_single, factor_using_pollard_rho)
(factor_using_pollard_rho2, factor_up, factor):
Put FACTORS arg first, for consistency.
This is just a refactoring and should not affect speed.
* src/factor.c (print_factors): Use single-precision code only
when word size is 32 or 64, as it isn't tested for other sizes and
is known to not work when it the size is 128. This change does
not affect any known practical platforms.
Suggested for consideration by Torbjörn Granlund in:
https://lists.gnu.org/r/coreutils/2025-01/msg00000.html
* src/factor.c (PROVE_PRIMALITY): Now defaults to false.
(mp_prime_p): Help the compiler by telling it mpz_prob_prime_p
returns nonnegative.
* tests/factor/create-test.sh (bigprime): Test 2^400 - 593,
since that’s now practical.
* tests/local.mk (factor_tests): Add new test.
* src/factor.c (USE_BAILLIE_PSW): New constant.
(prime_p, prime2_p): Use it, i.e., always use Baillie-PSW.
Do so by using mp_prime_p. Do not tell GCC these functions
are pure; the pure mark was present only to pacify GCC
and is no longer needed now that thes functions call mp_prime_p.
Inspired by a proposal by Torbjörn Granlund in:
https://lists.gnu.org/r/coreutils/2025-01/msg00000.html
* src/factor.c (mp_prime_p): Use mpz_probab_prime_p rather than doing
it by hand, as mpz_probab_prime_p uses Baillie-PSW which is a win.
This removes the need for the ret2 label and goto.
* src/factor.c (prime_p, prime2_p, mp_prime_p): Do not skip the
flag_prove_primality test for the last prime in the table, i.e.,
when r == PRIMES_PTAB_ENTRIES - 1.
(mp_prime_p): There is no longer a need for the ret1 label or goto.
mini-gmp lacs mpn_tdiv_qr, so supply an emulation of it
when using mini-gmp.
* src/factor.c (copy_mpn_from_mpz, mpn_tdiv_qr) [!mpn_tdiv_qr]:
New functions.
These changes are taken from a proposal by Torbjörn Granlund in:
https://lists.gnu.org/r/coreutils/2025-01/msg00000.html
On my x86-64 platform, they improve speed by more than 8× when
factoring 340282366920938463463374607431768211457.
* src/factor.c (mp_modadd, mp_modsub, mp_modadd_1, mp_mulredc):
New functions.
(MP_FACTOR_USING_POLLARD_RHO_N_MAX): New macro.
(mp_factor_using_pollard_rho): Act on mpn not mpz, and on
mp_limb_t not unsigned long int. Reorder args. All uses changed.
This is mostly to make the code a bit easier to read.
It shrinks the size of factor.o by 0.7% on my x86-64 platform
though it doesn’t affect CPU performance significantly.
* src/factor.c (mp_no_factors): Rename from mp_factor_init.
(mp_no_factors, mp_factor_using_division, mp_factor):
Return struct mp_factors rather than modifying one passed by reference.
All uses changed.
* src/factor.c (mp_factor_using_division):
When continuing in single precision, don’t divide by primes that
were already cast out in multiple precision.
On my platform this gave a 2.5% speedup when factoring
2**128 + 172261 = 4999 * 68070087401668026297934508388031283,
as W_TYPE_SIZE == 64 and 4999 is the last prime in the primes table.
* src/factor.c (mp_factor_using_division): Use index into
primes_diff consistently with other uses. This is mostly just a
style change, and likely doesn’t change the generated machine code.
This does not affect performance much, but it should allow future
performance improvements.
* src/factor.c (factor_using_division): Two new args I and P,
which generalize this function. All uses changed.
(mp_finish_up_in_single, factor_up): New functions, like the
non-*_up* versions but with two new args PRIME_IDX and PRIME.
They mostly just have the old body of the old non-*_up_ versions.
(mp_finish_in_single, factor): Rewrite in terms of the new functions.
This significantly improves performance when a number exceeds
2**(W_TYPE_SIZE - 1) and is the product of a prime less than
FIRST_OMITTED_PRIME and another prime less than 2**(W_TYPE_SIZE - 1).
On my platform, for example, it doubled the speed of factoring
4999 * (2**128 - 159).
* src/factor.c (mp_size, mp_finish_in_single): New functions.
(mp_factor_using_division, mp_factor_using_pollard_rho):
Finish using single precision when possible.
* tests/factor/factor.pl (lt-5000-times-128-bit): New test.
* src/factor.c (factor_insert_refind):
Use idx_t for indexes into primes_diff,
for consistency with other indexes into primes_diff.
This has no practical effect unless the primes_diff
table becomes unreasonably large.
Support a multiplicity argument in the mp case, too.
This helps keeps the two cases in sync, for maintenance.
* src/factor.c (mp_factor_insert, mp_factor_insert_ui):
New arg M, for multiplicity. All callers changed.
Use something other than a macro when that is easy and won’t hurt
performance.
* src/factor.c (__ll_B, __ll_lowpart, _ll_highpart) [!USE_LONGLONG_H]:
(MAX_NFACTS, highbit_to_mask, factor_insert, PRIMES_PTAB_ENTRIES):
Make these enums, or constants, or static functions instead of macros.
(highbit_to_mask): Rename from HIGHBIT_TO_MASK. All uses changed.