Repository navigation
Clamp an unrepresentable balanced split into range - #67
Merged
Merged
Conversation
A product of many factors can carry a declared scale far below the smallest Float64, and the balanced split of its multiplication block then aims at an eta beyond floatmax. balanced_split threw there, although every member of the block's cover family is valid. Where the clamped equalizing point exists but the pair (eta1, c12/eta1) would round to zero or infinity, balanced_split now moves eta1 into the window of clamp_representable: both entries normal and at most U = 2^(exponent(floatmax)/2). The headroom below floatmax is needed because each eta seeds a reverse sweep that adds to it; a seed at floatmax overflows at the first sum. Splits that are representable are returned unchanged, so covariance is lost only at blocks that have no representation. The three families with no ratio to read still throw, now with a message that says so. Also add a promote_rule between BigFloat and ExpScaled: Base claims every Real for BigFloat, and the conflicting rules made promote_type recurse. Assisted-by: Claude Opus 5.5 <noreply@anthropic.com>
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
A product of many factors can carry a declared scale far below the smallest Float64, and the balanced split of its multiplication block then aims at an eta beyond floatmax. balanced_split threw there, although every member of the block's cover family is valid.
Where the clamped equalizing point exists but the pair (eta1, c12/eta1) would round to zero or infinity, balanced_split now moves eta1 into the window of clamp_representable: both entries normal and at most U = 2^(exponent(floatmax)/2). The headroom below floatmax is needed because each eta seeds a reverse sweep that adds to it; a seed at floatmax overflows at the first sum. Splits that are representable are returned unchanged, so covariance is lost only at blocks that have no representation. The three families with no ratio to read still throw, now with a message that says so.
Also add a promote_rule between BigFloat and ExpScaled: Base claims every Real for BigFloat, and the conflicting rules made promote_type recurse.
Assisted-by: Claude Opus 5.5 noreply@anthropic.com