- Glasgow-haskell-users

compile-time evaluation
by David Feuer 06 Feb '02

06 Feb '02

>Just saying >'evaluate this CAF to WHNF' might only inline the outer combinator. I'm not sure quite what you mean. However, it seems to me that this sort of thing would match the way other things happen in Haskell fairly well (i.e., default lazy, strictify as necessary). If you want something to be evaluated more deeply, make a CAF to do so, and then tell the compiler to evaluate that... I suspect even this very limited control would be useful for some applications that would otherwise have long and useless start-up times. On a more adventurous note, I think it would be very interesting to be able to ask the compiler to make certain values persistent across program executions. This could solve the same problem, but may also help deal with long computations that may be interrupted, and should be re-started at the same point. (actually, I think this would not be all that great... however, if non-trivial memo tables ever get added to GHC, it could be _very_ useful to store those across program runs for simulations... that way, if the program is given input 12, and told to calculate something with it and print it out, if the program is interrupted, and re-started with the same input, calculation could presumably start at the point it left off...). Of course, I have no clue how realistic any of this is. David Feuer This message has been brought to you by the letter alpha and the number pi.

2 1

Boxing parameterized datatypes?
by Scott Williams 06 Feb '02

06 Feb '02

This one may be fairly basic, but I was wondering if there's a common idiom for boxing a parameterized datatype into a non-parameterized one, and accessing the components in a unified manner. (hope I said that right, pretend I mentioned polymorphism somewhere) > module VarExample where For example if I create a datatype to hold everything about a variable > data Base t = > Base { value :: t > , name :: String > -- , ... more components > } that I want to use as part of a larger data structure, say a heterogeneous list or an AST, I think I have to collect all of the variations under a parameterless datatype declaration. Is this true? > data Var = > IntegerT (Base Integer) > | DoubleT (Base Double) > -- | ... more constructors > type HeterogeneousList = [Var] If so, it looks like I then need to write something like the following in order to access the name of a given Var, and likewise for other components. > getName :: Var -> String > getName (IntegerT b) = name b > getName ( DoubleT b) = name b Doesn't seem bad, but if I have more components and type constructors it can become unwieldy. The following give compile-time errors, but represent what I've tried so far to simplify things. > getBase :: Var -> Base t > getBase (IntegerT b) = b > getBase ( DoubleT b) = b > getName2 v = name (getBase v) > callBaseFn :: Var -> (Base t -> a) -> a > callBaseFn (IntegerT b) fn = fn b > callBaseFn ( DoubleT b) fn = fn b > getName3 v = callBaseFn v name The other option I've thought about is to use the GHC Dynamic module, but it doesn't seem very standard for what there is likely a common idiom. As you can see I'm still somewhat new at this, but would appreciate any help. - Thanks, Scott Williams

1 0

RE: Binary Compatibility
by Julian Seward (Intl Vendor) 06 Feb '02

06 Feb '02

| -----Original Message----- | From: Sigbjorn Finne [mailto:sof@galois.com] | Sent: Wednesday, February 06, 2002 1:27 AM | To: Ashley Yakeley | Cc: GHC List | Subject: Re: Binary Compatibility | | | > Apologies if this has been covered before. What | compatibility is there | > between code compiled in different versions of GHC? | > | | As a rule, none. Always recompile your code for every new | release. Yes, even patchlevel ones. For your specific case we know for sure that the code produced by 5.02 is incompatible with both 5.02.2 and 5.03. J

1 0

Binary Compatibility
by Ashley Yakeley 06 Feb '02

06 Feb '02

Apologies if this has been covered before. What compatibility is there between code compiled in different versions of GHC? My JVM-Bridge was compiled under 5.02 and assembled as package 'javavm'. Will that package work if added to GHC 5.02.2? What about 5.03? -- Ashley Yakeley, Seattle WA

2 1

Re: ANNOUNCE: GHC 5.03.20020204 snapshot release
by Ashley Yakeley 06 Feb '02

06 Feb '02

At 2002-02-05 06:36, Julian Seward (Intl Vendor) wrote: > - newtypes support deriving *any* class for which the > underlying type is also an instance. How about multi-parameter classes where there are instances for the underlying type on more than one parameter: is it possible to specify which parameter the newtype is being derived on? > - Linear implicit parameters: a highly experimental feature. What are they? -- Ashley Yakeley, Seattle WA

1 0

the GCC snapshot
by David Feuer 05 Feb '02

05 Feb '02

What is arbitrarily ranked polymorphism? I don't understand anything past rank 2... Is this arbitrarily ranked polymorphism decidable? This message has been brought to you by the letter alpha and the number pi.

