Speedup for OA_from_quasi_difference_matrix

[nathanncohen]

Turns out that the use of a brain is underrated. I should really buy one someday.

This is what happens when you think before writing a loop:

sage -t --long database.py
    [340 tests, 3.20 s]

This is what happens when you do not

sage -t --long database.py
    [340 tests, 14.43 s]

CC: @videlec

Component: combinatorial designs

Author: Nathann Cohen

Branch/Commit: 41676cf

Reviewer: Vincent Delecroix

Issue created by migration from https://trac.sagemath.org/ticket/16720

[nathanncohen]

Branch: u/ncohen/16720

[sagetrac-git]

Commit: 26975dc

[sagetrac-git]

Branch pushed to git repo; I updated commit sha1. New commits:

26975dc

trac #16720: Speedup for OA_from_quasi_difference_matrix

[videlec]

comment:4

Hello,

I do not understand that

-    M = M[:len(M)/2] # only develop w.r.t the last two coordinates
+    M = [M[i*18+j] for i in range(len(M)/18) for j in range(9)] # only develop w.r.t the last two coordinates

Vincent

[nathanncohen]

comment:5

Yoooooooo !

I do not understand that

-    M = M[:len(M)/2] # only develop w.r.t the last two coordinates
+    M = [M[i*18+j] for i in range(len(M)/18) for j in range(9)] # only develop w.r.t the last two coordinates

Ahem :-PPPPP

Thaaaaaaaaat's because the rows of the matrix were listed as 111222333 before and now it is 123123123. You don't care about that in an OA or a difference matrix, so it does not matter and it is easier to get this speedup.

The function which I had to modify above called this routine, and only needed to remember the first half of the matrix. Formerly. But now that the rows are listed differently, well getting that "half" of the matrix is a bit different :-P

I know it is not "semantically correct". Please don't make me rewrite the function for that T_T

Nathann

[sagetrac-git]

Branch pushed to git repo; I updated commit sha1. New commits:

f5ea70d	trac #16720: Merged with 6.3.rc1
41676cf	trac #16720: A clearer line

[sagetrac-git]

Changed commit from 26975dc to 41676cf

[nathanncohen]

comment:7

This is clearer.

Nathann

[videlec]

Reviewer: Vincent Delecroix

[videlec]

comment:8

Ok, good!

Vincent

[nathanncohen]

comment:9

THaaaaaaaaaaaaaaaaaaaaaaanks !!

[vbraun]

Changed branch from u/ncohen/16720 to 41676cf