#14627: Make mod_int signed and speed up matrix_modn_dense_float
----------------------------------+--------------------------------
Reporter: vbraun | Owner: jason, was
Type: enhancement | Status: positive_review
Priority: major | Milestone: sage-5.11
Component: linear algebra | Resolution:
Keywords: | Merged in:
Authors: Volker Braun | Reviewers: Martin Albrecht
Report Upstream: N/A | Work issues:
Branch: | Dependencies: #12728
Stopgaps: |
----------------------------------+--------------------------------
Old description:
> As discussed on https://groups.google.com/d/msg/sage-
> devel/hnO8gT8VON4/YyZ5UCM5wUEJ, unsigned longs are much slower than
> signed longs when converted to a Python object.
>
> Without patch about 20x slower than naive list-of-list implementation:
> {{{
> $ sage test_matrix_modn_float.py
> setA:
> 5 loops, best of 50: 179 ms per loop
> setB:
> 25 loops, best of 50: 10.3 ms per loop
> }}}
> With patch no performance difference:
> {{{
> $ sage test_matrix_modn_float.py
> setA:
> 25 loops, best of 50: 11.5 ms per loop
> setB:
> 25 loops, best of 50: 12.8 ms per loop
> }}}
>
> Also, I collected some common extern cdef statements in pxd headers and
> cleaned up the `MAX_MODULUS` declarations.
>
> Apply
> * [attachment:trac_14627_signed_mod_int.patch]
> * [attachment:trac_14627_32bit.patch]
New description:
As discussed on https://groups.google.com/d/msg/sage-
devel/hnO8gT8VON4/YyZ5UCM5wUEJ, unsigned longs are much slower than signed
longs when converted to a Python object.
Without patch about 20x slower than naive list-of-list implementation:
{{{
$ sage test_matrix_modn_float.py
setA:
5 loops, best of 50: 179 ms per loop
setB:
25 loops, best of 50: 10.3 ms per loop
}}}
With patch no performance difference:
{{{
$ sage test_matrix_modn_float.py
setA:
25 loops, best of 50: 11.5 ms per loop
setB:
25 loops, best of 50: 12.8 ms per loop
}}}
Also, I collected some common extern cdef statements in pxd headers and
cleaned up the `MAX_MODULUS` declarations.
Apply
* [attachment:trac_14627_signed_mod_int.patch]
* [attachment:trac_14627_32bit.patch]
* [attachment:trac_14627_more_primes.patch]
--
Comment (by vbraun):
I noticed that #10281 intentionally uses 64-bit integers on all platforms
for `mod_int`, since otherwise you can run out of primes. This is checked
in one #long doctest that I missed. So the `trac_14627_more_primes.patch`
switches to signed `int_fast64_t` for `mod_int`, this is fast on 64-bit
Linux/OSX and sub-optimal (but with enough primes) on 64-bit Windows and
all 32-bit platforms.
--
Ticket URL: <http://trac.sagemath.org/ticket/14627#comment:21>
Sage <http://www.sagemath.org>
Sage: Creating a Viable Open Source Alternative to Magma, Maple, Mathematica,
and MATLAB
--
You received this message because you are subscribed to the Google Groups
"sage-trac" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
To post to this group, send email to [email protected].
Visit this group at http://groups.google.com/group/sage-trac.
For more options, visit https://groups.google.com/groups/opt_out.
