- ghc-devs - Haskell.org

vectorisation code?
by Richard Eisenberg 17 Jan '15

17 Jan '15

Hi devs, There's a sizable number of modules in the `vectorise` subdirectory of GHC. I'm sure these do all sorts of wonderful things. But what, exactly? And, does anyone make use of these wonderful things? A quick poking through the code shows a tiny link between the vectorise code and the rest of GHC -- the function `vectorise` exported from the module `Vectorise`, which is named in exactly one place from SimplCore. From what I can tell, the function will be called only when `-fvectorise` is specified, and then it seems to interact with a {-# VECTORISE #-} pragma. However, `{-# VECTORISE #-}` doesn't appear in the manual at all, and `-fvectorise` is given only a cursory explanation. It seems these work with DPH... which has been disabled, no? Searching online finds several hits, but nothing more recent than 2012. I hope this question doesn't offend -- it seems that vectorisation probably has amazing performance gains. Yet, the feature also seems unloved. In the meantime, compiling (and recompiling, and recompiling...) the modules takes time, as does going through them to propagate changes from elsewhere. If this feature is truly orphaned, unloved, and unused at the moment, is it reasonable to consider putting it on furlough? Thanks, Richard

5 6

Floating through small case analyses
by Ben Gamari 15 Jan '15

15 Jan '15

Hello Simon, Today I've been working on trying to whip bytestring's Builder into shape [1,2]. The last remaining issue is a performance problem that appears to be due to over-zealous floating of some small literals in every GHC version I've tested (7.6, 7.8, and 7.10). The test case can be found here [3] and is characterized by repeated monoidal appends (reproduced roughly here in half-Core/half-Haskell), loop s n | s `seq` n `seq` False = undefined loop _ 0 = mempty loop s n = singleton (s + (fromInteger @Word8 (__integer 0))) <> singleton (s + (fromInteger @Word8 (__integer 1))) <> singleton (s + (fromInteger @Word8 (__integer 2))) <> singleton (s + (fromInteger @Word8 (__integer 3))) <> ... singleton (s + (fromInteger @Word8 (__integer 15))) <> loop (s+16) (s-16) The initial float-out stage floats all of the (fromInteger ...) expressions out to the top level, resulting in definitions of the form, lvl_s1b3j :: Integer lvl_s1b3j = __integer 0 lvl_s1b3k :: Word8 lvl_s1b3k = $fBitsWord8_$cfromInteger lvl_s1b3j Simplifier phase 2 then does some inlining and in so doing introduces a boolean case analysis. Both alternatives of the case require the now-top-level literals. Unfortunately, the float-in pass is quite strict about when it will float expressions inside case analyses. Specifically, it requires the float to be both small [4] (which I believe these are) and used in more than one but not all alternatives [5]. Given that there are only two alternatives here this seems a bit too restrictive. The end result is that we end up producing a bunch of bindings outside of a boolean case analysis, case narrow8Word# sc_s1c39 of a_s1c3g { __DEFAULT -> case plusWord# sc_s1c39 (__word 1) of sat_s1c3i { __DEFAULT -> case narrow8Word# sat_s1c3i of a1_s1c3h { __DEFAULT -> case plusWord# sc_s1c39 (__word 2) of sat_s1c3k { __DEFAULT -> ... case plusWord# sc_s1c39 (__word 16) of sat_s1c3M { __DEFAULT -> case narrow8Word# sat_s1c3M of a16_s1c3L { __DEFAULT -> case tagToEnum# @ Bool sat_s1c3P of _ { False -> {-# write each of the values above to memory #-} True -> {-# write each of the values above to memory #-} } This ends up producing extremely poor assembly, with the compiler first computing all 16 store addresses, placing them on the stack, and then performing all 16 stores. I'm a bit unsure of what to blame this on. Perhaps it would make sense to always float small dupable expressions in to sufficiently small case expressions (say, three or fewer alternatives)? Perhaps these near-literals should never have been floated out at all? Perhaps there another option I've not thought of? Thoughts? Cheers, - Ben [1]: https://github.com/kolmodin/binary/pull/65 [2]: https://github.com/haskell/bytestring/pull/40 [3]: https://gist.github.com/bgamari/22d127779a48113c1153#file-test-hs [4]: https://github.com/ghc/ghc/blob/master/compiler/coreSyn/CoreUtils.hs#L710 [5]: https://github.com/ghc/ghc/blob/master/compiler/simplCore/FloatIn.hs#L527

1 0

API Annotations question
by Jan Stolarek 15 Jan '15

15 Jan '15

Alan, devs, since API annotations were added some data types in the source code have extra comments related to these annotations, eg. -- | A type or class declaration. data TyClDecl name = -- | @type/data family T :: *->*@ -- -- - 'ApiAnnotation.AnnKeywordId' : 'ApiAnnotation.AnnType', -- 'ApiAnnotation.AnnData', -- 'ApiAnnotation.AnnFamily','ApiAnnotation.AnnWhere', -- 'ApiAnnotation.AnnOpen','ApiAnnotation.AnnDcolon', -- 'ApiAnnotation.AnnClose' FamDecl { tcdFam :: FamilyDecl name } I'm totally clueless about what these comments mean. Where can I find an explanation? Couldn't find anything on the wiki. Janek

2 4

RE: [Diffusion] [Build Failed] rGHCe4cb8370eb91: Tiny refactoring (shorter, simpler code)
by Simon Peyton Jones 14 Jan '15

14 Jan '15

This failure message leads me to https://phabricator.haskell.org/harbormaster/build/2964/ which says that ghcirun003 is failing. But how can I see the failure log???? (It doesn't fail for me.) Simon | -----Original Message----- | From: noreply(a)phabricator.haskell.org | [mailto:noreply@phabricator.haskell.org] | Sent: 14 January 2015 17:37 | To: Simon Peyton Jones | Subject: [Diffusion] [Build Failed] rGHCe4cb8370eb91: Tiny refactoring | (shorter, simpler code) | | Harbormaster failed to build B2940: rGHCe4cb8370eb91: Tiny refactoring | (shorter, simpler code)! | | USERS | simonpj (Author) | GHC - Type checker/inferencer (Auditor) | | COMMIT | https://phabricator.haskell.org/rGHCe4cb8370eb91 | | EMAIL PREFERENCES | https://phabricator.haskell.org/settings/panel/emailpreferences/ | | To: simonpj, GHC - Type checker/inferencer

1 0

Ord instance for Data.Data.Constr
by Alan & Kim Zimmerman 13 Jan '15

13 Jan '15

Is there a reason there is no Ord instance for Data.Data.Constr? I want to use it as part of a key in a Map, and can't auto derive the instance as all the constructors are not in scope. Alan

1 1

Difference in Partial TypeFamily application between 7.8.3 & 7.8.4
by magesh b 13 Jan '15

13 Jan '15

Hi, {-# LANGUAGE TypeFamilies, ConstraintKinds #-} import GHC.Exts type family TyFun a data DictC (c :: * -> Constraint) data DictTF (tf :: * -> *) type Test1 = DictTF TyFun -- Fails here in 7.8.4 type Test2 = DictC Show When I compile the above code, I'm getting the following error in 7.8.4 and the same code works in 7.8.3. Test.hs:11:1: Type synonym ‘TyFun’ should have 1 argument, but has been given none In the type declaration for ‘Test1’ Is this a bug or a expected behavior? For reference, I could find two fixes to type family related bug in this release. https://ghc.haskell.org/trac/ghc/ticket/9433 https://ghc.haskell.org/trac/ghc/ticket/9316 Thanks & Regards, Magesh B

2 3

New performance dashboard front-end
by Joachim Breitner 13 Jan '15

13 Jan '15

Hi, over the holidays I’ve been working on a new and custom-made¹ front-end for our performance data, mainly to work around limitations of codespeed when it comes to understanding git, but also to add other features that I happen to want. I put a (not automatically updated) preview on http://deb.haskell.org/speed/ The code is not yet online, and there are some features missing (most notably: Working with multiple builders, the per-benchmark-graphs.) I’d like to have this hosted somewhere properly, so this is a request for the infrastructure team. I am a big fan of static content, so the system is a (shake-driven) batch process that generates a bunch of .json file. These can be served statically, together with the completely static index.html and a few javascript libraries. So it’s all very CDN-friendly. So all I need is * shell access to some machine, preferably with ghc installed * some disk space (actually quite a bit due to all the build logs, although that could be reduced by gzipping or reading them directly from the git repo²) * a new virtual host or a subdirectory of http://ghc.haskell.org/ where I can deploy my files to. * the possibility to either run a cronjob to poll for new logs, or maybe (later) some more sophisticated trigger. Would that be possible? Greetings, Joachim ¹ It is still a generic display of "values per git commit", and I hope I can keep it that way – maybe other projects can use it as well. ² currently at https://github.com/nomeata/ghc-speed-logs If the above becomes official, this probably also should move to git.haskell.org. The repo is 250M, but 7,2G checked out. I plan to make my code read the logs directly from the repo, and link to the cgit web interface to show the logs, so that it never has to be checked out. -- Joachim “nomeata” Breitner mail(a)joachim-breitner.de • http://www.joachim-breitner.de/ Jabber: nomeata(a)joachim-breitner.de • GPG-Key: 0xF0FBF51F Debian Developer: nomeata(a)debian.org

