Avril 2009 - Gambit-list - mailman.iro.umontreal.ca

Getting char*, uchar*, short*, unsigned short*, ... out of a uvector
by lowly coder 17 Avr '09

17 Avr '09

Hi! This is a different question from my previous silly question. All though it's really cool that I can take a C function with signature: void foo(int *); and in scheme land create a (scheme-foo) so that I can do: (define bar (make-u....vector ...)) (scheme-foo bar) There are a lot of C functions I need to use that take arguments int*, short*, ... I would prefer to not change all these C bindings, and somehow magically have a Scheme function that given: (uber-magic bar) returns me a int* corresponding to where the data is stored in u32-vector bar. Thanks!

3 3

Getting char*, uchar*, short*, unsigned short*, ...
by lowly coder 17 Avr '09

17 Avr '09

It's okay, this isn't urgent (as in done in the next 24 hours or die; more like I need to finish it in the next few days). Here's a concerete problem I'm running into:' ;(define gl/GenTextures (c-lambda ( GLsizei scheme-object ) void "glGenTextures(___arg1,\ ___CAST(GLuint*,&___FETCH_U32(___BODY(___arg2),___INT(0))));")) (define gl/TexImage2D (c-lambda ( GLenum GLint GLint GLsizei GLsizei GLint GLenum GLenum\ scheme-object ) void "glTexImage2D(___arg1, ___arg2, ___arg3, ___arg4, ___arg5, ___arg6\ , ___arg7, ___arg8, ___CAST(GLvoid*,&___FETCH_U32(___BODY(___arg9),___INT(0))));")) the commented out gl/GenTextures work fine the problem is with gl/TexImage2D ... I don't know, until run time, what type of *vector I want to pass to it. Do I need to create a separate gl/TexImage2D for each? I looked at gambit.h, grepping for ___FETCH ... and I got: #define ___FETCH_S64(base,i)*(___CAST(___S64*,base)+(i)) #define ___FETCH_U64(base,i)*(___CAST(___U64*,base)+(i)) #define ___FETCH_S32(base,i)*(___CAST(___S32*,base)+(i)) #define ___FETCH_U32(base,i)*(___CAST(___U32*,base)+(i)) #define ___FETCH_S32(base,i)___CAST_S32(___FETCH_U32(base,i)) #define ___FETCH_U32(base,i)(((i)&1) \ #define ___FETCH_U32(base,i)(((i)&1) \ #define ___FETCH_S16(base,i)*(___CAST(___S16*,base)+(i)) #define ___FETCH_U16(base,i)*(___CAST(___U16*,base)+(i)) #define ___FETCH_S16(base,i)___CAST_S16(___FETCH_U16(base,i)) #define ___FETCH_U16(base,i)(((i)&1) \ #define ___FETCH_U16(base,i)(((i)&1) \ #define ___FETCH_S16(base,i)(((i)&1) \ #define ___FETCH_U16(base,i) \ #define ___FETCH_U16(base,i) \ #define ___FETCH_S8(base,i)*(___CAST(___S8*,base)+(i)) #define ___FETCH_U8(base,i)*(___CAST(___U8*,base)+(i)) #define ___FETCH_S8(base,i)___CAST_S8(___FETCH_U8(base,i)) #define ___FETCH_U8(base,i) \ #define ___FETCH_U8(base,i) \ #define ___FETCH_U8(base,i) \ #define ___FETCH_U8(base,i) \ #define ___BIGAFETCH(base,i) ___FETCH_U32(base,i) #define ___BIGAFETCHSIGNED(base,i) ___FETCH_S32(base,i) #define ___BIGAFETCH(base,i) ___FETCH_U64(base,i) #define ___BIGAFETCHSIGNED(base,i) ___FETCH_S64(base,i) #define ___BIGMFETCH(base,i) ___FETCH_U16(base,i) #define ___BIGMFETCH(base,i) ___FETCH_U32(base,i) #define ___FRAME_FETCH_RA \ is there a generic FETCH instruction that says: "I don't care what kind of *vector you pass me, I'll just give you a void* pointing to the first element" ? Thanks! On Thu, Apr 16, 2009 at 9:47 PM, Marc Feeley <feeley(a)iro.umontreal.ca>wrote: > > On 17-Apr-09, at 12:41 AM, lowly coder wrote: > > I probably deserve to be shot for this, but ... >> >> I need this only when wrapping a single C call, and AFAIK, the gambit GC >> does not run while I'm inside a C function. >> >> Does gambit support things like: >> >> (disable-GC) >> (... unsafe call involving taking int* of a u32-vector) >> (enable-GC) >> >> Thanks! >> > > No it doesn't because just about everything (including function calls) can > allocate heap memory, and cause a GC. However, currently the GC does not > run concurrently with the main program, so it is possible (currently) to > pass a (possibly) movable u32vector to a c-lambda, as a "scheme-object", and > then cast that in the C code to a ___U32* using the ___BODY macro. It > shouldn't be too hard to simplify all of this with a c-define-type with > scheme-to-c and c-to-scheme converter functions. > > Too late for me to do that now. Jérémie can you help? > > Marc > >

2 1

debugging C errors in gambit
by lowly coder 17 Avr '09

17 Avr '09

I probably deserve this: $ gsi main.scm Segmentation fault this isn't an error in my scheme code, there's probably something broken with some weird C interface. Question is ... how do I load up gdb? (I just want a stack trace). Even if there are lots of weird gambit related functions that I have no idea about, it's okay; the only other cohice I have to sprinkle my code with lots of print statements, and I think this is a better way. :-) Thanks!

