[jit] Decouple ivalue.h from jit_type.h #70119
Closed
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
JIT type and IValue have a mutual dependency because of various reasons today. It makes things worse when we have `jit_type.h` and `ivalue.h` mutually include each other, causing non deterministic name resolutions at different translation units, preventing us safely use symbols from `jit_type.h` in `ivalue.h` . This diff doesn't address the mutual dependency between JIT type and IValue at linking level, but at header level. We choose to remove include of `ivalue.h` from `jit_type.h` because it's way harder to make a type-free header for IValue. The way we achieve this is by removing EnumType (which is the only type depending on IValue in JIT types) from `jit_type.h`, and let downstream users to specifiy an explicit `enum_type.h` as needed. We also move some IValue inline member function definitions back to `ivalue_inl.h` so that `jit_type.h` doesn't need IValue definition to be present. We also remove a seemingly accidental include of `jit_type.h` from `ATen/core/List_inl.h` so that `ivalue.h` can include `jit_type.h` directly, otherwise due to another mutual inclusion between `ivalue.h` and `List_inl.h` we can still get nondeterministic behavior. Differential Revision: [D33155792](https://our.internmc.facebook.com/intern/diff/D33155792/) [ghstack-poisoned]
CI Flow Status⚛️ CI FlowRuleset - Version:
You can add a comment to the PR and tag @pytorchbot with the following commands: # ciflow rerun, "ciflow/default" will always be added automatically
@pytorchbot ciflow rerun
# ciflow rerun with additional labels "-l <ciflow/label_name>", which is equivalent to adding these labels manually and trigger the rerun
@pytorchbot ciflow rerun -l ciflow/scheduled -l ciflow/slowFor more information, please take a look at the CI Flow Wiki. |
🔗 Helpful links
💊 CI failures summary and remediationsAs of commit 4c315e3 (more details on the Dr. CI page): 💚 💚 Looks good so far! There are no failures yet. 💚 💚 This comment was automatically generated by Dr. CI (expand for details).Please report bugs/suggestions to the (internal) Dr. CI Users group. |
facebook-github-bot
added
the
oncall: jit
This was referenced Dec 17, 2021
gmagogsfm
approved these changes
Dec 18, 2021
JIT type and IValue have a mutual dependency because of various reasons today. It makes things worse when we have `jit_type.h` and `ivalue.h` mutually include each other, causing non deterministic name resolutions at different translation units, preventing us safely use symbols from `jit_type.h` in `ivalue.h` . This diff doesn't address the mutual dependency between JIT type and IValue at linking level, but at header level. We choose to remove include of `ivalue.h` from `jit_type.h` because it's way harder to make a type-free header for IValue. The way we achieve this is by removing EnumType (which is the only type depending on IValue in JIT types) from `jit_type.h`, and let downstream users to specifiy an explicit `enum_type.h` as needed. We also move some IValue inline member function definitions back to `ivalue_inl.h` so that `jit_type.h` doesn't need IValue definition to be present. We also remove a seemingly accidental include of `jit_type.h` from `ATen/core/List_inl.h` so that `ivalue.h` can include `jit_type.h` directly, otherwise due to another mutual inclusion between `ivalue.h` and `List_inl.h` we can still get nondeterministic behavior. Differential Revision: [D33155792](https://our.internmc.facebook.com/intern/diff/D33155792/) [ghstack-poisoned]
JIT type and IValue have a mutual dependency because of various reasons today. It makes things worse when we have `jit_type.h` and `ivalue.h` mutually include each other, causing non deterministic name resolutions at different translation units, preventing us safely use symbols from `jit_type.h` in `ivalue.h` . This diff doesn't address the mutual dependency between JIT type and IValue at linking level, but at header level. We choose to remove include of `ivalue.h` from `jit_type.h` because it's way harder to make a type-free header for IValue. The way we achieve this is by removing EnumType (which is the only type depending on IValue in JIT types) from `jit_type.h`, and let downstream users to specifiy an explicit `enum_type.h` as needed. We also move some IValue inline member function definitions back to `ivalue_inl.h` so that `jit_type.h` doesn't need IValue definition to be present. We also remove a seemingly accidental include of `jit_type.h` from `ATen/core/List_inl.h` so that `ivalue.h` can include `jit_type.h` directly, otherwise due to another mutual inclusion between `ivalue.h` and `List_inl.h` we can still get nondeterministic behavior. Differential Revision: [D33155792](https://our.internmc.facebook.com/intern/diff/D33155792/) [ghstack-poisoned]
