forked from Mirrors/privacore-open-source-search-engine
try to fix performance slam when compiling
like 640k facet list from each of 32 shards. hashtable was not hashing well and destroying the complexity.
This commit is contained in:
@ -1032,7 +1032,12 @@ bool PosdbTable::allocTopTree ( ) {
|
||||
// stats. it has to be done at the aggregator node.
|
||||
if ( ! qt->m_facetHashTable.set ( 4,sizeof(FacetEntry),
|
||||
slots,NULL,0,false,
|
||||
0,"qfht" ) )
|
||||
0,"qfht"
|
||||
// use magic b/c keys seems
|
||||
// pretty similar in the
|
||||
// lower bits sometimes
|
||||
, true
|
||||
) )
|
||||
return false;
|
||||
// make it nongrowable because we'll be in a thread
|
||||
qt->m_facetHashTable.setNonGrow();
|
||||
|
Reference in New Issue
Block a user