Greetings,
This is the AMD Hammer architecture which should
start to be in stores next year (I think).

or

which one?

.NET = Too complicated, anyways Borland is already implementing the secured version of this, while TMT is i think implementing the non-secured version (the non-secured version allows untyped pointers). So that would be a lot of competition for Free Pascal.

JAVA VM = Impossible to implement without limiting a lot the pascal language, since untyped references are not allowed, and indirect calls are also not allowed.

The solution is a small vm that i designed and implemented, its like a virtual RISC CPU, but much much simpler and will work on any architecture which has glibc and gcc. The VM is done, the assembler and linker also, but i need to implement it in the compiler (started but not finished). Its also interesting because i designed so that a hardware version could also be created :)... If you wish more information and have suggestions for improvement i can send you the docs and code.

PS: as I think, today there is no DLL support for FPC under LINUX (link FPC produced *.so with FPC main module). Why? I suggested solution a one year before (november 2001). If my solution was wrong where is rigth solution?

I don't know if someone is working on the register allocator in the compiler currently, but i think that is the best solution.... There is another solution to use push/pop of the GOT register, but this will slow down the generated code in PIC mode A LOT, and its kind of a hack, but it would work...

Carl Eric Codere wrote

.NET = Too complicated,

What's too complicated with .NET? We discussed this some month ago and I still think that we should use .NET instead of VIS.

Non of them is open source :)

True, and keep in mind that this "shadowing" is already the case, and paying off. Most real serious/professional FPC users turn out to be somehow linked to Borland Delphi.

Alike, but slightly different implementation seems to be the "selling" point.

But Carl, why are you pursueing VIS? Only because it is not .NET, or does it have advantages in certain area's itself?

> Virtual machine -- is this a .NET virtaul machine, JAVA virtual machine > or which one? The solution is a small vm that i designed and implemented, its like a

virtual

improvement

i can send you the docs and code.

I don't think that I can give suggestions... I only try to trace current development
path. There is a words (inside sources) about VM but no words about which one :-(
But if You give a URL for some info about this VM...

About implemeting a linker inside FPC... What improvement this can give to current state? (I understand a gifts which give us an implementing of ASM inside
FPC)

Regards,
Sergey Korshunoff

About implemeting a linker inside FPC... What improvement this can give to current state? (I understand a gifts which give us an implementing of ASM inside FPC)

An internal linker gives the benefit of easier cross compiling. And it can also be faster, becuase the object data is already available in memory after compiling a unit.

I don't know what you mean by "ASM inside FPC". Because that is already available.

An internal linker gives the benefit of easier cross compiling. And it can also be faster, becuase the object data is already available in memory after compiling a unit.

Why not to implement linker as external program? This gives the benefit of easier cross compiling too. I think it is impossible to keep in memory all compiled objects for big programs as FPC itself.

Regards,
Sergey Korshunoff