diff options
author | shyouhei <shyouhei@b2dd03c8-39d4-4d8f-98ff-823fe69b080e> | 2017-03-06 06:04:52 +0000 |
---|---|---|
committer | shyouhei <shyouhei@b2dd03c8-39d4-4d8f-98ff-823fe69b080e> | 2017-03-06 06:04:52 +0000 |
commit | 3106f9412ab7bd89cd9fd4931bb7b4865b892d78 (patch) | |
tree | 508d56481df15bcd86ad634e7e25a19e6c55bc9a | |
parent | c8921b5ef07c3193d7b9ad6c329b8ade8bef8460 (diff) |
use HAVE_BUILTIN___BUILTIN_MUL_OVERFLOW
We already check for __builtin_mul_overflow in configure but never
actually referred it before. Why not call it if available, because
that should render supposedly-optimial assembly outputs.
Optionally if __builtin_mul_overflow_p is available, which is the case
for recent GCC, use that to detect fixnum overflow. This is much
faster than the previous. On my machine generated assembly of
numeric.c:int_pow reduces from 480 to 448 bytes, according to nm(1).
Also on my machine, following script boosts from 7.819 to 6.929 sec.
time ./miniruby -e 'i=0; while i < 30_000_000 do i += 1; 7 ** 23; end'
Signed-off-by: Urabe, Shyouhei <shyouhei@ruby-lang.org>
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@57784 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
-rw-r--r-- | configure.in | 1 | ||||
-rw-r--r-- | internal.h | 30 | ||||
-rw-r--r-- | process.c | 5 |
3 files changed, 32 insertions, 4 deletions
diff --git a/configure.in b/configure.in index 9eec2e9fa2..756f499387 100644 --- a/configure.in +++ b/configure.in @@ -2528,6 +2528,7 @@ RUBY_CHECK_BUILTIN_FUNC(__builtin_ctzll, [__builtin_ctzll(0)]) RUBY_CHECK_BUILTIN_FUNC(__builtin_add_overflow, [int x;__builtin_add_overflow(0,0,&x)]) RUBY_CHECK_BUILTIN_FUNC(__builtin_sub_overflow, [int x;__builtin_sub_overflow(0,0,&x)]) RUBY_CHECK_BUILTIN_FUNC(__builtin_mul_overflow, [int x;__builtin_mul_overflow(0,0,&x)]) +RUBY_CHECK_BUILTIN_FUNC(__builtin_mul_overflow_p, [__builtin_mul_overflow_p(0,0,(int)0)]) RUBY_CHECK_BUILTIN_FUNC(__builtin_constant_p, [__builtin_constant_p(0)]) RUBY_CHECK_BUILTIN_FUNC(__builtin_choose_expr, [ [int x[__extension__(__builtin_choose_expr(1, 1, -1))]]; diff --git a/internal.h b/internal.h index e74b91f667..eebd31073c 100644 --- a/internal.h +++ b/internal.h @@ -96,15 +96,41 @@ extern "C" { #endif #define TIMET_MAX_PLUS_ONE (2*(double)(TIMET_MAX/2+1)) +#ifdef HAVE_BUILTIN___BUILTIN_MUL_OVERFLOW_P +#define MUL_OVERFLOW_P(a, b) \ + __builtin_mul_overflow_p((a), (b), (__typeof__(a * b))0) +#elif defined HAVE_BUILTIN___BUILTIN_MUL_OVERFLOW +#define MUL_OVERFLOW_P(a, b) \ + ({__typeof__(a) c; __builtin_mul_overflow((a), (b), &c);}) +#endif + #define MUL_OVERFLOW_SIGNED_INTEGER_P(a, b, min, max) ( \ (a) == 0 ? 0 : \ (a) == -1 ? (b) < -(max) : \ (a) > 0 ? \ ((b) > 0 ? (max) / (a) < (b) : (min) / (a) > (b)) : \ ((b) > 0 ? (min) / (a) < (b) : (max) / (a) > (b))) + +#ifdef HAVE_BUILTIN___BUILTIN_MUL_OVERFLOW_P +/* __builtin_mul_overflow_p can take bitfield */ +/* and GCC permits bitfields for integers other than int */ +#define MUL_OVERFLOW_FIXNUM_P(a, b) ({ \ + struct { SIGNED_VALUE fixnum : SIZEOF_VALUE * CHAR_BIT - 1; } c; \ + __builtin_mul_overflow_p((a), (b), c.fixnum); \ +}) +#else #define MUL_OVERFLOW_FIXNUM_P(a, b) MUL_OVERFLOW_SIGNED_INTEGER_P(a, b, FIXNUM_MIN, FIXNUM_MAX) -#define MUL_OVERFLOW_LONG_P(a, b) MUL_OVERFLOW_SIGNED_INTEGER_P(a, b, LONG_MIN, LONG_MAX) -#define MUL_OVERFLOW_INT_P(a, b) MUL_OVERFLOW_SIGNED_INTEGER_P(a, b, INT_MIN, INT_MAX) +#endif + +#ifdef MUL_OVERFLOW_P +#define MUL_OVERFLOW_LONG_LONG_P(a, b) MUL_OVERFLOW_P(a, b) +#define MUL_OVERFLOW_LONG_P(a, b) MUL_OVERFLOW_P(a, b) +#define MUL_OVERFLOW_INT_P(a, b) MUL_OVERFLOW_P(a, b) +#else +#define MUL_OVERFLOW_LONG_LONG_P(a, b) MUL_OVERFLOW_SIGNED_INTEGER_P(a, b, LONG_LONG_MIN, LONG_LONG_MAX) +#define MUL_OVERFLOW_LONG_P(a, b) MUL_OVERFLOW_SIGNED_INTEGER_P(a, b, LONG_MIN, LONG_MAX) +#define MUL_OVERFLOW_INT_P(a, b) MUL_OVERFLOW_SIGNED_INTEGER_P(a, b, INT_MIN, INT_MAX) +#endif #ifndef swap16 # ifdef HAVE_BUILTIN___BUILTIN_BSWAP16 @@ -6916,11 +6916,13 @@ typedef LONG_LONG timetick_int_t; #define TIMETICK_INT_MIN LLONG_MIN #define TIMETICK_INT_MAX LLONG_MAX #define TIMETICK_INT2NUM(v) LL2NUM(v) +#define MUL_OVERFLOW_TIMETICK_P(a, b) MUL_OVERFLOW_LONG_LONG_P(a, b) #else typedef long timetick_int_t; #define TIMETICK_INT_MIN LONG_MIN #define TIMETICK_INT_MAX LONG_MAX #define TIMETICK_INT2NUM(v) LONG2NUM(v) +#define MUL_OVERFLOW_TIMETICK_P(a, b) MUL_OVERFLOW_LONG_P(a, b) #endif CONSTFUNC(static timetick_int_t gcd_timetick_int(timetick_int_t, timetick_int_t)); @@ -7036,8 +7038,7 @@ timetick2integer(struct timetick *ttp, timetick_int_t t = ttp->giga_count * 1000000000 + ttp->count; for (i = 0; i < num_numerators; i++) { timetick_int_t factor = numerators[i]; - if (MUL_OVERFLOW_SIGNED_INTEGER_P(factor, t, - TIMETICK_INT_MIN, TIMETICK_INT_MAX)) + if (MUL_OVERFLOW_TIMETICK_P(factor, t)) goto generic; t *= factor; } |