On March 25, 2009 21:38:55 Claus Reinke wrote: > If you don't have that kind of control, and generate code as if > there were no hardware-level optimizations, the resulting > mismatch will manifest in hard-to-predict variations in > performance, making it difficult to see how much or why > speed is lost. No fun.
I would think the ultimate for this would be to be able to feed profiling information into the compiler. No messy pragmas all over the code and I would guess there would be other benefits as well (e.g., inlining decisions). Of course it's easy to sit back and suggest things like this -- but I would image quite a whole other kettle of fish to build them. : ) Cheers! -Tyson
signature.asc
Description: This is a digitally signed message part.
_______________________________________________ Glasgow-haskell-users mailing list Glasgow-haskell-users@haskell.org http://www.haskell.org/mailman/listinfo/glasgow-haskell-users