2 1

Re: [gambit-list] Getting char*, uchar*, short*, unsigned short*, ...
by Marc Feeley 17 Avr '09

17 Avr '09

On 17-Apr-09, at 12:41 AM, lowly coder wrote: > I probably deserve to be shot for this, but ... > > I need this only when wrapping a single C call, and AFAIK, the > gambit GC does not run while I'm inside a C function. > > Does gambit support things like: > > (disable-GC) > (... unsafe call involving taking int* of a u32-vector) > (enable-GC) > > Thanks! No it doesn't because just about everything (including function calls) can allocate heap memory, and cause a GC. However, currently the GC does not run concurrently with the main program, so it is possible (currently) to pass a (possibly) movable u32vector to a c-lambda, as a "scheme-object", and then cast that in the C code to a ___U32* using the ___BODY macro. It shouldn't be too hard to simplify all of this with a c-define-type with scheme-to-c and c-to-scheme converter functions. Too late for me to do that now. Jérémie can you help? Marc

1 0

Gambit u32vector / C FFI
by lowly coder 17 Avr '09

17 Avr '09

Here's some C code: int num_textures = 10; int textures[10]; glGenTextures(10, textures); here's the current binding for gl/GenTextures: (define gl/GenTextures (c-lambda ( GLsizei GLuint* ) void "glGenTextures")) Now, what I'd like to do is: (define textures (make-u32vector 10)) and pass 'textures' to gl/GenTextures : A very smart friend suggested: (define gl/GenTextures2 (c-lambda ( GLsizei scmobj ) void "glGenTextures(___arg1,___CAST(int*,&___FETCH_U32(___BODY(___arg2),___INT(0)));")) $ ... *** ERROR IN ".../gl.scm"@515.45 -- Undefined C type identifier boo, probably due to scmobj; just to test that, let's change it to: (define gl/GenTextures2 (c-lambda ( GLsizei GLuint* ) void "glGenTextures(___arg1,___CAS\ T(int*,&___FETCH_U32(___BODY(___arg2),___INT(0)));")) $ ... gl.c: In function ‘long int ___H__20_gl_2e_o1(___processor_state_struct*)’: gl.c:23068: error: invalid operands of types ‘unsigned int*’ and ‘int’ to binary ‘operator&’ *** ERROR IN "ll.scm"@29.1 -- C compilation or link failed while compiling "gl" yay, we now get a different error so the 'scmobj' is the undefined type, but I don't know what to change it to; grepping through gambit-c.txt, there are a few entries of "scmobj" lying around (but it's embeeded in compilcated C code, and I'm not sure what the alternative should be). so ... what should I change this scmobj to? thanks!

1 1

gambit on the iphone
by lowly coder 16 Avr '09

16 Avr '09

has anyone here successfully written an iphone app in gambit? main points of concern are: (1) i'm not as good with gambit on mac os x as i am with gambit on linux (and haven't put in the time to figure out what i'm configuring incorrectly on the mac) (2) cocoa bindings / dealing with xcode / ... thanks!

4 3

Read from stdin
by Eduardo Cavazos 16 Avr '09

16 Avr '09

Hello, Other Schemes (like Chicken, Ypsilon, Larceny, and Ikarus) will read from stdin by default. For example: echo '(+ 1 1)' | csi However, this doesn't work with 'gsi'. Is there a way to get this to work? Ed

3 4

SRFI-13
by Alex Queiroz 15 Avr '09

15 Avr '09

Hallo, Doing a bit of search, I've found: http://www.math.purdue.edu/~lucier/gambit-srfis/ Is this the preferred implementation of this SRFI for Gambit-C? Cheers, -- -alex http://www.ventonegro.org/

3 4

___S32VECTORREF
by tom＠tmilford.com 05 Avr '09

05 Avr '09

I'm trying to use ___S32VECTORREF as ___F32VECTORREF was illustrated previously on this list. The rest of my code works fine, and if I use float homogenous arrays, that works. ___SCMOBJ shared_array; ... int ival = ___S32VECTORREF(shared_array, i); I've also tried int ival = ___S32VECTORREF(shared_array, ___FIX(i)); Both result in: Line Location scheme-bridge.c:66: error: '___hp' undeclared (first use in this function) Line Location scheme-bridge.c:66: error: '___s32_temp' undeclared (first use in this function) It's not clear to me from the macro definitions in gambit.h how to get around this. Any help appreciated. - Tom

2 1

Searching the archives
by tom＠tmilford.com 05 Avr '09

05 Avr '09

Is there a way to search the gambit mailing list archives? This would be really helpful as I try to get up to speed on the FFI. Thanks, Tom

2 1