darcs patch: UNDO: Inline shift in GHC's Bits instanc... (and 1 more)

Previous Topic Next Topic
classic Classic list List threaded Threaded
1 message Options
Reply | Threaded
Open this post in threaded view

darcs patch: UNDO: Inline shift in GHC's Bits instanc... (and 1 more)

Samuel Bronson
Sun Oct  8 22:09:06 EDT 2006  Samuel Bronson <[hidden email]>
  UNDO: Inline shift in GHC's Bits instances for {Int,Word}{,8,16,32,64}

Thu Oct 12 08:55:53 EDT 2006  Samuel Bronson <[hidden email]>
  * Use unchecked shifts to implement Data.Bits.rotate
  This should get rid of those cases, maybe lower the size enough that the inliner will like it?

New patches:

[Inline shift in GHC's Bits instances for {Int,Word}{,8,16,32,64}
Samuel Bronson <[hidden email]>*-20061009020906] {
hunk ./GHC/Word.hs 848
-    {-# INLINE shift #-}
hunk ./GHC/Word.hs 716
-    {-# INLINE shift #-}
hunk ./GHC/Word.hs 587
-    {-# INLINE shift #-}
hunk ./GHC/Word.hs 456
-    {-# INLINE shift #-}
hunk ./GHC/Word.hs 357
-    {-# INLINE shift #-}
hunk ./GHC/Word.hs 256
-    {-# INLINE shift #-}
hunk ./GHC/Word.hs 156
-    {-# INLINE shift #-}
hunk ./GHC/Int.hs 755
-    {-# INLINE shift #-}
hunk ./GHC/Int.hs 617
-    {-# INLINE shift #-}
hunk ./GHC/Int.hs 476
-    {-# INLINE shift #-}
hunk ./GHC/Int.hs 346
-    {-# INLINE shift #-}
hunk ./GHC/Int.hs 222
-    {-# INLINE shift #-}
hunk ./GHC/Int.hs 115
-    {-# INLINE shift #-}
hunk ./Data/Bits.hs 195
-    {-# INLINE shift #-}

[Use unchecked shifts to implement Data.Bits.rotate
Samuel Bronson <[hidden email]>**20061012125553
 This should get rid of those cases, maybe lower the size enough that the inliner will like it?
] {
hunk ./Data/Bits.hs 204
-        I# (word2Int# ((x'# `shiftL#` i'#) `or#`
-                       (x'# `shiftRL#` (wsib -# i'#))))
+        I# (word2Int# ((x'# `uncheckedShiftL#` i'#) `or#`
+                       (x'# `uncheckedShiftRL#` (wsib -# i'#))))
hunk ./GHC/Int.hs 126
-        = I8# (narrow8Int# (word2Int# ((x'# `shiftL#` i'#) `or#`
-                                       (x'# `shiftRL#` (8# -# i'#)))))
+        = I8# (narrow8Int# (word2Int# ((x'# `uncheckedShiftL#` i'#) `or#`
+                                       (x'# `uncheckedShiftRL#` (8# -# i'#)))))
hunk ./GHC/Int.hs 231
-        = I16# (narrow16Int# (word2Int# ((x'# `shiftL#` i'#) `or#`
-                                         (x'# `shiftRL#` (16# -# i'#)))))
+        = I16# (narrow16Int# (word2Int# ((x'# `uncheckedShiftL#` i'#) `or#`
+                                         (x'# `uncheckedShiftRL#` (16# -# i'#)))))
hunk ./GHC/Int.hs 481
-        = I32# (narrow32Int# (word2Int# ((x'# `shiftL#` i'#) `or#`
-                                        (x'# `shiftRL#` (32# -# i'#)))))
+        = I32# (narrow32Int# (word2Int# ((x'# `uncheckedShiftL#` i'#) `or#`
+                                         (x'# `uncheckedShiftRL#` (32# -# i'#)))))
hunk ./GHC/Int.hs 756
-        = I64# (word2Int# ((x'# `shiftL#` i'#) `or#`
-                           (x'# `shiftRL#` (64# -# i'#))))
+        = I64# (word2Int# ((x'# `uncheckedShiftL#` i'#) `or#`
+                           (x'# `uncheckedShiftRL#` (64# -# i'#))))
hunk ./GHC/Word.hs 165
-        | otherwise  = W# ((x# `shiftL#` i'#) `or#` (x# `shiftRL#` (wsib -# i'#)))
+        | otherwise  = W# ((x# `uncheckedShiftL#` i'#) `or#` (x# `uncheckedShiftRL#` (wsib -# i'#)))
hunk ./GHC/Word.hs 263
-        | otherwise  = W8# (narrow8Word# ((x# `shiftL#` i'#) `or#`
-                                          (x# `shiftRL#` (8# -# i'#))))
+        | otherwise  = W8# (narrow8Word# ((x# `uncheckedShiftL#` i'#) `or#`
+                                          (x# `uncheckedShiftRL#` (8# -# i'#))))
hunk ./GHC/Word.hs 362
-        | otherwise  = W16# (narrow16Word# ((x# `shiftL#` i'#) `or#`
-                                            (x# `shiftRL#` (16# -# i'#))))
+        | otherwise  = W16# (narrow16Word# ((x# `uncheckedShiftL#` i'#) `or#`
+                                            (x# `uncheckedShiftRL#` (16# -# i'#))))
hunk ./GHC/Word.hs 588
-        | otherwise  = W32# (narrow32Word# ((x# `shiftL#` i'#) `or#`
-                                            (x# `shiftRL#` (32# -# i'#))))
+        | otherwise  = W32# (narrow32Word# ((x# `uncheckedShiftL#` i'#) `or#`
+                                            (x# `uncheckedShiftRL#` (32# -# i'#))))
hunk ./GHC/Word.hs 845
-        | otherwise  = W64# ((x# `shiftL#` i'#) `or#`
-                             (x# `shiftRL#` (64# -# i'#)))
+        | otherwise  = W64# ((x# `uncheckedShiftL#` i'#) `or#`
+                             (x# `uncheckedShiftRL#` (64# -# i'#)))


[fix example in comment
Ross Paterson <[hidden email]>**20061023163925]
[Use the new Any type for dynamics (GHC only)
[add Data.Sequence to nhc98 build
[hidden email]**20061012135200]
[Remove Data.FiniteMap, add Control.Applicative, Data.Traversable, and
[hidden email]**20061012095605
 Data.Foldable to the nhc98 build.
[STM invariants
[hidden email]**20061007123253]
[Don't create GHC.Prim when bootstrapping; we can't, and we don't need it
Ian Lynagh <[hidden email]>**20061004165355]
[Data.ByteString: fix lazyness of take, drop & splitAt
Don Stewart <[hidden email]>**20061005011703
 ByteString.Lazy's take, drop and splitAt were too strict when demanding
 a byte string. Spotted by Einar Karttunen. Thanks to him and to Bertram
 Felgenhauer for explaining the problem and the fix.
[Fix syntax error that prevents building Haddock documentation on Windows
[hidden email]**20060917013530]
[Hugs only: unbreak typeRepKey
Ross Paterson <[hidden email]>**20060929102743]
[make hGetBufNonBlocking do something on Windows w/ -threaded
Simon Marlow <[hidden email]>**20060927145811
 hGetBufNonBlocking will behave the same as hGetBuf on Windows now, which
 is better than just crashing (which it did previously).
[add typeRepKey :: TypeRep -> IO Int
Simon Marlow <[hidden email]>**20060927100342
 See feature request #880
[fix header comment
Ross Paterson <[hidden email]>**20060926135843]
[Add strict versions of insertWith and insertWithKey (Data.Map)
[hidden email]**20060910162443]
[doc tweaks, including more precise equations for evaluate
Ross Paterson <[hidden email]>**20060910115259]
[Sync Data.ByteString with stable branch
Don Stewart <[hidden email]>**20060909050111
 This patch:
     * hides the LPS constructor (its in .Base if you need it)
     * adds functions to convert between strict and lazy bytestrings
     * and adds readInteger
[Typeable1 instances for STM and TVar
Ross Paterson <[hidden email]>**20060904231425]
[remove obsolete Hugs stuff
Ross Paterson <[hidden email]>**20060904223944]
[Cleaner isInfixOf suggestion from Ross Paterson
John Goerzen <[hidden email]>**20060901143654]
[New function isInfixOf that searches a list for a given sublist
John Goerzen <[hidden email]>**20060831151556
 isInfixOf "Haskell" "I really like Haskell." -> True
 isInfixOf "Ial" "I really like Haskell." -> False
 This function was first implemented in MissingH as MissingH.List.contains
[Better doc on Data.Map.lookup: explain what the monad is for
[hidden email]**20060903133440]
[fix hDuplicateTo on Windows
Simon Marlow <[hidden email]>**20060901150016
 deja vu - I'm sure I remember fixing this before...
[Improve documentation of atomically
[Add missing method genRange for StdGen (fixes #794)
 Trac #794 reports (correctly) that the implementation of StdGen
 only returns numbers in the range (0..something) rather than
 (minBound, maxBound), which is what StdGen's genRange claims.
 This commit fixes the problem, by implementing genRange for StdGen
 (previously it just used the default method).
[mark nhc98 import hack
Ross Paterson <[hidden email]>**20060831125219]
[remove some outdated comments
Simon Marlow <[hidden email]>**20060831104200]
[import Control.Arrow.ArrowZero to help nhc98's type checker
[hidden email]**20060831101105]
[remove Text.Regex(.Posix) from nhc98 build
[hidden email]**20060831101016]
[add Data.Foldable.{msum,asum}, plus tweaks to comments
Ross Paterson <[hidden email]>**20060830163521]
[fix doc typo
Ross Paterson <[hidden email]>**20060830134123]
[add Data.Foldable.{for_,forM_} and Data.Traversable.{for,forM}
Ross Paterson <[hidden email]>**20060830133805
 generalizing Control.Monad.{forM_,forM}
[Make length a good consumer
 Make length into a good consumer.  Fixes Trac bug #707.
 (Before length simply didn't use foldr.)
