[
https://issues.apache.org/jira/browse/NUMBERS-156?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17341499#comment-17341499
]
Matt Juntunen commented on NUMBERS-156:
---------------------------------------
bq. How are you generating the random vectors? Roughly what is the power of 2
exponent for each double, and the range of exponents over the numbers that are
summed?
I used a modified version of a random double generator from one of your
performance tests. For the benchmark posted above, I used {{maxExp = +550}} and
{{minExp = -550}}.
{code:java}
private double randomDouble() {
// Create random doubles using random bits in the sign bit and the
mantissa.
final long mask = ((1L << 52) - 1) | 1L << 63;
final long bits = rng.nextLong() & mask;
// The exponent must be unsigned so + 1023 to the signed exponent
final int expRange = Math.abs(maxExp - minExp);
final long exp = rng.nextInt(expRange) + minExp + 1023;
return Double.longBitsToDouble(bits | (exp << 52));
}
{code}
bq. I would favour the verbose manhatten, euclidian and maximum.
Same here. I'll start working toward that.
> SafeNorm 3D overload
> --------------------
>
> Key: NUMBERS-156
> URL: https://issues.apache.org/jira/browse/NUMBERS-156
> Project: Commons Numbers
> Issue Type: Improvement
> Reporter: Matt Juntunen
> Priority: Major
>
> We should create an overload of {{SafeNorm.value}} that accepts 3 arguments
> to potentially improve performance for 3D vectors.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
