[Ipopt] CUSolver with IPOPT
Stefan Vigerske
stefan at math.hu-berlin.de
Mon Mar 20 10:02:55 EDT 2017
Hi,
no, there isn't any documentation.
All existing interfaces are derived from SparseSymLinearSolverInterface:
https://www.coin-or.org/Ipopt/doxygen/classIpopt_1_1SparseSymLinearSolverInterface.html
For CUSolver, you might want to do the same.
I think the custom_solver is for cases where you want to exploit a
special structure of the linear system to be solved, e.g., some form of
block-structure.
Stefan
On 03/20/2017 01:44 PM, Amir Patel wrote:
> Hi
>
> Thanks for the reply.
>
> Is there any documentation how to incorporate your own linear solver? I see there is the option for 'custom' in the Ipopt solver selection (https://www.coin-or.org/Ipopt/documentation/node51.html ).
>
> Best Regards
> Amir
>
> From: Stefan Vigerske [mailto:stefan at math.hu-berlin.de]
> Sent: Monday, 20 March 2017 1:06 PM
> To: Amir Patel <a.patel at uct.ac.za>; ipopt at list.coin-or.org
> Subject: Re: [Ipopt] CUSolver with IPOPT
>
> Hi,
>
> I'm not aware of an interface from Ipopt to CUSolver, so it would need
> to be written.
> The linear solvers for which interfaces exist already (various HSL
> codes, Mumps, Wsmp, Pardiso) do not use GPUs.
>
> Stefan
>
> On 03/20/2017 11:32 AM, Amir Patel wrote:
>> Hi Everyone
>>
>> I would really like to leverage off the new GPU linear solvers to do the LU computations IPOPT requires. How can I go about getting IPOPT to use CUSolver? Would I need to write my own interface to it?
>>
>> Alternatively, do any of the existing solvers use GPUs at all? I found a paper, "General-Purpose Sparse Matrix Building Blocks using the NVIDIA CUDA Technology Platform", by Christen, Schenk, and Helmar which modified Paradiso and showed great success. But I cannot find this code anywhere...
>>
>> Any help would greatly be appreciated!
>>
>> Best Regards
>> Amir
>>
>> [Description: cid:GQFCUFUZWCCY.IMAGE_13.gif]
>>
>> AMIR PATEL
>> [Description: cid:NUFCNKRVYNWE.IMAGE_13.BMP]
>> DEPARTMENT OF ELECTRICAL ENGINEERING
>> Senior Lecturer
>> PhD (Mechatronics) Cape Town
>>
>> Room 4.08| Menzies Building, Upper Campus, Rondebosch
>> Email: a.patel at uct.ac.za<mailto:a.patel at uct.ac.za<mailto:a.patel at uct.ac.za%3cmailto:a.patel at uct.ac.za>>
>> Telephone: +27 (0)21 650 2803
>>
>>
>>
>>
>> Disclaimer - University of Cape Town This e-mail is subject to UCT policies and e-mail disclaimer published on our website at http://www.uct.ac.za/about/policies/emaildisclaimer/<https://protect-za.mimecast.com/s/9xNDBlfwzbdSz> or obtainable from +27 21 650 9111. If this e-mail is not related to the business of UCT, it is sent by the sender in an individual capacity. Please report security incidents or abuse via csirt at uct.ac.za<mailto:csirt at uct.ac.za>
>>
>>
>>
>> _______________________________________________
>> Ipopt mailing list
>> Ipopt at list.coin-or.org<mailto:Ipopt at list.coin-or.org>
>> https://list.coin-or.org/mailman/listinfo/ipopt<https://protect-za.mimecast.com/s/l1pRBnsqA53td>
>>
>
>
> --
> http://www.gams.com/~stefan<https://protect-za.mimecast.com/s/zK8dB2fMVrOc2>
> Disclaimer - University of Cape Town This e-mail is subject to UCT policies and e-mail disclaimer published on our website at http://www.uct.ac.za/about/policies/emaildisclaimer/ or obtainable from +27 21 650 9111. If this e-mail is not related to the business of UCT, it is sent by the sender in an individual capacity. Please report security incidents or abuse via csirt at uct.ac.za
>
--
http://www.gams.com/~stefan
More information about the Ipopt
mailing list