[Add Control.Monad.forM and forM_
Don Stewart <[hidden email]>**20060824081118
 flip mapM_ is more and more common, I find. Several suggestions have
 been made to add this, as foreach or something similar. This patch
 does just that:
     forM  :: (Monad m) => [a] -> (a -> m b) -> m [b]
     forM_ :: (Monad m) => [a] -> (a -> m b) -> m ()
 So we can write:
     Prelude Control.Monad> forM_ [1..4] $ \x -> print x
[Hide internal module from haddock in Data.ByteString
Don Stewart <[hidden email]>**20060828011515]
[add advice on avoiding import ambiguities
Ross Paterson <[hidden email]>**20060827170407]
[expand advice on importing these modules
Ross Paterson <[hidden email]>**20060827164044]
[add Haddock marker
Ross Paterson <[hidden email]>**20060827115140]
[Inline shift in GHC's Bits instances for {Int,Word}{,8,16,32,64}
Samuel Bronson <[hidden email]>**20061009020906]
[Clarify how one hides Prelude.catch
Don Stewart <[hidden email]>**20060826124346
 User feedback indicated that an example was required, of how to hide
 Prelude.catch, so add such an example to the docs
[Workaround for OSes that don't have intmax_t and uintmax_t
Ian Lynagh <[hidden email]>**20060825134936
 OpenBSD (and possibly others) do not have intmax_t and uintmax_t types:
 so substitute (unsigned) long long if we have them, otherwise
 (unsigned) long.
[add docs for par
Simon Marlow <[hidden email]>**20060825110610]
[document minimal complete definition for Bits
Ross Paterson <[hidden email]>**20060824140504]
[C regex library bits have moved to the regex-posix package
Simon Marlow <[hidden email]>**20060824132311]
[Add shared Typeable support (ghc only)
Esa Ilari Vuokko <[hidden email]>**20060823003126]
[this should have been removed with the previous patch
Simon Marlow <[hidden email]>**20060824121223]
[remove Text.Regx & Text.Regex.Posix
Simon Marlow <[hidden email]>**20060824094615
 These are subsumed by the new regex-base, regex-posix and regex-compat
[explicitly tag Data.ByteString rules with the FPS prefix.
Don Stewart <[hidden email]>**20060824041326]
[Add spec rules for sections in Data.ByteString
Don Stewart <[hidden email]>**20060824012611]
[Sync Data.ByteString with current stable branch, 0.7
Don Stewart <[hidden email]>**20060823143338]
[add notes about why copyFile doesn't remove the target
Simon Marlow <[hidden email]>**20060823095059]
[copyFile: try removing the target file before opening it for writing
Simon Marlow <[hidden email]>*-20060822121909]
[copyFile: try removing the target file before opening it for writing
Simon Marlow <[hidden email]>**20060822121909]
[add alternative functors and extra instances
Ross Paterson <[hidden email]>**20060821152151
 * Alternative class, for functors with a monoid
 * instances for Const
 * instances for arrows
[generate Haddock docs on all platforms
Simon Marlow <[hidden email]>**20060821131612]
[remove extra comma from import
Ross Paterson <[hidden email]>**20060819173954]
[fix docs for withC(A)StringLen
Ross Paterson <[hidden email]>**20060818170328]
[use Haskell'98 compliant indentation in do blocks
[hidden email]**20060818130810]
[use correct names of IOArray operations for nhc98
[hidden email]**20060818130714]
[add mapMaybe and mapEither, plus WithKey variants
Ross Paterson <[hidden email]>**20060817235041]
[remove Text.Html from nhc98 build
[hidden email]**20060817135502]
[eliminate more HOST_OS tests
Ross Paterson <[hidden email]>**20060815190609]
[Hugs only: disable unused process primitives
Ross Paterson <[hidden email]>**20060813184435
 These were the cause of Hugs bug #30, I think, and weren't used by Hugs anyway.
[markup fix to Data.HashTable
Ross Paterson <[hidden email]>**20060812103835]
[revert removal of ghcconfig.h from package.conf.in
Ross Paterson <[hidden email]>**20060812082702
 as it's preprocessed with -undef (pointed out by Esa Ilari Vuokko)
[fix Data.HashTable for non-GHC
Ross Paterson <[hidden email]>**20060811231521]
[remove deprecated 'withObject'
Simon Marlow <[hidden email]>**20060811152350]
[Jan-Willem Maessen's improved implementation of Data.HashTable
Simon Marlow <[hidden email]>**20060811151024
 Rather than incrementally enlarging the hash table, this version
 just does it in one go when the table gets too full.
