fix: preserve non-finite values in Float.scaleB - #14690
Open
felix314159 wants to merge 2 commits into
Open
Conversation
github-actions
Bot
added
the
toolchain-available
|
Mathlib CI status (docs):
|
Collaborator
|
Reference manual CI status:
|
This PR makes `Float.scaleB` and `Float32.scaleB` preserve infinities, NaN, and signed zero when scaling by exponents too large for the native `scalbn` interface. Handle boxed exponents using the limiting IEEE behavior and add regression coverage for both floating-point precisions.
github-actions
Bot
added
the
mathlib4-nightly-available
mathlib-lean-pr-testing
Bot
added
the
builds-mathlib
Contributor
|
You can add labels by writing a comment that just contains the name of the label (you can also delete the comment afterwards). You might need to edit your PR description back and forth for the check to notice the label change though. |
Rob23oba
reviewed
Aug 5, 2026
Comment on lines
+12
to
+22
| public def main : IO Unit := do | ||
| IO.println ((Float.inf.scaleB hugeNegativeExponent).isInf) | ||
| IO.println (((-Float.inf).scaleB hugeNegativeExponent).isInf) | ||
| IO.println ((Float.nan.scaleB hugeNegativeExponent).isNaN) | ||
| IO.println (((-1.0 : Float).scaleB hugeNegativeExponent).toBits == (-0.0 : Float).toBits) | ||
| IO.println (((-0.0 : Float).scaleB hugePositiveExponent).toBits == (-0.0 : Float).toBits) | ||
| IO.println ((Float32.inf.scaleB hugeNegativeExponent).isInf) | ||
| IO.println (((-Float32.inf).scaleB hugeNegativeExponent).isInf) | ||
| IO.println ((Float32.nan.scaleB hugeNegativeExponent).isNaN) | ||
| IO.println (((-1.0 : Float32).scaleB hugeNegativeExponent).toBits == (-0.0 : Float32).toBits) | ||
| IO.println (((-0.0 : Float32).scaleB hugePositiveExponent).toBits == (-0.0 : Float32).toBits) |
Contributor
There was a problem hiding this comment.
With the nature of this test, this should probably live in tests/elab instead and use #guard for testing (instead of IO.println).
Edit: Also, it might be more appropriate to use equality here for testing (which essentially compares toBits), i.e.
#guard Float.inf.scaleB hugeNegativeExponent = Float.inf
#guard (-Float.inf).scaleB hugeNegativeExponent = -Float.inf
#guard Float.nan.scaleB hugeNegativeExponent = Float.nan
#guard (-1.0 : Float).scaleB hugeNegativeExponent = -0.0
#guard (-0.0 : Float).scaleB hugeNegativeExponent = -0.0
Rob23oba
reviewed
Aug 5, 2026
| @@ -0,0 +1,22 @@ | |||
| module | |||
|
|
|||
| import Init.Data.Float.Float32 | |||
Contributor
There was a problem hiding this comment.
You don't need this import, declarations from Init are imported by default.
Author
|
Thanks for the feedback, changes have been implemented and label is applied |
mathlib-nightly-testing Bot
pushed a commit
to leanprover-community/batteries
that referenced
this pull request
Aug 5, 2026
mathlib-nightly-testing Bot
pushed a commit
to leanprover-community/mathlib4-nightly-testing
that referenced
this pull request
Aug 5, 2026
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
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.
This PR makes
Float.scaleBandFloat32.scaleBpreserve infinities, NaN, and signed zero when scaling by extremely large exponents.Previously, the boxed-exponent path returned positive zero for every huge negative exponent, including non-finite inputs and negative finite values. Handle non-finite inputs and zero directly, and preserve the sign of underflowed finite values.
The PR also adds regression coverage for both precisions in
tests/elab/floatScaleB.lean.