On 01/06/2013 06:59 PM, Remi Forax wrote:
On 01/06/2013 12:39 PM, Doug Lea wrote:
Also, I've done a micro-test on my laptop (a modesr 4 cores machine),
trying to calculate the maximum of one miliion values using
LongAccumulator by 4 threads and the calls to
Striped64::longAccumulate are never inlined,
longAccumulate is too big, a better segregation between the fast paths
and the slow ones (that should be pushed in another method) should
solve the issue.
answering myself because I have written something stupid,
we don't care if longAccumulate is inlined or not because it's not a
fast path.
Rémi