[Warning police: Make some prototypes from the RTS known
[hidden email]**20060811144629]
[Warning police: Removed useless catch-all clause
[hidden email]**20060811142208]
[reduce dependency on ghcconfig.h
Ross Paterson <[hidden email]>**20060811124030
 The only remaining use is in cbits/dirUtils.h, which tests solaris2_HOST_OS
 (Also System.Info uses ghcplatform.h and several modules import MachDeps.h
 to get SIZEOF_* and ALIGNMENT_* from ghcautoconf.h)
[(non-GHC only) track MArray interface change
Ross Paterson <[hidden email]>**20060810182902]
[move Text.Html to a separate package
Simon Marlow <[hidden email]>**20060810113017]
[bump version to 2.0
Simon Marlow <[hidden email]>**20060810112833]
[Remove deprecated Data.FiniteMap and Data.Set interfaces
Simon Marlow <[hidden email]>**20060809153810]
[move altzone test from ghc to base package
Ross Paterson <[hidden email]>**20060809124259]
[remove unnecessary #include "ghcconfig.h"
Ross Paterson <[hidden email]>**20060809123812]
[Change the API of MArray to allow resizable arrays
Simon Marlow <[hidden email]>**20060809100548
 See #704
 The MArray class doesn't currently allow a mutable array to change its
 size, because of the pure function
   bounds :: (HasBounds a, Ix i) => a i e -> (i,i)
 This patch removes the HasBounds class, and adds
   getBounds :: (MArray a e m, Ix i) => a i e -> m (i,i)
 to the MArray class, and
   bounds :: (IArray a e, Ix i) => a i e -> (i,i)
 to the IArray class.
 The reason that bounds had to be incorporated into the IArray class is
 because I couldn't make DiffArray work without doing this.  DiffArray
 acts as a layer converting an MArray into an IArray, and there was no
 way (that I could find) to define an instance of HasBounds for
[deprecate this module.
Simon Marlow <[hidden email]>**20060808100708]
[add traceShow (see #474)
Simon Marlow <[hidden email]>**20060807155545]
[remove spurious 'extern "C" {'
Simon Marlow <[hidden email]>**20060724160258]
[Fix unsafeIndex for large ranges
Simon Marlow <[hidden email]>**20060721100225]
[disambiguate uses of foldr for nhc98 to compile without errors
[hidden email]**20060711161614]
[make Control.Monad.Instances compilable by nhc98
[hidden email]**20060711160941]
Lemmih <[hidden email]>**20060708055528]
[UNDO: Merge "unrecognized long opt" fix from 6.4.2
Simon Marlow <[hidden email]>**20060705142537
 This patch undid the previous patch, "RequireOrder: do not collect
 unrecognised options after a non-opt".  I asked Sven to revert it, but
 didn't get an answer.
 See bug #473.
[Avoid strictness in accumulator for unpackFoldr
Don Stewart <[hidden email]>**20060703091806
 The seq on the accumulator for unpackFoldr will break in the presence of
 head/build rewrite rules. The empty list case will be forced, producing
 an exception. This is a known issue with seq and rewrite rules that we
 just stumbled on to.
[Disable unpack/build fusion
Don Stewart <[hidden email]>**20060702083913
 unpack/build on bytestrings seems to trigger a bug when interacting with
 head/build fusion in GHC.List. The bytestring001 testcase catches it.
 I'll investigate further, but best to disable this for now (its not
 often used anyway).
 Note that with -frules-off or ghc 6.4.2 things are fine. It seems to
 have emerged with the recent rules changes.
[Import Data.ByteString.Lazy, improve ByteString Fusion, and resync with FPS head
Don Stewart <[hidden email]>**20060701084345
 This patch imports the Data.ByteString.Lazy module, and its helpers,
 providing a ByteString implemented as a lazy list of strict cache-sized
 chunks. This type allows the usual lazy operations to be written on
 bytestrings, including lazy IO, with much improved space and time over
 the [Char] equivalents.
[Wibble in docs for new ForeignPtr functionsn
Don Stewart <[hidden email]>**20060609075924]
[comments for Applicative and Traversable
Ross Paterson <[hidden email]>**20060622170436]
[default to NoBuffering on Windows for a read/write text file
Simon Marlow <[hidden email]>**20060622144446
 Fixes (works around) #679
[remove dead code
Simon Marlow <[hidden email]>**20060622144433]
[clarify and expand docs
Simon Marlow <[hidden email]>**20060622112911]
[Add minView and maxView to Map and Set
[hidden email]**20060616180121]
[add signature for registerDelay
Ross Paterson <[hidden email]>**20060614114456]
[a few doc comments
Ross Paterson <[hidden email]>**20060613142704]
[Optimised foreign pointer representation, for heap-allocated objects
Don Stewart <[hidden email]>**20060608015011]
[Add the inline function, and many comments
[hidden email]**20060605115814
 This commit adds the 'inline' function described in the
 related patch in the compiler.
 I've also added comments about the 'lazy' function.
[small intro to exceptions
Ross Paterson <[hidden email]>**20060525111604]
[export breakpoint
Simon Marlow <[hidden email]>**20060525090456]
[Merge in changes from fps head. Highlights:
Don Stewart <[hidden email]>**20060525065012
     Wed May 24 15:49:38 EST 2006  [hidden email]
       * instance Monoid ByteString
     Wed May 24 15:04:04 EST 2006  Duncan Coutts <[hidden email]>
       * Rearange export lists for the .Char8 modules
     Wed May 24 14:59:56 EST 2006  Duncan Coutts <[hidden email]>
       * Implement mapAccumL and reimplement mapIndexed using loopU
     Wed May 24 14:47:32 EST 2006  Duncan Coutts <[hidden email]>
       * Change the implementation of the unfoldr(N) functions.
       Use a more compact implementation for unfoldrN and change it's behaviour
       to only return Just in the case that it actually 'overflowed' the N, so
       the boundary case of unfolding exactly N gives Nothing.
       Implement unfoldr and Lazy.unfoldr in terms of unfoldrN. Use fibonacci
       growth for the chunk size in unfoldr
     Wed May 24 08:32:29 EST 2006  [hidden email]
       * Add unfoldr to ByteString and .Char8
       A preliminary implementation of unfoldr.
     Wed May 24 01:39:41 EST 2006  Duncan Coutts <[hidden email]>
       * Reorder the export lists to better match the Data.List api
     Tue May 23 14:04:32 EST 2006  Don Stewart <[hidden email]>
       * pack{Byte,Char} -> singleton. As per fptools convention
     Tue May 23 14:00:51 EST 2006  Don Stewart <[hidden email]>
       * elemIndexLast -> elemIndexEnd
     Tue May 23 13:57:34 EST 2006  Don Stewart <[hidden email]>
       * In the search for a more orthogonal api, we kill breakFirst/breakLast,
         which were of dubious value
     Tue May 23 12:24:09 EST 2006  Don Stewart <[hidden email]>
       * Abolish elems. It's name implied it was unpack, but its type didn't. it made no sense
     Tue May 23 10:42:09 EST 2006  Duncan Coutts <[hidden email]>
       * Minor doc tidyup. Use haddock markup better.
     Tue May 23 11:00:31 EST 2006  Don Stewart <[hidden email]>
       * Simplify the join() implementation. Spotted by Duncan.
[add a way to ask the IO manager thread to exit
Simon Marlow <[hidden email]>**20060524121823]
[Sync with FPS head, including the following patches:
Don Stewart <[hidden email]>**20060520030436
     Thu May 18 15:45:46 EST 2006  [hidden email]
       * Export unsafeTake and unsafeDrop
     Fri May 19 11:53:08 EST 2006  Don Stewart <[hidden email]>
       * Add foldl1'
     Fri May 19 13:41:24 EST 2006  Don Stewart <[hidden email]>
       * Add fuseable scanl, scanl1 + properties
     Fri May 19 18:20:40 EST 2006  Don Stewart <[hidden email]>
       * Spotted another chance to use unsafeTake,Drop (in groupBy)
     Thu May 18 09:24:25 EST 2006  Duncan Coutts <[hidden email]>
       * More effecient findIndexOrEnd based on the impl of findIndex
     Thu May 18 09:22:49 EST 2006  Duncan Coutts <[hidden email]>
       * Eliminate special case in findIndex since it's handled anyway.
     Thu May 18 09:19:08 EST 2006  Duncan Coutts <[hidden email]>
       * Add unsafeTake and unsafeDrop
       These versions assume the n is in the bounds of the bytestring, saving
       two comparison tests. Then use them in varous places where we think this
       holds. These cases need double checking (and there are a few remaining
       internal uses of take / drop that might be possible to convert).
       Not exported for the moment.
     Tue May 16 23:15:11 EST 2006  Don Stewart <[hidden email]>
       * Handle n < 0 in drop and splitAt. Spotted by QC.
     Tue May 16 22:46:22 EST 2006  Don Stewart <[hidden email]>
       * Handle n <= 0 cases for unfoldr and replicate. Spotted by QC
     Tue May 16 21:34:11 EST 2006  Don Stewart <[hidden email]>
       * mapF -> map', filterF -> filter'
[haddock fix
Ross Paterson <[hidden email]>**20060518154723]
[simplify indexing in Data.Sequence
Ross Paterson <[hidden email]>**20060518154316]
[Move Eq, Ord, Show instances for ThreadId to GHC.Conc
Simon Marlow <[hidden email]>**20060518113339
 Eliminates orphans.
[Better error handling in the IO manager thread
Simon Marlow <[hidden email]>**20060518113303
 In particular, handle EBADF just like rts/posix/Select.c, by waking up
 all the waiting threads.  Other errors are thrown, instead of just
 being ignored.
[#define _REENTRANT 1  (needed to get the right errno on some OSs)
Simon Marlow <[hidden email]>**20060518104151
 Part 2 of the fix for threaded RTS problems on Solaris and possibly
 *BSD (Part 1 was the same change in ghc/includes/Rts.h).
[copyCString* should be in IO. Spotted by Tomasz Zielonka
Don Stewart <[hidden email]>**20060518012154]
[add import Prelude to get dependencies right for Data/Fixed.hs
Duncan Coutts <[hidden email]>**20060517222044
 Hopefully this fixes parallel builds.
[Fix negative index handling in splitAt, replicate and unfoldrN. Move mapF, filterF -> map', filter' while we're here
Don Stewart <[hidden email]>**20060517020150]
[Use our own realloc. Thus reduction functions (like filter) allocate on the Haskell heap. Makes around 10% difference.
Don Stewart <[hidden email]>**20060513051736]
[Last two CInt fixes for 64 bit, and bracket writeFile while we're here
Don Stewart <[hidden email]>**20060512050750]
[Some small optimisations, generalise the type of unfold
Don Stewart <[hidden email]>**20060510043309
     Tue May  9 22:36:29 EST 2006  Duncan Coutts <[hidden email]>
       * Surely the error function should not be inlined.
