id,summary,reporter,owner,description,type,status,priority,milestone,component,version,resolution,keywords,cc
39,./Data/Array/Accelerate/CUDA.hs:48 (unhandled): CUDA Exception: invalid argument,j.waldmann,tmcdonell,"Hello. 

I was trying to build/use the current accelerate version,
and run into some problems.

I am using ghc-6.12.3 on Linux 2.6.32-bpo.5-amd64 (debian 5.0.6)
with cudatoolkit_3.2.16_linux_64_ubuntu10.04  and gcc-4.5.1
and my hardware is GTX295

* cabal install accelerate  
is trying to build accelerate-0.8.1.0
and it starts with building the dependency cuda-0.2.2
which fails with 

{{{
c2hs: Errors during expansion of binding hooks:

./Foreign/CUDA/Driver/Context.chs:76: (column 15) [ERROR]  >>> Unknown identifier!
  Cannot find a definition for `cuCtxCreate' in the header file.
}}}

* cabal install cuda   
builds cuda-0.3.2  and the installation runs through.


* cabal unpack accelerate
and manually remove the ""cuda < 0.3"" dependency
then cabal install   runs through.

* cd accelerate-0.8.1.0/examples/simple ; make ; ./test
looks OK 

* cd accelerate-0.8.1.0/examples/rasterize ; ghc --make rasterize

{{{
[1 of 2] Compiling RasterizeAcc     ( RasterizeAcc.hs, RasterizeAcc.o )

RasterizeAcc.hs:26:9:
    Not in scope: type constructor or class `NFData'
}}}

I manually add ""import Control.DeepSeq"" to RasterizeAcc.hs
then ""ghc --make"" succeeds

./rasterize  prints 
4 * Haskell (pass), 4 * (Accelerate interpreted) pass

So I figure this isn't using cuda at all.
I change  Data.Array.Accelerate.Interpreter to
Data.Array.Accelerate.CUDA
in both  RasterizeAcc.hs  and   rasterize.hs

Then ""ghc --make rasterize"" is OK,
but running ./rasterize gives

{{{
rasterize-test1.txt (Haskell) - pass
rasterize-test2.txt (Haskell) - pass
rasterize-test3.txt (Haskell) - pass
rasterize-test4.txt (Haskell) - pass
rasterize: 
*** Internal error in package accelerate ***
*** Please submit a bug report at http://trac.haskell.org/accelerate
./Data/Array/Accelerate/CUDA.hs:48 (unhandled): CUDA Exception: invalid argument
}}}",defect,new,major,,CUDA backend,0.8.1.0,,,