I wrote a HLL compiler, at http://www1.webng.com/blackoil

[blackoil]

The syntax is showed as pictures there. You have not to register it. If you do, I am happy.

Any feedbacks are welcome!

[Brynet-Inc]

Hmm, For Windows users only and it's closed source? Interesting..

Good luck with that.....

EDIT: Haha this is just funny.. The mysterious l.com is a Token Lexer.. This was probably written in some funky BASIC compiler for Windows

[blackoil]

no... they're all made in assembly, and it's .COM file, not a Win32 app.

cmdline:

l sourcefile.src sourcefile.tok
cg32/cg64 sourcefile.tok output.asm

[Brynet-Inc]

Alright alright.. I was only speculating... Good luck.

[earlz]

I would really like to see more info on this before trying it out..

believe it or not, most people can't look at short examples and instantly know the language enough to right anything useful...

[blackoil]

ok. I updated some manual in the website. The type system is designed to be simple, with SIMD, multi-dimension array support. customed defined type and array can be transfered as parameter direclty, no pointer involved needed.

[Kevin McGuire]

I would imagine that making the basic type default to signed instead of unsigned would help reduce the number of bugs that could be introduced by a programmer used to using C/C++.

I do not know if this is some sort of internal limitation or just a preference you have for default to unsigned, but it might help it to become adopted for usage by someone if that was not such a inversion - unless you had a reason?

[blackoil]

ok, i changed the default integer type to signed now. I don't know C/C++ integer sign setting before. So the signed keyword has changed to unsigned.

[blackoil]

It may be a reason, integer has to be specified a signed or unsigned type, because for array operation, like:

array[ 2 ][ 3 ] = 100;

subscripts has to be unsigned integer type. I use a bit strict type checking, so it's boring to write:

array[ (unsigned) 2 ][ (unsigned) 3 ] = 100;

if integer number is specified as signed type.

Any advice?

[Kevin McGuire]

...subscripts has to be unsigned integer type.. .

You should be able to handle a signed integer for subscripts naturally on a 80386 and compatible, but I do not know about other processors.


mov ebp, ARRAY
mov ecx, ARRAY_INDEX
mov eax, dword [ebp+ecx*ARRAY_ITEM_SIZE]

[blackoil]

Done. Integer number constant is type dependant within expression evaluation.

int a;
unsigned int b;

a+100;
b+200;

[pcmattman]

EDIT: Haha this is just funny.. The mysterious l.com is a Token Lexer.. This was probably written in some funky BASIC compiler for Windows

Is there anything wrong with a token lexer? I've written one in C++ to 'compile' (tokenise) TI-BASIC programs to the 8XP file format for transferring to the calculator. You can view my source at http://www.sf.net/projects/tibasic (it's not done yet, but it works).

[Alboin]

EDIT: Haha this is just funny.. The mysterious l.com is a Token Lexer.. This was probably written in some funky BASIC compiler for Windows

Is there anything wrong with a token lexer? I've written one in C++ to 'compile' (tokenise) TI-BASIC programs to the 8XP file format for transferring to the calculator. You can view my source at http://www.sf.net/projects/tibasic (it's not done yet, but it works).

That reminds me of my favorite word: Nondeterministic finite automata.

Albeit, DFA's are better used, after being converted from the NFA's themselves, and all.

I really love state machines...

[pcmattman]

Nondeterministic finite automata

How can that be your favourite word, that's 3 words... Phrase on the other hand...

[Alboin]

Nondeterministic finite automata

How can that be your favourite word, that's 3 words... Phrase on the other hand...

Ahem, I stand corrected. Unless, of course, you use underscores: "Nondeterministic_finite_automata".