Revert "removed computing line coeffs (#643)"

[spapinistarkware]

Revert "removed computing line coeffs (#643)"

This reverts commit a150b8f.

Revert "integrated point vanishing in fri quotients (#619)"

This reverts commit a6eabdd.

---

This change is 

[spapinistarkware]

• FRI quotients y optimization #686
• Extension FRI optimization #685
• Revert "removed computing line coeffs (#643)" #684 👈
• dev

This stack of pull requests is managed by Graphite. Learn more about stacking.

Join @spapinistarkware and the rest of your teammates on Graphite

[codecov-commenter]

Codecov Report

Attention: Patch coverage is 98.81657% with 2 lines in your changes missing coverage. Please review.

Project coverage is 89.94%. Comparing base (a759ce0) to head (e33378a).

Files	Patch %	Lines
crates/prover/src/core/constraints.rs	96.96%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##              dev     #684      +/-   ##
==========================================
- Coverage   89.95%   89.94%   -0.02%     
==========================================
  Files          75       75              
  Lines        9657     9755      +98     
  Branches     9657     9755      +98     
==========================================
+ Hits         8687     8774      +87     
- Misses        889      900      +11     
  Partials       81       81              

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[shaharsamocha7]

Reviewed 2 of 4 files at r1, all commit messages.
Reviewable status: 2 of 4 files reviewed, 3 unresolved discussions (waiting on @spapinistarkware)

---

crates/prover/src/core/constraints.rs line 94 at r1 (raw file):

/// Evaluates the coefficients of a line between a point and its complex conjugate. Specifically,
/// `a, b, and c, s.t. a*x + b -c*y = 0` for (x,y) being (sample.y, sample.value) and

Can we change it so the equation will be
ax + by +-c = 0?

Seems more intuitive for me (non-blocking)

Code quote:

a*x + b -c*y = 0`

---

crates/prover/src/core/backend/cpu/quotients.rs line 59 at r1 (raw file):

    ) {
        let mut numerator = SecureField::zero();
        for ((column_index, _), (a, b, c)) in zip_eq(&sample_batch.columns_and_values, line_coeffs)

These already multiply by the constraint random element.
Can you document that or rename the variable names to contain the alpha?

We had exactly this bug here.

Code quote:

(a, b, c)

---

crates/prover/src/core/backend/simd/quotients.rs line 82 at r1 (raw file):

            // at sample_point and conj(sample_point) if the original polynomial had the values
            // sample_value and conj(sample_value) at these points.
            // TODO(AlonH): Use single point vanishing to save a multiplication.

Remove

Code quote:

            // TODO(AlonH): Use single point vanishing to save a multiplication.

[spapinistarkware]

Reviewable status: 2 of 4 files reviewed, 3 unresolved discussions (waiting on @shaharsamocha7)

---

crates/prover/src/core/constraints.rs line 94 at r1 (raw file):

Previously, shaharsamocha7 wrote…

Can we change it so the equation will be
ax + by +-c = 0?

Seems more intuitive for me (non-blocking)

That does sound better. Let's do it in another PR.

---

crates/prover/src/core/backend/cpu/quotients.rs line 59 at r1 (raw file):

Previously, shaharsamocha7 wrote…

These already multiply by the constraint random element.
Can you document that or rename the variable names to contain the alpha?

We had exactly this bug here.

Added in the last PR.

---

crates/prover/src/core/backend/simd/quotients.rs line 82 at r1 (raw file):

Previously, shaharsamocha7 wrote…

Remove

This gets removed in the following PRs. Better to keep this as a pure revert PR.

[shaharsamocha7]

Reviewable status: 2 of 4 files reviewed, 1 unresolved discussion (waiting on @spapinistarkware)

---

crates/prover/src/core/constraints.rs line 94 at r1 (raw file):

Previously, spapinistarkware (Shahar Papini) wrote…

That does sound better. Let's do it in another PR.

TODO?