Mercurial > octave

comparison NEWS @ 30228:39a4ab124fd0

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
factor.m: significant speedup for large input quantities (> 1e14) (bug #61129). - scripts/specfun/factor.m: overhaul function for performance. - NEWS: announce speedup. Co-authored-by: Michael Leitner <michael.leitner@frm2.tum.de>
author Arun Giridhar <arungiridhar@gmail.com>
date Wed, 06 Oct 2021 10:09:48 +0900
parents 50ab0178b6fc
children 72b6b886b9d0
comparison
equal deleted inserted replaced
30227:b00ff462e0f2 30228:39a4ab124fd0
73 avoid these kinds of problems either: 1) declare the first integer to be 73 avoid these kinds of problems either: 1) declare the first integer to be
74 of the desired size such as `[0x1u32; 0x100; 0x10000]`, or 2) pad 74 of the desired size such as `[0x1u32; 0x100; 0x10000]`, or 2) pad
75 constants in array expressions with leading zeros so that they use the 75 constants in array expressions with leading zeros so that they use the
76 same number of digits for each value such as 76 same number of digits for each value such as
77 `[0x00_00_01; 0x00_01_00; 0x01_00_00]`. 77 `[0x00_00_01; 0x00_01_00; 0x01_00_00]`.
78
79 - The `factor` function has been overhauled for speed. Especially, for
80 large input quantities (> 1e14) with small prime factors, e.g.
81 `intmax ("uint64")`, the function can be up to 10,000 times faster.
82 However, for large input quantities with large prime factors, the function
83 can be up to 4x faster.
78 84
79 - The `mldivide` function (i.e., the `\` operator) now uses an LU 85 - The `mldivide` function (i.e., the `\` operator) now uses an LU
80 decomposition to solve nearly singular full square matrices. This is 86 decomposition to solve nearly singular full square matrices. This is
81 Matlab-compatible and yields results which more nearly minimize `norm 87 Matlab-compatible and yields results which more nearly minimize `norm
82 (A*x - b)`. Previously, Octave computed a minimum-norm solution. 88 (A*x - b)`. Previously, Octave computed a minimum-norm solution.