Alexey Serbin has posted comments on this change. ( 
http://gerrit.cloudera.org:8080/14086 )

Change subject: [compression] Refactor unit tests and add simple benchmark test
......................................................................


Patch Set 5: Code-Review+2

(2 comments)

http://gerrit.cloudera.org:8080/#/c/14086/4/src/kudu/util/compression/compression-test.cc
File src/kudu/util/compression/compression-test.cc:

http://gerrit.cloudera.org:8080/#/c/14086/4/src/kudu/util/compression/compression-test.cc@120
PS4, Line 120: total_len >> 20
> On my platform, Intel Xeon E5-2620 v3, CentOS 7.3.1611, benchmark output:
Ah, great -- thank you for the information on the benchmark results.

Yes, the essential point here is that the throughput is in order of hundreds of 
{K,M}Bytes/sec, so the rounding would not hide any improvement north of 1%.

In other words, if the result numbers of {K,M}B/sec were low, rounding them 
down might hide performance improvement.  E.g., if the test were run just for 1 
second, then rounding down 19455 bytes to 18 Kbytes with the new implementation 
would result in 'hidden' almost 5% improvement assuming the decompression rate 
of prior implementation was 18432 bytes/sec.


http://gerrit.cloudera.org:8080/#/c/14086/4/src/kudu/util/compression/compression-test.cc@136
PS4, Line 136: total_len >> 20
> The same.
Good catch: yep, it's expected the decompression ratio should be higher than 
the compression ratio for those algos :)

At least, now result for LZ4 looks closer to the benchmark table at 
https://github.com/lz4/lz4



--
To view, visit http://gerrit.cloudera.org:8080/14086
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-Project: kudu
Gerrit-Branch: master
Gerrit-MessageType: comment
Gerrit-Change-Id: If09817d223b98c825d0c8276c8f663b5c5b9eb12
Gerrit-Change-Number: 14086
Gerrit-PatchSet: 5
Gerrit-Owner: Yingchun Lai <[email protected]>
Gerrit-Reviewer: Adar Dembo <[email protected]>
Gerrit-Reviewer: Alexey Serbin <[email protected]>
Gerrit-Reviewer: Kudu Jenkins (120)
Gerrit-Reviewer: Yingchun Lai <[email protected]>
Gerrit-Comment-Date: Tue, 20 Aug 2019 03:22:11 +0000
Gerrit-HasComments: Yes

Reply via email to