- Glasgow-haskell-users

Building Win32 DLLs: Manual is not up-to-date
by Michael Marte 01 Mar '06

01 Mar '06

Hello *, I am trying to turn a prototype written in Haskell into something I can call from a Visual Studio C++ 6.0 program. (The prototype is so clean and fast that there is no point in recoding everything.) So I followed the instructions in section 11.5 of the ghc manual ("Building and using Win32 DLLs") and ended up as follows: $ ghc -c dllMain.c dllMain.c:4: warning: type defaults to `int' in declaration of `EXTFUN' dllMain.c:4: warning: parameter names (without types) in function declaration dllMain.c:4: warning: data definition has no type or storage class dllMain.c: In function `DllMain': dllMain.c:18: error: `__stginit_Adder' undeclared (first use in this function) dllMain.c:18: error: (Each undeclared identifier is reported only once dllMain.c:18: error: for each function it appears in.) Section 8.2 proposes to add extern void __stginit_Adder(void) which actually works. So it seems that section 11.2 is not up-to-date!

2 1

Re[3]: factorial: let's get ahead of jhc! :)
by Bulat Ziganshin 01 Mar '06

01 Mar '06

Hello Bulat, Friday, February 24, 2006, 6:37:42 PM, you wrote: SPJ>> Perhaps you may consider doing this transformation on the C-- data type SPJ>> only, without involving the (already very complicated) STG -> C-- code SPJ>> generator? i have found my investiations in this area. that is the C-- code generated for fac worker: Fac_zdwfac_entry() { c1C0: _s1BD = I32[Sp + 0]; if (_s1BD != 1) goto c1C4; R1 = I32[Sp + 4]; Sp = Sp + 8; jump (I32[Sp + 0]); c1C4: _s1BI = _s1BD * I32[Sp + 4]; _s1BF = _s1BD - 1; I32[Sp + 4] = _s1BI; I32[Sp + 0] = _s1BF; jump c1C0; } first, we convert jump to the explicit loop: _s1BD = I32[Sp + 0]; while (_s1BD != 1) { _s1BI = _s1BD * I32[Sp + 4]; _s1BF = _s1BD - 1; I32[Sp + 4] = _s1BI; I32[Sp + 0] = _s1BF; _s1BD = I32[Sp + 0]; } R1 = I32[Sp + 4]; Sp = Sp + 8; jump (I32[Sp + 0]); then, we cache contents of sp[*] variables in the local ones: sp4 = I32[Sp + 4]; sp0 = I32[Sp + 0]; _s1BD = sp0; while (_s1BD != 1) { _s1BI = _s1BD * sp4; _s1BF = _s1BD - 1; sp4 = _s1BI; sp0 = _s1BF; _s1BD = sp0; } I32[Sp + 4] = sp4; I32[Sp + 0] = sp0; R1 = I32[Sp + 4]; Sp = Sp + 8; jump (I32[Sp + 0]); and then we wipe out all the superfluous variables: sp4 = I32[Sp + 4]; sp0 = I32[Sp + 0]; while (sp0 != 1) { sp4 = sp0 * sp4; sp0 = sp0 - 1; } R1 = sp4; Sp = Sp + 8; jump (I32[Sp + 0]); and it is even for simple fac() function!!! instead of straightforward generation of code that we need we should make all these nontrivial studying and hard transformations on already generated code. on the other side, we can just generate the following code right from the STG: int fac () { sp4 = I32[Sp + 4]; sp0 = I32[Sp + 0]; while (sp0 != 1) { sp4 = sp0 * sp4; sp0 = sp0 - 1; } R1 = sp4; Sp = Sp + 8; jump (I32[Sp + 0]); } i think that my way is 100x simpler -- Best regards, Bulat mailto:Bulat.Ziganshin@gmail.com

