-
Notifications
You must be signed in to change notification settings - Fork 447
Fixed sanitizer warnings in RadixSortScanBinsKernel #277
Conversation
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.
Question about the implementation.
Once the patch is finalized we should do a quick check to see if this has any significant perf impact.
@canonizer I'd like to get this in sometime this week to land in 1.13. Should I just remove the dead code mentioned in my review? Is there anything else that this needs to be ready? |
921fcaf
to
794a7c6
Compare
794a7c6
to
3aaeda3
Compare
DVS CL: 30066471 |
- correct per-pass spine length - correctly handling the last scan tile
3aaeda3
to
f0c5702
Compare
Fixed a new warning. DVS CL: 30073602 |
Found this PR from the release notes for the 1.13 RC. What is the practical effect of this change? Without it, we do OOB memory reads (and writes?), which might crash depending on page sizes? |
I don't think there's any practical effect, apart from fixing the sanitizer warning. There has been no out-of-bounds reads. Rather, uninitialized memory was fed into a scan, which caused it to be read, which triggered a warning. However, the results that were derived from uninitialized memory locations weren't used anywhere, so the original warning was benign. |
Fixed sanitizer warnings in
RadixSortScanBinsKernel
.