arnaucube/go-circom-prover-verifier
1
0
Fork 0
mirror of https://github.com/arnaucube/go-circom-prover-verifier.git synced 2026-02-06 19:06:43 +01:00
Code Issues Packages Projects Releases Wiki Activity

Added a description file

Browse Source
This commit is contained in:
druiz0992
2020-05-03 03:30:52 +02:00
parent 8c81f5041e
commit 72913bc801
4 changed files with 51 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

1
prover/gextra.go
View File

@@ -4,7 +4,6 @@ import (
"math/big"
bn256 "github.com/ethereum/go-ethereum/crypto/bn256/cloudflare"
cryptoConstants "github.com/iden3/go-iden3-crypto/constants"
//"fmt"
)
type TableG1 struct{

2
prover/gextra_test.go
View File

@@ -11,7 +11,7 @@ import (
)
const (
N = 5000
N = 50000
)
func randomBigIntArray(n int) []*big.Int{

25
prover/tables.md Normal file
View File

@@ -0,0 +1,25 @@
# Tables Pre-calculation
The most time consuming part of a ZKSnark proof calculation is the scalar multiplication of elliptic curve points. Direct mechanism accumulates each multiplication. However, prover only needs the total accumulation.
There are two potential improvements to the naive approach:
1. Apply Strauss-Shamir method (https://stackoverflow.com/questions/50993471/ec-scalar-multiplication-with-strauss-shamir-method).
2. Leave the doubling operation for the last step
Both options can be combined.
In the following table, we show the results of using the naive method, Srauss-Shamir and Strauss-Shamir + No doubling. These last two options are repeated for different table grouping order.
There are 5000 G1 Elliptical Curve Points, and the scalars are 254 bits (BN256 curve).
There may be some concern on the additional size of the tables since they need to be loaded into a smartphone during the proof, and the time required to load these tables may exceed the benefits. If this is a problem, another althernative is to compute the tables during the proof itself. Depending on the Group Size, timing may be better than the naive approach.
| Algorithm | GS / Time |
|---|---|---|
| Naive | 6.63s | | | | | | | |
| Strauss | 13.16s | 9.033s | 6.95s | 5.61s | 4.91s | 4.26s | 3.88s | 3.54 s | 1.44 s |
| Strauss + Table Computation | 16.13s | 11.32s | 8.47s | 7.10s | 6.2s | 5.94s | 6.01s | 6.69s |
| No Doubling | 3.74s | 3.00s | 2.38s | 1.96s | 1.79s | 1.54s | 1.50s | 1.44s|
| No Doubling + Table Computation | 6.83s | 5.1s | 4.16s | 3.52s| 3.22s | 3.21s | 3.57s | 4.56s |

25
tables.md Normal file
View File

@@ -0,0 +1,25 @@
# Tables Pre-calculation
The most time consuming part of a ZKSnark proof calculation is the scalar multiplication of elliptic curve points. Direct mechanism accumulates each multiplication. However, prover only needs the total accumulation.
There are two potential improvements to the naive approach:
1. Apply Strauss-Shamir method (https://stackoverflow.com/questions/50993471/ec-scalar-multiplication-with-strauss-shamir-method).
2. Leave the doubling operation for the last step
Both options can be combined.
In the following table, we show the results of using the naive method, Srauss-Shamir and Strauss-Shamir + No doubling. These last two options are repeated for different table grouping order.
There are 5000 G1 Elliptical Curve Points, and the scalars are 254 bits (BN256 curve).
There may be some concern on the additional size of the tables since they need to be loaded into a smartphone during the proof, and the time required to load these tables may exceed the benefits. If this is a problem, another althernative is to compute the tables during the proof itself. Depending on the Group Size, timing may be better than the naive approach.
| Algorithm | GS / Time |
|---|---|---|
| Naive | 6.63s | | | | | | | |
| Strauss | 13.16s | 9.033s | 6.95s | 5.61s | 4.91s | 4.26s | 3.88s | 3.54 s | 1.44 s |
| Strauss + Table Computation | 16.13s | 11.32s | 8.47s | 7.10s | 6.2s | 5.94s | 6.01s | 6.69s |
| No Doubling | 3.74s | 3.00s | 2.38s | 1.96s | 1.79s | 1.54s | 1.50s | 1.44s|
| No Doubling + Table Computation | 6.83s | 5.1s | 4.16s | 3.52s| 3.22s | 3.21s | 3.57s | 4.56s |