Optimize number of gates in folding verifier #849
Assignees
Comments
ledwards2225
added a commit
to AztecProtocol/aztec-packages
that referenced
this issue
Feb 8, 2024
Updates folding recursive verifier to use batch_mul for optimal goblin ec op efficiency. This reduces a single recursive verification from 1144 ECC ops to 264. (Note: 264 = 6*44 where 6 is the number of ecc op gate rows needed for two scalar muls (one for each instance) plus an "equals" op and 44 is the number of witnesses plus precomputed polys, not including shifts) Closes AztecProtocol/barretenberg#849 New benchmark result: ``` ----------------------------------------------------------------- Benchmark Time CPU Iterations ----------------------------------------------------------------- IvcBench/Full/6 54156 ms 51691 ms 1 ``` Old benchmark result: ``` ----------------------------------------------------------------- Benchmark Time CPU Iterations ----------------------------------------------------------------- IvcBench/Full/6 66891 ms 63569 ms 1 ```
AztecBot
pushed a commit
that referenced
this issue
Feb 9, 2024
Updates folding recursive verifier to use batch_mul for optimal goblin ec op efficiency. This reduces a single recursive verification from 1144 ECC ops to 264. (Note: 264 = 6*44 where 6 is the number of ecc op gate rows needed for two scalar muls (one for each instance) plus an "equals" op and 44 is the number of witnesses plus precomputed polys, not including shifts) Closes #849 New benchmark result: ``` ----------------------------------------------------------------- Benchmark Time CPU Iterations ----------------------------------------------------------------- IvcBench/Full/6 54156 ms 51691 ms 1 ``` Old benchmark result: ``` ----------------------------------------------------------------- Benchmark Time CPU Iterations ----------------------------------------------------------------- IvcBench/Full/6 66891 ms 63569 ms 1 ```
michaelelliot
pushed a commit
to Swoir/noir_rs
that referenced
this issue
Feb 28, 2024
…ol#4517) Updates folding recursive verifier to use batch_mul for optimal goblin ec op efficiency. This reduces a single recursive verification from 1144 ECC ops to 264. (Note: 264 = 6*44 where 6 is the number of ecc op gate rows needed for two scalar muls (one for each instance) plus an "equals" op and 44 is the number of witnesses plus precomputed polys, not including shifts) Closes AztecProtocol/barretenberg#849 New benchmark result: ``` ----------------------------------------------------------------- Benchmark Time CPU Iterations ----------------------------------------------------------------- IvcBench/Full/6 54156 ms 51691 ms 1 ``` Old benchmark result: ``` ----------------------------------------------------------------- Benchmark Time CPU Iterations ----------------------------------------------------------------- IvcBench/Full/6 66891 ms 63569 ms 1 ```
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
No description provided.
The text was updated successfully, but these errors were encountered: