- Haskell-Cafe - Haskell.org

Fwd: signficant improvements to the containers package
by Don Stewart 25 Jun '10

25 Jun '10

Some people might be quite excited by Milan's work on significant performance improvements to the containers package... ----- Forwarded message from Milan Straka <fox(a)ucw.cz> ----- Date: Thu, 24 Jun 2010 07:45:50 +0200 From: Milan Straka <fox(a)ucw.cz> To: Don Stewart <dons(a)galois.com> Cc: Claus Reinke <claus.reinke(a)talk21.com>, libraries(a)haskell.org Subject: Re: Map folds in containers: why no worker/wrapper split? > > Data.Map is the way it is for historical reasons. It needs a dedicated > performance maintainer to do detailed benchmarking and static analysis. I did quite a lot of benchmarking, strictness analysis and performance improvements to the containers package recently, see the draft of my paper http://fox.ucw.cz/papers/containers/containers.pdf. Next week I will start finalizing the patches and submit them. The 'not generating a worker/wrapper' issue is one of several on my radar, so this should be resolved. Cheers, Milan ----- End forwarded message -----

7 12

Read instance for GATD
by corentin.dupont＠ext.mpsa.com 25 Jun '10

25 Jun '10

Hello Haskellers, I'm having trouble writing a Read Instance for my GATD. Arg this GATD!! It causes me more problems that it solves ;) Especially with no automatic deriving, it adds a lot of burden to my code. >data Obs a where > ProposedBy :: Obs Int -- The player that proposed the tested rule > Turn :: Obs Turn -- The current turn > Official :: Obs Bool -- whereas the tested rule is official > Equ :: (Eq a, Show a, Typeable a) => Obs a -> Obs a -> Obs Bool > Plus :: (Num a) => Obs a -> Obs a -> Obs a > Time :: (Num a) => Obs a -> Obs a -> Obs a > Minus :: (Num a) => Obs a -> Obs a -> Obs a > And :: Obs Bool -> Obs Bool -> Obs Bool > Or :: Obs Bool -> Obs Bool -> Obs Bool > Not :: Obs Bool -> Obs Bool > Konst :: a -> Obs a > instance Read a => Read (Obs a) where > readPrec = (prec 10 $ do > Ident "ProposedBy" <- lexP > return (ProposedBy)) > +++ > (prec 10 $ do > Ident "Official" <- lexP > return (Official)) > (etc...) Observable.lhs:120:8: Couldn't match expected type `Int' against inferred type `Bool' Expected type: ReadPrec (Obs Int) Inferred type: ReadPrec (Obs Bool) Indeed "ProposedBy" does not have the same type that "Official". Mmh how to make it all gently mix altogether? Best, Corentin

3 2

ANN: ForSyDe DSL v3.1
by Seyed Hosein Attarzadeh Niaki 25 Jun '10

25 Jun '10

Dear All, A new version of the ForSyDe DSL is uploaded to the Hackage database. "The ForSyDe (Formal System Design) methodology has been developed with the objective to move system design to a higher level of abstraction and to bridge the abstraction gap by transformational design refinement. This library provides ForSyDe's implementation as a Haskell-embedded Domain Specific Language (DSL). For more information, please see ForSyDe's website: http://www.ict.kth.se/forsyde/." In addition to compatibility with base 4 and GHC 6.10.x, the new version provides more freedom for declaration of process functions (e.g., support for when and let blocks). BR/ Hosein Attarzadeh

1 1

ANN: parameterized-data library v0.1.4
by Seyed Hosein Attarzadeh Niaki 25 Jun '10

25 Jun '10

Dear Haskell developers, A new version of the parameterized-data library providing fixed-sized vectors (based on the type-level library) is pushed into the repository and uploaded to the Hackage database. This is only a minor update to fix the compatibility issues with base 4, GHC>=6.10 and cabal-install (not backward compatible). Bests, Hosein Attarzadeh

1 1

RE: [Haskell-cafe] proposal: HaBench, a Haskell Benchmark Suite
by Andy Georges 25 Jun '10

