The benchmark assumed that the first action in the bundle was for the
recipient it had added, but we've started shuffling recipients within
actions, and the builder pads to a minimum of two actions. This meant
that 50% of benchmark runs would fail to start, but would proceed fine
if they did start (as the bundle is cached for the entire test).
We now add two recipients, to cancel out the effect of the padding and
shuffle.
For almost all the sponge constructions defined in the Poseidon paper,
the domain can be defined completely statically. Variable-length hashing
requires knowledge of the message length, but that can be provided to
the fixed padding function in a subsequent commit, and in any case we
can't use variable-length inputs in a circuit.
Criterion's benchmark grouping does not match on group names; it only
groups benchmarks that are run prior to that specific benchmark group
instance being dropped. Since each benchmark group holds a mutable
reference to the criterion instance, this means we can't have multiple
active groups collecting measurements. Instead, we need to collect the
proving benchmarks for all recipient numbers, followed by verification
benchmarks.
As the underlying `Region` methods now return `AssignedCell` instead of
`Cell`, we can simplify all the places where we then constructed a
`CellValue` struct.
We change `CellValue` into a typedef of `AssignedCell` to simplify the
migration in this commit.
The migration from `CellValue` to `AssignedCell` requires several other
changes:
- `<CellValue as Var>::value()` returned `Option<F>`, whereas
`AssignedCell::<F, F>::value()` returns `Option<&F>`. This means we
need to dereference, use `Option::cloned`, or alter functions to take
`&F` arguments.
- `StateWord` in the Poseidon chip has been changed to a newtype around
`AssignedCell` (the chip was written before `CellValue` existed).