The new numbers show the performance on 4 partitions with 4 cores is as good or better than saxon. The two filter queries q00 and q01 are close to saxon's performance when running on 2 partitions.
Improvement Overview q00 >50% q01 >20% q02 25% to 35% q03 about the same q02 and q03 have issue with one of the rewrite rules that could be affecting their numbers. The improvements include short circuiting for conditional expressions, improved object reuse, child step optimization and rewrite rules to remove unnecessary functions. Here are results for the new VXQuery version compared to saxon (Only VXQuery numbers have been updated): Query q00 (500mb) --------------- 2m11.937s Saxon 4m25.269s VXQuery - 1 partition 2m19.062s VXQuery - 2 partitions 1m35.461s VXQuery - 4 partitions Query q01 (500mb) --------------- 2m07.096s Saxon 4m14.810s VXQuery - 1 partition 2m14.814s VXQuery - 2 partitions 1m34.537s VXQuery - 4 partitions Query q02 (500mb) --------------- 2m11.029s Saxon 6m04.867s VXQuery - 1 partition 3m06.819s VXQuery - 2 partitions 2m00.013s VXQuery - 4 partitions Query q03 (500mb) --------------- 1m58.784s Saxon 6m24.056s VXQuery - 1 partition 3m06.804s VXQuery - 2 partitions 2m01.431s VXQuery - 4 partitions Numbers for January 30 can be found here: http://mail-archives.apache.org/mod_mbox/incubator-vxquery-dev/201401.mbox/%3CCAO1XyR6eynnPVXTpXKdPJ%2BeXKMuhHApG8bd1suNqS4iQ9%3DCHHQ%40mail.gmail.com%3E
