Justin C. Walker on Thu, 18 Feb 2021 00:17:38 +0100


[Date Prev] [Date Next] [Thread Prev] [Thread Next] [Date Index] [Thread Index]

Lewis-Wester benchmarks


Hi,

I’m not a Pari developer, and I am not very familiar with the internals of Pari/gp.

I decided to see how the performance of 2.13.1 on my system, 2019 16” MBP (2.4 GHz, 8-core Core i9), with macOS 10.15.7n (Catalina).  I latched onto the benchmarks from the work of Lewis and Wester.

This all mostly worked well, and system performance (appended below) has improved since your last recorded results, a few generations back.  In the process, I noticed a few things that puzzled me.

First, three sets of tests (N, O1, O2) are disabled in the tests as posted.  The results for the last-run version actually show a value for O1, but for N it’s KD and for O2, it’s UN.

A bit of RTFMing implies that the disabling is due to the lack of support for (most?) linear algebra over polynomial rings.  The questions arise:
1. Does this mean that there is “clandestine” code supporting polynomial rings that is not ready for Prime Time?
2. And the “average” result for O1 comes about how?

In any case, your hard work on Pari/gp certainly shows.  Thanks!

	2.13.1

A:	2
B:	0
C:	5
D:	4
E:	3
F:	1
G:	736
Gp:	455
H:	377
I:	xs18
J:	2
K:	67
L:	11
M1:	16
M2:	8146
N:	KD
O1:	KD
O2:	UN
P:	33
P':	36
Pp:	2
Pp':	1
Q:	4
Q':	256
Qp:	4
Qp':	9
S:	1
T:	6
U:	1
V:	37
W1:	8
W2:	374
X:	1569
Y:	117