[LLVMdev] Re: question about gccld and external libraries

[Jakob Praher]

Chris Lattner wrote:
> On Tue, 1 Mar 2005, Jakob Praher wrote:
> 
>>> If you pass '-lrt' when linking your program, it should take care of 
>>> this for you.
>>>
>>
>> ah ok. so every library thet gccld can't find as a bytecode lib is 
>> added to the shell script then.
> 
> 
> Yup.  Note there are other options if you don't want to run your program 
> in the JIT.  In particular, you can use these commands:
> 
> llvm-gcc x.c -Wl,-native  -lrt -l...
> llvm-gcc x.c -Wl,-native-cbe ...
> 
> To produce native executables that do not need the JIT.  The first 
> option uses the native LLVM code generator, the second uses the C 
> backend and a system C compiler.

thanks for the info.
what is the advantage of doing so?
I suppose the at-link-time-optimization stuff, so I suppose that the 
nativization happens at link time? (haven't done the -v on that). 
Otherwise this wouldn't make much sense.

In my case I am strongly interested in llvm as a low level vm (jit) for 
all kinds of (bytecode/language) frontends. Like libjava and libmono. 
Plus I'd like to experiment with it with regards to analyzing large 
systems. But I am at the very start. I've also talked to gcj people at 
the FOSDEM and they would also be interested in that sort of thing at 
least they need a decent jit/dynamic compiler.

It should really be easier in the free world to bridge all those 
languages and llvm seems a viable solution for that.

I'll come with more info on that.

Thanks for your support.
-- Jakob

> 
> -Chris
>