     Tue May  9 22:35:53 EST 2006  Duncan Coutts <[hidden email]>
       * Reorder memory writes for better cache locality.
     Tue May  9 23:28:09 EST 2006  Duncan Coutts <[hidden email]>
       * Generalise the type of unfoldrN
       The type of unfoldrN was overly constrained:
       unfoldrN :: Int -> (Word8 -> Maybe (Word8, Word8)) -> Word8 -> ByteString
       if we compare that to unfoldr:
       unfoldr :: (b -> Maybe (a, b)) -> b -> [a]
       So we can generalise unfoldrN to this type:
       unfoldrN :: Int -> (a -> Maybe (Word8, a)) -> a -> ByteString
       and something similar for the .Char8 version. If people really do want to
       use it a lot with Word8/Char then perhaps we should add a specialise pragma.
     Wed May 10 13:26:40 EST 2006  Don Stewart <[hidden email]>
       * Add foldl', and thus a fusion rule for length . {map,filter,fold},
       that avoids creating an array at all if the end of the pipeline is a 'length' reduction
 Place the long patch description above the ***END OF DESCRIPTION*** marker.
 The first line of this file will be the patch name.
 This patch contains the following changes:
 M ./Data/ByteString.hs -8 +38
 M ./Data/ByteString/Char8.hs -6 +12
[portable implementation of WordPtr/IntPtr for non-GHC
Ross Paterson <[hidden email]>**20060510001826
 plus much tweaking of imports to avoid cycles
[add WordPtr and IntPtr types to Foreign.Ptr, with associated conversions
Simon Marlow <[hidden email]>**20060509092606
 As suggested by John Meacham.  
 I had to move the Show instance for Ptr into GHC.ForeignPtr to avoid
 recursive dependencies.
[add CIntPtr, CUIntPtr, CIntMax, CUIntMax types
Simon Marlow <[hidden email]>**20060509092427]
[add GHC.Dynamic
Simon Marlow <[hidden email]>**20060509082739]
[Two things. #if defined(__GLASGOW_HASKELL__) on INLINE [n] pragmas (for jhc). And careful use of INLINE on words/unwords halves runtime for those functions
Don Stewart <[hidden email]>**20060509023425]
[Make length a good consumer
 Make length into a good consumer.  Fixes Trac bug #707.
 (Before length simply didn't use foldr.)
[Trim imports
[Make unsafePerformIO lazy
 The stricteness analyser used to have a HACK which ensured that NOINLNE things
 were not strictness-analysed.  The reason was unsafePerformIO. Left to itself,
 the strictness analyser would discover this strictness for unsafePerformIO:
  unsafePerformIO:  C(U(AV))
 But then consider this sub-expression
  unsafePerformIO (\s -> let r = f x in
        case writeIORef v r s of (# s1, _ #) ->
        (# s1, r #)
 The strictness analyser will now find that r is sure to be eval'd,
 and may then hoist it out.  This makes tests/lib/should_run/memo002
 Solving this by making all NOINLINE things have no strictness info is overkill.
 In particular, it's overkill for runST, which is perfectly respectable.
  f x = runST (return x)
 This should be strict in x.
 So the new plan is to define unsafePerformIO using the 'lazy' combinator:
  unsafePerformIO (IO m) = lazy (case m realWorld# of (# _, r #) -> r)
 Remember, 'lazy' is a wired-in identity-function Id, of type a->a, which is
 magically NON-STRICT, and is inlined after strictness analysis.  So
 unsafePerformIO will look non-strict, and that's what we want.
[Sync with FPS head.
Don Stewart <[hidden email]>**20060508122322
 Mon May  8 10:40:14 EST 2006  Don Stewart <[hidden email]>
   * Fix all uses for Int that should be CInt or CSize in ffi imports.
   Spotted by Igloo, dcoutts
 Mon May  8 16:09:41 EST 2006  Don Stewart <[hidden email]>
   * Import nicer loop/loop fusion rule from ghc-ndp
 Mon May  8 17:36:07 EST 2006  Don Stewart <[hidden email]>
   * Fix stack leak in split on > 60M strings
 Mon May  8 17:50:13 EST 2006  Don Stewart <[hidden email]>
   * Try same fix for stack overflow in elemIndices
[Fix all uses for Int that should be CInt or CSize in ffi imports. Spotted by Duncan and Ian
Don Stewart <[hidden email]>**20060508010311]
[Fixed import list syntax
Sven Panne <[hidden email]>**20060507155008]
[Faster filterF, filterNotByte
[hidden email]**20060507042301]
[Much faster find, findIndex. Hint from sjanssen
[hidden email]**20060507033048]
[Merge "unrecognized long opt" fix from 6.4.2
Sven Panne <[hidden email]>**20060506110519]
[hidden email]**20060506061029
 Sat May  6 13:01:34 EST 2006  Don Stewart <[hidden email]>
   * Do loopU realloc on the Haskell heap. And add a really tough stress test
 Sat May  6 12:28:58 EST 2006  Don Stewart <[hidden email]>
   * Use simple, 3x faster concat. Plus QC properties. Suggested by sjanssen and dcoutts
 Sat May  6 15:59:31 EST 2006  Don Stewart <[hidden email]>
   * dcoutt's packByte bug squashed
   With inlinePerformIO, ghc head was compiling:
    packByte 255 `compare` packByte 127
   into roughly
    case mallocByteString 2 of
        ForeignPtr f internals ->
             case writeWord8OffAddr# f 0 255 of _ ->
             case writeWord8OffAddr# f 0 127 of _ ->
             case eqAddr# f f of
                    False -> case compare (GHC.Prim.plusAddr# f 0)
                                          (GHC.Prim.plusAddr# f 0)
   which is rather stunning. unsafePerformIO seems to prevent whatever
   magic inlining was leading to this. Only affected the head.
