harmony-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Ian Rogers <ian.rog...@manchester.ac.uk>
Subject Float/Double comparison performance
Date Tue, 28 Apr 2009 11:54:10 GMT
Hi,

with recent discussion of performance an area that hits our [1]
performance is in sorting arrays of floats. The current Harmony code
that does this uses similar logic to Float.compareTo/compare. The
Harmony code first always explicitly tests for the uncommon NaN cases,
uses none raw bit conversion to an integer and what looks like an
expensive way to test for NaN, and that the raw integer values of -0.0
and 0.0 are ordered. Perhaps someone could optimize this code or tell
me if I would be able to submit a patch (having already looked at and
submitted a similar patch to GNU Classpath) ?

Thanks,
Ian

[1] http://mrp.codehaus.org/

Mime
View raw message