Replying to [comment:90 carter]:

...

@lemming, Bool# is now Int# afaik, could you explain what you mean? even if we used a i1 rep, it still needs a full register! So theres no lost for the in register rep to be Int#, afaik

`Int#` represents a large range of integers, whereas `i1` in LLVM has only values 0 and 1. That is, LLVM can optimize based on the knowledge of

#6135: Unboxed Booleans ---------------------------------------------+----------------------------- Reporter: benl | Owner: jstolarek Type: feature request | Status: closed Priority: normal | Milestone: 7.8.1 Component: Compiler | Version: 7.4.1 Resolution: fixed | Keywords: Operating System: Unknown/Multiple | Architecture: Type of failure: None/Unknown | Unknown/Multiple Test Case: primops/should_run/T6135 | Difficulty: Unknown Blocking: | Blocked By: 8103, | 8103 | Related Tickets: #605 ---------------------------------------------+----------------------------- Comment (by altaic): Replying to [comment:91 Lemming]: the restricted value set of `i1`, but it certainly fails to do so on `i32` or `i64`. Two of the advantages in representing Bool# as an i1 is that llvm can vectorize them, and arrays of i1 can fit in an i32 or i64. Off the top of my head, I don't know of any software that makes use of large arrays of booleans, but in that case this sort of optimization would likely be very effective. -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/6135#comment:92 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler