[REVIEW] Avoid inserting null elements into join hash table when nulls are treated as unequal #6943
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
Please update the changelog in order to start CI tests. View the gpuCI docs here. |
|
Here is the null benchmark running without this change. I will update this as it progresses, but the benchmark is laughable. And with this change: |
jrhemstad
reviewed
Dec 8, 2020
jrhemstad
reviewed
Dec 8, 2020
Thanks for the suggestions! Co-authored-by: Jake Hemstad <jhemstad@nvidia.com>
|
The reason it is so slow without this change is because when nulls are not equal but are inserted into the map, you end up with a huge chain in the hash map. Every time you do an insert of a null value, you end up having to serially iterate that chain. Each slot you visit in the chain requires an |
harrism
requested changes
Dec 8, 2020
Codecov Report
@@ Coverage Diff @@
## branch-0.18 #6943 +/- ##
===============================================
+ Coverage 81.58% 82.02% +0.43%
===============================================
Files 96 96
Lines 15920 16275 +355
===============================================
+ Hits 12989 13350 +361
+ Misses 2931 2925 -6
Continue to review full report at Codecov.
|
hyperbolic2346
added
bug
jrhemstad
requested changes
Dec 9, 2020
jrhemstad
changed the title
jrhemstad
reviewed
Dec 10, 2020
jrhemstad
approved these changes
Dec 10, 2020
nvdbaranec
requested changes
Dec 10, 2020
Thanks for the suggestions. Co-authored-by: nvdbaranec <56695930+nvdbaranec@users.noreply.github.com>
jrhemstad
reviewed
Dec 10, 2020
harrism
requested changes
Dec 11, 2020
Co-authored-by: Mark Harris <mharris@nvidia.com>
nvdbaranec
approved these changes
Dec 11, 2020
harrism
approved these changes
Dec 11, 2020
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This change mirrors what is done in
groupbyto eliminate null-containing columns from the join hash table if nulls not equal is set. This prevents absolute runaway of the process. I added benchmarks for joins with nulls and I can't even get it to finish without these changes. The 195ms test without nulls takes 2,000,000ms to complete and the larger tests I haven't had the patience to even see complete. With this change, the timings are faster than without nulls proportional to the % of nulls. Meaning half the table is nulls means the query is twice as fast as the non-null version, which makes sense.closes #6052