Currently, langcc appears to use SLR followsets for initializing the partitions in the LR(0) NFA.
It looks to me like LALR followsets could be used instead.
Of course, calculating LALR followsets is more tricky than the standard followsets, but
LPG is a LALR(k) parser generator written in C++ and maybe adopting its approach wouldn't be too difficult.
I think it would be really interesting to see how big of an effect this could have on the size of the final tables.
Currently, langcc appears to use SLR followsets for initializing the partitions in the LR(0) NFA.
It looks to me like LALR followsets could be used instead.
Of course, calculating LALR followsets is more tricky than the standard followsets, but LPG is a LALR(k) parser generator written in C++ and maybe adopting its approach wouldn't be too difficult.
I think it would be really interesting to see how big of an effect this could have on the size of the final tables.