25 Jun '10

Hi Simon et al, I've picked up the HaBench/nofib/nobench issue again, needing a decent set of real applications to do some exploring of what people these days call split-compilation. We have a framework that was able to explore GCC optimisations [1] -- the downside there was the dependency of these optimisations on each other, requiring them to be done in certain order -- for a multi-objective search space, and extended this to exploring a JIT compiler [2] for Java in our case -- which posed its own problems. Going one step further, we'd like to explore the tradeoffs that can be made when compiling on different levels: source to bytecode (in some sense) and bytecode to native. Given that LLVM is quicly becoming a state-of-the-art framework and with the recent GHC support, we figured that Haskell would be an excellent vehicle to conduct our exploration and research (and the fact that some people at our lab have a soft spot for Haskell helps too). Which brings me back to benchmarks. Are there any inputs available that allow the real part of the suite to run for a sufficiently long time? We're going to use criterion in any case given our own expertise with rigorous benchmarking [3,4], but since we've made a case in the past against short running apps on managed runtime systems [5], we'd love to have stuff that runs at least in the order of seconds, while doing useful things. All pointers are much appreciated. Or if any of you out there have (recent) apps with inputs that are open source ... let us know. -- Andy [1] COLE: Compiler Optimization Level Exploration, Kenneth Hoste and Lieven Eeckhout, CGO 2008 [2] Automated Just-In-Time Compiler Tuning, Kenneth Hoste, Andy Georges and Lieven Eeckhout, CGO 2010 [3] Statistically Rigorous Java Performance Evaluation, Andy Georges, Dries Buytaert and Lieven Eeckhout, OOPSLA 2007 [4] Java Performance Evaluation through Rigorous Replay Compilation, Andy Georges, Lieven Eeckhout and Dries Buytaert, OOPSLA 2008 [5] How Java Programs Interact with Virtual Machines at the Microarchitectural Level, Lieven Eeckhout, Andy Georges, Koen De Bosschere, OOPSLA 2003

5 6

How does one get off haskell?
by Günther Schmidt 25 Jun '10

25 Jun '10

Hi list, I'm facing a really tough problem. About 3 years ago I stopped doing freelance and quite nicely paid projects in Java, PHP and C#. Now I'm dire straits, again, and need to get back into the project market which seems to have picked up again, quite a lot of projects out there and it looks like I could ask again for decent rates. (I personally call them compensation because I never ever enjoyed doing Java etc. but the money was good.) Anyway the problem is that I am totally reluctant to code in anything else but haskell. It has always been a problem to me getting up early in the morning, taking a train to work and coming back in the evening totally exhausted. But I think I could manage that again, at least for 3 or 6 months and then my bank account will be fine again and I can take it easy for another year or so. But this time all this is much harder. I really cannot see myself writing such huge amounts of code over and over again not doing much, well you know the story. BTW this is not meant as a fun post, I'm actually quite serious, ie. I need money, only way of getting it is doing Java, C# or PHP. So how does one get off haskell? Are there people in similar situations that have managed? How did you do it? Günther

27 35

AusHac2010, the first Australian Haskell Hackathon is growing near, and we need you!
by Alex Mason 25 Jun '10

25 Jun '10

AusHac2010 will be held from the 16th to the 18th of July at UNSW, where we'll be working on various haskell projects such as: The LLVM backend for GHC (I'm going to see if we can somehow get in contact with David T during the hackathon so we can discuss what can be done) A Generic graph class The Accelerate and Repa libraries for fast array computations Hubris, the Ruby-Haskell bridge (Mark's done some really awesome work here, but I'm sure could use some help) Leksah, the Haskell IDE written in Haskell MPI bindings, to either bring the previous package alive from its bitrotten state, or create a new one from scratch A Notification library for things like Growl and libnotify A Library re-builder, extending haskell-updater for use with cabal-install, Arch, etc. If you can come for any of the days we'll be hacking, we'd love to see you! But you'll need to sign up first (so you can get access to the UNSW wireless network). If you're not sure if you can make it, sign up anyway, you're not obliged to come, but mark yourself as only possibly being able to come. The signup page can be found at: Sign up!: [http://axman6.wufoo.com/forms/aushac-2010-sign-up-20/] See the AusHac2010 wiki page [http://www.haskell.org/haskellwiki/AusHac2010] for more details on what we'll be doing, and various other details. We've already got 20-30 people coming hopefully, but the more we have the better. If you've already signed up using the previous form, you we still need you to fill out this new 2.0 version of the sign up form If you have questions, please feel free to email me or Ivan (our addresses are the reply-to addresses for this email... I hope), or talk to us (Axman6, ivanm) on #haskell. Cheers, Alex Mason