3 3

ANNOUNCE: GHC version 7.8.4
by Austin Seipp 13 Jan '15

13 Jan '15

============================================================== The (Interactive) Glasgow Haskell Compiler -- version 7.8.4 ============================================================== The GHC Team is pleased to announce a new patchlevel release of GHC, 7.8.4. This is an important bugfix release relative to 7.8.3 (with over 30 defects fixed), so we highly recommend upgrading from the previous 7.8 releases. The full release notes are here: https://downloads.haskell.org/~ghc/7.8.4/docs/html/users_guide/release-7-8-… How to get it ~~~~~~~~~~~~~ The easy way is to go to the web page, which should be self-explanatory: https://www.haskell.org/ghc/ We supply binary builds in the native package format for many platforms, and the source distribution is available from the same place. Packages will appear as they are built - if the package for your system isn't available yet, please try again later. Background ~~~~~~~~~~ Haskell is a standard lazy functional programming language. GHC is a state-of-the-art programming suite for Haskell. Included is an optimising compiler generating good code for a variety of platforms, together with an interactive system for convenient, quick development. The distribution includes space and time profiling facilities, a large collection of libraries, and support for various language extensions, including concurrency, exceptions, and foreign language interfaces (C, whatever). GHC is distributed under a BSD-style open source license. A wide variety of Haskell related resources (tutorials, libraries, specifications, documentation, compilers, interpreters, references, contact information, links to research groups) are available from the Haskell home page (see below). On-line GHC-related resources ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Relevant URLs on the World-Wide Web: GHC home page http://www.haskell.org/ghc/ GHC developers' home page http://ghc.haskell.org/trac/ghc/ Haskell home page http://www.haskell.org/ Supported Platforms ~~~~~~~~~~~~~~~~~~~ The list of platforms we support, and the people responsible for them, is here: http://ghc.haskell.org/trac/ghc/wiki/Platforms http://ghc.haskell.org/trac/ghc/wiki/CodeOwners Ports to other platforms are possible with varying degrees of difficulty. The Building Guide describes how to go about porting to a new platform: http://ghc.haskell.org/trac/ghc/wiki/Building Developers ~~~~~~~~~~ We welcome new contributors. Instructions on accessing our source code repository, and getting started with hacking on GHC, are available from the GHC's developer's site run by Trac: http://ghc.haskell.org/trac/ghc/ Mailing lists ~~~~~~~~~~~~~ We run mailing lists for GHC users and bug reports; to subscribe, use the web interfaces at http://www.haskell.org/mailman/listinfo/glasgow-haskell-users http://www.haskell.org/mailman/listinfo/glasgow-haskell-bugs There are several other haskell and ghc-related mailing lists on www.haskell.org; for the full list, see http://www.haskell.org/mailman/listinfo/ Some GHC developers hang out on #haskell on IRC, too: http://www.haskell.org/haskellwiki/IRC_channel Please report bugs using our bug tracking system. Instructions on reporting bugs can be found here: http://www.haskell.org/ghc/reportabug Hashes & Signatures ~~~~~~~~~~~~~~~~~ On https://downloads.haskell.org/~ghc/7.8.4/ you will find a signed copy of the SHA256 hashes for the tarballs, using my GPG key (keyid 0x3B58D86F). -- Regards, Austin Seipp, Haskell Consultant Well-Typed LLP, http://www.well-typed.com/

