Re: [Capstone-users] LLVM Bitcode as Input File

[Nguyen A. Q. <aq...@gm...>]

On Tue, Sep 16, 2014 at 11:20 PM, Peter Lucia <pe...@ro...> wrote:

> Yes, I want disassemble LLVM bitcode to LLVM assembly.  Basically I want
> to do exactly what llvm-dis does, but I'd like to have
> the programmatic bindings that capstone provides; ie: get list of all the
> instructions with each instruction object having it's opcode,
> source registers, and destination registers broken out in nice
> Python/Ruby/C#/Java/etc. objects.  Ideally, I'd like to use some of the
> same tools already built (or that I build in the future) using capstone but
> with LLVM bitcode files as source instead of *insert architecture here*
> machine code files.  Sorry for not being clear at first.
>
> I'll take a look at the llvm-dis source, the question really is how hard
> will it be to integrate that source into capstone so we have API parity
> with the other supported input machine languages.  Thanks for any further
> thoughts you have now or in the future regarding this!
>

the difference is that you would have to feed Capstone the whole LLVM
bitcode binary file, not just a chunk of binary bytes like with machine
code.

i havent looked that the code of llm-dis, but i dont think this is too
hard, as it does not involve any backend here.

if you make it, please share the code. i am curious ....

thanks.



> Peter
>
> On Tue, Sep 16, 2014 at 10:10 AM, Nguyen Anh Quynh <aq...@gm...>
> wrote:
>
>>
>>
>> On Tue, Sep 16, 2014 at 10:06 PM, Nguyen Anh Quynh <aq...@gm...>
>> wrote:
>>
>>>
>>>
>>> On Tue, Sep 16, 2014 at 9:51 PM, Peter Lucia <pe...@ro...>
>>> wrote:
>>>
>>>> Any thoughts on how hard it would be to implement?  Note that I'm only
>>>> trying to go to LLVM assembly code, not some other architecture's assembly
>>>> code.  Thanks
>>>>
>>>
>>> do you mean generating machine code from LLVM bitcode? this involves the
>>> backends of all related archs, and is quite different from what Capstone is
>>> doing. so extending Capstone to do this is not the choice, but you need to
>>> do that from scratch.
>>>
>>> to have a better idea, look at the source of llvm-dis. that is certainly
>>> a lot of works.
>>>
>>
>> i think i misunderstood your question: you actually want to disassemble
>> LLVM bitcode to LLVM assembly, right?
>>
>> then it is much easier than generating machine code from LLVM bitcode,
>> but again you should look at the source of llvm-dis.
>>
>>
>> cheers,
>> Q
>>
>>
>>>
>>>
>>>> Peter
>>>>
>>>> On Tue, Sep 16, 2014 at 8:24 AM, Nguyen Anh Quynh <aq...@gm...>
>>>> wrote:
>>>>
>>>>>
>>>>>
>>>>> On Tue, Sep 16, 2014 at 6:20 AM, Peter Lucia <pe...@ro...>
>>>>> wrote:
>>>>>
>>>>>> Hello,
>>>>>>
>>>>>> Is LLVM Bitcode itself a supported input architecture?  I have some
>>>>>> files that have not yet been fully compiled to PowerPC or MIPS, but instead
>>>>>> are still in LLVM Bitcode format (in other words, clang with the -
>>>>>> emit-llvm option) and I'd love to use capstone to disassemble them
>>>>>> pragmatically.
>>>>>>
>>>>>> If this is not already implemented, how hard would it be to do?  Is
>>>>>> there some other better way to do this?  I'm using llvm-dis now, but it's
>>>>>> not ideal for all the reasons capstone would seem to excel.
>>>>>>
>>>>>>
>>>>> unfortunately no. the main idea of Capstone is to disassemble machine
>>>>> binary code back to assembly code. so Capstone does not, and will never
>>>>> convert LLVM bitcode back to assembly for you.
>>>>>
>>>>>
>>>>> thanks.
>>>>> Q
>>>>>
>>>>
>>>>
>>>
>