A distinct address for each `n` (more than the 256 `addr(u8)` affords), for fixtures large enough to exercise the multilevel coarsening path. Big-endian so that address-sort order (which fixes block ids) matches numeric order, keeping cluster members id-contiguous in the fixtures below.
(n: u32)
| 2028 | } |
| 2029 | shard_of[b as usize] = cur; |
| 2030 | if members.len() == 1 && predicted(member_cost, foreign_bytes) > max_cycles |
| 2031 | { |
| 2032 | infeasible = true; // this block alone overflows the cap |
| 2033 | } |
| 2034 | if i == order.len() - 1 { |
| 2035 | max_shard_steps = |
| 2036 | max_shard_steps.max(predicted(member_cost, foreign_bytes)); |
| 2037 | } |
no outgoing calls