inference: Remove special casing for !
#55271
Merged
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
Keno
added
the
needs pkgeval
|
@nanosoldier |
Keno
added a commit
that referenced
this pull request
Jul 28, 2024
As mentioned in #55271, I'd like to get rid of `istopfunction` for various reasons. Rather than hardcoding constprop overrides based on binding name, this annotates the relevant TypeNames (like the existing `max_methods` override), removing the binding lookups from inference.
Keno
added a commit
that referenced
this pull request
Jul 28, 2024
As mentioned in #55271, I'd like to get rid of `istopfunction` for various reasons. Rather than hardcoding constprop overrides based on binding name, this annotates the relevant TypeNames (like the existing `max_methods` override), removing the binding lookups from inference.
Keno
added a commit
that referenced
this pull request
Jul 28, 2024
As mentioned in #55271, the `istopfunction` binding-based comparisons are problematic. In #55272 and #55273, I attempted to remove the inference special cases for `>:` and `typename` (respectively) entirely, but for differing reasons (`>:` gets too many extra specializations, `typename` loses precision), those PRs are suboptimal. As discussed in #55273, this PR instead moves these functions to Core, so that both `Core.Compiler` and `Base` share the function object, allowing inference to detect them and apply the special handling by simple value-comparison.
Keno
added a commit
that referenced
this pull request
Jul 28, 2024
As mentioned in #55271, the `istopfunction` binding-based comparisons are problematic. In #55272 and #55273, I attempted to remove the inference special cases for `>:` and `typename` (respectively) entirely, but for differing reasons (`>:` gets too many extra specializations, `typename` loses precision), those PRs are suboptimal. As discussed in #55273, this PR instead moves these functions to Core, so that both `Core.Compiler` and `Base` share the function object, allowing inference to detect them and apply the special handling by simple value-comparison.
Keno
added a commit
that referenced
this pull request
Jul 28, 2024
As mentioned in #55271, the `istopfunction` binding-based comparisons are problematic. In #55272 and #55273, I attempted to remove the inference special cases for `>:` and `typename` (respectively) entirely, but for differing reasons (`>:` gets too many extra specializations, `typename` loses precision), those PRs are suboptimal. As discussed in #55273, this PR instead moves these functions to Core, so that both `Core.Compiler` and `Base` share the function object, allowing inference to detect them and apply the special handling by simple value-comparison.
|
The package evaluation job you requested has completed - possible new issues were detected. |
aviatesk
reviewed
Jul 29, 2024
aviatesk
force-pushed
the
kf/nospecialbang
branch
from
a3b3f55
to
73d4666
Compare
|
@nanosoldier |
We have a handful of cases in inference where we look up functions by name (using `istopfunction`) and give them special behavior. I'd like to remove these. They're not only aesthetically ugly, but because they depend on binding lookups, rather than values, they have unclear semantics as those bindings change. They are also unsound should a user use the same name for something different in their own top modules (of course, it's unlikely that a user would do such a thing, but it's bad that they can't). This particular PR removes the special case for `!`, which was there to strengthen the inference result for `!` on Conditional. However, with a little bit of strengthening of the rest of the system, this can be equally well evaluated through the existing InterConditional mechanism.
|
The package evaluation job you requested has completed - possible new issues were detected. |
Keno
removed
the
needs pkgeval
Keno
added a commit
that referenced
this pull request
Aug 1, 2024
…alue As mentioned in #55271, the `istopfunction` binding-based comparisons are problematic. In #55272 and #55273, I attempted to remove the inference special cases for `>:` and `typename` (respectively) entirely, but for differing reasons (`>:` gets too many extra specializations, `typename` loses precision), those PRs are suboptimal. As discussed in #55273, this PR instead moves these functions to Core, so that both `Core.Compiler` and `Base` share the function object, allowing inference to detect them and apply the special handling by simple value-comparison.
Keno
added a commit
that referenced
this pull request
Aug 2, 2024
…55289) As mentioned in #55271, the `istopfunction` binding-based comparisons are problematic. In #55272 and #55273, I attempted to remove the inference special cases for `>:` and `typename` (respectively) entirely, but for differing reasons (`>:` gets too many extra specializations, `typename` loses precision), those PRs are suboptimal. As discussed in #55273, this PR instead moves these functions to Core, so that both `Core.Compiler` and `Base` share the function object, allowing inference to detect them and apply the special handling by simple value-comparison. --- - closes #55273
lazarusA
pushed a commit
to lazarusA/julia
that referenced
this pull request
Aug 17, 2024
…JuliaLang#55288) As mentioned in JuliaLang#55271, I'd like to get rid of `istopfunction` for various reasons. Rather than hardcoding constprop overrides based on binding name, this annotates the relevant TypeNames (like the existing `max_methods` override), removing the binding lookups from inference.
lazarusA
pushed a commit
to lazarusA/julia
that referenced
this pull request
Aug 17, 2024
We have a handful of cases in inference where we look up functions by name (using `istopfunction`) and give them special behavior. I'd like to remove these. They're not only aesthetically ugly, but because they depend on binding lookups, rather than values, they have unclear semantics as those bindings change. They are also unsound should a user use the same name for something differnt in their own top modules (of course, it's unlikely that a user would do such a thing, but it's bad that they can't). This particular PR removes the special case for `!`, which was there to strengthen the inference result for `!` on Conditional. However, with a little bit of strengthening of the rest of the system, this can be equally well evaluated through the existing InterConditional mechanism.
lazarusA
pushed a commit
to lazarusA/julia
that referenced
this pull request
Aug 17, 2024
…uliaLang#55289) As mentioned in JuliaLang#55271, the `istopfunction` binding-based comparisons are problematic. In JuliaLang#55272 and JuliaLang#55273, I attempted to remove the inference special cases for `>:` and `typename` (respectively) entirely, but for differing reasons (`>:` gets too many extra specializations, `typename` loses precision), those PRs are suboptimal. As discussed in JuliaLang#55273, this PR instead moves these functions to Core, so that both `Core.Compiler` and `Base` share the function object, allowing inference to detect them and apply the special handling by simple value-comparison. --- - closes JuliaLang#55273
Keno
added a commit
that referenced
this pull request
Oct 23, 2024
This commit teaches to compiler to update its world bounds whenever it looks at a binding partition, making the compiler sound in the presence of a partitioned binding. The key adjustment is that the compiler is no longer allowed to directly query the binding table without recording the world bounds, so all the various abstract evaluations that look at bindings need to be adjusted and are no longer pure tfuncs. We used to look at bindings a lot more, but thanks to earlier prep work to remove unnecessary binding-dependent code (#55288, #55289 and #55271), these changes become relatively straightforward. Note that as before, we do not create any binding partitions by default, so this commit is mostly preperatory.
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.
We have a handful of cases in inference where we look up functions by name (using
istopfunction) and give them special behavior. I'd like to remove these. They're not only aesthetically ugly, but because they depend on binding lookups, rather than values, they have unclear semantics as those bindings change. They are also unsound should a user use the same name for something differnt in their own top modules (of course, it's unlikely that a user would do such a thing, but it's bad that they can't).This particular PR removes the special case for
!, which was there to strengthen the inference result for!on Conditional. However, with a little bit of strengthening of the rest of the system, this can be equally well evaluated through the existing InterConditional mechanism.