Project Proposal: PowerPC/AIX port

[Steve Poole]

On 08/05/12 14:07, Volker Simonis wrote:
> On Tue, May 8, 2012 at 12:56 PM, Alan Bateman<Alan.Bateman at oracle.com>  wrote:
>> On 07/05/2012 16:03, Simonis, Volker wrote:
>>> :
>>>
>>> To assist with project bootstrapping and maintain momentum of VM porting
>>> issues
>>> independently from class library issues the project will evolve an
>>> interface
>>> between the VM and class libraries that allows alternative implementations
>>> to be substituted.  The project will start by porting the stable JDK 7
>>> codebase
>>> with a view to moving onto the JDK 8 stream as soon as practical.
>> Volker - can you say anymore about the interface between the VM and class
>> libraries? In particular, do you envisage that this project will want to
>> evolove the existing interface as part of some future push into the main
>> line? I may be reading too much into this but I would think that any changes
>> would require wider discussion beyond the porting project.
>>
> The "VM interface" part of the project will be mainly a playground for our
> IBM colleagues as they have a natural interest to interface the OpenJDK
> class library with their alternative VM implementation. We as SAP are interested
> in this topic because it may help us to more easily support old Java releases
> with newer VM versions.
>
> We are aware of the  former "Common VM Interface" project
> (http://openjdk.java.net/projects/cvmi/) which was started as part of the
> OpenJDK Community Innovators' Challenge. But as far as we can see, this
> project has been discontinued.
>
> While defining such a new interface is surely not the main purpose of this
> project it may very well help to push things into the right direction - perhaps
> in the same way as the MacOS X port restarted the discussion about the
> refactoring of the native class library code.
It's important that we do take this unique new opportunity to explore 
the boundaries between the VM and the class libraries.
I see this project as being a place where some of the edges and gray 
areas get revealed and maybe tweaked a little.
I'm stealing Neil Richards thunder (sorry Neil) but he and I have been 
talking to Andrew Hughes and Mikael Vidstedt about revitalizing
The CVMI project as a place to document and discover the interface 
between the VM and the classes.

So although this project is focused on porting, we would expect it feed 
into the CVMI work and consume from it as appropriate.

I'll let Neil say more elsewhere.

>> For the linux-ppc port, and based on your experiences porting to this
>> target, do you envisage bug fixes, say for memory ordering issues, in many
>> areas of the libraries? If there are non-ppc specific
>> reliability/portability fixes then maybe they should be examined on a
>> case-by-case basis to see whether they should go into jdk8 anyway,
>> irrespective of when the ports come up for review to go into the main line.
>>
> With regards to the linux-ppc port, we don't see many changes in the class
> library area - they are mainly in the shared HotSpot code.
>
> But we are of course highly interested in bringing such changes as
> fast as possible
> into the jdk7/8 main branches. We think that smaller, isolated changes
> in shared code
> may be integrated from the porting repository into the main line at
> any time if this seems
> reasonable (i.e. if they don't affect the stability and performance on
> the existing platforms
> while fixing potential problems or have a positive impact on new platforms).

Yes, agree - if we find changes that aid platform portability I'd expect 
to be be trying to argue them into the mainstream.

Cheers
Steve

>
> Regards,
> Volker
>
>> -Alan.