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

[BUG] Sum inaccurate result #2205

Open
tfeher opened this issue Feb 28, 2024 · 0 comments
Open

[BUG] Sum inaccurate result #2205

tfeher opened this issue Feb 28, 2024 · 0 comments
Assignees
@mfoerste4
Labels
bug Something isn't working

Comments

@tfeher
Copy link
Contributor

tfeher commented Feb 28, 2024

Describe the bug
The sum kernel does not handle underflows correctly, and that leads to inaccurate results.

Steps/Code to reproduce bug

As reported by @lijinf2:

We also did an experiment of summing up a column of 0.1 across 130,382,757 rows. Observed raft returns 12,528,200.

Expected behavior
Compensate for underflow e.g. by Kahan summation or a similar method.
@tfeher tfeher added the bug Something isn't working label Feb 28, 2024
@mfoerste4 mfoerste4 self-assigned this Mar 12, 2024
@mfoerste4 mfoerste4 mentioned this issue Mar 13, 2024
Merged
rapids-bot bot pushed a commit that referenced this issue Mar 16, 2024
@mfoerste4
Fix illegal acces mean/stdev, sum add Kahan Summation (#2223)
7335267 
This PR addresses #2204 and #2205. 


* fixes illegal access / test coverage for mean row-wise kernel
* fixes illegal access / test coverage for stdev row-wise kernel
* modified sum kernels to utilize Kahan/Neumaier summation per thread, also increase load per thread to benefit from this

FYI, @tfeher

Authors:
  - Malte Förster (https://github.com/mfoerste4)

Approvers:
  - Tamas Bela Feher (https://github.com/tfeher)

URL: #2223
@mfoerste4 mfoerste4 mentioned this issue Jun 25, 2024
Open
@mfoerste4 mfoerste4 mentioned this issue Jul 16, 2024
Merged
rapids-bot bot pushed a commit that referenced this issue Jul 24, 2024
@mfoerste4
Consolidate SUM reductions (#2381)
759095a 
This PR consists of multiple parts:

1. redirect custom reduction kernels within `stats `namespace to `linalg::reduce`
2. Specialize reduction kernels for addition utilizing the _Kahan-Babushka-Neumaier-Sum_ [link](https://en.wikipedia.org/wiki/Kahan_summation_algorithm)
3. Slightly adjust kernel heuristics for coalesced reductions

This should address #2366 and #2205. With the kernel heuristics adjusted the maximum performance drop is 4%.

FYI, @tfeher

Authors:
  - Malte Förster (https://github.com/mfoerste4)

Approvers:
  - Tamas Bela Feher (https://github.com/tfeher)
  - Corey J. Nolet (https://github.com/cjnolet)

URL: #2381
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees

@mfoerste4 mfoerste4

Labels
bug Something isn't working
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
2 participants
@tfeher @mfoerste4