This was referenced Dec 20, 2021
JIT type and IValue have a mutual dependency because of various reasons today. It makes things worse when we have `jit_type.h` and `ivalue.h` mutually include each other, causing non deterministic name resolutions at different translation units, preventing us safely use symbols from `jit_type.h` in `ivalue.h` . This diff doesn't address the mutual dependency between JIT type and IValue at linking level, but at header level. We choose to remove include of `ivalue.h` from `jit_type.h` because it's way harder to make a type-free header for IValue. The way we achieve this is by removing EnumType (which is the only type depending on IValue in JIT types) from `jit_type.h`, and let downstream users to specifiy an explicit `enum_type.h` as needed. We also move some IValue inline member function definitions back to `ivalue_inl.h` so that `jit_type.h` doesn't need IValue definition to be present. We also remove a seemingly accidental include of `jit_type.h` from `ATen/core/List_inl.h` so that `ivalue.h` can include `jit_type.h` directly, otherwise due to another mutual inclusion between `ivalue.h` and `List_inl.h` we can still get nondeterministic behavior. Differential Revision: [D33155792](https://our.internmc.facebook.com/intern/diff/D33155792/) [ghstack-poisoned]
JIT type and IValue have a mutual dependency because of various reasons today. It makes things worse when we have `jit_type.h` and `ivalue.h` mutually include each other, causing non deterministic name resolutions at different translation units, preventing us safely use symbols from `jit_type.h` in `ivalue.h` . This diff doesn't address the mutual dependency between JIT type and IValue at linking level, but at header level. We choose to remove include of `ivalue.h` from `jit_type.h` because it's way harder to make a type-free header for IValue. The way we achieve this is by removing EnumType (which is the only type depending on IValue in JIT types) from `jit_type.h`, and let downstream users to specifiy an explicit `enum_type.h` as needed. We also move some IValue inline member function definitions back to `ivalue_inl.h` so that `jit_type.h` doesn't need IValue definition to be present. We also remove a seemingly accidental include of `jit_type.h` from `ATen/core/List_inl.h` so that `ivalue.h` can include `jit_type.h` directly, otherwise due to another mutual inclusion between `ivalue.h` and `List_inl.h` we can still get nondeterministic behavior. Differential Revision: [D33155792](https://our.internmc.facebook.com/intern/diff/D33155792/) [ghstack-poisoned]
JIT type and IValue have a mutual dependency because of various reasons today. It makes things worse when we have `jit_type.h` and `ivalue.h` mutually include each other, causing non deterministic name resolutions at different translation units, preventing us safely use symbols from `jit_type.h` in `ivalue.h` . This diff doesn't address the mutual dependency between JIT type and IValue at linking level, but at header level. We choose to remove include of `ivalue.h` from `jit_type.h` because it's way harder to make a type-free header for IValue. The way we achieve this is by removing EnumType (which is the only type depending on IValue in JIT types) from `jit_type.h`, and let downstream users to specifiy an explicit `enum_type.h` as needed. We also move some IValue inline member function definitions back to `ivalue_inl.h` so that `jit_type.h` doesn't need IValue definition to be present. We also remove a seemingly accidental include of `jit_type.h` from `ATen/core/List_inl.h` so that `ivalue.h` can include `jit_type.h` directly, otherwise due to another mutual inclusion between `ivalue.h` and `List_inl.h` we can still get nondeterministic behavior. Differential Revision: [D33155792](https://our.internmc.facebook.com/intern/diff/D33155792/) [ghstack-poisoned]
JIT type and IValue have a mutual dependency because of various reasons today. It makes things worse when we have `jit_type.h` and `ivalue.h` mutually include each other, causing non deterministic name resolutions at different translation units, preventing us safely use symbols from `jit_type.h` in `ivalue.h` . This diff doesn't address the mutual dependency between JIT type and IValue at linking level, but at header level. We choose to remove include of `ivalue.h` from `jit_type.h` because it's way harder to make a type-free header for IValue. The way we achieve this is by removing EnumType (which is the only type depending on IValue in JIT types) from `jit_type.h`, and let downstream users to specifiy an explicit `enum_type.h` as needed. We also move some IValue inline member function definitions back to `ivalue_inl.h` so that `jit_type.h` doesn't need IValue definition to be present. We also remove a seemingly accidental include of `jit_type.h` from `ATen/core/List_inl.h` so that `ivalue.h` can include `jit_type.h` directly, otherwise due to another mutual inclusion between `ivalue.h` and `List_inl.h` we can still get nondeterministic behavior. Differential Revision: [D33155792](https://our.internmc.facebook.com/intern/diff/D33155792/) [ghstack-poisoned]
