Jim Apple has posted comments on this change. ( http://gerrit.cloudera.org:8080/7438 )
Change subject: IMPALA-4939, IMPALA-4940: Decimal V2 multiplication ...................................................................... Patch Set 6: > > I totally agree about doing anything fancy as a followup. In > > particular, I did not find anything at all useful in AVX to help. > > There is no vectored multiply large enough for this, and FMA > > operations won't help either as they basically only help with > > precision loss on floating point types. > > > > Considering the perf difference is so extreme in some cases, I > > think we should strongly consider either living with the broken > > behavior until we can come up with a solution, or else verifying > > with users making use of the DECIMAL_V2 feature that this will > not > > be a problem. > > AVX2 has 32x8 times 32x8 multiplication. I have elsewhere simulated > 64-bit multiplication with this, combines with additions and > shifts, and received a speedup on 64x4 times 64x4 operations. Oh, and AVX512 has vpmullq for multiplying vectors of 64-bit values. -- To view, visit http://gerrit.cloudera.org:8080/7438 To unsubscribe, visit http://gerrit.cloudera.org:8080/settings Gerrit-Project: Impala-ASF Gerrit-Branch: master Gerrit-MessageType: comment Gerrit-Change-Id: I37ad6232d7953bd75c18dc86e665b2b501a1ebe1 Gerrit-Change-Number: 7438 Gerrit-PatchSet: 6 Gerrit-Owner: Taras Bobrovytsky <tbobrovyt...@cloudera.com> Gerrit-Reviewer: Dan Hecht <dhe...@cloudera.com> Gerrit-Reviewer: Jim Apple <jbapple-imp...@apache.org> Gerrit-Reviewer: Michael Ho <k...@cloudera.com> Gerrit-Reviewer: Taras Bobrovytsky <tbobrovyt...@cloudera.com> Gerrit-Reviewer: Tim Armstrong <tarmstr...@cloudera.com> Gerrit-Reviewer: Zach Amsden <zams...@cloudera.com> Gerrit-Comment-Date: Thu, 28 Sep 2017 01:56:55 +0000 Gerrit-HasComments: No