You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
#variables
Size of the MSM required to compute the witness of a synthesized circuit
#constraints
Size of the $T$ vector that is committed to during the folding step.
|A| + |B| + |C|
Complexity of creating a compressed SNARK for the circuit, and the corresponding proving/verifying key.
These three quantities are the main parameters which affect performance. While we are pretty good about reporting the two first ones, the last one will be important when working with CycleFold and recursive verification. In particular, the number of coefficients of the R1CS matrices will usually be higher than the other two quantities.
As a priority, we should report these figures for:
Native scalar multiplication
For both 128 and 256 bit exponents
Native group mul-add
Used for the main CycleFold non-native group operation
Non-native field mul-add with 128-bit
Used for the random-linear combination of the public inputs $X$ during the folding over the BN circuit, as well as the SNARK-based PCS for avoiding IPA.
The text was updated successfully, but these errors were encountered:
The preprocessing SNARK's run-time is determined by
#variables
Size of the MSM required to compute the witness of a synthesized circuit
#constraints$T$ vector that is committed to during the folding step.
Size of the
|A| + |B| + |C|
Complexity of creating a compressed SNARK for the circuit, and the corresponding proving/verifying key.
These three quantities are the main parameters which affect performance. While we are pretty good about reporting the two first ones, the last one will be important when working with CycleFold and recursive verification. In particular, the number of coefficients of the R1CS matrices will usually be higher than the other two quantities.
As a priority, we should report these figures for:
For both 128 and 256 bit exponents
mul-add
Used for the main CycleFold non-native group operation
mul-add
with 128-bitUsed for the random-linear combination of the public inputs
The text was updated successfully, but these errors were encountered: