[
https://issues.apache.org/jira/browse/MATH-1290?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15248891#comment-15248891
]
Gilles commented on MATH-1290:
------------------------------
bq. I don't agree that anything has changed, the Math.ulp will just cast the 1
to 1.0 or 1.0f
{noformat}
Math.ulp(1.0) != Math.ulp(1.0f)
{noformat}
The former is {{2.220446049250313E-16}} and the latter is
{{1.1920928955078125E-7}}.
And
{noformat}
Math.ulp(1) == Math.ulp(1.0f)
{noformat}
When testing the conversion between {{Complex}} and {{double[]}}, you don't
want to not notice a loss of 9 orders of magnitude.
Actually, the tolerances should probably be set to 0 (zero) since there is no
computation involved, just reading a field and writing it back into an array
and the other way around.
> Additions to Complex Utils; creation of conversion methods between Complex[],
> double[], float[], and interleaved arrays
> -----------------------------------------------------------------------------------------------------------------------
>
> Key: MATH-1290
> URL: https://issues.apache.org/jira/browse/MATH-1290
> Project: Commons Math
> Issue Type: Improvement
> Affects Versions: 3.5
> Reporter: Eric Barnhill
> Priority: Minor
> Labels: newbie, patch
> Fix For: 4.0
>
> Attachments: ComplexUtils.java, ComplexUtils.java, ComplexUtils.java,
> ComplexUtils.java, ComplexUtilsTest.java, ComplexUtilsTest.java,
> ComplexUtilsTest.java, ComplexUtilsTest.java, LaguerreSolver.java
>
> Original Estimate: 24h
> Remaining Estimate: 24h
>
> I propose to add several new methods to the ComplexUtils class enabling easy
> conversion between Complex[] arrays, real or imaginary double[] or float[]
> arrays, and interleaved complex double[] or float[] arrays. The last two in
> particular are beneficial to have for OpenCL implementations and the
> JTransforms library.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
