-
-
Notifications
You must be signed in to change notification settings - Fork 8.8k
Revert #5755 #6145
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.
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
Revert #5755 #6145
Conversation
I'm still looking into it. Expect some results today. Reverting it might break quantile dmatrix completely. |
@trivialfis At least it does not break the gtest. Let's see whether the Python tests would pass too. |
The relevant PR removes the specialised copy for column wise data and instead uses this:
I wonder if the copy_if is responsible for the extra memory allocation. |
Thrust is using n words to store predicates for the output location of copy_if: https://github.com/thrust/thrust/blob/94e0b1c63fadd5c574c45f05267e45ac136e78ef/thrust/system/detail/generic/copy_if.inl#L64 We can write this more efficiently by redirecting the output of an exclusive scan purely with iterators. Does someone want to try this? Otherwise I can implement myself. |
Let me give a try. |
Codecov Report
@@ Coverage Diff @@
## master #6145 +/- ##
=======================================
Coverage 78.83% 78.83%
=======================================
Files 12 12
Lines 3090 3090
=======================================
Hits 2436 2436
Misses 654 654 Continue to review full report at Codecov.
|
I don't think this is necessary anymore. @hcho3 Thanks for bisecting the code. |
Temporarily revert #5755 to prevent OOM error in the e2e mortgage notebook.