[PyTorch] Fix out-of-bounds array access in atomic_add_vec #138744
Conversation
There is no guarantee that `len` here is enough for a full vector. This was causing at least one test failure on #137426. Differential Revision: [D64857786](https://our.internmc.facebook.com/intern/diff/D64857786/) [ghstack-poisoned]
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/138744
Note: Links to docs will display an error until the docs builds have been completed. ✅ You can merge normally! (1 Unrelated Failure)As of commit b0e8635 with merge base fb0da32 ( FLAKY - The following job failed but was likely due to flakiness present on trunk:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
|
This pull request was exported from Phabricator. Differential Revision: D64857786 |
|
@pytorchbot label "topic: not user facing" |
|
@pytorchbot merge -f "Not need to wait any longer" |
Merge startedYour change will be merged immediately since you used the force (-f) flag, bypassing any CI checks (ETA: 1-5 minutes). Please use Learn more about merging in the wiki. Questions? Feedback? Please reach out to the PyTorch DevX Team |
The correct vector length for ARM64 is 128 bits (16 bytes). We were previously using double this, apparently just because that would be the same length as AVX2. Differential Revision: [D63984039](https://our.internmc.facebook.com/intern/diff/D63984039/) Pull Request resolved: #137426 Approved by: https://github.com/jgong5, https://github.com/malfet ghstack dependencies: #138486, #138542, #138655, #138716, #138744
Stack from ghstack (oldest at bottom):
There is no guarantee that
lenhere is enough for a full vector. This was causing at least one test failure on #137426.Differential Revision: D64857786
cc @voznesenskym @penguinwu @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @wenzhe-nrv @jiayisunx @ipiszy @yf225 @chenyang78 @kadeng @muchulee8 @ColinPeppler @amjames @desertfire @chauhang @aakhundov