5.9: [FieldSensitivePL] Fix vectorization. #66735
Merged
Conversation
This file contains hidden or 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
According to language rules, such lifetimes are fixed and the relative order of their deinits is guaranteed. rdar://110913116
The members were declared but undefined.
Its storage vector is intended to be of some type like `std::vector<std::pair<Key, Optional<Value>>>`, i.e., some collection of pairs whose `second` is an `Optional<Value>`. So when constructing a default instance of that pair, just construct an Optional in the None case.
FieldSensitivePrunedLiveness is used as a vectorization of
PrunedLiveness. An instance of FSPL with N elements needs to be able to
represent the same states as N instances of PL.
Previously, it failed to do that in two significant ways:
(1) It attempted to save space for which elements were live by using
a range. This failed to account for instructions which are users of
non-contiguous fields of an aggregate.
apply(
@owned (struct_element_addr %s, #S.f1),
@owned (struct_element_addr %s, #S.f3)
)
(2) It used a single bit to represent whether the instruction was
consuming. This failed to account for instructions which consumed
some fields and borrowed others.
apply(
@owned (struct_element_addr %s, #S.f1),
@guaranteed (struct_element_addr %s, #S.f2)
)
The fix for (1) is to use a bit vector to represent which elements
are used by the instruction. The fix for (2) is to use a second bit
vector to represent which elements are _consumed_ by the instruction.
Adapted the move-checker to use the new representation.
rdar://110909290
|
@swift-ci please test |
tbkka
approved these changes
Jun 19, 2023
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.
Description: Fix (and enable fixing) a class of miscompiles in the move-only address checker.
The move-only address checker relies heavily on FieldSensitivePrunedLiveness (FSPL), a vectorization of PrunedLiveness (PL). FSPL of width N needs to be able to represent exactly the same information as N copies of PL. In particular, for any instruction, must be able to be a non-user, a non-consuming user, or a consuming user of each of the N fields.
Previously, however, that was not true. FSPL stored a range
[i, i+k)of elements at which a given instruction was live. Additionally it stored only a single bit for whether the instruction did or did not consume that range.The result was incorrect liveness computations in cases where disjoint fields were used, such as
and in cases where not all the fields had the same consumed-ness
(In the above,
Sis a move-only struct with three fields, f1, f2, f3 in that order.)The fix is to change the record corresponding to an instruction to have enough information to store all these possibilities. Here, the representation of two bit vectors of length N, one for live-at and a second for consumed-at, is implemented.
Risk: Low. The fix is involved, but the current behavior is definitely wrong.
Scope: Narrow. This only affects move-only types.
Original PR: #66690
Reviewed By: Andrew Trick ( @atrick )
Testing: Current unit tests and two new tests which were previously miscompiled--though note that a second patch will be necessary to fix the second of these two.
Resolves: rdar://110909290