diff --git a/Data/Array/Accelerate/Math/DFT.hs b/Data/Array/Accelerate/Math/DFT.hs --- a/Data/Array/Accelerate/Math/DFT.hs +++ b/Data/Array/Accelerate/Math/DFT.hs @@ -2,7 +2,8 @@ {-# LANGUAGE TypeOperators #-} -- | -- Module : Data.Array.Accelerate.Math.DFT --- Copyright : [2012] Manuel M T Chakravarty, Gabriele Keller, Trevor L. McDonell +-- Copyright : [2012..2014] Manuel M T Chakravarty, Gabriele Keller, Trevor L. McDonell +-- [2013..2014] Robert Clifton-Everest -- License : BSD3 -- -- Maintainer : Manuel M T Chakravarty diff --git a/Data/Array/Accelerate/Math/DFT/Centre.hs b/Data/Array/Accelerate/Math/DFT/Centre.hs --- a/Data/Array/Accelerate/Math/DFT/Centre.hs +++ b/Data/Array/Accelerate/Math/DFT/Centre.hs @@ -1,7 +1,9 @@ -{-# LANGUAGE TypeOperators #-} +{-# LANGUAGE FlexibleContexts #-} +{-# LANGUAGE TypeOperators #-} -- | -- Module : Data.Array.Accelerate.Math.DFT.Centre --- Copyright : [2012..2013] Manuel M T Chakravarty, Gabriele Keller, Trevor L. McDonell, Robert Clifton-Everest +-- Copyright : [2012..2014] Manuel M T Chakravarty, Gabriele Keller, Trevor L. McDonell +-- [2013..2014] Robert Clifton-Everest -- License : BSD3 -- -- Maintainer : Manuel M T Chakravarty @@ -67,7 +69,7 @@ where p ix = let Z:.x = unlift ix :: Z :. Exp Int - in index1 (x <* mw ? (x + mw, x - mw)) + in index1 (x A.<* mw ? (x + mw, x - mw)) Z:.w = unlift (A.shape arr) mw = w `div` 2 @@ -80,8 +82,8 @@ where p ix = let Z:.y:.x = unlift ix :: Z :. Exp Int :. Exp Int - in index2 (y <* mh ? (y + mh, y - mh)) - (x <* mw ? (x + mw, x - mw)) + in index2 (y A.<* mh ? (y + mh, y - mh)) + (x A.<* mw ? (x + mw, x - mw)) Z:.h:.w = unlift (A.shape arr) (mh,mw) = (h `div` 2, w `div` 2) @@ -94,10 +96,10 @@ where p ix = let Z:.z:.y:.x = unlift ix :: Z :. Exp Int :. Exp Int :. Exp Int - in index3 (z <* md ? (z + md, z - md)) - (y <* mh ? (y + mh, y - mh)) - (x <* mw ? (x + mw, x - mw)) - Z:.h:.w:.d = unlift (A.shape arr) - (mh,mw,md) = (h `div` 2, w `div` 2, d `div` 2) + in index3 (z A.<* md ? (z + md, z - md)) + (y A.<* mh ? (y + mh, y - mh)) + (x A.<* mw ? (x + mw, x - mw)) + Z:.h:.w:.d = unlift (A.shape arr) + (mh,mw,md) = (h `div` 2, w `div` 2, d `div` 2) index3 i j k = lift (Z:.i:.j:.k) diff --git a/Data/Array/Accelerate/Math/DFT/Roots.hs b/Data/Array/Accelerate/Math/DFT/Roots.hs --- a/Data/Array/Accelerate/Math/DFT/Roots.hs +++ b/Data/Array/Accelerate/Math/DFT/Roots.hs @@ -1,7 +1,8 @@ {-# LANGUAGE TypeOperators #-} -- | -- Module : Data.Array.Accelerate.Math.DFT.Roots --- Copyright : [2012] Manuel M T Chakravarty, Gabriele Keller, Trevor L. McDonell +-- Copyright : [2012..2014] Manuel M T Chakravarty, Gabriele Keller, Trevor L. McDonell +-- [2013..2014] Robert Clifton-Everest -- License : BSD3 -- -- Maintainer : Manuel M T Chakravarty diff --git a/Data/Array/Accelerate/Math/FFT.hs b/Data/Array/Accelerate/Math/FFT.hs --- a/Data/Array/Accelerate/Math/FFT.hs +++ b/Data/Array/Accelerate/Math/FFT.hs @@ -8,7 +8,8 @@ {-# LANGUAGE ViewPatterns #-} -- | -- Module : Data.Array.Accelerate.Math.FFT --- Copyright : [2012..2013] Manuel M T Chakravarty, Gabriele Keller, Trevor L. McDonell, Robert Clifton-Everest +-- Copyright : [2012..2013] Manuel M T Chakravarty, Gabriele Keller, Trevor L. McDonell +-- [2013..2014] Robert Clifton-Everest -- License : BSD3 -- -- Maintainer : Manuel M T Chakravarty @@ -31,7 +32,6 @@ ) where -import Prelude as P import Data.Array.Accelerate as A import Data.Array.Accelerate.Array.Sugar ( showShape ) import Data.Array.Accelerate.Data.Complex @@ -45,10 +45,12 @@ import System.Mem.Weak import System.IO.Unsafe import Foreign.CUDA.FFT -import qualified Foreign.CUDA.Driver as CUDA hiding (free) +import qualified Foreign.CUDA.Types as CUDA +import qualified Foreign.CUDA.Driver as CUDA #endif import Data.Bits +import Prelude as P data Mode = Forward | Reverse | Inverse deriving (Eq, Show) @@ -301,14 +303,16 @@ pure = interleave . p . deinterleave sh sign = signOfMode mode :: Int - foreignFFT :: Array DIM1 e -> CIO (Array DIM1 e) - foreignFFT arr' = do + foreignFFT :: CUDA.Stream -> Array DIM1 e -> CIO (Array DIM1 e) + foreignFFT stream arr' = do output <- allocateArray (S.shape arr') iptr <- floatingDevicePtr arr' optr <- floatingDevicePtr output --Execute - liftIO $ execute iptr optr + liftIO $ do + setStream hndl stream + execute iptr optr return output @@ -345,7 +349,7 @@ in generate (lift (sh :. n+m)) (\ix -> let sz :. i = unlift ix :: Exp sh :. Exp Int - in i <* n ? (xs ! lift (sz:.i), ys ! lift (sz:.i-n) )) + in i A.<* n ? (xs ! lift (sz:.i), ys ! lift (sz:.i-n) )) #ifdef ACCELERATE_CUDA_BACKEND diff --git a/accelerate-fft.cabal b/accelerate-fft.cabal --- a/accelerate-fft.cabal +++ b/accelerate-fft.cabal @@ -1,7 +1,7 @@ Name: accelerate-fft -Version: 0.15.0.0 +Version: 0.15.1.0 Cabal-version: >= 1.6 -Tested-with: GHC >= 7.6 +Tested-with: GHC >= 7.8 Build-type: Simple Synopsis: FFT using the Accelerate library @@ -26,13 +26,12 @@ Stability: Experimental Flag cuda - Description: Enable support for using CUFFT via the CUDA backend's - FFI + Description: Enable support for using CUFFT via the CUDA backend's FFI Default: True Library Build-depends: accelerate == 0.15.*, - base >= 4.6 && < 4.8 + base >= 4.7 && < 4.9 Exposed-modules: Data.Array.Accelerate.Math.FFT Data.Array.Accelerate.Math.DFT @@ -45,7 +44,7 @@ CPP-options: -DACCELERATE_CUDA_BACKEND Build-depends: accelerate-cuda == 0.15.*, cuda >= 0.5, - cufft >= 0.1 + cufft >= 0.1.2 -- Don't add the extensions list here. Instead, place individual LANGUAGE -- pragmas in the files that require a specific extension. This means the @@ -53,7 +52,9 @@ -- -- Extensions: -Source-repository head - Type: git - Location: git://github.com/AccelerateHS/accelerate-fft.git +Source-repository this + type: git + location: git://github.com/AccelerateHS/accelerate-fft.git + branch: release/0.15 + tag: 0.15.1.0