[REVIEW] Fix incorrect output from averages with filters in partial only mode #612
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
revans2
previously approved these changes
Aug 27, 2020
abellina
requested changes
Aug 27, 2020
sql-plugin/src/main/scala/org/apache/spark/sql/rapids/AggregateFunctions.scala
Outdated
Show resolved
Hide resolved
|
Addressed review comments @abellina and additionally took out a test that was redundant after this fix. Please take a look. |
kuhushukla
changed the title
|
build |
abellina
reviewed
Aug 27, 2020
sql-plugin/src/main/scala/org/apache/spark/sql/rapids/AggregateFunctions.scala
Outdated
Show resolved
Hide resolved
abellina
previously approved these changes
Aug 27, 2020
|
build |
|
Waiting for approvals after which, I will rebase-squash and sign off. |
Signed-off-by: Kuhu Shukla <[email protected]>
kuhushukla
force-pushed
the
filter_partial
branch
from
a89891c
to
f291564
Compare
|
Squashed and signed off, re-runing CI |
abellina
approved these changes
Aug 27, 2020
|
build |
revans2
approved these changes
Aug 27, 2020
nartal1
pushed a commit
to nartal1/spark-rapids
that referenced
this pull request
Jun 9, 2021
…DIA#612) Signed-off-by: Kuhu Shukla <[email protected]> Co-authored-by: Kuhu Shukla <[email protected]>
nartal1
pushed a commit
to nartal1/spark-rapids
that referenced
this pull request
Jun 9, 2021
…DIA#612) Signed-off-by: Kuhu Shukla <[email protected]> Co-authored-by: Kuhu Shukla <[email protected]>
tgravescs
pushed a commit
to tgravescs/spark-rapids
that referenced
this pull request
Nov 30, 2023
…IDIA#612) Signed-off-by: spark-rapids automation <[email protected]> Signed-off-by: spark-rapids automation <[email protected]>
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.
Fixes #155 .
This is WIP for a couple reasons and also as I want to get more comments on the approach shown in the PR.
As the associated issue explains in some detail, how we are ending up with nulls being sent down to the CPU's final aggregation. The fix tries to circumvent this by special casing how averages are handled by the case-when of the filter. One thing to note is, we need this special case since we don't have a clean way to pass the filter down to the
GpuAverageDeclarativeAggregate itself. When anullcomes into the filter we want it not pass down any further but instead default to(0.0, 0). This seems to be specific for averages while other aggregates likecountrequire nulls. I had to use another case-when withisnotnullto make this happen. I have some concerns on the else condition added in this PR and would like to know what others think is a better way to do this. I tried some games with having initialValues be reused in the filter it was not straight forward what do to with their data types. At the end, I am also open to use this PR to morph the solution to put the hammer on filters+avg+partial_only_conf to fall back on the CPU, but that is not my first preference. Tagging @abellina for some discussion to get this PR going.