Refactor get_scalar_constant_value
#643
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
brandonwillard
added
enhancement
brandonwillard
force-pushed
the
refactor-get_scalar_constant_value
branch
4 times, most recently
from
b208cec
to
090661c
Compare
Codecov Report
@@ Coverage Diff @@
## main #643 +/- ##
==========================================
+ Coverage 77.10% 77.15% +0.04%
==========================================
Files 156 156
Lines 46909 47007 +98
Branches 10259 10281 +22
==========================================
+ Hits 36170 36269 +99
+ Misses 8158 8151 -7
- Partials 2581 2587 +6
|
brandonwillard
force-pushed
the
refactor-get_scalar_constant_value
branch
from
980e944
to
2d1ff8a
Compare
kc611
reviewed
Nov 3, 2021
kc611
reviewed
Nov 3, 2021
brandonwillard
force-pushed
the
refactor-get_scalar_constant_value
branch
2 times, most recently
from
a341c88
to
ddd9ebe
Compare
* Convert tests to pytest `parametrized` tests * Use `pytest.raises` * Remove unused timing code and methods
These changes "flatten" the nested conditions that lead to the replacement logic. They also clarify the tests that are labeled as being for `local_add_specialize`, which actually aren't. Also, the unrelated tests implicitly relied on the canonicalizations built into `get_scalar_constant_value`; now, the actual canonicalizations are required in anticipation of `get_scalar_constant_value`'s replacement.
brandonwillard
force-pushed
the
refactor-get_scalar_constant_value
branch
from
ddd9ebe
to
4c4eb86
Compare
* Use `pytest.mark.parameterize` * Remove reliance on `get_scalar_constant_value`
brandonwillard
force-pushed
the
refactor-get_scalar_constant_value
branch
from
4c4eb86
to
3a3b122
Compare
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.
This PR adds missing canonicalizations and misc. updates that relate to
get_scalar_constant_value.get_scalar_constant_valueuses a confusing and poorly defined abstraction: it attempts to return the non-symbolic value of a symbolic scalar graph and/or the "underlying" scalar value of a graph. The latter concept of an "underlying" scalar value is ambiguous and very limited, since it only applies in a few cases—like the outputs ofAllocOps, because they use a scalar input (i.e. the "underlying" scalar value) to construct an array/tensor.get_scalar_constant_valueforAllocandSecondIts implementation is a hard to follow (and maintain) set of conditions, iterations, and recursions—all of which are a mix of existing/missing constants-based canonicalizations and re-implemented constant folding steps.
The goal of this PR is to further the replacement/removal of
get_scalar_constant_valuewith a more complete set of optimizations, and, when/if necessary, explicit use of constant folding.This PR also implements some of the missing
Shape-related canonicalizations mentioned in #642.