Am Donnerstag, 6. Dezember 2018 11:41:23 UTC+1 schrieb Bill Hart:
> All the other systems are using modular methods here, so Roman use the
> modstd library (the command is modStd) in Singular to get those timings.
> Indeed cyclic7 over Q takes about 20s on my laptop in Singular using this
> approach.
>
Thanks for the hint. With that, I get the following times over ℚ.
ℚ | singular (4) singular (16)
cyclic7 | 28.79 21.27
cyclic8 | 1109.03 623.98
katsura9 | 28.07 25.14
katsura10 | 329.86 172.15
alea6 | 3053.70 2549.42
eco12 | 4879.15 1888.11
jason210 | 122.07 106.33
noon9 | 2339.98 1502.74
Am Donnerstag, 6. Dezember 2018 20:00:52 UTC+1 schrieb parisse:
> From giac, call gbasis with the list of polynomial and the list of the
> first block indets.
>
That worked. Thanks.
The timings on Q do not reflect the timings of native giac, there must be
> some misconfiguration or something wrong with the sage interface. Higher
> timings with 16 threads instead of 4 for a modular algorithm is suspicious.
>
While there will be some overhead due to the conversion from and to Sage,
it is the same in both cases. In fact, I observe similar times with the
native Giac that is installed into the Sage environment, when applied to
the Giac input file cyclic8 you linked above [1]: 0m27.309s using 4 threads
and 1m43.493s using 16. Indeed, CPU usage rarely surpasses 400% during the
computation. I am not sure what could be wrong here. Perhaps it is
something related to the patches of the Giac in Sage [2]?
[1]
https://www-fourier.ujf-grenoble.fr/~parisse/giac/examples/groebner/cyclic8
[2] https://github.com/sagemath/sage/tree/8.5.beta6/build/pkgs/giac/patches
--
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 https://groups.google.com/group/sage-devel.
For more options, visit https://groups.google.com/d/optout.