2 1

RE: GHC 6.4.1 crash on Windows XP
by Simon Marlow 01 Mar '06

01 Mar '06

Can you tell which process is crashing? Is it the GHC process that is interpreting Setup.hs, or the process invoked by Setup.hs to build the package? One thing you could try is compiling Setup.hs to a binary, and running that instead. Also, if you could run Setup like this and send us the output: $ ./setup build -v (or the runhaskell version, if that's the only one that crashes) Cheers, Simon On 28 February 2006 10:57, Cyril Schmidt wrote: > Simon Marlow wrote: > >> Does it always crash in the same way? > > Yes, as far as I can see. Unfortunately, most of the context is > already lost when the error message pops up. The error code is > always the same. > >> Can you fire up GHCi and perform a small evaluation or two? > > Yes, this works just fine. > >> Can you compile small programs, do they run? > > Also works fine. > >> >> Please send us the complete output from commands that crash: eg. add >> -v to the build command above. >> > > I attach the stderr and stdout from > runhaskell -v5 Setup.hs build > > I could also send the sources of the package being compiled, but I do > not think they are relevant here: it crashes on *any* package. > > BTW, I had installed Cabal-1-1.3, but it did not help, so I reverted > to the version of Cabal that came with the compiler. > > It is strange that the same version of GHC works fine on my PC and > breaks on that of my colleague. To the best of my knowledge, these > two PCs are not different from each other in any special way. I did > have some strange crashes on my PC as well, but these were cured by > upgrading from 6.4 to > 6.4.1. > > Cheers, > > Cyril > > >> Cyril Schmidt wrote: >>> A freshly installed GHC 6.4.1 on my colleague's PC crashes when I >>> try to build a package: >>> >>> runhaskell Setup.hs build >>> >>> The effect is easily reproduceable (it shows up on *any* package >>> that I try to build). >>> >>> Does anyone have any idea of what might be wrong here?

1 0

GHC 6.4.1 crash on Windows XP
by Cyril Schmidt 28 Feb '06

28 Feb '06

A freshly installed GHC 6.4.1 on my colleague's PC crashes when I try to build a package: runhaskell Setup.hs build The effect is easily reproduceable (it shows up on *any* package that I try to build). Does anyone have any idea of what might be wrong here? Cyril ___ For the record, the information given by Windows at the point of crash: Exception Information Code: 0xc0000005 Flags: 0x0 Record: 0x0 Address: 0x0 System Information Windows NT 5.1 Build: 2600 Module 1 ghc.exe Image Base: 0x00400000 Image Size: 0x0 Checksum: 0x0088d2bd Time Stamp: 0x433058b8 (The IP and SP already point somewhere inside Dr. Watson, as far as I can see, so I don't think the registers and stack contents are of any use at this point.)

2 2

How to bind a window library (in C) to Haskell?
by Brian Hulley 27 Feb '06

27 Feb '06

Hi - I'd like to be able to use Haskell for the project I'm working on but the problem is that I've already written a lot of code for a nice GUI using DirectX in Visual C++. I thought it might be possible to make up some kind of simple API for it which I could call from Haskell, so I started with the following simple Haskell module to try and see if I could understand the FFI: module Main where data Window = Window_Edit | Window_List | Window_Tree data Layout = Layout_Left | Layout_Right type Callback = () -> Bool foreign import ccall gui_init :: [(Window, Layout, Callback)] -> IO () main = gui_init [(Window_Edit, Layout_Left, \_->True)] Of course my "init" function is far too simple, but I was thinking this captures the essence of the kinds of things that would need to be passed to the C++ code ie a list of windows and call back functions etc. However I've immediately run into a major problem. When I try to compile with ghc -fglasgow-exts --make main.hs I get an error: "unacceptable argument type in foreign declaration..." My questions are: 1) Does this mean that the FFI can only pass Int, Char etc and not user defined data types or compound data types? 2) I'm also really confused by the different kinds of pointers available, and how to safely store a function closure in some C data structure, and how to use this to call the function from within C and access the result safely 3) When does GHC do garbage collection? Is the garbage collection done in the same thread as the executing program? Does GHC run a normal Haskell program using multiple threads? Would I need to link my C DLL with a multithreaded version of the C runtime to avoid problems? Alternatively, has anyone managed to use DirectX or COM from within a Haskell program? (because then I could perhaps rewrite all my gui code from scratch in Haskell...) (I'm loathe to switch to OpenGL because OpenGL is very poorly supported on windows - a default WinXP installation does not come with hardware accelerated OpenGL drivers, afaik, and also I can't find any documentation on the Haskell OpenGL bindings except the Haddock type signatures which are just not nearly enough to understand how to use it, and some out of date docs) Thanks, Brian.

