<div dir="ltr">Hi,<br><br>I already commited them to CVS. This timing is after fixing the include statements. Without the intrinsic (the default) non-bonded kernel I get only .3Gflop/s (/2 of with).<br><br>Because this is a compile time problem it should as well not compile on BlueGene/L with the special kernel enabled. <br>
The Gromacs Wiki on BlueGene explained how to compile with the default non-bonded kernel. Thus I'm expecting that people were running with the default (non-instrinsic) kernel. I added an remark to the wiki that one should compile with --enable-bluegene.<br>
<br>The configure had a bug too. The configure help said that the option is colled --ennable-bluegene-asm but it was --enable-bluegene. This I also commited.<br><br>Roland<br><br><div class="gmail_quote">On Thu, Jul 31, 2008 at 3:35 AM, <span dir="ltr"><<a href="mailto:hessb@mpip-mainz.mpg.de">hessb@mpip-mainz.mpg.de</a>></span> wrote:<br>
<blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">Hi,<br>
<br>
So there is really something wrong in the include statements?<br>
Is this only for bluegene/P, or probably also for bluegene/L?<br>
I know people who are running Gromacs on a bluegene/L and have<br>
not heard about these issues.<br>
Could you mail the fixes?<br>
<font color="#888888"><br>
Berk.<br>
</font><div><div></div><div class="Wj3C7c"><br>
> Hi,<br>
><br>
> it was mostly wrong names in the #include statements.<br>
><br>
> It works on BlueGene/P and is much faster than the C non-bonded kernel<br>
> (about 2x). With this intrinsic C kernel one gets 18% of peak. Because<br>
> BlueGene doesn't have SIMD commands specific for single precision (thus it<br>
> does only 2 at a time instead of 4) the performance difference to Opteron<br>
> is<br>
> greater than for double precision code. The performance difference to a<br>
> Barcelona 2.1Ghz core is 5.7x (at 20% peak).<br>
><br>
> Roland<br>
><br>
> On Wed, Jul 30, 2008 at 12:29 PM, Roland Schulz <<a href="mailto:roland@utk.edu">roland@utk.edu</a>> wrote:<br>
><br>
>> Hi,<br>
>><br>
>> it seems that the blugene non-bonded kernel in the CVS version is<br>
>> incomplete. E.g. the nb_kernel_bluegene_test_asm.h is missing as far as<br>
>> I<br>
>> can see. The gromacs wiki for bluegene explains how to get the kernel<br>
>> generated by mknb to work. Also on the mailinglist all the messages I<br>
>> found<br>
>> about BlueGene were about mknb and not abou the bluegene specifik<br>
>> kernel.<br>
>><br>
>> Could it be that it is incomplete do one has to do somethng to generate<br>
>> these files?<br>
>><br>
>> Roland<br>
>><br>
>><br>
>> On Wed, Jul 30, 2008 at 11:45 AM, Berk Hess<br>
>> <<a href="mailto:hessb@mpip-mainz.mpg.de">hessb@mpip-mainz.mpg.de</a>>wrote:<br>
>><br>
>>> Hi,<br>
>>><br>
>>> Mathias Puetz from IBM wrote non-bonded kernels for BlueGene/L.<br>
>>> I don't know if these are also used for BlueGene/P.<br>
>>> If there are problems, he would be the person to contact.<br>
>>><br>
>>> Berk.<br>
>>><br>
>>><br>
>>> Roland Schulz wrote:<br>
>>><br>
>>>> Hi,<br>
>>>><br>
>>>> has someone looked at the non-bonded kernels for BlueGene/P? Based on<br>
>>>> the<br>
>>>> assembler produced by the compiler, it does no SIMD, runs out of<br>
>>>> registers<br>
>>>> and does unnecessary rounding.<br>
>>>><br>
>>>> Are there plans for a PPC450 assembler kernel?<br>
>>>><br>
>>>> Roland<br>
>>>> This email was Anti Virus checked by Astaro Security Gateway.<br>
>>>> <a href="http://www.astaro.com" target="_blank">http://www.astaro.com</a><br>
>>>> ------------------------------------------------------------------------<br>
>>>><br>
>>>> _______________________________________________<br>
>>>> gmx-developers mailing list<br>
>>>> <a href="mailto:gmx-developers@gromacs.org">gmx-developers@gromacs.org</a><br>
>>>> <a href="http://www.gromacs.org/mailman/listinfo/gmx-developers" target="_blank">http://www.gromacs.org/mailman/listinfo/gmx-developers</a><br>
>>>> Please don't post (un)subscribe requests to the list. Use the www<br>
>>>> interface or send it to <a href="mailto:gmx-developers-request@gromacs.org">gmx-developers-request@gromacs.org</a>.<br>
>>>><br>
>>><br>
>>> _______________________________________________<br>
>>> gmx-developers mailing list<br>
>>> <a href="mailto:gmx-developers@gromacs.org">gmx-developers@gromacs.org</a><br>
>>> <a href="http://www.gromacs.org/mailman/listinfo/gmx-developers" target="_blank">http://www.gromacs.org/mailman/listinfo/gmx-developers</a><br>
>>> Please don't post (un)subscribe requests to the list. Use the www<br>
>>> interface or send it to <a href="mailto:gmx-developers-request@gromacs.org">gmx-developers-request@gromacs.org</a>.<br>
>>><br>
>><br>
>><br>
><br>
> This email was Anti Virus checked by Astaro Security Gateway.<br>
> <a href="http://www.astaro.com" target="_blank">http://www.astaro.com</a><br>
><br>
> _______________________________________________<br>
> gmx-developers mailing list<br>
> <a href="mailto:gmx-developers@gromacs.org">gmx-developers@gromacs.org</a><br>
> <a href="http://www.gromacs.org/mailman/listinfo/gmx-developers" target="_blank">http://www.gromacs.org/mailman/listinfo/gmx-developers</a><br>
> Please don't post (un)subscribe requests to the list. Use the<br>
> www interface or send it to <a href="mailto:gmx-developers-request@gromacs.org">gmx-developers-request@gromacs.org</a>.<br>
<br>
<br>
_______________________________________________<br>
gmx-developers mailing list<br>
<a href="mailto:gmx-developers@gromacs.org">gmx-developers@gromacs.org</a><br>
<a href="http://www.gromacs.org/mailman/listinfo/gmx-developers" target="_blank">http://www.gromacs.org/mailman/listinfo/gmx-developers</a><br>
Please don't post (un)subscribe requests to the list. Use the<br>
www interface or send it to <a href="mailto:gmx-developers-request@gromacs.org">gmx-developers-request@gromacs.org</a>.<br>
</div></div></blockquote></div><br></div>