This PR does two things: first it renames the "corchuelo" recoverer to "cpctplus" to match the paper. Second, it adds a new recoverer "CPCTPlusDynDist" (not a snappy name, I admit) which is a mix of cpctplus and MF: it uses dyn_dist to speed up cpctplus. My quick experiments suggest that CPCTPlusDynDist is quite a bit faster than CPCTPlus, but still a bit slower than MF.
The cpctplusdyndist.rs file is, basically, a copy of mf.rs (minus the Dist struct) with the reduction rules taken from cpctplus.rs with the addition of the dyn_dist calls. It's less scary than it looks (I hope).
ptersilie
commented
6 years ago
This PR does two things: first it renames the "corchuelo" recoverer to "cpctplus" to match the paper. Second, it adds a new recoverer "CPCTPlusDynDist" (not a snappy name, I admit) which is a mix of cpctplus and MF: it uses dyn_dist to speed up cpctplus. My quick experiments suggest that CPCTPlusDynDist is quite a bit faster than CPCTPlus, but still a bit slower than MF.
The
cpctplusdyndist.rsfile is, basically, a copy ofmf.rs(minus theDiststruct) with the reduction rules taken fromcpctplus.rswith the addition of thedyn_distcalls. It's less scary than it looks (I hope).