All pages
Powered by GitBook
1 of 1

Coordinate Forms

Affine (x,y)(x, y)(x,y)

Affine form refers to the regular (x,y)(x,y)(x,y) form of coordinates on an xyxyxy-plane.

Problem

In affine form, every addition or doubling requires field inversion when computing fractions (e.g., 11+dx1x2y1y2\frac{1}{1 + dx_1x_2y_1y_2}1+dx1​x2​y1​y2​1​). Like with Weierstrass curves, this becomes the bottleneck for performance.

Extended Affine (x,y,xy)(x, y, xy)(x,y,xy)

Extended Affine form is a lightweight optimization over the basic affine coordinates. It explicitly stores the product xyxyxy, which is used repeatedly in twisted Edwards point addition formulas. This avoids redundant multiplications and slightly speeds up scalar multiplication loops.

Point Representation:

A point is stored as:

(x,y,xy)(x, y, xy)(x,y,xy)

This means we store:

  • xxx: the affine x-coordinate

  • yyy: the affine y-coordinate

  • t=xyt = xyt=xy: the product of x and y

Problem

In standard affine form, when computing twisted Edwards additions, the term xyxyxy appears frequently. Without optimization, it must be recomputed each time.

Solution

By storing t=xyt = xyt=xy explicitly as a third coordinate, we reduce the number of field multiplications per operation. Though this doesn't eliminate field inversions like projective coordinates do, it provides a minimal memory overhead with noticeable performance gain in systems where full projective coordinates are unnecessary or costly.

Affine ⟹\Longrightarrow⟹ Extended Affine

(x,y)⟹(x,y,xy)(x, y) \Longrightarrow (x, y, xy)(x,y)⟹(x,y,xy)

Extended Affine ⟹\Longrightarrow⟹ Affine

Simply discard the third value:

(x,y,t)⟹(x,y)(x, y, t) \Longrightarrow (x, y)(x,y,t)⟹(x,y)

Extended Projective (X:Y:Z:T)(X : Y : Z : T)(X:Y:Z:T)

To eliminate the need for inverses during addition and doubling, extended projective coordinates are widely used. This is the most efficient coordinate system for twisted Edwards curves.

Point Representation:

x=XZ,y=YZ,T=XYZx = \frac{X}{Z}, \quad y = \frac{Y}{Z}, \quad T = \frac{XY}{Z}x=ZX​,y=ZY​,T=ZXY​

So, the point is stored as:

(X,Y,Z,T)(X, Y, Z, T)(X,Y,Z,T)

The extra coordinate TTT allows for more efficient addition formulas and better reuse of intermediate terms.

Elliptic Curve Equation:

In this form, the twisted Edwards curve becomes:

(aX2+Y2)Z2=Z4+dX2Y2(aX^2 + Y^2)Z^2 = Z^4 + dX^2Y^2(aX2+Y2)Z2=Z4+dX2Y2

However, in practice, point operations are implemented through optimized formulas that directly compute the result without explicitly evaluating this equation.

Affine ⟹\Longrightarrow⟹ Extended Projective

(x,y)⟹(X,Y,Z,T)=(x,y,1,xy)(x, y) \Longrightarrow (X, Y, Z, T) = (x, y, 1, xy)(x,y)⟹(X,Y,Z,T)=(x,y,1,xy)

Extended Projective ⟹\Longrightarrow⟹ Affine

(X,Y,Z,T)⟹(XZ,YZ)(X, Y, Z, T) \Longrightarrow \left( \frac{X}{Z}, \frac{Y}{Z} \right)(X,Y,Z,T)⟹(ZX​,ZY​)

Calculating Operations

Refer to Hyperelliptic for Extended Projective points on the best practices on how to calculate addition and double operations on Projective coordinates.

Summary of all forms:

Form
Point Form in Affine Form
Storage of values
Elliptic Curve Equation

Affine

(x,y)(x, y)(x,y)

(x,y)(x, y)(x,y)

ax2+y2=1+dx2y2ax^2 + y^2 = 1 + dx^2y^2ax2+y2=1+dx2y2

Extended Affine

(x,y)(x, y)(x,y)

(x,y,t)(x, y, t)(x,y,t)

ax2+y2=1+dx2y2ax^2 + y^2 = 1 + dx^2y^2ax2+y2=1+dx2y2

Extended Projective

(XZ,YZ)\left( \frac{X}{Z}, \frac{Y}{Z}\right)(ZX​,ZY​)

(X,Y,Z,T)(X, Y, Z, T)(X,Y,Z,T)

(aX2+Y2)Z2=Z4+dX2Y2(aX^2 + Y^2)Z^2 = Z^4 + dX^2Y^2(aX2+Y2)Z2=Z4+dX2Y2

References

  • https://www.hyperelliptic.org/EFD/g1p/auto-twisted.html