As reported in https://github.com/TimelyDataflow/differential-dataflow/issues/479, rhh.rs too casually assumes that usize has 64 bits of representation. Chasing that down, we manipulate hashes, wired to be u64 inconsistently and this PR means to fix the most obvious forms of that. The new code has the potential to error at runtime if usize has more bits than u64, but I'm happy to make addressing that a problem for tomorrow.
The RHH code is also pre-production, and mostly we want to avoid bad habits and unblock folks with 32 bit usize platforms.
As reported in https://github.com/TimelyDataflow/differential-dataflow/issues/479,
rhh.rs
too casually assumes thatusize
has 64 bits of representation. Chasing that down, we manipulate hashes, wired to beu64
inconsistently and this PR means to fix the most obvious forms of that. The new code has the potential to error at runtime ifusize
has more bits thanu64
, but I'm happy to make addressing that a problem for tomorrow.The RHH code is also pre-production, and mostly we want to avoid bad habits and unblock folks with 32 bit
usize
platforms.cc: @oli-w