Hi everybody,
it's the last day of Sage Days 32 and I push Burcin's and Clément's patch far
enough to have all doctests pass on sage.math, cf.
http://trac.sagemath.org/sage_trac/ticket/4260
With respect to performance the difference is like night and day:
== Night (without #4260)==
sage: A = random_matrix(GF(previous_prime(2^23)),1000,1000)
sage: B = random_matrix(GF(previous_prime(2^23)),1000,1000)
sage: %time A*B
CPU times: user 326.12 s, sys: 1.27 s, total: 327.39 s
Wall time: 328.79 s
1000 x 1000 dense matrix over Finite Field of size 8388593
sage: %time A.echelonize()
CPU times: user 92.59 s, sys: 0.64 s, total: 93.23 s
Wall time: 93.67 s
== Day (with #4260) ==
sage: A = random_matrix(GF(previous_prime(2^23)),1000,1000)
sage: B = random_matrix(GF(previous_prime(2^23)),1000,1000)
sage: %time A*B
CPU times: user 0.26 s, sys: 0.02 s, total: 0.28 s
Wall time: 0.28 s
1000 x 1000 dense matrix over Finite Field of size 8388593
sage: %time A.echelonize()
CPU times: user 0.41 s, sys: 0.00 s, total: 0.41 s
Wall time: 0.42 s
Rob Beezer (sitting next to me :)) is currently looking at the patch, adding
more doctests/documentation and stress testing the whole thing. It would be
great if people could give #4260 a spin on different platforms to check
whether it works for them. Especially
- big endian machines to test whether pickling works
- 32-bit machines and
- non-Linux machines
would be greatly appreciated. Also, people who care deeply about matrices over
ZZ and QQ should give it a spin and check for performance regressions or
improvements (because of multi-modular stuff) There shouldn't be any, but who
knows.
Cheers,
Martin
PS: After SD32 I probably wont' have time to push this patch much further so
if you care about switching from very slow to very very fast for operations
with dense matrices mod n (n < 2^23) then feel free to take charge.
--
name: Martin Albrecht
_pgp: http://pgp.mit.edu:11371/pks/lookup?op=get&search=0x8EF0DC99
_otr: 47F43D1A 5D68C36F 468BAEBA 640E8856 D7951CCF
_www: http://martinralbrecht.wordpress.com/
_jab: martinralbre...@jabber.ccc.de
--
To post to this group, send an email to sage-devel@googlegroups.com
To unsubscribe from this group, send an email to
sage-devel+unsubscr...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/sage-devel
URL: http://www.sagemath.org
