-
Hi, Does anyone know how to accomplish modulo operations (%) and still use the remainder in complex assertions? I have looked at emulated (https://pkg.go.dev/github.com/consensys/[email protected]/std/math/emulated) and I'm having issues understanding how I can use my result in complex assertions afterwards. I am able to get the remainder using emulated field elements, like so:
I was wondering maybe if there is a way to turn a *emulated.Element[T] into a frontend.Variable? The issue is that I am trying to do somewhat complex series of assertions with the remainder, but I have the remainder as a frontend.Variable (see func below). This is how I am trying to use the remainder:
Thank you! |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment 2 replies
-
What are the value ranges you are performing operations in? Keep in mind that If the operations you define are small (and in your example it seems they may be as you work with a grid and want to use it for the grid size?), then you can essentially compute using native arithmetic (as you never over or underflow). There is a very similar discussion happening in #1361. You could for example do: package smallmod
import (
"errors"
"fmt"
"math/big"
"testing"
"github.com/consensys/gnark/constraint/solver"
"github.com/consensys/gnark/frontend"
"github.com/consensys/gnark/test"
)
func smallModHint(mod *big.Int, inputs []*big.Int, outputs []*big.Int) error {
// computes a % r = b
// inputs[0] = a -- input
// inputs[1] = r -- modulus
// outputs[0] = b -- remainder
// outputs[1] = (a-b)/r -- quotient
if len(outputs) != 2 {
return errors.New("expected 2 outputs")
}
if len(inputs) != 2 {
return errors.New("expected 2 inputs")
}
outputs[1].QuoRem(inputs[0], inputs[1], outputs[0])
fmt.Println(outputs[0], outputs[1], inputs[0], inputs[1])
return nil
}
func SmallMod(api frontend.API, a, r frontend.Variable) (quo, rem frontend.Variable) {
res, err := api.Compiler().NewHint(smallModHint, 2, a, r)
if err != nil {
panic(err)
}
rem = res[0]
quo = res[1]
// to prevent against overflows, we assume that the inputs are small relative to the native field
nbBits := api.Compiler().Field().BitLen()/2 - 2
bound := new(big.Int).Lsh(big.NewInt(1), uint(nbBits))
api.AssertIsLessOrEqual(rem, bound)
api.AssertIsLessOrEqual(quo, bound)
api.AssertIsEqual(a, api.Add(api.Mul(quo, r), rem))
return
}
type Circuit struct {
A, R frontend.Variable
}
func (c *Circuit) Define(api frontend.API) error {
quo, rem := SmallMod(api, c.A, c.R)
api.Println(c.A)
api.Println(c.R)
api.Println(quo)
api.Println(rem)
return nil
}
func TestSmallMod(t *testing.T) {
assert := test.NewAssert(t)
// hint needs to be provided manually to the solver. Otherwise it doesn't know how to compute the value
solver.RegisterHint(smallModHint)
assert.CheckCircuit(&Circuit{}, test.WithValidAssignment(&Circuit{A: 123000, R: 1000}))
} |
Beta Was this translation helpful? Give feedback.
What are the value ranges you are performing operations in? Keep in mind that
frontend.Variable
value are defined in a finite field defined by the elliptic curve used for generating the proof. For example when use use BN254 for generating the proof then the operations are in the scalar field of BN254, which is mod 21888242871839275222246405745257275088548364400416034343698204186575808495617.If the operations you define are small (and in your example it seems they may be as you work with a grid and want to use it for the grid size?), then you can essentially compute using native arithmetic (as you never over or underflow).
There is a very similar discussion happening in #1361. You could f…