On Tue, 22 Aug 2006, Maynard Johnson wrote:
> John Levon wrote:
>> On Mon, Aug 21, 2006 at 02:34:29PM -0700, Hussam Mousa wrote:
>>> symbols from a JVM, specifically jikes RVM.
>>> There are two distinct set of symbols associated with the RVM, the VMs
>>> symbols that are available in a file called RVM.map with a format
>>> similar to system.map.
>> Why aren't these just normal symbols?
>>> The second set of symbols are the applications
>>> and those are dynamically compiled in the common RVM setup (JITed).
>> This topic has been discussed heavily on the list before. I believe IBM
>> are still working on something. Check the list archives for more.
> It is our intent to build on the effort (started on Sept 24, 2004) of
> "providing support in OProfile to allow profiling applications that
> generate code dynamically, such as Java, C#, and other interpreted
> languages". However, we're currently focused on contributing XML
> formatted output to OProfile. We will return to working on the
> interpreted code support as the XML work winds down.
> As John said, there's a lot of info about this topic on the mailing
> list. See postings with subject "oprofile and jitted code" between Sept
> 2004 and May 2005. Also, see postings from Jason Yeh (AMD) with subject
> "Oprofile JIT Code Proposal" in May 2006. Be our guest -- join in the fun!
I am still on the process of polishing this code and this discussion turns
up. I have created a set of patches that follows the original proposal I
send in May. Even though the code is still missing several features, and
not in perfect condition, I would like to sent the code out to continue
The patches are made against Oprofile 0.9.1. They are separated into 4
separate files for easier reading.
I do not have an agent using the libopagent that I can post to the mail
list at ths time. I will send a follow-up email including the agent
using the libopagent and more details tomorrow.