-
Notifications
You must be signed in to change notification settings - Fork 915
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
Add row conversion code from spark-rapids-jni #14664
Conversation
Signed-off-by: Nghia Truong <[email protected]>
Signed-off-by: Nghia Truong <[email protected]>
Signed-off-by: Nghia Truong <[email protected]>
Signed-off-by: Nghia Truong <[email protected]>
cpp/include/cudf/row_conversion.hpp
Outdated
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Looks like this is just copied/moved from java/src/main/native/src/row_conversion.hpp
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
The style diff here is because the code in java/
was not formatted with clang-format.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
That accounts for the style diffs. There were material differences in the spark-rapids-jni
version of row_conversion.cu
that's not the easiest to spot in this diff. This remains a temp fix, of course, until the CCCL issue is resolved.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Let's update the description to mention that this is a workaround for the failure in spark-rapids-jni
.
Signed-off-by: Nghia Truong <[email protected]>
|
||
namespace cudf { | ||
namespace jni { |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I admire your thoroughness in this, @ttnghia.
Seems there was a cmake file missed |
Signed-off-by: Nghia Truong <[email protected]>
/merge |
A call to a deprecated `cudf::make_strings_column` factory function was introduced in #14664. This call had been previously fixed in #14461 but was lost when the source file was moved in #14664. Authors: - David Wendt (https://github.com/davidwendt) Approvers: - Bradley Dice (https://github.com/bdice) - Nghia Truong (https://github.com/ttnghia) URL: #14695
This reverts commit 36f56c9. # Conflicts: # cpp/include/cudf/row_conversion.hpp # cpp/tests/transform/row_conversion.cpp # java/src/main/native/src/row_conversion.cu
This is to remove the row conversion code from libcudf. It was move from spark-rapids-jni (by #14664) to temporarily workaround the issue due to conflict of kernel names that causes invalid memory access when calling to `thrust::in(ex)clusive_scan` (NVIDIA/spark-rapids-jni#1567). Now we have fixes for the namespace visibility issue (by marking all libcudf kenels private in rapidsai/rapids-cmake#523 and NVIDIA/cuCollections#422) and need to move back the code. Closes #14853. Authors: - Nghia Truong (https://github.com/ttnghia) Approvers: - David Wendt (https://github.com/davidwendt) - Bradley Dice (https://github.com/bdice) URL: #15234
This temporarily moves the row conversion code from spark-rapids-jni into libcudf. It is necessary to have the row conversion code compiled in a static library to overcome a CCCL issue that triggers invalid memory access when calling to
thrust::in(ex)clusive_scan
(NVIDIA/spark-rapids-jni#1567).This also removes the existing (outdated)
row_conversion
code under thejava
folder in cudf repository.In the future, when we have CCCL updated to fix the issue (1567), we may need to move the code back into spark-rapids-jni.