Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

[FEA] Improve copying decimal data from CPU columnar data #4784

Closed
sperlingxx opened this issue Feb 15, 2022 · 1 comment · Fixed by #4872
Closed

[FEA] Improve copying decimal data from CPU columnar data #4784

sperlingxx opened this issue Feb 15, 2022 · 1 comment · Fixed by #4872
Assignees
@sperlingxx
Labels
feature request New feature or request performance A performance related task/issue task Work required that improves the product but is not user facing

Comments

@sperlingxx
Copy link
Collaborator

Is your feature request related to a problem? Please describe.
For now, when we copy decimal column vectors during HostColumnarToGpu, it involves a round trip of Decimal encoding and decoding for each value: cv.getDecimal(i, precision, scale).toUnscaledLong(), which can be avoided under some circumstances.

Describe the solution you'd like
Ideally, we can bypass the round trip through feeding cuDF.HostColumnBuilder with the underlying storage(int32/int64/binary), if we know the exact type of upstreaming ColumnVector, such as: WritableColumnVector.
@sperlingxx sperlingxx added feature request New feature or request ? - Needs Triage Need team to review and classify performance A performance related task/issue task Work required that improves the product but is not user facing labels Feb 15, 2022
@sperlingxx sperlingxx self-assigned this Feb 15, 2022
@sameerz sameerz removed the ? - Needs Triage Need team to review and classify label Feb 15, 2022
@sperlingxx
Copy link
Collaborator Author

This issue can be regarded as a sub issue of #4393.

@sperlingxx sperlingxx mentioned this issue Feb 21, 2022
Merged
rapids-bot bot pushed a commit to rapidsai/cudf that referenced this issue Feb 22, 2022
@sperlingxx
JNI: Support appending DECIMAL128 into ColumnBuilder in terms of byte…
cf65ac3 
… array (#10338)

This PR is to add new append method for `HostColumnVector.ColumnBuilder`. This newly-added method `appendDecimal128` allows users to add binary data, which represents the unscaled value of BigDecimal, as cuDF Decimal128.
This PR serves as a prerequisite of NVIDIA/spark-rapids#4784.

Authors:
  - Alfred Xu (https://github.com/sperlingxx)

Approvers:
  - Robert (Bobby) Evans (https://github.com/revans2)

URL: #10338
@sperlingxx sperlingxx mentioned this issue Feb 25, 2022
Merged
sperlingxx added a commit that referenced this issue Mar 7, 2022
@sperlingxx
Speed up copying decimal column from parquet buffer to GPU buffer (#4872
f22795d 
)

Signed-off-by: sperlingxx <[email protected]>

Closes #4784

Adds specialized support for the columnar copy of WritableColumnVector on decimal type. The new implementation copies the unscaled values directly to avoid the round trip of Decimal encoding/decoding
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees

@sperlingxx sperlingxx

Labels
feature request New feature or request performance A performance related task/issue task Work required that improves the product but is not user facing
Projects
None yet
Milestone
No milestone
Development

Successfully merging a pull request may close this issue.

Speed up copying decimal column from parquet buffer to GPU buffer sperlingxx/spark-rapids
2 participants
@sperlingxx @sameerz