4 8

Build failure under Fedora 21
by Dominick Samperi 12 Jan '15

12 Jan '15

Hello, I'm trying to build ghc-7.8 branch under Fedora 21 and I get the failure diagnostics appended below. The problem seems to be that the build process cannot satisfy deepseq >=1.2 && <1.4, but I explicitly installed deepseq-1.3.0.2, and this did not help (same error?). Under Fedora 21 Haskell Platform contains ghc-7.6.3, so this is the boot compiler. The ghc binary for centos65 installs provided I define /usr/lib64/libgmp.so.3 as a symbolic link to /usr/lib64/libgmp.so.10, a risky move. Any tips would be much appreciated. Thanks, Dominick ===--- building phase 0 make -r --no-print-directory -f ghc.mk phase=0 phase_0_builds make[1]: Nothing to be done for 'phase_0_builds'. ===--- building phase 1 make -r --no-print-directory -f ghc.mk phase=1 phase_1_builds "inplace/bin/ghc-cabal" check libraries/containers 'ghc-options: -O2' is rarely needed. Check that it is giving a real benefit and not just imposing longer compile times on your users. "inplace/bin/ghc-cabal" configure libraries/containers dist-install "" --with-ghc="/home/dsamperi/install/git/ghc/inplace/bin/ghc-stage1" --with-ghc-pkg="/home/dsamperi/install/git/ghc/inplace/bin/ghc-pkg" --disable-library-for-ghci --enable-library-vanilla --disable-library-profiling --enable-shared --configure-option=CFLAGS=" -fno-stack-protector " --configure-option=LDFLAGS=" " --configure-option=CPPFLAGS=" " --gcc-options=" -fno-stack-protector " --with-gcc="/usr/bin/gcc" --with-ld="/usr/bin/ld" --configure-option=--with-cc="/usr/bin/gcc" --with-ar="/usr/bin/ar" --with-ranlib="/usr/bin/ranlib" --with-alex="/home/dsamperi/.cabal/bin/alex" --with-happy="/home/dsamperi/.cabal/bin/happy" Configuring containers-0.5.5.1... ghc-cabal: At least the following dependencies are missing: deepseq >=1.2 && <1.4 libraries/containers/ghc.mk:4: recipe for target 'libraries/containers/dist-install/package-data.mk' failed make[1]: *** [libraries/containers/dist-install/package-data.mk] Error 1 Makefile:71: recipe for target 'all' failed make: *** [all] Error 2

3 4

Clarification of HsBang and isBanged
by Johan Tibell 11 Jan '15

11 Jan '15

HsBang is defined as: -- HsBang describes what the *programmer* wrote -- This info is retained in the DataCon.dcStrictMarks field data HsBang = HsUserBang -- The user's source-code request (Maybe Bool) -- Just True {-# UNPACK #-} -- Just False {-# NOUNPACK #-} -- Nothing no pragma Bool -- True <=> '!' specified | HsNoBang -- Lazy field -- HsUserBang Nothing False means the same as HsNoBang | HsUnpack -- Definite commitment: this field is strict and unboxed (Maybe Coercion) -- co :: arg-ty ~ product-ty | HsStrict -- Definite commitment: this field is strict but not unboxed This data type is a bit unclear to me: * What are the reasons for the following constructor overlaps? * `HsNoBang` and `HsUserBang Nothing False` * `HsStrict` and `HsUserBang Nothing True` * `HsUnpack mb_co` and `HsUserBang (Just True) True` * Why is there a coercion in `HsUnpack` but not in `HsUserBang (Just True) True`? * Is there a difference in what the user wrote in the case of HsUserBang and HsNoBang/HsUnpack/HsStrict e.g are the latter three generated by the compiler as opposed to being written by the user (the function documentation notwithstanding)? A very related function is isBanged: isBanged :: HsBang -> Bool isBanged HsNoBang = False isBanged (HsUserBang Nothing bang) = bang isBanged _ = True What's the meaning of this function? Is it intended to communicate what the user wrote or whether result of what the user wrote results in a strict function? Context: I'm adding a new StrictData language pragma [1] that makes fields strict by default and a '~' annotation of fields to reverse the default behavior. My intention is to change HsBang like so: - Bool -- True <=> '!' specified + (Maybe Bool) -- True <=> '!' specified, False <=> '~' + -- specified, Nothing <=> unspecified 1. https://ghc.haskell.org/trac/ghc/wiki/StrictPragma -- Johan

3 11