On 2013-10-01 12:30, Erik Joelsson wrote:
Hello again,
I have now regenerated the classlists @b107 and run new comparisons.
To me it looks like the differences are essentially non existant
between new and old classlists and in the previous comparison we saw
no difference between random order and old classlists.
It's very possible that it has an impact on cold starts, as Staffan
suggests, but we have no way (that I know of) of measuring that, and
is that really a valid requirement if we can't measure it?
I'd also like to reiterate that getting rid of the classlists will allow
for rt.jar to be generated by the "normal" build system rules, which
will allow blazingly fast incremental jar updates -- and thus making the
cost of keeping the rt.jar up-to-date in the jdk target negligible, so
we can move generation of the rt.jar from the images target to the jdk
target.
Given this clear advantage if we get rid of it, I think we need some
clear proof of the advantage of keeping it.
/Magnus
