Eric,

I think that it's better to not make assumptions, so I 
check for unistd.h, where sysconf() is defined, and qualify each 
_SC_LEVEL*_SIZE as you suggested.

Please, review the new patch in attach.

Thanks, 


-- 
__________________________________________________________________
Evandro Menezes    Austin, TX      http://rosarynovice.stblogs.com





>________________________________
> From: Eric J Korpela <[email protected]>
>To: Evandro Menezes <[email protected]> 
>Cc: Joachim Fritzsch <[email protected]>; David Anderson (UCBerkeley) 
><[email protected]>; "[email protected]" 
><[email protected]> 
>Sent: Monday, December 17, 2012 10:43 AM
>Subject: Re: [boinc_dev] BOINC on ARM
> 
>
>I think you'll need to preface the block with both 
>#ifdef HAVE_SYSCONF and each if sysconf call with "#ifdef  
>_SC_LEVEL1_DCACHE_SIZE" or similar.
>
>
>Of you could go with "#if defined(__arm__) && defined (__linux__)" if you're 
>sure every arm linux will have the right sysconf macros.
>
>
>
>
>
>On Mon, Dec 17, 2012 at 8:03 AM, Evandro Menezes <[email protected]> wrote:
>
>Joachim,
>>
>>After I posted the patch I wondered if I should have qualified the calls to 
>>sysconf() to figure the cache out with HAVE_SYSCONF or some such autoconf 
>>variable.
>>
>>
>>Thanks for the review.
>>
>>
>>
>>--
>>__________________________________________________________________
>>Evandro Menezes    Austin, TX      http://rosarynovice.stblogs.com
>>
>>
>>
>>
>>
>>>________________________________
>>> From: Joachim Fritzsch <[email protected]>
>>>To: Evandro Menezes <[email protected]>; David Anderson (UCBerkeley) 
>>><[email protected]>
>>>Cc: "[email protected]" <[email protected]>; Rom Walton 
>>><[email protected]>
>>>Sent: Monday, December 17, 2012 3:40 AM
>>>Subject: Re: [boinc_dev] BOINC on ARM
>>
>>>
>>>
>>>Thank you for sharing this! I had those changes on my agenda as well.
>>>
>>>
>>>The parsing of the cache size breaks the build for me, so I created a patch 
>>>file that omits these parts[1].
>>>
>>>
>>>I tested it on two devices: the Client gets the proper CPU description now. 
>>>This can help targeting specific ARM features (e.g. NEON) with the project 
>>>scheduler.
>>>
>>>
>>>David: Could you please review the patch and commit it to the Client code?
>>>
>>>
>>>[1] https://github.com/novarow/AndroidBOINC/blob/master/native/diffs_android/gitmaster171212_arm_host-info.patch
>>>
>>>
>>>
>>>On Sun, Dec 16, 2012 at 7:19 AM, Evandro Menezes <[email protected]> wrote:
>>>
>>>It seems that ARM is the little CPU that could and is become ubiquitous, 
>>>which makes it a candidate to run BOINC.  Now, granted that it doesn't have 
>>>a stunning single-thread performance, yet BOINC used to run on processors of 
>>>similar performance not too long ago.  However, it may be the ideal procesor 
>>>to run NCI BOINC apps.
>>>>
>>>>Anyways, I've been playing with BOINC on the ARM that powers my NAS and it 
>>>>works quite alright and there already are a handful of projects supporting 
>>>>ARM ( http://bit.ly/Zxnv4W ).
>>>>
>>>>However, it seems that a few simple changes to the code are necessary at 
>>>>first to deal with properly identifying the host.  I propose this patch in 
>>>>attach to address this shortcoming.
>>>>
>>>>I haven't tested it on many ARM hosts, actually, just my one such host, but 
>>>>I based it on what the information provided by the Linux kernel for ARM 
>>>>that's very well defined and established.
>>>>
>>>>HTH
>>>>
>>>>
>>>>--
>>>>__________________________________________________________________
>>>>Evandro Menezes    Austin, TX      http://rosarynovice.stblogs.com
>>>>_______________________________________________
>>>>boinc_dev mailing list
>>>>[email protected]
>>>>http://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev
>>>>To unsubscribe, visit the above URL and
>>>>(near bottom of page) enter your email address.
>>>>
>>>
>>>
>>>
>>_______________________________________________
>>boinc_dev mailing list
>>[email protected]
>>http://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev
>>To unsubscribe, visit the above URL and
>>(near bottom of page) enter your email address.
>>
>>
>
>
>

Attachment: hostinfo_unix.diff
Description: Binary data

_______________________________________________
boinc_dev mailing list
[email protected]
http://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev
To unsubscribe, visit the above URL and
(near bottom of page) enter your email address.

Reply via email to