[ImageJ-devel] Lock-free bit fields, was Re: What's left to do for ImgLib2

[Johannes Schindelin]

Hi Tobias,

On Thu, 30 Oct 2014, Tobias Pietzsch wrote:

> We'll have to see what the overhead of the “read-value-again-to-check”
> operation is.

For the record: synchronization is *always* more expensive than a simple
read-value-again check. I am sure that it is relatively easy to use the
junit-benchmarks [*1*] to prove that.

Ciao,
Johannes

Footnote *1*: Please note that we forked junit-benchmarks because it
requires a relatively ugly fix to work with @Before/@After methods, a fix
so ugly that upstream preferred to keep the bug rather than accepting
something as Gollum-like. The artifact was already deployed as
org.scijava:junit-benchmarks:0.7.3-scijava and it requires an upgrade to
JUnit >= 4.11.