Francois Bissey wrote:
On 09/05/13 14:27, leif wrote:
P.S.: I strongly doubt GCC's version matters here.
If I were you I wouldn't be so sure. I became involved in that bug
when I stumbled on a similar one porting to power7
http://trac.sagemath.org/sage_trac/ticket/14098
the version of gcc on that platform had an impact.
Well, I tried a couple of GCC versions and different compiler flags,
always with exactly the same result (see below), on x86_64 that is.
The main
problem here is that zn_poly tuning parameters are completely
wrong.
At least in this thread (as opposed to the ticket's title, or to some
comments there), we're not talking about segfaults.
/Comparisons/ of zn_poly's results to those of a reference
implementation fail -- deterministically -- if you only "tweak" the
tuning parameters appropriately, and btw. only when *squaring* (but not
in all of these cases IIRC) using Nussbaumer multiplication, apparently
just because(?) then test_nuss_mul() lets the inputs alias. [As
mentioned on the ticket, the test failures vanish for me if I only
remove the argument aliasing.]
"Random" values from, or segfaults (I couldn't reproduce) in zn_poly's
tuning are a separate issue, and apparently indeed OS-dependent (MacOS
X, Cygwin).
Anyway, "random" (or unreasonable/unexpected/invalid) tuning parameters
should cause assertions to fail, which are enabled at least when
zn_poly's test suite is built. (The whole library is [re]built with and
without assertions enabled.)
-leif
P.S.: I had put a link to #13947 on a (IIRC hardly related) ticket
David Harvey was involved in, asking for help, so I assumed he was aware
of the former.
--
() The ASCII Ribbon Campaign
/\ Help Cure HTML E-Mail
--
You received this message because you are subscribed to the Google Groups
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to sage-devel+unsubscr...@googlegroups.com.
To post to this group, send email to sage-devel@googlegroups.com.
Visit this group at http://groups.google.com/group/sage-devel?hl=en.
For more options, visit https://groups.google.com/groups/opt_out.
