o1vm/pickles: absorb quotient polynomial and evaluate #2655
Conversation
dannywillems
force-pushed
the
dw/continue-quotient-polynomial
branch
from
0cd8e3f to
aa2fdc6
Compare
Codecov ReportAttention: Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## master #2655 +/- ##
==========================================
- Coverage 72.57% 72.54% -0.04%
==========================================
Files 246 246
Lines 58105 58132 +27
==========================================
+ Hits 42171 42173 +2
- Misses 15934 15959 +25 ☔ View full report in Codecov by Sentry. |
dannywillems
force-pushed
the
o1vm/quotient-polynomial
branch
from
3e694db to
e675ff1
Compare
dannywillems
force-pushed
the
dw/continue-quotient-polynomial
branch
from
aa2fdc6 to
57a6adb
Compare
It will be better to have them instead of having randomly the value popping up in the code at different places.
dannywillems
force-pushed
the
dw/continue-quotient-polynomial
branch
from
57a6adb to
f573ef6
Compare
| /// have a degree 7 for the quotient polynomial. | ||
| /// Used to keep track of the number of chunks we do have when we commit to the | ||
| /// quotient polynomial. | ||
| pub const DEGREE_QUOTIENT_POLYNOMIAL: u64 = 7; |
There was a problem hiding this comment.
it's not because we evaluate on d8 that the degree is 8
it should be degree of the constraints - 1
| @@ -225,7 +230,9 @@ where | |||
| quotient | |||
| }; | |||
|
|
|||
| let _t_comm = srs.commit_non_hiding("ient_poly, 7); | |||
| let t_comm = srs.commit_non_hiding("ient_poly, DEGREE_QUOTIENT_POLYNOMIAL as usize); | |||
There was a problem hiding this comment.
Not a part of this PR, but do we want to do chunking ?
We do not have any limitation on the SRS
| // Compute ft(X) = \ | ||
| // (1 - ζ^n) \ | ||
| // (t_0(X) + ζ^n t_1(X) + ... + ζ^{kn} t_{k}(X)) | ||
| // where Σ_i t_i(X) X^{i n} = t(X), and t(X) is the quotient polynomial. | ||
| // At the end, we get the (partial) evaluation of the constraint polynomial | ||
| // in ζ. | ||
| let ft: DensePolynomial<G::ScalarField> = { | ||
| let evaluation_point_to_domain_size = zeta.pow([domain.d1.size]); | ||
| // Compute Σ_i t_i(X) ζ^{i n} | ||
| // First we split t in t_i, and we reduce to degree (n - 1) after using `linearize` | ||
| let t_chunked: DensePolynomial<G::ScalarField> = quotient_poly | ||
| .to_chunked_polynomial(DEGREE_QUOTIENT_POLYNOMIAL as usize, domain.d1.size as usize) | ||
| .linearize(evaluation_point_to_domain_size); | ||
| // -Z_H = (1 - ζ^n) | ||
| let minus_vanishing_poly_at_zeta = -domain.d1.vanishing_polynomial().evaluate(&zeta); | ||
| // Multiply the polynomial Σ_i t_i(X) ζ^{i n} by -Z_H(ζ) | ||
| // (the evaluation in ζ of the vanishing polynomial) | ||
| t_chunked.scale(minus_vanishing_poly_at_zeta) | ||
| }; |
There was a problem hiding this comment.
I am confused, why are we evaluating this ?
We should evaluate the chunks of t, not scale by the vanishsing poly ?
Also, if we do chunking, at this point we should have access to the chunks (since we commited to them), so we should not call to chunk one more time
And if we want to evaluate the constraint poly, we should not go through t, we have a direct access to it
|
Closing in favor of #2692 |
No description provided.