2 2

Re: inside the GHC code generator
by Zdenek Dvorak 27 Feb '06

27 Feb '06

Hello, > > SPJ> tail calls is another, > > > > nowadays gcc optimize tail calls > > I found this thesis on the subject by Andreas Bauer: > > "Compilation of Functional Programming Languages using GCC -- Tail > Calls" > > Do you know if it is his work that was incorporated into GCC? no. Also, it is based on now obsolete version of the GCC (most of the tail call optimization is currently handled in a completely different way). If anyone has some concrete proposals on improvements of handling of tail calls in GCC that could help GHC, please use the gcc bugzilla (http://gcc.gnu.org/bugzilla/) to file an improvement request, and add me to CC, I will look what I can do. Zdenek

1 0

RE: inside the GHC code generator
by Simon Peyton-Jones 24 Feb '06

24 Feb '06

| The closest thing I've seen to a solution is the technique used in Urban | Boquist's thesis, which is to put a static table in the executable with | information about register and stack usage indexed by procedure return | point, and use that to unwind the stack and find roots. Every accurate garbage collector (including GHC's) uses a technique like this, but the solution is invariably tightly-coupled to the garbage collector, compiler and run-time system. A major feature of C-- is that it squarely addresses this problem in a *reusable* way. Simon

2 1

inside the GHC code generator
by Rene de Visser 24 Feb '06

24 Feb '06

The following link gives reasons for not generating via C http://groups.google.com/groups?hl=en&lr=&ie=UTF-8&selm=4zp8kn7xe.fsf_-_%40… Naturally a number of these are common lisp specific, however I think that Haskell and GCC are quite semantically different, so using GCC might prevent a lot of optimizations, that aren't possible in C, but would be in GHC. The OCAML-OPT backend is supposed to produce i386 assembly that is competitive with GCC. Maybe this could be ported to GHC? Rene.

5 16

RE: Re[3]: factorial: let's get ahead of jhc! :)
by Simon Peyton-Jones 24 Feb '06

24 Feb '06

| i have found my investiations in this area. that is the C-- code | generated for fac worker: | | Fac_zdwfac_entry() { | c1C0: | _s1BD = I32[Sp + 0]; | if (_s1BD != 1) goto c1C4; | R1 = I32[Sp + 4]; | Sp = Sp + 8; | jump (I32[Sp + 0]); | c1C4: | _s1BI = _s1BD * I32[Sp + 4]; | _s1BF = _s1BD - 1; | I32[Sp + 4] = _s1BI; | I32[Sp + 0] = _s1BF; | jump c1C0; | } Once we do the A/B split of the code generator that I referred to, we should get something like this from the A part fac(n,m) { if n!=1 goto L return(m) L: jump fac( n-1, n*m ) } The B part will introduce the Sp nonsense. I think you'll find this a much easier optimisation target. Simon -users

1 0

UTF-8 decoding error
by Christian Maeder 20 Feb '06

20 Feb '06

Hi, with ghc-6.5.20060201 I get a "UTF-8 decoding error" for latin1 characters in my string literals. Do I have to change my sources or can I set a certain environment variable? I have LANG=de_DE@euro and LC_CTYPE not set (which is ok for hugs) Cheers Christian

4 10