JIT type and IValue have a mutual dependency because of various reasons today. It makes things worse when we have `jit_type.h` and `ivalue.h` mutually include each other, causing non deterministic name resolutions at different translation units, preventing us safely use symbols from `jit_type.h` in `ivalue.h` . This diff doesn't address the mutual dependency between JIT type and IValue at linking level, but at header level. We choose to remove include of `ivalue.h` from `jit_type.h` because it's way harder to make a type-free header for IValue. The way we achieve this is by removing EnumType (which is the only type depending on IValue in JIT types) from `jit_type.h`, and let downstream users to specifiy an explicit `enum_type.h` as needed. We also move some IValue inline member function definitions back to `ivalue_inl.h` so that `jit_type.h` doesn't need IValue definition to be present. We also remove a seemingly accidental include of `jit_type.h` from `ATen/core/List_inl.h` so that `ivalue.h` can include `jit_type.h` directly, otherwise due to another mutual inclusion between `ivalue.h` and `List_inl.h` we can still get nondeterministic behavior. Differential Revision: [D33155792](https://our.internmc.facebook.com/intern/diff/D33155792/) [ghstack-poisoned]
JIT type and IValue have a mutual dependency because of various reasons today. It makes things worse when we have `jit_type.h` and `ivalue.h` mutually include each other, causing non deterministic name resolutions at different translation units, preventing us safely use symbols from `jit_type.h` in `ivalue.h` . This diff doesn't address the mutual dependency between JIT type and IValue at linking level, but at header level. We choose to remove include of `ivalue.h` from `jit_type.h` because it's way harder to make a type-free header for IValue. The way we achieve this is by removing EnumType (which is the only type depending on IValue in JIT types) from `jit_type.h`, and let downstream users to specifiy an explicit `enum_type.h` as needed. We also move some IValue inline member function definitions back to `ivalue_inl.h` so that `jit_type.h` doesn't need IValue definition to be present. We also remove a seemingly accidental include of `jit_type.h` from `ATen/core/List_inl.h` so that `ivalue.h` can include `jit_type.h` directly, otherwise due to another mutual inclusion between `ivalue.h` and `List_inl.h` we can still get nondeterministic behavior. Differential Revision: [D33155792](https://our.internmc.facebook.com/intern/diff/D33155792/) [ghstack-poisoned]
JIT type and IValue have a mutual dependency because of various reasons today. It makes things worse when we have `jit_type.h` and `ivalue.h` mutually include each other, causing non deterministic name resolutions at different translation units, preventing us safely use symbols from `jit_type.h` in `ivalue.h` . This diff doesn't address the mutual dependency between JIT type and IValue at linking level, but at header level. We choose to remove include of `ivalue.h` from `jit_type.h` because it's way harder to make a type-free header for IValue. The way we achieve this is by removing EnumType (which is the only type depending on IValue in JIT types) from `jit_type.h`, and let downstream users to specifiy an explicit `enum_type.h` as needed. We also move some IValue inline member function definitions back to `ivalue_inl.h` so that `jit_type.h` doesn't need IValue definition to be present. We also remove a seemingly accidental include of `jit_type.h` from `ATen/core/List_inl.h` so that `ivalue.h` can include `jit_type.h` directly, otherwise due to another mutual inclusion between `ivalue.h` and `List_inl.h` we can still get nondeterministic behavior. Differential Revision: [D33155792](https://our.internmc.facebook.com/intern/diff/D33155792/) [ghstack-poisoned]