[Add array fusion versions of map, filter and foldl
[hidden email]**20060505060858
 This patch adds fusable map, filter and foldl, using the array fusion
 code for unlifted, flat arrays, from the Data Parallel Haskell branch,
 after kind help from Roman Leshchinskiy,
 Pipelines of maps, filters and folds should now need to walk the
 bytestring once only, and intermediate bytestrings won't be constructed.
[fix for non-GHC
Ross Paterson <[hidden email]>**20060504093044]
[use bracket in appendFile (like writeFile)
Ross Paterson <[hidden email]>**20060504091528]
[writeFile: close the file on error
Simon Marlow <[hidden email]>**20060504084505
 Suggested by Ross Paterson, via Neil Mitchell
[Sync with FPS head
[hidden email]**20060503105259
 This patch brings Data.ByteString into sync with the FPS head.
 The most significant of which is the new Haskell counting sort.
 Sun Apr 30 18:16:29 EST 2006  [hidden email]
   * Fix foldr1 in Data.ByteString and Data.ByteString.Char8
 Mon May  1 11:51:16 EST 2006  Don Stewart <[hidden email]>
   * Add group and groupBy. Suggested by conversation between sjanssen and petekaz on #haskell
 Mon May  1 16:42:04 EST 2006  [hidden email]
   * Fix groupBy to match Data.List.groupBy.
 Wed May  3 15:01:07 EST 2006  [hidden email]
   * Migrate to counting sort.
   Data.ByteString.sort used C's qsort(), which is O(n log n).  The new algorithm
   is O(n), and is faster for strings larger than approximately thirty bytes.  We
   also reduce our dependency on cbits!
[improve performance of Integer->String conversion
Simon Marlow <[hidden email]>**20060503113306
 Submitted by: [hidden email]
[inline withMVar, modifyMVar, modifyMVar_
Simon Marlow <[hidden email]>**20060503111152]
[Fix string truncating in hGetLine -- it was a pasto from Simon's code
Simon Marlow <[hidden email]>**20060503103504
 (from Don Stewart)
[Merge in Data.ByteString head. Fixes ByteString+cbits in hugs
Don Stewart <[hidden email]>**20060429040733]
[Import Data.ByteString from fps 0.5.
Don Stewart <[hidden email]>**20060428130718
 Fast, packed byte vectors, providing a better PackedString.
[fix previous patch
Ross Paterson <[hidden email]>**20060501154847]
[fixes for non-GHC
Ross Paterson <[hidden email]>**20060501144322]
[fix imports for mingw32 && !GHC
Ross Paterson <[hidden email]>**20060427163248]
[RequireOrder: do not collect unrecognised options after a non-opt
Simon Marlow <[hidden email]>**20060426121110
 The documentation for RequireOrder says "no option processing after
 first non-option", so it doesn't seem right that we should process the
 rest of the arguments to collect the unrecognised ones.  Presumably
 the client wants to know about the unrecognised options up to the
 first non-option, and will be using a different option parser for the
 rest of the command line.
 eg. before:
 Prelude System.Console.GetOpt> getOpt' RequireOrder [] ["bar","--foo"]
 Prelude System.Console.GetOpt> getOpt' RequireOrder [] ["bar","--foo"]
[fix for Haddock 0.7
Ashley Yakeley <[hidden email]>**20060426072521]
[add Data.Fixed module
Ashley Yakeley <[hidden email]>**20060425071853]
[add instances
Ross Paterson <[hidden email]>**20060424102146]
[add superclasses to Applicative and Traversable
Ross Paterson <[hidden email]>**20060411144734
 Functor is now a superclass of Applicative, and Functor and Foldable
 are now superclasses of Traversable.  The new hierarchy makes clear the
 inclusions between the classes, but means more work in defining instances.
 Default definitions are provided to help.
[add Functor and Monad instances for Prelude types
Ross Paterson <[hidden email]>**20060410111443]
Lemmih <[hidden email]>**20060407125827]
[Track the GHC source tree reorganisation
Simon Marlow <[hidden email]>**20060407041631]
[in the show instance for Exception, print the type of dynamic exceptions
Simon Marlow <[hidden email]>**20060406112444
 Unfortunately this requires some recursve module hackery to get at
 the show instance for Typeable.
