Yao Xu has posted comments on this change. ( http://gerrit.cloudera.org:8080/13703 )
Change subject: KUDU-2483 (part 1/3) Add bloom filter predicate for Java API ...................................................................... Patch Set 1: (2 comments) http://gerrit.cloudera.org:8080/#/c/13703/1/java/kudu-client/src/main/java/org/apache/kudu/util/BloomFilter.java File java/kudu-client/src/main/java/org/apache/kudu/util/BloomFilter.java: http://gerrit.cloudera.org:8080/#/c/13703/1/java/kudu-client/src/main/java/org/apache/kudu/util/BloomFilter.java@31 PS1, Line 31: space-efficient filter > unrelated to this patch, but for best performance, maybe we should > consider one of the cacheline-aware bloom alternatives? > http://www.vldb.org/pvldb/vol12/p502-lang.pdf is a recent reference > on this topic that specifically pertains to this scenario (bloom > pushdown across joins) > > Given we haven't made use of the bloom stuff yet in Impala or Spark > (afaik outside of experimental branches) maybe it's not too late to > change this? It’s cool. I will try to implement it in kudu (maybe next week or later). http://gerrit.cloudera.org:8080/#/c/13703/1/java/kudu-client/src/main/java/org/apache/kudu/util/BloomFilter.java@292 PS1, Line 292: long bitPos = h % nBits; > nit: can we constrain the number of bits to be a power of two, so > that we can optimize this to be a bit mask operation? or is that > not feasible? I think this is feasible, but it may waste some memory. I think memory should not be a serious problem. -- To view, visit http://gerrit.cloudera.org:8080/13703 To unsubscribe, visit http://gerrit.cloudera.org:8080/settings Gerrit-Project: kudu Gerrit-Branch: master Gerrit-MessageType: comment Gerrit-Change-Id: If96ec164e86091f8fb3b7f92b3623ca3e728edfd Gerrit-Change-Number: 13703 Gerrit-PatchSet: 1 Gerrit-Owner: Yao Xu <[email protected]> Gerrit-Reviewer: Grant Henke <[email protected]> Gerrit-Reviewer: Kudu Jenkins (120) Gerrit-Reviewer: Todd Lipcon <[email protected]> Gerrit-Reviewer: Yao Xu <[email protected]> Gerrit-Comment-Date: Tue, 25 Jun 2019 15:45:47 +0000 Gerrit-HasComments: Yes
