[mclinker] Garbage Collection

41 views
Skip to first unread message

Ryan Sherlock

unread,
Oct 28, 2013, 2:43:14 PM10/28/13
to mcli...@googlegroups.com
Diana,

I've taken a look at your branch for Garbage Collection and noticed you were using an adjacency matrix for determining which fragments are used by other fragments.
Is there a particular reason you chose this design for Garbage Collection? Are you not concerned with the possibility of high memory and CPU usage for large programs?


Ryan Sherlock

-- 
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, hosted by the Linux Foundation 

Diana Chen

unread,
Oct 29, 2013, 5:35:18 AM10/29/13
to Ryan Sherlock, mcli...@googlegroups.com
Hi Ryan,
Yes, as you said, consider to the memory and CPU usage, we won't use matrix approach. We've done some analysis and experiments and do show that the link time and memory usage are not acceptable when using matrix. In the formal release, we'll use the traditional mark-and-sweep to implement GC. 

Thanks,
Diana 


--
You received this message because you are subscribed to the Google Groups "MCLinker" group.
To unsubscribe from this group and stop receiving emails from it, send an email to mclinker+u...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.



--
柏聿

Ryan Sherlock

unread,
Nov 1, 2013, 9:51:58 AM11/1/13
to Diana Chen, mcli...@googlegroups.com

Hey Diana,

 

Is there design document or specification detailing how you’ll be adding these features to MCLinker? Do you have an estimate or time frame when those features will be available or when the work will begin?

Thanks for your help.

 

Ryan

Diana Chen

unread,
Nov 1, 2013, 11:37:23 PM11/1/13
to Ryan Sherlock, mcli...@googlegroups.com
Hi Ryan,

Currently we don't have the document for garbage collection. We plan to begin the work at the end of this year, and will be available at Feb. 2014. We focus on AArch64 backend support now, so the garbage collection work has been delayed. Do you have any requirement of gc? Thank you.

Diana

Shankar Easwaran

unread,
Nov 3, 2013, 9:34:38 PM11/3/13
to Diana Chen, Ryan Sherlock, mcli...@googlegroups.com
Hi Diana,

Thanks for sharing your plans.

Garbage collection is one of the most used features with our Customers
currently.

Thanks

Shankar Easwaran

On 11/1/2013 10:37 PM, Diana Chen wrote:
> Hi Ryan,
>
> Currently we don't have the document for garbage collection. We plan to
> begin the work at the end of this year, and will be available at Feb. 2014.
> We focus on AArch64 backend support now, so the garbage collection work has
> been delayed. Do you have any requirement of gc? Thank you.
>
> Diana
>
> On Fri, Nov 1, 2013 at 9:51 PM, Ryan Sherlock <rshe...@codeaurora.org>wrote:
>
>> Hey Diana,****
>>
>> ** **
>>
>> Is there design document or specification detailing how you’ll be adding
>> these features to MCLinker? Do you have an estimate or time frame when
>> those features will be available or when the work will begin?****
>>
>> Thanks for your help.****
>>
>> ** **
>>
>> Ryan****
>>
>> ** **
>>
>> *From:* Diana Chen [mailto:mys...@gmail.com]
>> *Sent:* Tuesday, October 29, 2013 4:35 AM
>> *To:* Ryan Sherlock
>> *Cc:* mcli...@googlegroups.com
>> *Subject:* Re: [MCLinker] [mclinker] Garbage Collection****
>>
>> ** **
>>
>> Hi Ryan,****
>>
>> Yes, as you said, consider to the memory and CPU usage, we won't use
>> matrix approach. We've done some analysis and experiments and do show that
>> the link time and memory usage are not acceptable when using matrix. In the
>> formal release, we'll use the traditional mark-and-sweep to implement GC.
>> ****
>>
>> ** **
>>
>> Thanks,****
>>
>> Diana ****
>>
>> ** **
>>
>> On Tue, Oct 29, 2013 at 2:43 AM, Ryan Sherlock <rshe...@codeaurora.org>
>> wrote:****
>>
>> Diana,****
>>
>> ** **
>>
>> I've taken a look at your branch for Garbage Collection and noticed you
>> were using an adjacency matrix for determining which fragments are used by
>> other fragments.****
>>
>> Is there a particular reason you chose this design for Garbage Collection?
>> Are you not concerned with the possibility of high memory and CPU usage for
>> large programs?****
>>
>> ** **
>>
>>
>> Ryan Sherlock
>>
>> --
>> Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, hosted
>> by the Linux Foundation ****
>>
>> --
>> You received this message because you are subscribed to the Google Groups
>> "MCLinker" group.
>> To unsubscribe from this group and stop receiving emails from it, send an
>> email to mclinker+u...@googlegroups.com.
>> For more options, visit https://groups.google.com/groups/opt_out.****
>>
>>
>>
>> ****
>>
>> ** **
>>
>> --
>> 柏聿 ****
Reply all
Reply to author
Forward
0 new messages