The general rule is that anytime you do division, it pushes itself out to
the ring where it makes sense. You can even see this with integers. The
reason for this is because checking if something is a unit or is exactly
divisible by the denominator is generally expensive.
Best,
Travis
On Satur
Thanks! AVX512 dgemm is indeed currently disabled on OpenBLAS.
On Saturday, June 15, 2019 at 9:15:19 PM UTC+2, Clement Pernet wrote:
>
> Hi,
> I just successfully compiled and tested upstream develop (at
> 8df690fd2a4e7a0c36e7dbc05c22d4d977217065 ) on a Intel(R) Xeon(R) Gold 6126
> CPU @ 2.60GH
I would guess that `libsingular' is the C++ kernel of Singular which does
not implement all strategies for GB's over Q. There are additional
strategies implemented both in Singular library code and in the Singular
interpreter. Presumably the `singular' interface is using the interpreter,
if not
Hi,
I just successfully compiled and tested upstream develop (at
8df690fd2a4e7a0c36e7dbc05c22d4d977217065 ) on a Intel(R) Xeon(R) Gold 6126
CPU @ 2.60GHz server with AVX512.
All tests passed.
Is there any specific compilation to test for #27961 ?
I'll give a look at the ticket later on.
As a side n
If I compute a Gröbner basis using libSingular, it is much slower than
using the Singular interface. Could this be related to the memory leaks or
is it a different issue?
sage: I = sage.rings.ideal.Katsura(PolynomialRing(QQ, 'x', 9))
sage: %time gb1 = I.groebner_basis(algorithm='singular')
CPU t
Has anybody except for John Palmieri tested the latest Sage beta releases
on a machine with avx512 support? I.e. at least Core i7 7820X (the X being
important):
https://en.wikipedia.org/wiki/AVX-512#CPUs_with_AVX-512
Trying to get some data for https://trac.sagemath.org/ticket/27961
--
You re