1 0

Re: [Haskell-cafe] Question regarding deepseq (Control.DeepSeq)
by Frank Moore 25 Jun '10

25 Jun '10

Daniel, >This means *when meanOver2 is evaluated*, then evaluate (mean as). >Binding it in a let is lazy, so it won't be evaluated until it's needed >(for printing in this case). >Also note that (mean as) is a Double, so deepseq is just seq in this case >(but I suppose this is just a boiled down example and you also want to time >computations with results where deepseq does strictly more than seq). Thanks for the help, I think I understand now. This is indeed a boiled down example. I have some data that I am trying to compute in parallel, and I want to control how much evaluation to force. To do this, I was trying to use NFData and deepseq, since I thought bang patterns were a little clunky (but maybe they are less clunky than forcing evaluations with seq and deepseqs). Are there best practices for keeping track of which variables are bound where, when writing a complex program? >There are two standard ways to achieve what you want, >1. > let meanOver2 = ... > end <- meanOver2 `deepseq` getCurrentTime >2. put {-# LANGUAGE BangPatterns #-} at the top of the file and write > let !meanOver2 = ... > end <- getCurrentTime >The bang on meanOver2 means "evaluate this expression now (to weak head >normal form, i.e. to the outermost constructor)". Sometimes for me, whnf is not quite enough. For example, if I have a newtype P (with type constructor P) that is a for Map M Int. Is NFData and deepseq the way to do this, or should I be trying to implement my own Strategy along the lines of rwhnf in Control.Parallel.Strategies? >Another thing, for timing computations, wall-clock time is not appropriate, >better use >System.CPUTime.getCPUTime >to get only the CPU-time the process took, and not also what your browser >or whatever used in the meantime. Thanks for the tip. Frank > On Thu, Jun 24, 2010 at 9:38 PM, Daniel Fischer <daniel.is.fischer(a)web.de> wrote: >> >> On Friday 25 June 2010 02:57:31, Frank Moore wrote: >> > Hello Haskellers, >> > >> > I am new to programming in Haskell and I am having trouble understanding >> > exactly when statements become evaluated. My goal is to try and measure >> > how long a computation takes without having to use a show function. The >> > code I am trying to use is below (taken in part from RWH chapter 25) >> > >> > ---------------------------------- >> > import Data.List (foldl') >> > import Data.Time.Clock (diffUTCTime, getCurrentTime) >> > import Control.DeepSeq (deepseq) >> > >> > mean :: [Double] -> Double >> > mean xs = s / fromIntegral n where >> > (n,s) = foldl' k (0,0) xs >> > k (n,s) x = n `seq` s `seq` (n+1,s+x) >> > >> > main = do >> > let as = [1..1e7] :: [Double] >> > start <- getCurrentTime >> > let meanOver2 = deepseq (mean as) `seq` (mean as) / fromIntegral 2 >> >> This means *when meanOver2 is evaluated*, then evaluate (mean as). >> Binding it in a let is lazy, so it won't be evaluated until it's needed >> (for printing in this case). >> Also note that (mean as) is a Double, so deepseq is just seq in this case >> (but I suppose this is just a boiled down example and you also want to time >> computations with results where deepseq does strictly more than seq). >> >> There are two standard ways to achieve what you want, >> >> 1. >> let meanOver2 = ... >> end <- meanOver2 `deepseq` getCurrentTime >> >> 2. put {-# LANGUAGE BangPatterns #-} at the top of the file and write >> >> let !meanOver2 = ... >> end <- getCurrentTime >> >> The bang on meanOver2 means "evaluate this expression now (to weak head >> normal form, i.e. to the outermost constructor)". >> >> > end <- getCurrentTime >> > putStrLn (show (end `diffUTCTime` start)) >> > putStrLn (show meanOver2) >> > ------------------------------------- >> >> Another thing, for timing computations, wall-clock time is not appropriate, >> better use >> >> System.CPUTime.getCPUTime >> >> to get only the CPU-time the process took, and not also what your browser >> or whatever used in the meantime. >> >> > >> > My understanding of deepseq was that it evaluates (mean as) completely >> > before continuing, and then the show would not take any time, but >> >> No, it evaluates (mean as) completely *when meanOver2 is demanded*, not >> before. >> >> > instead all the time is spent in the show meanOver2 function. I feel >> > like I am missing something fundamental here. Any suggestions? Thanks >> > for your help. >> > >> > Frank >> > > > > -- > Dr. W. Frank Moore > H. C. Wang Assistant Professor > Department of Mathematics, Cornell University > 310 Malott Hall, Ithaca NY 14853-4201, USA > > Office: Malott 587, Phone: +1 607 255 4030 > Email: frankmoore(a)math.cornell.edu -- Dr. W. Frank Moore H. C. Wang Assistant Professor Department of Mathematics, Cornell University 310 Malott Hall, Ithaca NY 14853-4201, USA Office: Malott 587, Phone: +1 607 255 4030 Email: frankmoore(a)math.cornell.edu

