RFR (M): 8027476: Improve performance of Stringtable unlink, 8027455: Improve symbol table scan times during gc pauses

Thomas Schatzl thomas.schatzl at oracle.com
Thu Jan 16 13:07:43 PST 2014


Hi,

  thanks for the review.

On Thu, 2014-01-16 at 13:14 -0500, Coleen Phillimore wrote:
> Thomas,
> 
> In symbolTable.cpp shouldn't this be buckets_unlink_or_oops_do() ?
> 
> +void StringTable::buckets_unlink_or_do(BoolObjectClosure* is_alive, OopClosure* f, int start_idx, int end_idx, int* processed, int* removed) {
> 
> Actually, I see where you got this.  StringTable::buckets_do should be 
> buckets_oops_do too.

I fixed both cases in a new webrev:
http://cr.openjdk.java.net/~tschatzl/8027476/webrev.1/

> The symbolTable.cpp/hpp changes look fine.  The code for SymbolTable and 
> StringTable is more similar now but there isn't enough common code to 
> add to the superclass HashTable<>. 

I know. I thought about that too.

>  We should separate these into their 
> own files later.  I only clicked on the gc files, but you have reviews 
> for these.

Thanks a lot,
  Thomas




More information about the hotspot-runtime-dev mailing list