[Paraview] CUDA dependency issue when building ParaView 4.1.0-RC2 on Mac

[Utkarsh Ayachit]

Ben,

Any updates? We should clean this up before the release.

Utkarsh


On Fri, Jan 3, 2014 at 4:02 PM, Burlen Loring <bloring at lbl.gov> wrote:

> that's how it was handled it in the past, prior the recent refactoring,
> and I think that's the way to go. Where does the refactoring stand is Ben B
> finished? It would be great if Ben B could clean this up, but if not I can
> make a pass next week.
>
> Burlen
>
>
> On 01/03/2014 11:47 AM, Utkarsh Ayachit wrote:
>
>> Burlen,
>>
>> CUDA is confusing more people than not. Also it's adding too many
>> variables. A cleanup to FindCUDA.cmake is certainly in the order.
>> Until then, I propose we swicth SQTK_CUDA variable to do a
>> find_package(CUDA) only if SQTK_CUDA is enabled. What do you think?
>> BTW, I want to tag the release soon, so we should try to keep changes
>> to a minimal at this point.
>>
>> Utkarsh
>>
>>
>> On Thu, Jan 2, 2014 at 5:33 PM, Burlen Loring <bloring at lbl.gov> wrote:
>>
>>> Hey Patrick and Ben, Thanks for the feedback.
>>>
>>> In this case I think it's confusing to be dumped into a separate ccmake
>>> screen. I'd say it's really an issue in the FindCUDA.cmake module, which
>>> does
>>>
>>>   message("CUDA_TOOLKIT_ROOT_DIR not found or specified")
>>>
>>> according to the documentation a naked message command is supposed to
>>> mean
>>> "Important information". In this case it's not. I think this message
>>> should
>>> probably be downgraded to STATUS type. just my 2 cents.
>>>
>>> At any rate, we could avoid the problem in PV if in the plugin we used
>>>
>>> find_package(CUDA QUIET)
>>>
>>> but initially all the cuda stuff, including find_package, was controlled
>>> by
>>> a cache variable so that this confusing behavior could be avoided and I
>>> wouldn't have to use the QUIET option because when you're actually
>>> trying to
>>> configure for cuda the output is helpful if things go wrong.
>>>
>>> Burlen
>>>
>>>
>>> On 01/02/2014 02:01 PM, Benjamin Spencer wrote:
>>>
>>> You're right.  I didn't realize this was just an informational message,
>>> and
>>> not an error that has to be resolved.  After I hit that message the first
>>> time, then configure a second time and hit that message again, I can
>>> successfully generate the makefile.
>>>
>>> Thanks,
>>> Ben
>>>
>>> On Thu, Jan 2, 2014 at 2:37 PM, Patrick O'Leary <
>>> patrick.oleary at kitware.com>
>>> wrote:
>>>
>>>> Dear Ben and Burlen,
>>>>
>>>> On my Mac Book Pro with NVidia card, I received the same informational
>>>> message but it builds if I exit the informational screen with 'e' and
>>>> hit
>>>> 'g'.
>>>>
>>>> Best regards,
>>>> Patrick
>>>>
>>>>
>>>> On Thu, Jan 2, 2014 at 2:33 PM, Burlen Loring <bloring at lbl.gov> wrote:
>>>>
>>>>> Hi Benjamin,
>>>>>
>>>>> This is of interest to me as I haven't been involved in the recent
>>>>> plugin
>>>>> changes and need to look through them more closely.
>>>>>
>>>>> You reported an informational message that is not an error. Could you
>>>>> report the actual error?
>>>>>
>>>>> btw, I also see the message on my ATI based system, but it doesn't
>>>>> prevent me from building. For example in ccmake I can press "e" to
>>>>> exit the
>>>>> informational screen where the message is displayed and then hit "g" to
>>>>> generate the make files and go on to compile without issue.
>>>>>
>>>>> Burlen
>>>>>
>>>>>
>>>>> On 01/02/2014 12:41 PM, Benjamin Spencer wrote:
>>>>>
>>>>> I'm trying to build ParaView 4.1.0-RC2 from source on Mac OS (10.8.5).
>>>>> This machine does not have CUDA installed (and has no nvidia graphics
>>>>> hardware).
>>>>>
>>>>> When running ccmake, I get this error:
>>>>>
>>>>> CUDA_TOOLKIT_ROOT_DIR not found or specified
>>>>>
>>>>> This seems to be coming from the call:
>>>>> find_package(CUDA)
>>>>> in
>>>>> Plugins/SciberQuestToolKit/SciberQuest/CMakeLists.txt
>>>>>
>>>>> If I comment that line out, I can successfully compile.  I'm no cmake
>>>>> guru, but it seems like if I don't have CUDA installed,
>>>>> find_package(CUDA)
>>>>> should just set CUDA_FOUND to OFF and not require any paths to CUDA to
>>>>> be
>>>>> set.  Maybe this is an issue with the CUDA module in CMake rather than
>>>>> ParaView.  I'm using the latest CMake (2.8.12.1).
>>>>>
>>>>> Thanks,
>>>>> Ben
>>>>>
>>>>> Benjamin Spencer
>>>>> Fuels Modeling and Simulation
>>>>> Idaho National Laboratory
>>>>>
>>>>>
>>>>> _______________________________________________
>>>>> Powered by www.kitware.com
>>>>>
>>>>> Visit other Kitware open-source projects at
>>>>> http://www.kitware.com/opensource/opensource.html
>>>>>
>>>>> Please keep messages on-topic and check the ParaView Wiki at:
>>>>> http://paraview.org/Wiki/ParaView
>>>>>
>>>>> Follow this link to subscribe/unsubscribe:
>>>>> http://www.paraview.org/mailman/listinfo/paraview
>>>>>
>>>>>
>>>>>
>>>>> _______________________________________________
>>>>> Powered by www.kitware.com
>>>>>
>>>>> Visit other Kitware open-source projects at
>>>>> http://www.kitware.com/opensource/opensource.html
>>>>>
>>>>> Please keep messages on-topic and check the ParaView Wiki at:
>>>>> http://paraview.org/Wiki/ParaView
>>>>>
>>>>> Follow this link to subscribe/unsubscribe:
>>>>> http://www.paraview.org/mailman/listinfo/paraview
>>>>>
>>>>>
>>>
>>> _______________________________________________
>>> Powered by www.kitware.com
>>>
>>> Visit other Kitware open-source projects at
>>> http://www.kitware.com/opensource/opensource.html
>>>
>>> Please keep messages on-topic and check the ParaView Wiki at:
>>> http://paraview.org/Wiki/ParaView
>>>
>>> Follow this link to subscribe/unsubscribe:
>>> http://www.paraview.org/mailman/listinfo/paraview
>>>
>>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.paraview.org/pipermail/paraview/attachments/20140107/87dc557b/attachment.htm>