Fix undefined behavior in integer advanced indexing and indexing functions #1894
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
…ent `ssize_t` Previously, indices were directly cast to `ssize_t` before being clipped or wrapped, causing values to overflow or underflow and giving unreliable results ClipIndex and WrapIndex remade as structs which check the bounds of `ssize_t` against the bounds of the indices type then choose to cast to the appropriate type, before performing clipping and/or wrapping ClipIndex and WrapIndex structs have also been moved to a separate header file, `libtensor/include/utils/indexing_utils.hpp`
Moved common constexpr variables out of branches. Replaced `static constexpr` with `constexpr`. Since these are defined in procedure scope, `static` is not required. Introduced typed temporary variables, so that type deduction for `sycl::min`, `sycl::max`, `sycl::clamp` can work and removed explicit use of their template parameter. Added explicit static_cast on value of `projected` variable computed as IndT type.
This is possible because ProjectorT is literal type (no state and default constructor).
|
Deleted rendered PR docs from intelpython.github.com/dpctl, latest should be updated shortly. 🤞 |
|
Array API standard conformance tests for dpctl=0.19.0dev0=py310hdf72452_209 ran successfully. |
oleksandr-pavlyk
approved these changes
Nov 15, 2024
There was a problem hiding this comment.
LGMT! Thank you for fixing this and adding tests @ndgrigorian
|
Array API standard conformance tests for dpctl=0.19.0dev0=py310hdf72452_212 ran successfully. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This PR proposes a solution to undefined behavior that could occur in some edge cases with integer advanced indexing, where indices OOB for
ssize_t(akastd::ptrdiff_t) would be cast directly tossize_tand overflow or underflow.As
ssize_t/std::ptrdiff_tis defined to be a signed type with the same size assize_t, this means that on 32-bit systems, overflow/underflow could occur for even smaller values.This PR also re-organizes
integer_advanced_indexing.hppby reducing namespace clutter, and moves the rewrittenClipIndexandWrapIndexstructs into a separate header file. This enables them to be re-used more easily in extensions.