Date: 2009-02-16 06:36
Sender: Akinori MUSHA
The implementation needs to follow such principles as the coerce
architecture. (You have to call rb_num_coerce_relop instead of
calling super, etc.)
I also think that Bignum should be supported within the core
for functional completeness, so I'll try to look into 1.9 and
see if there's stuff there I can bring in.
Maybe I'll add #lcm and #gcdlcm as well when #gcd should be. |
Date: 2008-10-30 19:20
Sender: Roger Pack
Looks like the current difference for 1.9 is
=> #<Benchmark::Tms:0x118aa0 @label="",
@real=2.26090502738953, @cstime=0.0, @cutime=0.0, @stime=0.02,
@utime=2.23, @total=2.25>
to
=> #<Benchmark::Tms:0x10e44c @label="",
@real=2.2900869846344, @cstime=0.0, @cutime=0.0, @stime=0.01,
@utime=2.25, @total=2.26>
so at least the current one isn't awful, though this patch is
still slightly faster. |