Tue, 26 Sep 2000 13:29:39 +0200 (MET DST), John Hughes <[EMAIL PROTECTED]> pisze:
> Of course, in principle, an optimizer *could* replace x==x by
> x`seq`True (if x is known to be of base type), and the x`seq` might
> well be removed by later transformations (if x can be shown to be
> defined, something compilers do analyses to discover). Who knows,
> maybe this happens in the innards of ghc...
It gets optimized for Bool. It does not get optimized for Int, because
(==) on Int is defined in terms of a primitive operation (==#) and
the compiler does not know that x ==# x is always True.
But the programmer can teach it :-)
{-# RULES
"x == x" forall (x :: Int). (==) x x = x `seq` True
#-}
Well, it's probably better to tell this on a lower level, because
some other operations might use (==#) directly instead of through (==):
import PrelGHC
{-# RULES
"x ==# x" forall (x :: Int#). (==#) x x = True
#-}
Here it is not necessary to use seq because Int# values are always
evaluated.
Now a function like
f:: Int -> Int
f x = if x == x then x else 10
compiles to exactly the same code as id.
--
__("< Marcin Kowalczyk * [EMAIL PROTECTED] http://qrczak.ids.net.pl/
\__/
^^ SYGNATURA ZASTĘPCZA
QRCZAK
