64-bit support in Harmony

[Andrey Chernyshev <a....@gmail.com>]

The README for HARMONY-14 class libraries contribution says that the
currently supported configurations are "win.IA32" or "linux.IA32".
Some classes have also dependencies on ICU4C and ICU4JNI libraries
which are included as 32-bit binaries with the contribution.

I wonder what people think how we could extend it to support any
64-bit platforms? More specifically:

 - are there any plans to commit 64-bit binaries of ICU4C and ICU4JNI,
say, for EM64T or IPF (Itanium Platform Family)?

- will it be possible to compile and run the current implementation of
Harmony class libraries on 64-bit platforms, provided that the 64-bit
binaries of ICU and 64-bit VM are in place? At least the file
'native-src\linux.IA32\port\hysignal.c' seems to be IA32-specific as
it contains information about IA32 registers. The files 'thrhelp.s'
and 'thrspinlock.s' in the 'thread' directory are also the examples of
IA32-specific code.

Thank you,
Andrey Chernyshev
Intel Managed Runtime Division

Re: 64-bit support in Harmony

[Geir Magnusson Jr <ge...@4quarters.com>]

Andrey Chernyshev wrote:
> The README for HARMONY-14 class libraries contribution says that the
> currently supported configurations are "win.IA32" or "linux.IA32".
> Some classes have also dependencies on ICU4C and ICU4JNI libraries
> which are included as 32-bit binaries with the contribution.
> 
> I wonder what people think how we could extend it to support any
> 64-bit platforms? More specifically:
> 
>  - are there any plans to commit 64-bit binaries of ICU4C and ICU4JNI,
> say, for EM64T or IPF (Itanium Platform Family)?

Seems like we have them now.  Thanks for volunteering!  There is no 
reason why we wouldn't do this.

> 
> - will it be possible to compile and run the current implementation of
> Harmony class libraries on 64-bit platforms, provided that the 64-bit
> binaries of ICU and 64-bit VM are in place? At least the file
> 'native-src\linux.IA32\port\hysignal.c' seems to be IA32-specific as
> it contains information about IA32 registers. The files 'thrhelp.s'
> and 'thrspinlock.s' in the 'thread' directory are also the examples of
> IA32-specific code.

This is up to us, and I think the answer is "certainly".  Someone just 
needs to do the work.

You know where to post the code :)

geir

> 
> Thank you,
> Andrey Chernyshev
> Intel Managed Runtime Division
> 
> 

Re: 64-bit support in Harmony

[Tim Ellison <t....@gmail.com>]

Andrey Chernyshev wrote:
> The README for HARMONY-14 class libraries contribution says that the
> currently supported configurations are "win.IA32" or "linux.IA32".

Right, we figured that they were the most popular development platforms.

> Some classes have also dependencies on ICU4C and ICU4JNI libraries
> which are included as 32-bit binaries with the contribution.

To be clear, the ICU code is an external dependency, not part of the
contribution.  It comes from an established open source project (with
compatible license) http://icu.sourceforge.net/

ICU covers a number of platforms (processor/OS combinations) beyond the
two mentioned above.

> I wonder what people think how we could extend it to support any
> 64-bit platforms? More specifically:

Absolutely.

>  - are there any plans to commit 64-bit binaries of ICU4C and ICU4JNI,
> say, for EM64T or IPF (Itanium Platform Family)?
> 
> - will it be possible to compile and run the current implementation of
> Harmony class libraries on 64-bit platforms, provided that the 64-bit
> binaries of ICU and 64-bit VM are in place? At least the file
> 'native-src\linux.IA32\port\hysignal.c' seems to be IA32-specific as
> it contains information about IA32 registers. The files 'thrhelp.s'
> and 'thrspinlock.s' in the 'thread' directory are also the examples of
> IA32-specific code.

There are a number of places where 64-bit support will have to be
explicitly enabled.  Once we agree on a repository structure that
reflects all the dimensions we want, it will become another work item.
I'd hope that Intel can bring expertise in this area to bear on the problem.

Regards,
Tim

> Thank you,
> Andrey Chernyshev
> Intel Managed Runtime Division
> 

-- 

Tim Ellison (t.p.ellison@gmail.com)
IBM Java technology centre, UK.