SA and disassemblers

Tom Rodriguez Thomas.Rodriguez at Sun.COM
Thu Aug 27 11:24:45 PDT 2009

Yes I still have to write the Disassembler native methods for those.   
I may try to write a shared version of the them so that I don't have  
to duplicate the sources.


On Aug 27, 2009, at 11:10 AM, Vladimir Kozlov wrote:

> Tom,
> It is very nice cleanup. I assume you will fix it for linux and win  
> also.
> Thanks,
> Vladimir
> Tom Rodriguez wrote:
>> One of the limitations of the current SA is that it doesn't have a  
>> disassembler for amd64 which makes looking at 64 bit cores somewhat  
>> challenging.  Instead of trying to implement a full disassembler  
>> for x64 in Java I think we should switch the SA to using hsdis-arch  
>> for the decoding.  It requires a minor extension to the  
>> decode_instructions interface so that we can disassemble a buffer  
>> of code that's not physically located where the code came from.   
>> I've got a working version of this that also fixes the printing to  
>> match up better with what the JVM itself prints.  I was planning on  
>> blowing away the other disassemblers completely because they don't  
>> really fit into the new model.  I thought I'd float this before  
>> finishing it to see if there are any concerns.  The initial webrev  
>> is at  This includes  
>> building the SA with source 1.5 to allow use of printf which will  
>> probably be pushed separately.
>> tom

More information about the hotspot-dev mailing list