On Tue, Feb 22, 2011 at 12:15 AM, Simon Peyton-Jones <simo...@microsoft.com> wrote: > | 2) I've gotten really interested in array copying performance lately. > > Do the new primops help? (I'm not sure if they've made it in yet, and Simon > M is on hol.)
They do, inserts get faster and are on par with Data.Map with the new primops. I think we can take it much further though. C compilers go to great length at optimizing memcpy but we're not getting much of the benefit here because 1) we get extra overhead from two function calls (first to the C-- primop and then to C memcpy) and 2) memcpy cannot be inlined into the C-- code so the C compiler doesn't have much static information to go on when compiling memcpy (e.g. it cannot unroll the loop). Johan _______________________________________________ Cvs-ghc mailing list Cvs-ghc@haskell.org http://www.haskell.org/mailman/listinfo/cvs-ghc