1 0

Linking Problem, ghc 5.02: cannot find -lreadline
by Marc van Dongen 05 Feb '02

05 Feb '02

Dear all, After a long time I decided to start using Haskell again:-) I've downloaded rpm sources (RPMs for SuSE Linux 7.3 (i386, glibc 2.2) from Ralf Hinze: base and docs RPM, and profiling libraries RPM.) from the ghc site and installed them. The only problem I had was that I had to re-install gmp using rpm. So that was great. When I tried to compile some large program I was delighted to see that all object files got created smoothly. At the linking stage I got an error: ''cannot find -lreadline collect2: ld returned 1 exit status'' I couldn't find anything about this on the web or in the mini faq. Anybody there to give me a quick hint how to get around this? Regards, Marc van Dongen -- Marc van Dongen | dongen(a)cs.ucc.ie | Computer Science Department | Western Road | () ASCII ribbon campaign University College Cork | Cork, Ireland | /\ against HTML mail phone: +353 (0)21 4903578 | fax: 4903978 |

2 2

RE: partial evaluation
by Simon Peyton-Jones 04 Feb '02

04 Feb '02

If you say {-# INLINE f #-} on a fn defn for f, it'll get inlined at all its call sites, which is sort of what you want. Just saying 'evaluate this CAF to WHNF' might only inline the outer combinator. Generally, some precise way of controlling what gets evaluated at compile time, a bit less heavyweight than MetaML, would be a fine thing. After all that's what C++ templates are. But the details of a workable design aren't clear to me. Ideas? Simon | -----Original Message----- | From: David Feuer [mailto:David_Feuer@brown.edu] | Sent: 01 February 2002 07:41 | To: glasgow-haskell-users(a)haskell.org | Subject: partial evaluation | Sensitivity: Confidential | | | It seems to me that it might be useful (and probably | fairly easy) to add to GHC a limited sort of partial | evaluation: programmer-designated top-level CAFs could be | evaluated to WHNF during compilation. Termination of that | compiler phase would obviously not be guaranteed, but if | this were possible, I am guessing it would be very useful | for some applications, such as combinator-based parsers. | The parsers would then be generated at compiler time | rather than re-generated every time the parser is | executed. If GHC in fact supports something of this | nature, I would love to hear about it. If it does not, I | would guess it would probably be fairly easy to add to the | end of the code generation phase. | | David Feuer | | This message has been brought to you by the letter alpha and | the number pi. | | | _______________________________________________ | Glasgow-haskell-users mailing list | Glasgow-haskell-users(a)haskell.org | http://www.haskell.org/mailman/listinfo/glasgow-| haskell-users |

2 1

Re: compiling hslibs/xlib
by Jens Petersen 04 Feb '02

04 Feb '02

Jens Petersen <juhp(a)01.246.ne.jp> writes: > "Julian Seward (Intl Vendor)" <v-julsew(a)microsoft.com> writes: > > > Precisely what version of gcc are you using here? > > (gcc -v) > > > > This has been known to happen if you use the default > > gcc on RedHat 6.1 and 6.2. > > Good question. Nothing so old, is the short answer I think. > > The ghc I'm using is from Manuel's ghc-5.02.2-1 rpm (built > on RH72 I believe) is probably with gcc-2.96. > > I am building green-card and xlib with gcc from cvs! > > gcc version 3.1 20020131 (Red Hat Linux Rawhide 3.1-0.20) > > Actually I wanted to build with gcc296, but it (rightly?) > doesn't seem to like "-o" and "-c" in the same command: Ok, scratch that! (I was doing "make CC=gcc296".) Seems it was a gcc-3.1 related problem. :-( When I make green-card and xlib correctly with gcc-2.96 using "ghc -pgmc/usr/bin/gcc296", everything goes through fine. Sorry for the trouble, Jens

1 0

RE: Ticky-ticky profiling (was: GHC 5.02.2 installation/compilationproblems)
by Simon Marlow 25 Jan '02

25 Jan '02

> On Fri, 25 Jan 2002, Nicholas Nethercote wrote: > > > > > Also, when I try to compile for ticky-ticky profiling... > > Further in the ticky-ticky adventure... > > Having compiled the libraries and RTS and other bits for > ticky-ticky, when > I try the -ticky option with ghc I get loads of messages like this at > link-time: > > /homes/njn25/ws/build/ghc/driver/../rts/libHSrts_t.a(GC.t_o): > In function `threadSqueezeStack': > GC.t_o(.text+0x2fe0): undefined reference to > `LDV_recordDead_FILL_SLOP_DYNAMIC' > GC.t_o(.text+0x3010): undefined reference to `LDV_recordCreate' > GC.t_o(.text+0x304d): undefined reference to `LDV_recordCreate' It looks like ticky-ticky has rotted a bit. Some calls to LDV_recordDead_FILL_SLOP_DYNAMIC() and LDV_recordCreate() in Storage.h need to be additionally wrapped in #ifdef TICKY_TICKY. There may be more. If you send me a patch, I'll commit it. Cheers, Simon

1 0