diff --git a/Foreign/CUDA/FFT.hs b/Foreign/CUDA/FFT.hs --- a/Foreign/CUDA/FFT.hs +++ b/Foreign/CUDA/FFT.hs @@ -2,11 +2,14 @@ module Foreign.CUDA.FFT ( module Foreign.CUDA.FFT.Error, - module Foreign.CUDA.FFT.Plan, module Foreign.CUDA.FFT.Execute, + module Foreign.CUDA.FFT.Plan, + module Foreign.CUDA.FFT.Stream ) where import Foreign.CUDA.FFT.Error -import Foreign.CUDA.FFT.Plan import Foreign.CUDA.FFT.Execute +import Foreign.CUDA.FFT.Plan +import Foreign.CUDA.FFT.Stream + diff --git a/Foreign/CUDA/FFT/Error.chs b/Foreign/CUDA/FFT/Error.chs --- a/Foreign/CUDA/FFT/Error.chs +++ b/Foreign/CUDA/FFT/Error.chs @@ -22,22 +22,26 @@ -- Describe each error code -- describe :: Result -> String -describe Success = "success" -describe InvalidPlan = "invalid plan handle" -describe AllocFailed = "resource allocation failed" -describe InvalidType = "no longer used" -describe InvalidValue = "unsupported value or parameter passed to a function" -describe InternalError = "internal error" -describe ExecFailed = "failed to execute an FFT on the GPU" -describe SetupFailed = "the CUFFT library failed to initialize" -describe InvalidSize = "invalid transform size" -describe UnalignedData = "no longer used" +describe Success = "success" +describe InvalidPlan = "invalid plan handle" +describe AllocFailed = "resource allocation failed" +describe InvalidType = "no longer used" +describe InvalidValue = "unsupported value or parameter passed to a function" +describe InternalError = "internal error" +describe ExecFailed = "failed to execute an FFT on the GPU" +describe SetupFailed = "the CUFFT library failed to initialize" +describe InvalidSize = "invalid transform size" +describe UnalignedData = "no longer used" #if CUDA_VERSION >= 6000 describe IncompleteParameterList = "missing parameters in call" -describe InvalidDevice = "execution of a plan was on different GPU than plan creation" -describe ParseError = "internal plan database error" -describe NoWorkspace = "no workspace has been provided prior to plan execution" +describe InvalidDevice = "execution of a plan was on a different GPU than plan creation" +describe ParseError = "internal plan database error" +describe NoWorkspace = "no workspace has been provided prior to plan execution" #endif +#if CUDA_VERSION >= 6050 +describe NotImplemented = "not implemented" +describe LicenseError = "cufft license error" +#endif -- Exceptions ------------------------------------------------------------------ @@ -78,3 +82,4 @@ case status of Success -> return () _ -> throwIO (ExitCode status) + diff --git a/Foreign/CUDA/FFT/Internal/C2HS.hs b/Foreign/CUDA/FFT/Internal/C2HS.hs --- a/Foreign/CUDA/FFT/Internal/C2HS.hs +++ b/Foreign/CUDA/FFT/Internal/C2HS.hs @@ -8,6 +8,7 @@ -- System import Foreign +import Foreign.C -- Conversions ----------------------------------------------------------------- @@ -15,42 +16,46 @@ -- | Integral conversion -- -{-# INLINE cIntConv #-} +{-# INLINE [1] cIntConv #-} cIntConv :: (Integral a, Integral b) => a -> b cIntConv = fromIntegral -- | Floating conversion -- -{-# INLINE[1] cFloatConv #-} +{-# INLINE [1] cFloatConv #-} cFloatConv :: (RealFloat a, RealFloat b) => a -> b cFloatConv = realToFrac -- As this conversion by default goes via `Rational', it can be very slow... {-# RULES - "cFloatConv/Float->Float" forall (x::Float). cFloatConv x = x; - "cFloatConv/Double->Double" forall (x::Double). cFloatConv x = x + "cFloatConv/Float->Float" forall (x::Float). cFloatConv x = x; + "cFloatConv/Double->Double" forall (x::Double). cFloatConv x = x; + "cFloatConv/Float->CFloat" forall (x::Float). cFloatConv x = CFloat x; + "cFloatConv/CFloat->Float" forall (x::Float). cFloatConv CFloat x = x; + "cFloatConv/Double->CDouble" forall (x::Double). cFloatConv x = CDouble x; + "cFloatConv/CDouble->Double" forall (x::Double). cFloatConv CDouble x = x #-} -- | Obtain C value from Haskell 'Bool'. -- -{-# INLINE cFromBool #-} +{-# INLINE [1] cFromBool #-} cFromBool :: Num a => Bool -> a cFromBool = fromBool -- | Obtain Haskell 'Bool' from C value. -- -{-# INLINE cToBool #-} +{-# INLINE [1] cToBool #-} cToBool :: (Eq a, Num a) => a -> Bool cToBool = toBool -- | Convert a C enumeration to Haskell. -- -{-# INLINE cToEnum #-} +{-# INLINE [1] cToEnum #-} cToEnum :: (Integral i, Enum e) => i -> e cToEnum = toEnum . cIntConv -- | Convert a Haskell enumeration to C. -- -{-# INLINE cFromEnum #-} +{-# INLINE [1] cFromEnum #-} cFromEnum :: (Enum e, Integral i) => e -> i cFromEnum = cIntConv . fromEnum diff --git a/Foreign/CUDA/FFT/Plan.chs b/Foreign/CUDA/FFT/Plan.chs --- a/Foreign/CUDA/FFT/Plan.chs +++ b/Foreign/CUDA/FFT/Plan.chs @@ -124,3 +124,4 @@ {# fun unsafe cufftDestroy { useHandle `Handle' } -> `Result' cToEnum #} + diff --git a/Foreign/CUDA/FFT/Stream.chs b/Foreign/CUDA/FFT/Stream.chs new file mode 100644 --- /dev/null +++ b/Foreign/CUDA/FFT/Stream.chs @@ -0,0 +1,35 @@ +{-# LANGUAGE CPP #-} +{-# LANGUAGE ForeignFunctionInterface #-} + +module Foreign.CUDA.FFT.Stream ( + + -- * Streamed transforms + setStream, + +) where + +-- friends +import Foreign.CUDA.Types +import Foreign.CUDA.FFT.Plan +import Foreign.CUDA.FFT.Error +import Foreign.CUDA.FFT.Internal.C2HS + +-- system +import Foreign +import Foreign.C + +#include +{# context lib="cufft" #} + +-- | Associates a CUDA stream with a CUFFT plan. All kernel launches made during +-- plan execution are now done through the associated stream, enabling overlap +-- with activity in other streams (e.g. data copying). The association remains +-- until the plan is destroyed or the stream is changed. +-- +setStream :: Handle -> Stream -> IO () +setStream ctx st = nothingIfOk =<< cufftSetStream ctx st + +{# fun unsafe cufftSetStream + { useHandle `Handle' + , useStream `Stream' } -> `Result' cToEnum #} + diff --git a/cbits/wrap.h b/cbits/wrap.h --- a/cbits/wrap.h +++ b/cbits/wrap.h @@ -13,9 +13,11 @@ #define __AVAILABILITY__ #define __OSX_AVAILABLE_STARTING(_mac, _iphone) #define __OSX_AVAILABLE_BUT_DEPRECATED(_macIntro, _macDep, _iphoneIntro, _iphoneDep) +#define __OSX_AVAILABLE_BUT_DEPRECATED_MSG(_osxIntro, _osxDep, _iosIntro, _iosDep, _msg) #endif #include #include #endif + diff --git a/cufft.cabal b/cufft.cabal --- a/cufft.cabal +++ b/cufft.cabal @@ -1,24 +1,24 @@ --- Initial cufft.cabal generated by cabal init. For further documentation, --- see http://haskell.org/cabal/users-guide/ +name: cufft +version: 0.1.2.0 +synopsis: Haskell bindings for the CUFFT library +description: + The CUFFT library is part of the CUDA developer toolkit. + . + + . + It contains a highly optimised, FFTW compatible, Fast-Fourier Transform + implementation that can execute on NVIDIA GPUs. + . -name: cufft -version: 0.1.1.0 -synopsis: Haskell bindings for the CUFFT library -description: The CUFFT library is part of the CUDA developer toolkit. - . - - . - It contains a highly optimised, FFTW compatible, Fast-Fourier Transform - implementation that can execute on NVIDIA GPUs. -homepage: http://github.com/robeverest/cufft -license: BSD3 -license-file: LICENSE -author: Robert Clifton-Everest -maintainer: Robert Clifton-Everest +homepage: http://github.com/robeverest/cufft +license: BSD3 +license-file: LICENSE +author: Robert Clifton-Everest +maintainer: Robert Clifton-Everest -category: Foreign -build-type: Custom -cabal-version: >=1.8 +category: Foreign +build-type: Custom +cabal-version: >=1.8 Extra-tmp-files: cufft.buildinfo config.status config.log Extra-source-files: configure @@ -33,12 +33,13 @@ exposed-modules: Foreign.CUDA.FFT other-modules: Foreign.CUDA.FFT.Error - Foreign.CUDA.FFT.Plan Foreign.CUDA.FFT.Execute + Foreign.CUDA.FFT.Plan + Foreign.CUDA.FFT.Stream Foreign.CUDA.FFT.Internal.C2HS build-depends: base == 4.*, - cuda >= 0.4.1 + cuda >= 0.6.6 build-tools: c2hs >= 0.16 ghc-options: -Wall -O2 -funbox-strict-fields -fwarn-tabs