diff --git a/Data/Array/Accelerate/CUDA/CodeGen.hs b/Data/Array/Accelerate/CUDA/CodeGen.hs --- a/Data/Array/Accelerate/CUDA/CodeGen.hs +++ b/Data/Array/Accelerate/CUDA/CodeGen.hs @@ -506,10 +506,10 @@ codegenIntegralScalar ty x | IntegralDict <- integralDict ty = [cexp| ( $ty:(codegenIntegralType ty) ) $exp:(cintegral x) |] codegenFloatingScalar :: FloatingType a -> a -> C.Exp -codegenFloatingScalar (TypeFloat _) x = C.Const (FloatConst (shows x "f") (toRational x) noSrcLoc) noSrcLoc -codegenFloatingScalar (TypeCFloat _) x = C.Const (FloatConst (shows x "f") (toRational x) noSrcLoc) noSrcLoc -codegenFloatingScalar (TypeDouble _) x = C.Const (DoubleConst (show x) (toRational x) noSrcLoc) noSrcLoc -codegenFloatingScalar (TypeCDouble _) x = C.Const (DoubleConst (show x) (toRational x) noSrcLoc) noSrcLoc +codegenFloatingScalar (TypeFloat _) x = C.Const (FloatConst (shows x "f") (toRational x) noLoc) noLoc +codegenFloatingScalar (TypeCFloat _) x = C.Const (FloatConst (shows x "f") (toRational x) noLoc) noLoc +codegenFloatingScalar (TypeDouble _) x = C.Const (DoubleConst (show x) (toRational x) noLoc) noLoc +codegenFloatingScalar (TypeCDouble _) x = C.Const (DoubleConst (show x) (toRational x) noLoc) noLoc codegenNonNumScalar :: NonNumType a -> a -> C.Exp codegenNonNumScalar (TypeBool _) x = cbool x diff --git a/Data/Array/Accelerate/CUDA/CodeGen/Base.hs b/Data/Array/Accelerate/CUDA/CodeGen/Base.hs --- a/Data/Array/Accelerate/CUDA/CodeGen/Base.hs +++ b/Data/Array/Accelerate/CUDA/CodeGen/Base.hs @@ -71,7 +71,7 @@ ccall fn args = [cexp|$id:fn ($args:args)|] typename :: String -> Type -typename name = Type (DeclSpec [] [] (Tnamed (Id name noSrcLoc) noSrcLoc) noSrcLoc) (DeclRoot noSrcLoc) noSrcLoc +typename name = Type (DeclSpec [] [] (Tnamed (Id name noLoc) noLoc) noLoc) (DeclRoot noLoc) noLoc cchar :: Char -> Exp cchar c = [cexp|$char:c|] @@ -178,7 +178,7 @@ -- Turn a plain type into a ptr type -- cptr :: Type -> Type -cptr t | Type d@(DeclSpec _ _ _ _) r@(DeclRoot _) lb <- t = Type d (Ptr [] r noSrcLoc) lb +cptr t | Type d@(DeclSpec _ _ _ _) r@(DeclRoot _) lb <- t = Type d (Ptr [] r noLoc) lb | otherwise = t diff --git a/Data/Array/Accelerate/CUDA/Compile.hs b/Data/Array/Accelerate/CUDA/Compile.hs --- a/Data/Array/Accelerate/CUDA/Compile.hs +++ b/Data/Array/Accelerate/CUDA/Compile.hs @@ -36,8 +36,6 @@ import Numeric import Prelude hiding ( exp, catch ) import Control.Applicative hiding ( Const ) -import Blaze.ByteString.Builder -import Blaze.ByteString.Builder.Char8 import Control.Exception import Control.Monad import Control.Monad.Trans @@ -52,11 +50,12 @@ import System.IO import System.IO.Unsafe import System.Process -import Text.PrettyPrint.Mainland ( RDoc(..), ppr, renderCompact ) -import Data.ByteString.Internal ( w2c ) +import Text.PrettyPrint.Mainland ( ppr, renderCompact, displayLazyText ) import qualified Data.HashSet as Set import qualified Data.ByteString as B -import qualified Data.ByteString.Lazy as L +import qualified Data.Text.Lazy as T +import qualified Data.Text.Lazy.IO as T +import qualified Data.Text.Lazy.Encoding as T import qualified Foreign.CUDA.Driver as CUDA import qualified Foreign.CUDA.Analysis as CUDA @@ -360,12 +359,13 @@ compile table dev acc fvar = do exists <- isJust `fmap` liftIO (KT.lookup table key) unless exists $ do - message $ unlines [ show key, map w2c (L.unpack code) ] + message $ unlines [ show key, T.unpack code ] nvcc <- fromMaybe (error "nvcc: command not found") <$> liftIO (findExecutable "nvcc") (file,hdl) <- openTemporaryFile "dragon.cu" -- rawr! flags <- compileFlags file (_,_,_,pid) <- liftIO $ do - L.hPut hdl code `finally` hClose hdl + message $ "execute: " ++ nvcc ++ " " ++ unwords flags + T.hPutStr hdl code `finally` hClose hdl createProcess (proc nvcc flags) `onException` removeFile file -- liftIO $ KT.insert table key (CompileProcess file pid) @@ -374,15 +374,8 @@ where cunit = codegenAcc dev acc fvar entry = show cunit - key = (CUDA.computeCapability dev, hashlazy code) - code = toLazyByteString - . layout . renderCompact $ ppr cunit - -- - layout (RText _ s next) = fromString s `mappend` layout next - layout (RChar c next) = fromChar c `mappend` layout next - layout (RLine _ next) = fromChar '\n' `mappend` layout next -- no indenting - layout (RPos _ next) = layout next -- no line markers - layout REmpty = mempty -- done + key = (CUDA.computeCapability dev, hashlazy (T.encodeUtf8 code) ) + code = displayLazyText . renderCompact $ ppr cunit -- Wait for the compilation process to finish diff --git a/Data/Array/Accelerate/CUDA/Execute.hs b/Data/Array/Accelerate/CUDA/Execute.hs --- a/Data/Array/Accelerate/CUDA/Execute.hs +++ b/Data/Array/Accelerate/CUDA/Execute.hs @@ -1,6 +1,15 @@ -{-# LANGUAGE UndecidableInstances, OverlappingInstances, IncoherentInstances #-} -{-# LANGUAGE BangPatterns, CPP, GADTs, ScopedTypeVariables, FlexibleInstances #-} -{-# LANGUAGE RankNTypes, TupleSections, TypeOperators, TypeSynonymInstances #-} +{-# LANGUAGE BangPatterns #-} +{-# LANGUAGE CPP #-} +{-# LANGUAGE FlexibleInstances #-} +{-# LANGUAGE GADTs #-} +{-# LANGUAGE IncoherentInstances #-} +{-# LANGUAGE OverlappingInstances #-} +{-# LANGUAGE RankNTypes #-} +{-# LANGUAGE ScopedTypeVariables #-} +{-# LANGUAGE TupleSections #-} +{-# LANGUAGE TypeOperators #-} +{-# LANGUAGE TypeSynonymInstances #-} +{-# LANGUAGE UndecidableInstances #-} {-# OPTIONS -fno-warn-orphans #-} -- | -- Module : Data.Array.Accelerate.CUDA.Execute diff --git a/accelerate-cuda.cabal b/accelerate-cuda.cabal --- a/accelerate-cuda.cabal +++ b/accelerate-cuda.cabal @@ -1,5 +1,5 @@ Name: accelerate-cuda -Version: 0.12.1.1 +Version: 0.12.1.2 Cabal-version: >= 1.6 Tested-with: GHC >= 7.4 Build-type: Configure @@ -89,25 +89,25 @@ Build-depends: accelerate >= 0.12.1 && < 0.13, array >= 0.3 && < 0.5, base == 4.*, - binary == 0.5.*, - blaze-builder == 0.3.*, - bytestring == 0.9.*, - containers == 0.4.*, - cryptohash == 0.7.*, + binary >= 0.5, + bytestring >= 0.9, + containers >= 0.4, + cryptohash >= 0.7, cuda >= 0.4.1 && < 0.5, - directory >= 1.0 && < 1.2, - fclabels >= 1.0 && < 1.2, - filepath >= 1.0 && < 1.4, - hashable == 1.1.*, - hashtables >= 1.0.1 && < 1.1, - language-c-quote == 0.3.*, - mainland-pretty == 0.1.*, - mtl >= 2.0 && < 2.2, - pretty >= 1.0 && < 1.2, - process >= 1.0 && < 1.2, - srcloc == 0.1.