[Paraview] Multi-core custom applications

[Burlen Loring]

right, but just as the current multicore makes it easy for the user, 
this would all happen automatically as PV starts. So when multicore is 
on, the baked in multicore pvsc dialog is presented as pv starts, the 
user reviews the defaults, and hits OK, and pv starts up connected to a 
server. same as multicore is now, but with the possibility to review the 
run parameters.

I think that even the most basic user will need to modify the number of 
cores used or some encounter performance problems. My sense is that this 
makes it even easier to use than it currently is. at least I find it a 
bit of a hassle to change the number of cores. for example PV 
automatically starts a server using my previous settings, which often 
not what I want. I have to wait for it to start the server I don't want, 
then I have to open up the settings menu, change the number of cores to 
what I do want, then restart it by disconnecting from the current 
server. before I learned that trick I was restarting pv, which was even 
more of a pain. I'm afraid some users don't know about the disconnect 
trick.

On 02/10/2014 12:18 PM, Utkarsh Ayachit wrote:
> All makes sense, but then why not just do that as a new server
> configuration named multicore? The main thing that multicore does is
> that it hides the "builtin" connection to be parallel server
> connection. MutliCore is really for people who are confused by remote
> server to launch a remote server without them seeing it. If one wants
> to do anything fancy, I am still not convinced it makes sense to hide
> it under the "multi-core" umbrella. BTW, you can change the number of
> cores to use from the settings dialog.
>
> Utkarsh
>
> On Mon, Feb 10, 2014 at 2:57 PM, Burlen Loring <bloring at lbl.gov> wrote:
>> for me every run is different, sometimes I want to use 2 cores others I want
>> to use 16. my thinking is that we need a way to set the number of cores
>> before the server starts up. It sounds like others are encountering
>> situations where they need to tweak the startup parameters as well.  and why
>> not just make use of the existing infrastructure for specifying those run
>> parameters? The connect to mechanism isn't just for remote servers, also
>> works fine for local ones, and it's easy to add dialog entries for
>> parameters by pvsc. Perhaps the multicore just starts the connect to process
>> at the pvsc dialog, based on a baked in multicore pvsc, which could
>> potentially be configured during the build/install? Also internally, for
>> example in the memory inspector, we currently need to handle the multicore
>> server differently than any other server (it's actually currently broken for
>> multicore because of this). even though it's really no different than any
>> other server, at least judging from the command line displayed in the
>> terminal.
>>
>>
>> On 02/10/2014 10:51 AM, Utkarsh Ayachit wrote:
>>> Burlen,
>>>
>>> Won't it just be easier to add support for a "default" server to
>>> connect to? I am not sure why connecting to a remote-server in the
>>> guise of the multicore would be any better.
>>>
>>> Utkarsh
>>>
>>> On Mon, Feb 10, 2014 at 12:11 PM, Burlen Loring <bloring at lbl.gov> wrote:
>>>> Hi Utkarsh,
>>>>
>>>> I'm wondering what you think about the idea that multicore use the same
>>>> mechanism as the normal connect to server? Is there a strong reason for
>>>> it
>>>> to be different? I'm think that when multicore featured is enabled a
>>>> dialog
>>>> defined by an internal pvsc with some basic options such as number of
>>>> cores
>>>> and perhaps pvserver location etc is used. This would allow the user to
>>>> make
>>>> some basic parameter changes and internally make the built in multicore
>>>> runs
>>>> look the same as any other pvserver run. We had discussed this before and
>>>> you asked me to file a feature request.
>>>> http://paraview.org/Bug/view.php?id=14312
>>>>
>>>> Burlen
>>>>
>>>>
>>>> On 02/10/2014 06:40 AM, Utkarsh Ayachit wrote:
>>>>
>>>> Makes sense. May providing static API on the class so that custom
>>>> application to change the location of pvserver may be helpful. What do
>>>> you think? If you want to contribute a patch, I'll be more than happy
>>>> to merge it in.
>>>>
>>>> Utkarsh
>>>>
>>>> On Mon, Feb 10, 2014 at 6:28 AM,  <mathieu.westphal at gmail.com> wrote:
>>>>
>>>> Hello
>>>>
>>>> it appears, from vtkProcessModuleAutoMPI.cxx n taht pvserver MUST be in
>>>> the
>>>> binary dir to be able to use Multi-core option.
>>>>
>>>> Shouldn't it be configurable in cmake at least ? and even automatically
>>>> found by cmake ?
>>>>
>>>> Mathieu
>>>> _______________________________________________
>>>> 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