All subblock reuse runs so far are, at a specific point, ignoring parallelization:
In the beginning of an iteration, the parity matrix is checked for each block, adding the block if it is not known. The problem is that in an iteration, block 0 is also taken into consideration to check if block 1 is known, then block 0 and 1 are taken into consideration for block 2... etc
This is in part good: The parity of the last block never needs to be sent after the first iteration and that happens with this implementation and this might even improve the reconciliation efficiency. However, the number of channel uses should be taking into consideration that the initial parity exchange for each iteration is not one message with N block parities but N messages with 1 block parity each. (this will be fixed by applying a manual patch on the channel usages results)
All subblock reuse runs so far are, at a specific point, ignoring parallelization:
In the beginning of an iteration, the parity matrix is checked for each block, adding the block if it is not known. The problem is that in an iteration, block 0 is also taken into consideration to check if block 1 is known, then block 0 and 1 are taken into consideration for block 2... etc
This is in part good: The parity of the last block never needs to be sent after the first iteration and that happens with this implementation and this might even improve the reconciliation efficiency. However, the number of channel uses should be taking into consideration that the initial parity exchange for each iteration is not one message with N block parities but N messages with 1 block parity each. (this will be fixed by applying a manual patch on the channel usages results)