Re: mixing lush with clush

On Thu, 16 Mar 2006, Ralf Juengling wrote:

> On Thu, 16 Mar 2006, Yann LeCun wrote:
>> 
>>> Sorry, I'm not clear on the differences between DX, DY, and DH.
>>> What does it mean to generate DXs instead of DHs?
>> 
>> DX and DYs are are interpreter functions written in C.
>> A DH is a function generated by the Lush compiler.

The compiler actually generates a pair of functions, a C function
that other C functions can call, and a DH, a little wrapper for
that C function. The DH is what the interpreter gets to see; if
the interpreter wants to evaluate a function that is a DH, it
calls dh_listeval (in src/lisp_c).

A DX is another kind of wrapper for C functions. If the interpreter
wants to evaluate a DX, it calls dx_listeval (defined in src/functions;
same thing for DYs). Currently, there are C functions for many DX/DY
functions, but not for all. If you look into src/string.c, for example,
you find the pair str_index (the C function), xstr_index (the DX
wrapper for str_index), which implements the function 'index'. The
implementation of function 'concat' OTH is not broken down into a
pair of C function + DX wrapper, there is only one function,
xstr_concat (so in this case the "wrapper" includes everything).

This is what I meant with "C API" earlier: some_cfunc for every
pair of some_cfunc/xsome_cfunc or some_func/ysome_func in the files
in src.

The DX wrappers are all hand-coded. They essentially do runtime
type-checking and then call the wrapped C function (if it exits).
When you say "the compiler should produce marshalling code" for
C API functions, I guess you mean it should do the type checking
at compile time if possible, perhaps emit C code for runtime type
checking where it doesn't have enough information at compile time,
and finally emit the C function call. It currently cannot do compile
time type-checking because the DXs do not carry information about
the C functions they wrap. DHs do carry that information.

I hope that makes sense.
Ralf



-------------------------------------------------------
This SF.Net email is sponsored by xPML, a groundbreaking scripting language
that extends applications into web and mobile media. Attend the live webcast
and join the prime developer group breaking into this new coding territory!
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=110944&bid=241720&dat=121642