[Gc] Re[2]: Aarch64 GC support

Ivan Maidanski ivmai at mail.ru
Wed Nov 7 09:40:44 PST 2012


Hi Yvan,

1. to submit code please fork https://github.com/ivmai/bdwgc and https://github.com/ivmai/libatomic_ops (master branches) and do pull requests when ready (I think it's easier to start with libatomic_ops)
2. ARM 64-bit code of libatomic_ops should go to https://github.com/ivmai/libatomic_ops/blob/master/src/atomic_ops/sysdeps/gcc/arm.h (similar to x86 - now it hold both 32-bit and 64-bit code)
3. I think it's better to add new arch (e.g., ARM64) to BDWGC (there are only 2-3 place there you should probably write like defined(ARM32) || defined(ARM64)

Regards,
Ivan


Wed, 7 Nov 2012 17:23:26 +0000 "Boehm, Hans" <hans.boehm at hp.com>:
>	
>
>
	
	
>
		
		
			
>
>
>Yvan –
> 
>It’s better to ask this question on the gc list.  Ivan Maidanski is now the primary maintainer of the collector, and reads that list, as do many other people
 who contribute patches.
> 
>I hadn’t realized that the gc list archive is still  stuck.  I will have to look at that once I return from my current trip.  The gmane archive is fine, which
 is probably the reason I haven’t been getting a lot of complaints.
> 
>I think this is a judgment call.  Is there much that has to be different in the two cases, other than macros that are directly related to word length?  If not,
 I would consider it as two variants of the same architecture.  If there are other major differences, I would split it out as a separate architecture.
> 
>Libatomic_ops will have to deal with it quite differently, but that’s a separable issue, I think.
> 
>Hans
> 
>
>
>
>From: Yvan Roux [mailto:yvan.roux at linaro.org]
>Sent: Wednesday, November 07, 2012 5:17 AM
>To: Boehm, Hans
>Subject: Aarch64 GC support
> 
>Hi Hans,
>
> 
>
>I'm in the Linaro toolchain working group, and will start to work on adding the
>
>Aarch64 support to your GC. I don't know if the mailing list is still in use (as the
>
>archive hosted on HP site are stuck in March 2012) and this is why I write to you
>
>directly. 
>
> 
>
>You give a particular advice in your porting wiki page to 64-bit architectures which 
>
>are not treated as new architectures. In the Aarch64 case, the ARM guys consider
>
>it as a new one and implement its support as a new target in GCC for instance, so
>
>do you think I can treat it as an exception like x86_64 ?
>
> 
>
>The porting page seems clear but if you have some advices don't hesitate, I'm a bit
>
>new in the garbage collection world !
>
> 
>
>Regards,
>
>Yvan
>
> 
			
		
		
	

	
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://napali.hpl.hp.com/pipermail/gc/attachments/20121107/c1b64dbe/attachment.htm


More information about the Gc mailing list