diff options
| author | Paul Eggert <eggert@cs.ucla.edu> | 2025-07-05 11:04:54 -0700 |
|---|---|---|
| committer | Paul Eggert <eggert@cs.ucla.edu> | 2025-07-09 17:12:40 -0700 |
| commit | d70e07668ca92b812ea31d0fd21f0bace4edeef8 (patch) | |
| tree | b33cd990515dacbec6427b479d40f2692d91afe6 | |
| parent | factor: streamline gcd2_odd (diff) | |
| download | coreutils-d70e07668ca92b812ea31d0fd21f0bace4edeef8.tar.gz coreutils-d70e07668ca92b812ea31d0fd21f0bace4edeef8.zip | |
factor: speed up umul_ppmm when !USE_LONGLONG_H
* 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.
| -rw-r--r-- | src/factor.c | 27 |
1 files changed, 20 insertions, 7 deletions
diff --git a/src/factor.c b/src/factor.c index b62ec32e3..96ef8e828 100644 --- a/src/factor.c +++ b/src/factor.c @@ -179,13 +179,6 @@ typedef uint64_t UDItype; # define HAVE_HOST_CPU_FAMILY_powerpc 1 # endif # include "longlong.h" - -#else /* not USE_LONGLONG_H */ - -static mp_limb_t const __ll_B = (mp_limb_t) 1 << (W_TYPE_SIZE / 2); -static mp_limb_t __ll_lowpart (mp_limb_t t) { return t & (__ll_B - 1); } -static mp_limb_t __ll_highpart (mp_limb_t t) { return t >> (W_TYPE_SIZE / 2); } - #endif /* 2*3*5*7*11...*101 fits in 128 bits, and has 26 prime factors. @@ -307,6 +300,26 @@ static void factor_up (struct factors *, mp_limb_t, mp_limb_t, /* Set (w1,w0) = u * v. */ #ifndef umul_ppmm +/* Speed things up if there is an unsigned type uuroom_t that is wide + enough to hold two words. */ +# if W_TYPE_SIZE <= UINTMAX_WIDTH / 2 +# define uuroom_t uintmax_t +# elif W_TYPE_SIZE <= 64 && defined __SIZEOF_INT128__ +# define uuroom_t unsigned __int128 +# endif +# ifdef uuroom_t +# define umul_ppmm(w1, w0, u, v) \ + do { \ + uuroom_t _u = u, _w = _u * (v); \ + (w1) = _w >> W_TYPE_SIZE; \ + (w0) = _w; \ + } while (0) +# endif +#endif +#ifndef umul_ppmm +static mp_limb_t const __ll_B = (mp_limb_t) 1 << (W_TYPE_SIZE / 2); +static mp_limb_t __ll_lowpart (mp_limb_t t) { return t & (__ll_B - 1); } +static mp_limb_t __ll_highpart (mp_limb_t t) { return t >> (W_TYPE_SIZE / 2); } # define umul_ppmm(w1, w0, u, v) \ do { \ mp_limb_t __u = u, __v = v, \ |