Re: [VirtualGL-Users] Bug with Fluent inside Workbench 16.2

[DRC <dco...@us...>]

This bug should be readily reproducible with any version of VNC.  It's
an issue with the interaction between the VirtualGL interposer and the
ANSYS Workbench launcher.  The X proxy should not figure into it at all.
 There is basically no difference in how RealVNC and TurboVNC present
themselves to applications.  Both appear to be X.org servers as far as
the application is concerned.

However, one thing to bear in mind is that, if you are running
VirtualGL+ANSYS in RealVNC or TigerVNC, those VNC implementations have
an (unaccelerated) GLX extension, so it is possible that the application
(Fluent) would appear to work when running with VirtualGL+RealVNC, but
the rendering wouldn't be hardware-accelerated.  And it's even more
possible that, if ANSYS has in fact tested their product with VirtualGL
(I will give them the benefit of the doubt if they are saying that they
"support" that configuration), they aren't aware of the VGL_LOGO trick.

<brief rant>
I love how these companies say "we only support VirtualGL if you use it
with {Product X that the VirtualGL developers don't in fact support or
have anything to do with}."  Cracks me up.  Even the enterprise version
of RealVNC, with their whiz-bang proprietary compression technology, is
about 2X slower than TurboVNC or TigerVNC on a LAN and something like 5X
slower on a LAN (because they don't support the RFB flow control
extensions, and in fact they couldn't support those unless they spent a
lot of time and money to back-engineer them, because the code is under
the GPL.)
</brief rant>

If they are saying that they officially support VirtualGL + RealVNC
Enterprise, then that's easy.  Just download a 30-day free trial of
RealVNC Enterprise 5.3.0 and verify that you can reproduce the issue in
that X server with VGL_LOGO=1 (although, per above, the issue in RealVNC
may not be that OpenGL rendering is disabled.  It may be that it is not
accelerated, but it's really the same issue.)

Now if for some reason it *does* work properly in RealVNC, then that is
useful info as well, because it tells me that there is something
different about that VNC implementation that ANSYS is detecting.


On 2/17/16 8:28 AM, Quantum HPC wrote:
> Hi Gents,
> 
> Thanks for taking a look again at this. I have tried to work with the 
> Ansys team as well on this and their developers were not of much help 
> since they told me VirtualGl is supported but only with the commercial 
> version of RealVNC....
> I have tried for now the beta of the version 2.5 and also with Ansys v17 
> but still the same issue.
> I cannot find a way to override the LD_PRELOAD inside workbench since it 
> is a program calling another program.
> 
> I am interested if you find something.
> 
> Cheers
> 
> Quentin
> 
> ------ Original Message ------
> From: "Nathan Kidd" <nat...@sp...>
> To: "VirtualGL Users" <vir...@li...>
> Sent: 2016-02-16 2:12:41 PM
> Subject: Re: [VirtualGL-Users] Bug with Fluent inside Workbench 16.2
> 
>> On 16/02/16 12:17 PM, DRC wrote:
>>>  Was there ever a resolution to this?  I found this old tracker item:
>>>
>>>  https://sourceforge.net/p/virtualgl/bugs/22/
>>>
>>>  that seems to confirm that, at least in older versions of ANSYS, 
>>> running
>>>  applications through the launcher stomps on LD_PRELOAD somehow.  It 
>>> was
>>>  supposed to be fixed in 14.5, but maybe it got reintroduced, or 
>>> perhaps
>>>  it was never really fixed for Fluent.
>>>
>>>  The real mystery, though, is why Nathan can't repro the issue.  
>>> Nathan,
>>>  is it possible that the application is falling back to client-side
>>>  OpenGL rendering, which I'm given to understand is supported in ETX 
>>> (but
>>>  isn't supported in TurboVNC)?  In your testing, did you enable 
>>> VGL_LOGO
>>>  to verify that VirtualGL was active?
>>
>> For some reason that email with the screencast link never made it to my
>> inbox; this is the first I've seen it. I don't have cycles to look at
>> this right now, but I've harassed someone else to test it exactly the
>> same way as the screen cast.  (BTW, I always use VGL_LOGO to avoid
>> rendering confusion.)
>>
>> -Nathan
>>
>> ------------------------------------------------------------------------------
>> Site24x7 APM Insight: Get Deep Visibility into Application Performance
>> APM + Mobile APM + RUM: Monitor 3 App instances at just $35/Month
>> Monitor end-to-end web transactions and take corrective actions now
>> Troubleshoot faster and improve end-user experience. Signup Now!
>> http://pubads.g.doubleclick.net/gampad/clk?id=272487151&iu=/4140
>> _______________________________________________
>> VirtualGL-Users mailing list
>> Vir...@li...
>> https://lists.sourceforge.net/lists/listinfo/virtualgl-users
> 
> 
> ------------------------------------------------------------------------------
> Site24x7 APM Insight: Get Deep Visibility into Application Performance
> APM + Mobile APM + RUM: Monitor 3 App instances at just $35/Month
> Monitor end-to-end web transactions and take corrective actions now
> Troubleshoot faster and improve end-user experience. Signup Now!
> http://pubads.g.doubleclick.net/gampad/clk?id=272487151&iu=/4140
> _______________________________________________
> VirtualGL-Users mailing list
> Vir...@li...
> https://lists.sourceforge.net/lists/listinfo/virtualgl-users
>