new benchmarks

This commit is contained in:
John Guibas
2023-03-28 12:01:57 -07:00
parent cf16f0f595
commit a6e3319c5a

View File

@@ -22,12 +22,17 @@ go run benchmark.go
Here are relevant numbers from a benchmark ran on an M1 Max with 10 CPU cores. Here are relevant numbers from a benchmark ran on an M1 Max with 10 CPU cores.
``` ```
# of constraints: 6740784 11:04:08 INF compiling circuit
circuit compilation time: 19 minutes, 33 sec. 11:04:08 INF parsed circuit inputs nbPublic=0 nbSecret=0
witness generation time: < 1 second 11:12:30 INF building constraint system nbConstraints=6740784
groth16 setup: 8 minutes 23 seconds Generating witness 2023-03-28 11:12:42.702566 -0700 PDT m=+514.333410376
proof generation: 23 seconds Running circuit setup 2023-03-28 11:12:42.702666 -0700 PDT m=+514.333509834
proof verification: < 1 second Creating proof 2023-03-28 11:18:58.881518 -0700 PDT m=+890.519971543
11:18:59 DBG constraint system solver done backend=groth16 nbConstraints=6740784 took=675.361625
11:19:10 DBG prover done backend=groth16 curve=bn254 nbConstraints=6740784 took=10512.664584
Verifying proof 2023-03-28 11:19:10.169636 -0700 PDT m=+901.808314709
11:19:10 DBG verifier done backend=groth16 curve=bn254 took=6.288792
bn254 2023-03-28 11:19:10.175992 -0700 PDT m=+901.814670834
``` ```
The circuit can be significantly optimized by using more efficient arithmetic for Goldilocks, among other things. The circuit can be significantly optimized by using more efficient arithmetic for Goldilocks, among other things.