From: Raymond Hettinger Date: Thu, 19 Feb 2009 05:48:05 +0000 (+0000) Subject: Add an example for math.fsum() and elaborate on the accurary note. X-Git-Tag: v3.1a1~140 X-Git-Url: https://granicus.if.org/sourcecode?a=commitdiff_plain;h=f3936f875a94596d2e105838897c933fe9daeb8e;p=python Add an example for math.fsum() and elaborate on the accurary note. --- diff --git a/Doc/library/math.rst b/Doc/library/math.rst index ab39424373..55e03bfa55 100644 --- a/Doc/library/math.rst +++ b/Doc/library/math.rst @@ -80,14 +80,18 @@ Number-theoretic and representation functions .. function:: fsum(iterable) Return an accurate floating point sum of values in the iterable. Avoids - loss of precision by tracking multiple intermediate partial sums. The - algorithm's accuracy depends on IEEE-754 arithmetic guarantees and the - typical case where the rounding mode is half-even. - - .. note:: - - The accuracy of fsum() may be impaired on builds that use - extended precision addition and then double-round the results. + loss of precision by tracking multiple intermediate partial sums:: + + >>> sum([.1, .1, .1, .1, .1, .1, .1, .1, .1, .1]) + 0.99999999999999989 + >>> fsum([.1, .1, .1, .1, .1, .1, .1, .1, .1, .1]) + 1.0 + + The algorithm's accuracy depends on IEEE-754 arithmetic guarantees and the + typical case where the rounding mode is half-even. On some non-Windows + builds, the underlying C library uses extended precision addition and may + occasionally double-round an intermediate sum causing it to be off in its + least significant bit. .. function:: isinf(x)