JIT type and IValue have a mutual dependency because of various reasons today. It makes things worse when we have `jit_type.h` and `ivalue.h` mutually include each other, causing non deterministic name resolutions at different translation units, preventing us safely use symbols from `jit_type.h` in `ivalue.h` . This diff doesn't address the mutual dependency between JIT type and IValue at linking level, but at header level. We choose to remove include of `ivalue.h` from `jit_type.h` because it's way harder to make a type-free header for IValue. The way we achieve this is by removing EnumType (which is the only type depending on IValue in JIT types) from `jit_type.h`, and let downstream users to specifiy an explicit `enum_type.h` as needed. We also move some IValue inline member function definitions back to `ivalue_inl.h` so that `jit_type.h` doesn't need IValue definition to be present. We also remove a seemingly accidental include of `jit_type.h` from `ATen/core/List_inl.h` so that `ivalue.h` can include `jit_type.h` directly, otherwise due to another mutual inclusion between `ivalue.h` and `List_inl.h` we can still get nondeterministic behavior. Differential Revision: [D33155792](https://our.internmc.facebook.com/intern/diff/D33155792/) [ghstack-poisoned]
JIT type and IValue have a mutual dependency because of various reasons today. It makes things worse when we have `jit_type.h` and `ivalue.h` mutually include each other, causing non deterministic name resolutions at different translation units, preventing us safely use symbols from `jit_type.h` in `ivalue.h` . This diff doesn't address the mutual dependency between JIT type and IValue at linking level, but at header level. We choose to remove include of `ivalue.h` from `jit_type.h` because it's way harder to make a type-free header for IValue. The way we achieve this is by removing EnumType (which is the only type depending on IValue in JIT types) from `jit_type.h`, and let downstream users to specifiy an explicit `enum_type.h` as needed. We also move some IValue inline member function definitions back to `ivalue_inl.h` so that `jit_type.h` doesn't need IValue definition to be present. We also remove a seemingly accidental include of `jit_type.h` from `ATen/core/List_inl.h` so that `ivalue.h` can include `jit_type.h` directly, otherwise due to another mutual inclusion between `ivalue.h` and `List_inl.h` we can still get nondeterministic behavior. Differential Revision: [D33155792](https://our.internmc.facebook.com/intern/diff/D33155792/) [ghstack-poisoned]
JIT type and IValue have a mutual dependency because of various reasons today. It makes things worse when we have `jit_type.h` and `ivalue.h` mutually include each other, causing non deterministic name resolutions at different translation units, preventing us safely use symbols from `jit_type.h` in `ivalue.h` . This diff doesn't address the mutual dependency between JIT type and IValue at linking level, but at header level. We choose to remove include of `ivalue.h` from `jit_type.h` because it's way harder to make a type-free header for IValue. The way we achieve this is by removing EnumType (which is the only type depending on IValue in JIT types) from `jit_type.h`, and let downstream users to specifiy an explicit `enum_type.h` as needed. We also move some IValue inline member function definitions back to `ivalue_inl.h` so that `jit_type.h` doesn't need IValue definition to be present. We also remove a seemingly accidental include of `jit_type.h` from `ATen/core/List_inl.h` so that `ivalue.h` can include `jit_type.h` directly, otherwise due to another mutual inclusion between `ivalue.h` and `List_inl.h` we can still get nondeterministic behavior. Differential Revision: [D33155792](https://our.internmc.facebook.com/intern/diff/D33155792/) [ghstack-poisoned]
JIT type and IValue have a mutual dependency because of various reasons today. It makes things worse when we have `jit_type.h` and `ivalue.h` mutually include each other, causing non deterministic name resolutions at different translation units, preventing us safely use symbols from `jit_type.h` in `ivalue.h` . This diff doesn't address the mutual dependency between JIT type and IValue at linking level, but at header level. We choose to remove include of `ivalue.h` from `jit_type.h` because it's way harder to make a type-free header for IValue. The way we achieve this is by removing EnumType (which is the only type depending on IValue in JIT types) from `jit_type.h`, and let downstream users to specifiy an explicit `enum_type.h` as needed. We also move some IValue inline member function definitions back to `ivalue_inl.h` so that `jit_type.h` doesn't need IValue definition to be present. We also remove a seemingly accidental include of `jit_type.h` from `ATen/core/List_inl.h` so that `ivalue.h` can include `jit_type.h` directly, otherwise due to another mutual inclusion between `ivalue.h` and `List_inl.h` we can still get nondeterministic behavior. Differential Revision: [D33155792](https://our.internmc.facebook.com/intern/diff/D33155792/) [ghstack-poisoned]
