#5470: The DPH library needs to support PData and PRepr instances for more than
15-tuples
---------------------------------+------------------------------------------
    Reporter:  chak              |       Owner:  chak                 
        Type:  bug               |      Status:  new                  
    Priority:  normal            |   Component:  Data Parallel Haskell
     Version:  7.3               |    Keywords:                       
    Testcase:                    |   Blockedby:                       
          Os:  Unknown/Multiple  |    Blocking:                       
Architecture:  Unknown/Multiple  |     Failure:  Compile-time crash   
---------------------------------+------------------------------------------
Changes (by chak):

  * owner:  benl => chak


Old description:

> Currently, `Data.Array.Parallel.PArray.PDataInstances` only generates
> tuple instances up to 6-tuple.  This is not sufficient as these instances
> are used for environments of closures by the vectoriser — i.e., once a
> closures has more than 6 free variables, the compiler fails with
> `VectMonad.lookupFamInst: not found`.
>
> To fix this in the DPH library, `vector` needs support for `zip` and
> `unzip` beyond 6-tuples as well.

New description:

 Currently, `Data.Array.Parallel.PArray.PDataInstances` only generates
 tuple instances up to 6-tuple.  This is not sufficient as these instances
 are used for environments of closures by the vectoriser — i.e., once a
 closures has more than 15 free variables, the compiler fails with
 `VectMonad.lookupFamInst: not found`.

--

Comment:

 I bumped the limit up from 6 to 15.  In the long term, we will need to
 break big tuples up as Roman suggested.

-- 
Ticket URL: <http://hackage.haskell.org/trac/ghc/ticket/5470#comment:2>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler

_______________________________________________
Glasgow-haskell-bugs mailing list
[email protected]
http://www.haskell.org/mailman/listinfo/glasgow-haskell-bugs

Reply via email to