Change early-bound default args in Python bindings to late-bound #7347
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
In PyBind11, if you specify a default argument for a method, it is evaluated when the Python module is initialized, *not* when the method is called (as you might expect in C++). For defaults that are just constants/literals, this is no big deal, but when calling get_*_target_from_environment, this means it is called at module init time -- also normally not a big deal (since the values ~never change at runtime anyway), with one big exception (no pun intended): if the function throws an exception (e.g. via calling user_assert() or similar), that exception is thrown at Module-initialization time, which is a much more inscrutable crash, and one that is very hard to recover from. This may seem unlikely, but can happen pretty easily if you set (say) HL_JIT_TARGET=host-cuda (or other gpu) and the given GPU runtime isn't present on the given system; the current behavior is basically "make if impossible for the libHalidePython bindings to run", whereas what we want is "runtime exception thrown when you call the method". This changes the relevant methods to use `Target()` as the default value, and inside the method wrapper, if the value passed equals `Target()`, it replaces the value with the righ `get_*_target_from_environment()` call. (This turned up while doing some testing of #6924 on a system without Vulkan available)
steven-johnson
added
the
backport me
|
All clean, PTAL |
abadams
approved these changes
Feb 14, 2023
steven-johnson
added a commit
that referenced
this pull request
Feb 14, 2023
…ython bindings to late-bound (#7348) Change early-bound default args in Python bindings to late-bound In PyBind11, if you specify a default argument for a method, it is evaluated when the Python module is initialized, *not* when the method is called (as you might expect in C++). For defaults that are just constants/literals, this is no big deal, but when calling get_*_target_from_environment, this means it is called at module init time -- also normally not a big deal (since the values ~never change at runtime anyway), with one big exception (no pun intended): if the function throws an exception (e.g. via calling user_assert() or similar), that exception is thrown at Module-initialization time, which is a much more inscrutable crash, and one that is very hard to recover from. This may seem unlikely, but can happen pretty easily if you set (say) HL_JIT_TARGET=host-cuda (or other gpu) and the given GPU runtime isn't present on the given system; the current behavior is basically "make if impossible for the libHalidePython bindings to run", whereas what we want is "runtime exception thrown when you call the method". This changes the relevant methods to use `Target()` as the default value, and inside the method wrapper, if the value passed equals `Target()`, it replaces the value with the righ `get_*_target_from_environment()` call. (This turned up while doing some testing of #6924 on a system without Vulkan available)
ardier
pushed a commit
to ardier/Halide-mutation
that referenced
this pull request
Mar 3, 2024
…ide#7347) In PyBind11, if you specify a default argument for a method, it is evaluated when the Python module is initialized, *not* when the method is called (as you might expect in C++). For defaults that are just constants/literals, this is no big deal, but when calling get_*_target_from_environment, this means it is called at module init time -- also normally not a big deal (since the values ~never change at runtime anyway), with one big exception (no pun intended): if the function throws an exception (e.g. via calling user_assert() or similar), that exception is thrown at Module-initialization time, which is a much more inscrutable crash, and one that is very hard to recover from. This may seem unlikely, but can happen pretty easily if you set (say) HL_JIT_TARGET=host-cuda (or other gpu) and the given GPU runtime isn't present on the given system; the current behavior is basically "make if impossible for the libHalidePython bindings to run", whereas what we want is "runtime exception thrown when you call the method". This changes the relevant methods to use `Target()` as the default value, and inside the method wrapper, if the value passed equals `Target()`, it replaces the value with the righ `get_*_target_from_environment()` call. (This turned up while doing some testing of halide#6924 on a system without Vulkan available)
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.
In PyBind11, if you specify a default argument for a method, it is evaluated when the Python module is initialized, not when the method is called (as you might expect in C++). For defaults that are just constants/literals, this is no big deal, but when calling get_*_target_from_environment, this means it is called at module init time -- also normally not a big deal (since the values ~never change at runtime anyway), with one big exception (no pun intended): if the function throws an exception (e.g. via calling user_assert() or similar), that exception is thrown at Module-initialization time, which is a much more inscrutable crash, and one that is very hard to recover from.
This may seem unlikely, but can happen pretty easily if you set (say) HL_JIT_TARGET=host-cuda (or other gpu) and the given GPU runtime isn't present on the given system; the current behavior is basically "make if impossible for the libHalidePython bindings to run", whereas what we want is "runtime exception thrown when you call the method".
This changes the relevant methods to use
Target()as the default value, and inside the method wrapper, if the value passed equalsTarget(), it replaces the value with the righget_*_target_from_environment()call.(This turned up while doing some testing of #6924 on a system without Vulkan available)
This change should probably be backported to the 15.x branch as well.