feat: field/goldilocks (more efficient 1-limb modulus arith)#177
Merged
feat: field/goldilocks (more efficient 1-limb modulus arith)#177
Conversation
Tabaie
requested changes
May 3, 2022
| {{- end}} | ||
|
|
||
| {{ template "reduce" .}} | ||
| {{ template "reduce" .}} |
Contributor
There was a problem hiding this comment.
Should reduce be in an else clause? There's no situation where we would need two reductions. Does it matter?
Collaborator
Author
There was a problem hiding this comment.
I don't understand the two reductions part / does it matter --> both code path must exist at the template level, when all bits are set and you have a carry to the last addition, then you need to reduce, but if not, you still need to check that z < q.
Tabaie
approved these changes
May 4, 2022
Merged
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
2 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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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 adds
field/goldilocks(q = 0xffffffff00000001) .Mostly tunes the code generation for field arithmetic to produce more efficient code when len(modulus) == 1 limb; no assembly is generated (it would be counter productive) and no "goldilocks" specific tricks are used.
plonky2 paper mentions some "arithmetic tricks" that makes modular reduction efficient. In practice, perf is in the same range than the CIOS Montgomery reduction for 1-liimb.
closes #174