Path: csiph.com!usenet.pasdenom.info!news.etla.org!nntp-feed.chiark.greenend.org.uk!ewrotcd!feeds.news.ox.ac.uk!news.ox.ac.uk!zen.net.uk!hamilton.zen.co.uk!reader01.nrc01.news.zen.net.uk.POSTED!not-for-mail
From: Nobody <nobody@nowhere.com>
Subject: Re: Short-circuit Logic
Date: Thu, 30 May 2013 23:55:25 +0100
User-Agent: Pan/0.14.2 (This is not a psychotic episode. It's a cleansing moment of clarity.)
Message-Id: <pan.2013.05.30.22.55.17.876000@nowhere.com>
Newsgroups: comp.lang.python
References: <5f101d70-e51f-4531-9153-c92ee2486fd9@googlegroups.com> <51a1fc7b$0$30002$c3e8da3$5496439d@news.astraweb.com> <2abf4e9c-8c3b-4e2f-80c9-50c1f1d75c9d@googlegroups.com> <51a4b5a1$0$29966$c3e8da3$5496439d@news.astraweb.com> <04b90c02-833a-4bad-88ad-ab71178b8f79@googlegroups.com> <48519aa0-d0cd-4ffc-a2f5-2107465321d8@qn4g2000pbc.googlegroups.com> <mailman.2365.1369847746.3114.python-list@python.org> <51a6b969$0$29966$c3e8da3$5496439d@news.astraweb.com> <mailman.2392.1369885517.3114.python-list@python.org> <51a6e6b8$0$11118$c3e8da3@news.astraweb.com> <qotk3mhx78l.fsf@ruuvi.it.helsinki.fi> <51a70df4$0$11118$c3e8da3@news.astraweb.com> <qotbo7sygwz.fsf@ruuvi.it.helsinki.fi>
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
Lines: 19
Organization: Zen Internet
NNTP-Posting-Host: 79bcb7e2.news.zen.co.uk
X-Trace: DXC=@@Q<9VLcJ`j]TR9CI`A?^ja0UP_O8AJol=dR0\ckLKG`WeZ<[7LZNRfd8ZBjHUl7NmM2Z^cWRFGAk50YC[@beFHo
X-Complaints-To: abuse@zen.co.uk
Xref: csiph.com comp.lang.python:46551

On Thu, 30 May 2013 12:07:40 +0300, Jussi Piitulainen wrote:

> I suppose this depends on the complexity of the process and the amount
> of data that produced the numbers of interest. Many individual
> floating point operations are required to be within an ulp or two of
> the mathematically correct result, I think, and the rounding error
> when parsing a written representation of a number should be similar.

Elementary operations (+, -, *, /, %, sqrt) are supposed to be within
+/- 0.5 ULP (for round-to-nearest), i.e. the actual result should be the
closest representable value to the exact result.

Transcendental functions should ideally be within +/- 1 ULP, i.e. the
actual result should be one of the two closest representable values to the
exact result. Determining the closest value isn't always feasible due to
the "table-maker's dilemma", i.e. the fact that regardless of the number
of digits used for intermediate results, the upper and lower bounds
can remain on opposite sides of the dividing line.