[implement ForeignEnvPtr, newForeignPtrEnv, addForeignPtrEnv for GHC
Simon Marlow <[hidden email]>**20060405155448]
[add  forkOnIO :: Int -> IO () -> IO ThreadId
Simon Marlow <[hidden email]>**20060327135018]
[Rework previous: not a gcc bug after all
Simon Marlow <[hidden email]>**20060323161229
 It turns out that we were relying on behaviour that is undefined in C,
 and undefined behaviour in C means "the compiler can do whatever the
 hell it likes with your entire program".  So avoid that.
[work around a gcc 4.1.0 codegen bug in -O2 by forcing -O1 for GHC.Show
Simon Marlow <[hidden email]>**20060323134514
 See http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26824
[commit mysteriously missing parts of "runIOFastExit" patch
Simon Marlow <[hidden email]>**20060321101535]
[add runIOFastExit :: IO a -> IO a
Simon Marlow <[hidden email]>**20060320124333
 Similar to runIO, but calls stg_exit() directly to exit, rather than
 shutdownHaskellAndExit().  Needed for running GHCi in the test suite.
[Fix a broken invariant
Simon Marlow <[hidden email]>**20060316134151
 Patch from #694,  for the problem "empty is an identity for <> and $$" is
 currently broken by eg. isEmpty (empty<>empty)"
[Add unsafeSTToIO :: ST s a -> IO a
Simon Marlow <[hidden email]>**20060315160232
 Implementation for Hugs is missing, but should be easy.  We need this
 for the forthcoming nested data parallelism implementation.
[Added 'alter'
[hidden email]**20060315143539
 Added 'alter :: (Maybe a -> Maybe a) -> k -> Map k a -> Map k a' to IntMap and Map
 This addresses ticket #665
[deprecate FunctorM in favour of Foldable and Traversable
Ross Paterson <[hidden email]>**20060315092942
 as discussed on the libraries list.
[Simplify Eq, Ord, and Show instances for UArray
Simon Marlow <[hidden email]>**20060313142701
 The Eq, Ord, and Show instances of UArray were written out longhand
 with one instance per element type.  It is possible to condense these
 into a single instance for each class, at the expense of using more
 extensions (non-std context on instance declaration).
 Suggestion by: Frederik Eaton <[hidden email]>
[Oops typo in intSet notMember
[hidden email]**20060311224713]
[IntMap lookup now returns monad instead of Maybe.
[hidden email]**20060311224502]
[Added notMember to Data.IntSet and Data.IntMap
[hidden email]**20060311085221]
[add Data.Set.notMember and Data.Map.notMember
John Meacham <[hidden email]>**20060309191806]
[addToClockTime: handle picoseconds properly
Simon Marlow <[hidden email]>**20060310114532
 fixes #588
[make head/build rule apply to all types, not just Bool.
John Meacham <[hidden email]>**20060303045753]
[Avoid overflow when normalising clock times
Ian Lynagh <[hidden email]>**20060210144638]
[Years have 365 days, not 30*365
Ian Lynagh <[hidden email]>**20060210142853]
[declare blkcmp() static
Simon Marlow <[hidden email]>**20060223134317]
[typo in comment in Foldable class
Ross Paterson <[hidden email]>**20060209004901]
[simplify fmap
Ross Paterson <[hidden email]>**20060206095048]
[update ref in comment
Ross Paterson <[hidden email]>**20060206095139]
[Give -foverlapping-instances to Data.Typeable
 For some time, GHC has made -fallow-overlapping-instances "sticky":
 any instance in a module compiled with -fallow-overlapping-instances
 can overlap when imported, regardless of whether the importing module
 allows overlap.  (If there is an overlap, both instances must come from
 modules thus compiled.)
 Instances in Data.Typeable might well want to be overlapped, so this
 commit adds the flag to Data.Typeable (with an explanatory comment)
[Add -fno-bang-patterns to modules using both bang and glasgow-exts
[hidden email]**20060203175759]
[When splitting a bucket, keep the contents in the same order
Simon Marlow <[hidden email]>**20060201130427
 To retain the property that multiple inserts shadow each other
 (see ticket #661, test hash001)
[add foldr/build optimisation for take and replicate
Simon Marlow <[hidden email]>**20060126164603
 This allows take to be deforested, and improves performance of
 replicate and replicateM/replicateM_.  We have a separate problem that
 means expressions involving [n..m] aren't being completely optimised
 because eftIntFB isn't being inlined but otherwise the results look
 Sadly this has invalidated a number of the nofib benchmarks which were
 erroneously using take to duplicate work in a misguided attempt to
 lengthen their runtimes (ToDo).
[Generate PrimopWrappers.hs with Haddock docs
Simon Marlow <[hidden email]>**20060124131121
 Patch originally from Dinko Tenev <[hidden email]>, modified
 to add log message by me.
[[project @ 2006-01-19 14:47:15 by ross]
 backport warning avoidance from Haddock
[[project @ 2006-01-18 11:45:47 by malcolm]
 Fix import of Ix for nhc98.
[[project @ 2006-01-17 09:38:38 by ross]
 add Ix instance for GeneralCategory.
[TAG Initial conversion from CVS complete
John Goerzen <[hidden email]>**20060112154126]
Patch bundle hash:

Libraries mailing list
[hidden email]