*, - transformers >= 0.2 && < 0.4, - unordered-containers >= 0.1.4 && < 0.3 + directory >= 1.0, + fclabels >= 1.0, + filepath >= 1.0, + hashable >= 1.1, + hashtables >= 1.0.1, + language-c-quote >= 0.4, + mainland-pretty >= 0.2, + mtl >= 2.0, + pretty >= 1.0, + process >= 1.0, + srcloc >= 0.2, + text >= 0.11, + transformers >= 0.2, + unordered-containers >= 0.1.4 if os(windows) build-depends: Win32 >= 2.2.1 diff --git a/configure b/configure --- a/configure +++ b/configure @@ -1,6 +1,6 @@ #! /bin/sh # Guess values for system-dependent variables and create Makefiles. -# Generated by GNU Autoconf 2.69 for accelerate-cuda 0.12.1.1. +# Generated by GNU Autoconf 2.69 for accelerate-cuda 0.13.0.0. # # Report bugs to . # @@ -579,8 +579,8 @@ # Identity of this package. PACKAGE_NAME='accelerate-cuda' PACKAGE_TARNAME='accelerate-cuda' -PACKAGE_VERSION='0.12.1.1' -PACKAGE_STRING='accelerate-cuda 0.12.1.1' +PACKAGE_VERSION='0.13.0.0' +PACKAGE_STRING='accelerate-cuda 0.13.0.0' PACKAGE_BUGREPORT='accelerate-haskell@googlegroups.com' PACKAGE_URL='' @@ -591,7 +591,6 @@ ghc_flags type_hs_char type_hs_int -GHC OBJEXT EXEEXT ac_ct_CXX @@ -599,6 +598,7 @@ LDFLAGS CXXFLAGS CXX +GHC target_alias host_alias build_alias @@ -640,6 +640,9 @@ ac_subst_files='' ac_user_opts=' enable_option_checking +with_compiler +with_nvcc +with_gcc ' ac_precious_vars='build_alias host_alias @@ -1190,7 +1193,7 @@ # Omit some internal or obsolete options to make the list less imposing. # This message is too long to be a string in the A/UX 3.1 sh. cat <<_ACEOF -\`configure' configures accelerate-cuda 0.12.1.1 to adapt to many kinds of systems. +\`configure' configures accelerate-cuda 0.13.0.0 to adapt to many kinds of systems. Usage: $0 [OPTION]... [VAR=VALUE]... @@ -1251,10 +1254,17 @@ if test -n "$ac_init_help"; then case $ac_init_help in - short | recursive ) echo "Configuration of accelerate-cuda 0.12.1.1:";; + short | recursive ) echo "Configuration of accelerate-cuda 0.13.0.0:";; esac cat <<\_ACEOF +Optional Packages: + --with-PACKAGE[=ARG] use PACKAGE [ARG=yes] + --without-PACKAGE do not use PACKAGE (same as --with-PACKAGE=no) +Haskell compiler +CUDA compiler +C compiler + Some influential environment variables: CXX C++ compiler command CXXFLAGS C++ compiler flags @@ -1330,7 +1340,7 @@ test -n "$ac_init_help" && exit $ac_status if $ac_init_version; then cat <<\_ACEOF -accelerate-cuda configure 0.12.1.1 +accelerate-cuda configure 0.13.0.0 generated by GNU Autoconf 2.69 Copyright (C) 2012 Free Software Foundation, Inc. @@ -1385,7 +1395,7 @@ This file contains any messages produced by compilers while running configure, to aid debugging if configure makes a mistake. -It was created by accelerate-cuda $as_me 0.12.1.1, which was +It was created by accelerate-cuda $as_me 0.13.0.0, which was generated by GNU Autoconf 2.69. Invocation command line was $ $0 $@ @@ -1737,9 +1747,59 @@ ac_config_files="$ac_config_files accelerate-cuda.buildinfo cubits/accelerate_cuda_shape.h" -# This doesn't pick up a user specified '--with-compiler' flag -# -ac_ext=cpp + +# Check whether --with-compiler was given. +if test "${with_compiler+set}" = set; then : + withval=$with_compiler; GHC=$withval +else + # Extract the first word of "ghc", so it can be a program name with args. +set dummy ghc; ac_word=$2 +{ $as_echo "$as_me:${as_lineno-$LINENO}: checking for $ac_word" >&5 +$as_echo_n "checking for $ac_word... " >&6; } +if ${ac_cv_path_GHC+:} false; then : + $as_echo_n "(cached) " >&6 +else + case $GHC in + [\\/]* | ?:[\\/]*) + ac_cv_path_GHC="$GHC" # Let the user override the test with a path. + ;; + *) + as_save_IFS=$IFS; IFS=$PATH_SEPARATOR +for as_dir in $PATH +do + IFS=$as_save_IFS + test -z "$as_dir" && as_dir=. + for ac_exec_ext in '' $ac_executable_extensions; do + if as_fn_executable_p "$as_dir/$ac_word$ac_exec_ext"; then + ac_cv_path_GHC="$as_dir/$ac_word$ac_exec_ext" + $as_echo "$as_me:${as_lineno-$LINENO}: found $as_dir/$ac_word$ac_exec_ext" >&5 + break 2 + fi +done + done +IFS=$as_save_IFS + + ;; +esac +fi +GHC=$ac_cv_path_GHC +if test -n "$GHC"; then + { $as_echo "$as_me:${as_lineno-$LINENO}: result: $GHC" >&5 +$as_echo "$GHC" >&6; } +else + { $as_echo "$as_me:${as_lineno-$LINENO}: result: no" >&5 +$as_echo "no" >&6; } +fi + + +fi + + +# Check whether --with-nvcc was given. +if test "${with_nvcc+set}" = set; then : + withval=$with_nvcc; NVCC=$withval +else + ac_ext=cpp ac_cpp='$CXXCPP $CPPFLAGS' ac_compile='$CXX -c $CXXFLAGS $CPPFLAGS conftest.$ac_ext >&5' ac_link='$CXX -o conftest$ac_exeext $CXXFLAGS $CPPFLAGS $LDFLAGS conftest.$ac_ext $LIBS >&5' @@ -2246,50 +2306,18 @@ ac_link='$CC -o conftest$ac_exeext $CFLAGS $CPPFLAGS $LDFLAGS conftest.$ac_ext $LIBS >&5' ac_compiler_gnu=$ac_cv_c_compiler_gnu -# Extract the first word of "ghc", so it can be a program name with args. -set dummy ghc; ac_word=$2 -{ $as_echo "$as_me:${as_lineno-$LINENO}: checking for $ac_word" >&5 -$as_echo_n "checking for $ac_word... " >&6; } -if ${ac_cv_path_GHC+:} false; then : - $as_echo_n "(cached) " >&6 -else - case $GHC in - [\\/]* | ?:[\\/]*) - ac_cv_path_GHC="$GHC" # Let the user override the test with a path. - ;; - *) - as_save_IFS=$IFS; IFS=$PATH_SEPARATOR -for as_dir in $PATH -do - IFS=$as_save_IFS - test -z "$as_dir" && as_dir=. - for ac_exec_ext in '' $ac_executable_extensions; do - if as_fn_executable_p "$as_dir/$ac_word$ac_exec_ext"; then - ac_cv_path_GHC="$as_dir/$ac_word$ac_exec_ext" - $as_echo "$as_me:${as_lineno-$LINENO}: found $as_dir/$ac_word$ac_exec_ext" >&5 - break 2 - fi -done - done -IFS=$as_save_IFS - - ;; -esac fi -GHC=$ac_cv_path_GHC -if test -n "$GHC"; then - { $as_echo "$as_me:${as_lineno-$LINENO}: result: $GHC" >&5 -$as_echo "$GHC" >&6; } -else - { $as_echo "$as_me:${as_lineno-$LINENO}: result: no" >&5 -$as_echo "no" >&6; } -fi +# Check whether --with-gcc was given. +if test "${with_gcc+set}" = set; then : + withval=$with_gcc; CC=$withval +fi + { $as_echo "$as_me:${as_lineno-$LINENO}: checking size of Int" >&5 $as_echo_n "checking size of Int... " >&6; } @@ -2870,7 +2898,7 @@ # report actual input values of CONFIG_FILES etc. instead of their # values after options handling. ac_log=" -This file was extended by accelerate-cuda $as_me 0.12.1.1, which was +This file was extended by accelerate-cuda $as_me 0.13.0.0, which was generated by GNU Autoconf 2.69. Invocation command line was CONFIG_FILES = $CONFIG_FILES @@ -2923,7 +2951,7 @@ cat >>$CONFIG_STATUS <<_ACEOF || ac_write_fail=1 ac_cs_config="`$as_echo "$ac_configure_args" | sed 's/^ //; s/[\\""\`\$]/\\\\&/g'`" ac_cs_version="\\ -accelerate-cuda config.status 0.12.1.1 +accelerate-cuda config.status 0.13.0.0 configured by $0, generated by GNU Autoconf 2.69, with options \\"\$ac_cs_config\\"