Fwd: Request for review(M): 6923991: G1: improve scalability of RSet scanning

[Igor Veresov]

-------- Original Message --------
Subject: Request for review(M): 6923991: G1: improve scalability of RSet 
scanning
Date: Fri, 05 Feb 2010 20:29:15 -0800
From: Igor Veresov <Igor.Veresov at Sun.COM>
To: g1 at sun.com

G1 showed poor load balancing and high contention behavior during RSet
scanning on CMT machines as well as on many-core x86. Previously
implemented exponential skipping had unacceptable performance when it
came to a fairly large processor count. The fix is to implement
block-based work stealing in the rset scanning phase. It was also
required to get rid of copying during this phase and postpone it until
the main copying phase, where we have a fine-grain load balancing on the 
object granularity.

I've also made the size of the sparse PRT entry and number of entries in 
the fine-grain PRT depend on the region size.

Webrev: http://cr.openjdk.java.net/~iveresov/6923991/webrev.00/


igor