OpenJDK / jigsaw / jake / hotspot
changeset 2891:d23d2b18183e
7118202: G1: eden size unnecessarily drops to a minimum
Summary: An integer underflow can cause the RSet lengths to be massively overpredicted which forces the eden size to the minimum.
Reviewed-by: brutisso, johnc
author | tonyp |
---|---|
date | Wed, 07 Dec 2011 12:54:51 -0500 |
parents | f4414323345f |
children | e9b91fd07263 |
files | src/share/vm/gc_implementation/g1/g1CollectorPolicy.cpp |
diffstat | 1 files changed, 9 insertions(+), 3 deletions(-) [+] |
line wrap: on
line diff
--- a/src/share/vm/gc_implementation/g1/g1CollectorPolicy.cpp Mon Nov 28 14:58:31 2011 +0100 +++ b/src/share/vm/gc_implementation/g1/g1CollectorPolicy.cpp Wed Dec 07 12:54:51 2011 -0500 @@ -1549,9 +1549,15 @@ _partially_young_cards_per_entry_ratio_seq->add(cards_per_entry_ratio); } - size_t rs_length_diff = _max_rs_lengths - _recorded_rs_lengths; - if (rs_length_diff >= 0) - _rs_length_diff_seq->add((double) rs_length_diff); + // It turns out that, sometimes, _max_rs_lengths can get smaller + // than _recorded_rs_lengths which causes rs_length_diff to get + // very large and mess up the RSet length predictions. We'll be + // defensive until we work out why this happens. + size_t rs_length_diff = 0; + if (_max_rs_lengths > _recorded_rs_lengths) { + rs_length_diff = _max_rs_lengths - _recorded_rs_lengths; + } + _rs_length_diff_seq->add((double) rs_length_diff); size_t copied_bytes = surviving_bytes; double cost_per_byte_ms = 0.0;