Changes between Version 10 and Version 11 of DataParallel/Vectorisation

Timestamp:

05/28/07 00:31:57 (6 years ago)

Author:

chak

Comment:

--

DataParallel/Vectorisation

@@ -78 +78 @@
 }}}
 
-Hmm, thinking about this, a data type will get us into trouble with unboxed types.  Maybe
+Hmm, thinking about this, a data type will get us into trouble with unboxed types, but we also can't use a type synonym, as we can have partial applications of function arrows.  So, we have to split this up.
+
+A data type to combine the scalar and lifted version of  a function:
 {{{
-type a ->> b = (a -> b) :*: (a^ -> b^)
-vect :: (a -> b) -> (a ->> b)
-vect f = f :*: mapP f
+data VFun f = VFun { vfunS :: !f
+                   , vfunP :: !(PArr f)
+                   }
 }}}
-But that doesn't make any sense due to the `a^` and `b^`.  Seems like we have to hardcode this representation into the type translation scheme.
+On top of this we can define a vectorised function space constructor:
+{{{
+newtype a :-> b = Fun (VFun (a -> b))
+funS (Fun (VFun f _)) = f
+funP (Fun (VFun _ f)) = f
+}}}
+So, we have `(->_v) = (:->)`.  Note that the type arguments to `(:->)` will be vectorised types (not vanilla types).  Consequently, instances of `PA` and `PArr` need to be defined for vectorised types.
 ----