[LLVMdev] Re: Re: How could I get memory address for each assemble

[Zhang Qiuyu]

Thanks Andrew. 

I still do not understand your method. Would you mind telling me more? Thanks.
But for hash problem,  it is common problem for hash. We cannot make sure the mapping is 1 to 1. 

I used a way to grab the BB address, but there is some limitation. What I did is that I rewrote some part of disassemble tool. During the disassembling,  I check all the conditional/unconditional instruction, usually with 'j' and call instruction. Once disassemble tool hits the instruction I interested in,  I calculate the target address from the instruction. So the target address must be a entry  of BB and I also need calculate the address which follow the 'j' instruction, that is another address for BB address. In other words, for each 'j' instruction I can extract two entries for BB. But for call instruction, it is different and I can grab one address of entry for BB. That's what I did.

The limitation is that for some case, for example, the instruction look like 
        call %eax
        jmp %eax
I cannot grab the entry address of BB :(. 

When I use llvm, llvm can get the entry for this case, how could llvm get. I haven't read the code. Any opinion is welcome.

Hope it useful.
Qiuyu

> This can be done with a quick hack.  Just modify the assembly printer to
> emit a symbol that encodes the basic block uniquely.  I have had to do
> something similar, and I take the function name (with full type
> information) and the label of the BB and hash them with crc32.  This
> gives a unique int which you can use in a label (which you output) to
> uniquely identify the entry point of a basic block.  There are problems
> if the hash is not unique or if there is not a 1-1 mapping of BB to
> machine BB or if a function is defined in multiple modules with the
> exact same type.
> 
> Anyway, it's a start.
> 
> Andrew
> 
> 
> On Mon, 2004-09-20 at 17:02, Zhang Qiuyu wrote:
> >  
> > Thanks John and Chris. 
> >  
> > I could get the address information by the way you guys mentioned. But
> > it is not what I want, because it will lose information about the
> > entry of basic block. Actually what I want to get is the address of
> > each entry of basic block. Now I am trying to declare the label of
> > basic block as global type, hopefully I can get it from symbol table.
> > Can I get it by this way?
> >  
> > Thanks
> >  
> > >For example,
> > 
> > >llc -o file.s file.bc
> > >gcc -o file.o file.s
> > >objdump -d file.o
> > 
> > >This would produce the list of native code instructions generated by
> > >llc, including the addresses of the native code instructions.
> > 
> > >Is this the sort of answer you're looking for?
> > 
> > >Hi all,
> > 
> > >I am trying to disassemble *.bc to assemble code by using llvm-dis
> > command, but what I got is like the following. So how >could I get the
> > assemble code like objdump? I mean the memory address for each
> > instruction.
> > 
> > >Thanks
> > 
> > >Qiuyu
> >