RFR (M): JDK-8030177: G1: Enable TLAB resizing

[Thomas Schatzl]

On Thu, 2014-01-23 at 10:27 +0100, Bengt Rutisson wrote:
> Hi all,
> 
> I went through this change with Stefan Karlsson. Here is an updated 
> webrev based on his suggestions:
> 
> http://cr.openjdk.java.net/~brutisso/8030177/webrev.01/
> 
> Here is a webrev with just the changes that Stefan suggested:
> 
> http://cr.openjdk.java.net/~brutisso/8030177/webrev.01_delta/

A few more cleanup requests:

 - G1CollectedHeap::unsafe_max_tlab_alloc(Thread* ignored) const could
use the new max_tlab_size() method too instead of doing the same
calculation. It seems to return a just too large value too..

 - the comment for G1CollectedHeap::max_tlab_size() could probably
improved, e.g. "TLABs should not contain..." instead of "TLABs should
not be containing..."

 - in G1CollectedHeap::tlab_capacity(Thread), why not use
g1_policy->young_list_target_length() - young_list()->length() as a
better approximation of the remaining amount of space?

 - ThreadLocalAllocBuffer::set_max_size() does not seem to be used
anywhere any more. Which makes the assert() in max_size() fail always. 
(I grepped
http://cr.openjdk.java.net/~brutisso/8030177/webrev.01/hs-gc9.patch and
the only occurrence of set_max_size() is its declaration).

Is the patch complete?

 - not sure why the change makes G1CollectedHeap::young_list() const
while adding young_list_target_lenght() non-const. Any reasons?

 - could the patch initialize the other members of
ThreadLocalAllocBuffer in
ThreadLocalAllocBuffer::ThreadLocalAllocBuffer() too?

Thanks,
  Thomas