1 0

Question regarding deepseq (Control.DeepSeq)
by Frank Moore 25 Jun '10

25 Jun '10

Hello Haskellers, I am new to programming in Haskell and I am having trouble understanding exactly when statements become evaluated. My goal is to try and measure how long a computation takes without having to use a show function. The code I am trying to use is below (taken in part from RWH chapter 25) ---------------------------------- import Data.List (foldl') import Data.Time.Clock (diffUTCTime, getCurrentTime) import Control.DeepSeq (deepseq) mean :: [Double] -> Double mean xs = s / fromIntegral n where (n,s) = foldl' k (0,0) xs k (n,s) x = n `seq` s `seq` (n+1,s+x) main = do let as = [1..1e7] :: [Double] start <- getCurrentTime let meanOver2 = deepseq (mean as) `seq` (mean as) / fromIntegral 2 end <- getCurrentTime putStrLn (show (end `diffUTCTime` start)) putStrLn (show meanOver2) ------------------------------------- My understanding of deepseq was that it evaluates (mean as) completely before continuing, and then the show would not take any time, but instead all the time is spent in the show meanOver2 function. I feel like I am missing something fundamental here. Any suggestions? Thanks for your help. Frank

4 4

Tiger compiler: variable escaping analysis phase
by José Romildo Malaquias 24 Jun '10

24 Jun '10

Hello. I have been teaching an introductory course on compiler construction to our undergraduates students using Appel's "Modern Compiler Implementation in Java". There are also versions of the book in ML and C. The books explain how to write a compiler for the Tiger programming language. Now I want to implement a Tiger compiler in Haskell. The lexer and parser (built with the help of alex and happy) and the type checker are already working. Next step is implementing a variable escape analysis phase, needed for the intermediate code generator. The objective of this phase is to find out if each variable escapes or not. In general an escaping variable should be allocated in the stack frame (in memory). Non escaping variables may be allocated in the stack frame or in a register. Generally, a variable escapes if it is passed by reference, its address is taken (using C's & operator, for instance), or it is accessed from a nested function. Only the last is possible with Tiger. The approach adopted by Appel in his books is easy: a muttable field in the abstract syntax of variable declarations, of for expressions, and of function formal parameters, which introduce new variables, is used to collect the escaping information of the variable. In the ML version of the book this is a reference to boolean (bool ref). Initially, in a conservative approach, the reference is initialized to true. In the variable escaping analysis phase of the compiler, a function findEscape looks for escaping variables and record this information in the escape fields of the abstract syntax. To do this the entire abstract syntax tree is traversed looking for escaping uses of every variable, and, when found, the field is set to true. findEscape uses a symbol table to accomplish its work, binding variable names to a reference to boolean (the same reference used in the abstract syntax tree). When processing a variable declaraction, for instance, it inserts the variable name into the symbol table, binding it to its escaping field. When processing an expression which is a simple variable, if the variable occurs in a nested function, its binding in the symbol table is set to true. This reflects directly in the abstract syntax tree of the variable declaration, as the escape field in the variable declaration and the binding in the symbol table are the same reference to bool. I am look for good ways to implement the variable escaping analysis phase in Haskell, which is a pure language. I see two alternatives: 1) adopt the same approach as Appel used in his ML version of the compiler: use a mutable reference in the IO monad (Data.IORef) to hold the variable escaping information, and write everything inside the IO monad 2) build a new abstract syntax tree with updated information regarding variable escaping The second option is more elegant in my point of view, but would be much less efficient than the first one. So I want to know what advices Haskell programmers has to me about implementing this phase of the Tiger compiler in Haskell. Regards, Romildo -- Computer Science Department Universidade Federal de Ouro Preto, Brazil

5 10