I streamlined the Go STARK implementation with goroutines, channels + big.Int optimizations and on an ordinary laptop (2014 MacBook Pro 2.2 GHz Intel Core i7 with 8 “logical” cores) got to this basic level of temporal performance:
NUM_CORES |
STARK Proof Generation | STARK Verification |
|---|---|---|
| 1 | 3.12s | 46.11ms |
| 2 | 2.01s | 24.70ms |
| 4 | 1.49s | 15.64ms |
| 8 | 1.42s | 16.58ms |
| 16 | 1.49s | 18.30ms |
| 32 | 1.48s | 20.18ms |
| ethereum/research python | 3.78s | 52.10ms |