RFR(M): 6921087: G1: remove per-GC-thread expansion tables from the fine-grain remembered sets

[John Cuthbertson]

Hi Bengt,

On 06/20/12 11:12, John Cuthbertson wrote:
> Hi Bengt,
>
> A similar question regarding PRT_COUNT_OCCUPIED was asked by Thomas.
>
> It looks like the code under SAMPLE_FOR_EVICTION describe alternative 
> strategies for determining which PerRegionTable to evict from the fine 
> grain entries. If SAMPLE_FOR_EVICTION is defined TRUE then the PRT 
> selected for eviction is chosen from a small subset of the PRTs; 
> otherwise we look at all the PRTs and select the one with most number 
> of entries. Using the non SAMPLE_FOR_EVICTION code will most likely 
> lead to regressions during RSet updating except when _n_fine_entries 
> is less than _fine_eviction_sample_size.

The last sentence should read "... most likely lead to *performance* 
regressions...".

JohnC