HyperLogLog made thread-safe using atomic operations

[assaron]

HyperLogLog and underlying RegisterSet use atomic operations, so HyperLogLog.offer() can be called from different threads.

[tea-dragon]

I can't see much of a reason for a substantial increase in memory use but that might be a good thing to test also. You might be interested in the library http://code.google.com/p/caliper/ which has some helpful functions and guidence on robust benchmarking. I would also be curious to see how well it scales with multiple threads but I suppose performance might not be the only motivation for thread safety.

[tdunning]

Wouldn't it be better to simply keep a separate HyperLogLog accumulator per
thread? Memory increase would give far better performance than using
atomic objects.

On Fri, Feb 15, 2013 at 10:14 AM, Ian Barfield notifications@github.comwrote:

I can't see much of a reason for a substantial increase in memory use but
that might be a good thing to test also. You might be interested in the
library http://code.google.com/p/caliper/ which has some helpful
functions and guidence on robust benchmarking. I would also be curious to
see how well it scales with multiple threads but I suppose performance
might not be the only motivation for thread safety.

—
Reply to this email directly or view it on GitHubhttps://github.com//pull/17#issuecomment-13617058.

[assaron]

Ted, it's seems reasonable. I'll try that.