Following a discussion at LHC
http://lhcathomeclassic.cern.ch/sixtrack/forum_thread.php?id=3457
I've been looking (superficially) at the plan_class code.
 
Seems that both sched_customize and plan_class_spec offer a sample SSE3 plan 
class, tested by finding the substring sse3 in p_model or p_features, depending 
on the age of the client.
 
BUT: the clients don't return sse3 as a feature, they return pni instead.
 
Intel processors (later than the Prescott CPU after which pni is named) support 
ssse3 (triple-s), but AMD processors don't.
 
Because of the sub-string detection, it seems that the sse3 plan_class is 
operational for Intel processors, but blocked for AMD.


>________________________________
>From: Bernd Machenschalk <[email protected]>
>To: Eric J Korpela <[email protected]> 
>Cc: [email protected] 
>Sent: Wednesday, 18 July 2012, 8:27
>Subject: Re: [boinc_dev] plan_class_spec
>
>Hi Eric!
>
>As the original author of that plan_class_spec code I have to admit that the 
>flops scaling etc. is still something of a mystery to me, and although I 
>changed it a few times I am pretty sure I didn't get this right. Apparently 
>the latest rework by David didn't fix it either, although the latest 
>version should use the functions provided in sched_customize.cpp and thus have 
>more code in common with it.
>
>Could you send me (or point me to) SETIs sched_customize.cpp? I hope it could 
>help me to understand and fix the problem.
>
>Best,
>Bernd
>
>On 14.07.12 18:32, Eric J Korpela wrote:
>> I've spent a few weeks in the SETI@home beta trying to move from some
>> highly customized plan classes to the plan_class_spec mechanism for some
>> OpenCL apps.  It's not working, and at this point I'm going to go back to
>> the highly customized plan classes even though they are difficult the keep
>> synced with changes to the source tree.
>>
>> The best I can figure out is that a lot of the scheduling and credit
>> mechanisms are having problems.  Many of the problems seem to have
>> something to do with flop scaling.  In sched_customize.cpp where
>> customizable GPU plan classes live, a flops scale is applied to the GPUs
>> processing rate.  In plan_class_sched, it's applied to the CPUs processing
>> rate even for GPU plans.  The "sample" plan_class_spec.xml file has values
>> that would be suitable for the sched_customize method do not work for
>> plan_class_spec at all.  The scheduler itself is very confused about what
>> it means.  If you put a value of 10 for flops_scale in plan_class_spec.xml,
>> it will multiply the cpu benchmarks by 10 to get the processing rate, but
>> then it does an incorrect calculation of processing rate based on PFC.
>>  Most GPUs seem to get estimated at about 100MFLOP from the bad PFC
>> calculation.  Credits granted for GPU work done come in about 1/500th of
>> what they should be.  Even beta testers get annoyed when the see 0.86
>> credits for something that should have been 700.
>>
>> Unfortunately there's not a a lot of source commonality between the two
>> mechanisms, so anything I learned writing plan classes is lost in
>> plan_class_spec.  So I'm giving up on fixing it myself.
>>
>> If you're planning to use the plan_class_spec mechanism, I would hold off
>> until it's fixed.
>> _______________________________________________
>> 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.
>
>
>
_______________________________________________
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