optimizations: better modeling and codegen for apply and svec calls #59548
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
vtjnash
changed the title
aviatesk
reviewed
Sep 14, 2025
Comment on lines
+1423
to
+1425
| elseif is_known_call(stmt, Core._apply_iterate, compact) | ||
| length(stmt.args) >= 4 || continue | ||
| lift_apply_args!(compact, idx, stmt, 𝕃ₒ) |
There was a problem hiding this comment.
Just commenting for reference, and we don't have to do this in this PR, but I startd to think it'd be better to make this kind of optimization independent of sroa_pass!.
There was a problem hiding this comment.
We might want to instead just rename the pass GVN or MemSSAOpt, since doing everything in one pass is probably a lot more efficient, and either alternative name would reflect that this does general memory-value-replacement optimizations
vtjnash
force-pushed
the
jn/svec-opt-apply
branch
from
fde4599 to
66687c1
Compare
- Use svec instead of tuple for arguments (better match for ABI which will require boxes) - Directly forward single svec argument, both runtime and codegen, without copying. - Optimize all consistant builtin functions of constant arguments, not just ones with special tfuncs. Reducing code duplication and divergence. - Codegen for `svec()` directly, so optimizer can see each store (and doesn't have to build the whole thing on the stack first).
vtjnash
force-pushed
the
jn/svec-opt-apply
branch
from
66687c1 to
19ad3be
Compare
Without a release store, it seems LLVM considers it a data race to have read the initial state on another thread. Marking this as a release store seems sufficient to prevent that optimization. It is also more consistent with how we initialize and write to most other structs, particularly since #55767. Fixes #59547 (more)
aviatesk
pushed a commit
that referenced
this pull request
Sep 19, 2025
…59548) - Use svec instead of tuple for arguments (better match for ABI which will require boxes) - Directly forward single svec argument, both runtime and codegen, without copying. - Optimize all consistant builtin functions of constant arguments, not just ones with special tfuncs. Reducing code duplication and divergence. - Codegen for `svec()` directly, so optimizer can see each store (and doesn't have to build the whole thing on the stack first). Written with help by Claude
|
Backported to 1.12. |
25 tasks
aviatesk
added a commit
that referenced
this pull request
Sep 19, 2025
Further improves the implementation from #59548. Specifically, uses `widenconst` to enable conversion of `tuple` calls that have become `PartialStruct`, and removes incorrect comments and unused arguments. Also adds some Julia-IR level tests.
aviatesk
added a commit
that referenced
this pull request
Sep 30, 2025
Adds a dedicated `_svec_len_nothrow` function that does more precise `:nothrow` modeling of `Core._svec_len` implemented in #59548.
aviatesk
added a commit
that referenced
this pull request
Sep 30, 2025
Adds a dedicated `_svec_len_nothrow` function that does more precise `:nothrow` modeling of `Core._svec_len` implemented in #59548.
xal-0
pushed a commit
to xal-0/julia
that referenced
this pull request
Sep 30, 2025
…Lang#59548 (JuliaLang#59601) Further improves the implementation from JuliaLang#59548. Specifically, uses `widenconst` to enable conversion of `tuple` calls that have become `PartialStruct`, and removes incorrect comments and unused arguments. Also adds some Julia-IR level tests.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
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.
svec()directly, so optimizer can see each store (and doesn't have to build the whole thing on the stack first).Written by Claude