JIT type and IValue have a mutual dependency because of various reasons today. It makes things worse when we have `jit_type.h` and `ivalue.h` mutually include each other, causing non deterministic name resolutions at different translation units, preventing us safely use symbols from `jit_type.h` in `ivalue.h` . This diff doesn't address the mutual dependency between JIT type and IValue at linking level, but at header level. We choose to remove include of `ivalue.h` from `jit_type.h` because it's way harder to make a type-free header for IValue. The way we achieve this is by removing EnumType (which is the only type depending on IValue in JIT types) from `jit_type.h`, and let downstream users to specifiy an explicit `enum_type.h` as needed. We also move some IValue inline member function definitions back to `ivalue_inl.h` so that `jit_type.h` doesn't need IValue definition to be present. We also remove a seemingly accidental include of `jit_type.h` from `ATen/core/List_inl.h` so that `ivalue.h` can include `jit_type.h` directly, otherwise due to another mutual inclusion between `ivalue.h` and `List_inl.h` we can still get nondeterministic behavior. Differential Revision: [D33155792](https://our.internmc.facebook.com/intern/diff/D33155792/) [ghstack-poisoned]
JIT type and IValue have a mutual dependency because of various reasons today. It makes things worse when we have `jit_type.h` and `ivalue.h` mutually include each other, causing non deterministic name resolutions at different translation units, preventing us safely use symbols from `jit_type.h` in `ivalue.h` . This diff doesn't address the mutual dependency between JIT type and IValue at linking level, but at header level. We choose to remove include of `ivalue.h` from `jit_type.h` because it's way harder to make a type-free header for IValue. The way we achieve this is by removing EnumType (which is the only type depending on IValue in JIT types) from `jit_type.h`, and let downstream users to specifiy an explicit `enum_type.h` as needed. We also move some IValue inline member function definitions back to `ivalue_inl.h` so that `jit_type.h` doesn't need IValue definition to be present. We also remove a seemingly accidental include of `jit_type.h` from `ATen/core/List_inl.h` so that `ivalue.h` can include `jit_type.h` directly, otherwise due to another mutual inclusion between `ivalue.h` and `List_inl.h` we can still get nondeterministic behavior. Differential Revision: [D33155792](https://our.internmc.facebook.com/intern/diff/D33155792/) [ghstack-poisoned]
JIT type and IValue have a mutual dependency because of various reasons today. It makes things worse when we have `jit_type.h` and `ivalue.h` mutually include each other, causing non deterministic name resolutions at different translation units, preventing us safely use symbols from `jit_type.h` in `ivalue.h` . This diff doesn't address the mutual dependency between JIT type and IValue at linking level, but at header level. We choose to remove include of `ivalue.h` from `jit_type.h` because it's way harder to make a type-free header for IValue. The way we achieve this is by removing EnumType (which is the only type depending on IValue in JIT types) from `jit_type.h`, and let downstream users to specifiy an explicit `enum_type.h` as needed. We also move some IValue inline member function definitions back to `ivalue_inl.h` so that `jit_type.h` doesn't need IValue definition to be present. We also remove a seemingly accidental include of `jit_type.h` from `ATen/core/List_inl.h` so that `ivalue.h` can include `jit_type.h` directly, otherwise due to another mutual inclusion between `ivalue.h` and `List_inl.h` we can still get nondeterministic behavior. Differential Revision: [D33155792](https://our.internmc.facebook.com/intern/diff/D33155792/) [ghstack-poisoned]
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.
Stack from ghstack:
JIT type and IValue have a mutual dependency because of various reasons today. It makes things worse when we have
jit_type.handivalue.hmutually include each other, causing non deterministic name resolutions at different translation units, preventing us safely use symbols fromjit_type.hinivalue.h. This diff doesn't address the mutual dependency between JIT type and IValue at linking level, but at header level.We choose to remove include of
ivalue.hfromjit_type.hbecause it's way harder to make a type-free header for IValue. The way we achieve this is by removing EnumType (which is the only type depending on IValue in JIT types) fromjit_type.h, and let downstream users to specifiy an explicitenum_type.has needed. We also move some IValue inline member function definitions back toivalue_inl.hso thatjit_type.hdoesn't need IValue definition to be present.We also remove a seemingly accidental include of
jit_type.hfromATen/core/List_inl.hso thativalue.hcan includejit_type.hdirectly, otherwise due to another mutual inclusion betweenivalue.handList_inl.hwe can still get nondeterministic behavior.Differential Revision: D33155792