VMD-L Mailing List

From: Cara Kreck (cara_a_k_at_hotmail.com)
Date: Wed Aug 27 2014 - 11:37:39 CDT

We're trying to get VMD to run remotely on a GPU cluster (RHEL) from our Windows 7 workstations using MobaXTerm. VMD was compiled from source on the GPU cluster in order to get it to run with CUDA 5. Remote visualisation using MobaXTerm works on a CPU cluster we have access to and this is the info I get there:

Info) VMD for LINUXAMD64, version 1.9.1 (November 19, 2013)
Info) http://www.ks.uiuc.edu/Research/vmd/
Info) Email questions and bug reports to vmd_at_ks.uiuc.edu
Info) Please include this reference in published work using VMD:
Info) Humphrey, W., Dalke, A. and Schulten, K., `VMD - Visual
Info) Molecular Dynamics', J. Molec. Graphics 1996, 14.1, 33-38.
Info) -------------------------------------------------------------
Info) Multithreading available, 12 CPUs detected.
Info) Free system memory: 13839MB (77%)
Warning) Detected X11 'Composite' extension: if incorrect display occurs
Warning) try disabling this optional X server feature.
Info) OpenGL renderer: Quadro 600/PCIe/SSE2
Info) Features: STENCIL MSAA(16) MDE MTX NPOT PP PS
Info) GLSL rendering mode is NOT available.
Info) Textures: 2-D (18x17), 3-D (18x17x18), Multitexture (4)

The only things that seems special about the setup here is the unusually low texture numbers (is this due to the limited graphics ability of the remote CPUs?) and setting environment variable LIBGL_ALWAYS_INDIRECT=y. However, if I try to use the same Windows workstation to run VMD on the GPU cluster I get this:

Info) VMD for LINUXAMD64, version 1.9.1 (August 19, 2014)
Info) http://www.ks.uiuc.edu/Research/vmd/
Info) Email questions and bug reports to vmd_at_ks.uiuc.edu
Info) Please include this reference in published work using VMD:
Info) Humphrey, W., Dalke, A. and Schulten, K., `VMD - Visual
Info) Molecular Dynamics', J. Molec. Graphics 1996, 14.1, 33-38.
Info) -------------------------------------------------------------
Info) Multithreading available, 4 CPUs detected.
Info) Free system memory: 7062MB (89%)
Info) Creating CUDA device pool and initializing hardware...
Info) Detected 2 available CUDA accelerators:
Info) [0] Tesla C1060 30 SM_1.3 @ 1.30 GHz, 4.0GB RAM, OIO, ZCP
Info) [1] Tesla C1060 30 SM_1.3 @ 1.30 GHz, 4.0GB RAM, OIO, ZCP
Warning) Detected X11 'Composite' extension: if incorrect display occurs
Warning) try disabling this optional X server feature.
XRequest.146: GLXBadRenderRequest 0x0
Info) OpenGL renderer: Quadro 600/PCIe/SSE2
Info) Features: STENCIL MSAA(16) MDE MTX NPOT PP PS GLSL(OVFG)
XRequest.146: BadRequest (invalid request code or no such operation) 0xa00001
XRequest.146: BadRequest (invalid request code or no such operation) 0xa00001
XRequest.146: BadRequest (invalid request code or no such operation) 0xa00001
XRequest.146: GLXBadRenderRequest 0x0
XRequest.146: GLXBadLargeRequest 0xfd
XRequest.146: GLXBadLargeRequest 0x2
XRequest.146: GLXBadRenderRequest 0x0
XRequest.146: BadRequest (invalid request code or no such operation) 0x0
XRequest.146: GLXBadRenderRequest 0x0
XRequest.146: BadRequest (invalid request code or no such operation) 0x0
Warning) GPU driver failed to compile shader:
Warning) /usr/local/vmd-1.9.1/shaders/vmd
Warning) This GPU/driver is buggy, or doesn't fully implement GLSL.
Warning) Set environment variable VMDGLSLVERBOSE for more info.
Info) GLSL rendering mode is NOT available.
Info) Textures: 2-D (16384x16384), 3-D (2048x2048x2048), Multitexture (4)

Although the menu and OpenGL display open, it can not display any structure file I try to load, instead giving me a long list of XRequest.146: GLXBadLargeRequest and XRequest.146: GLXBadRenderRequest errors. Setting LIBGL_ALWAYS_INDIRECT=y or LIBGL_ALWAYS_SOFTWARE=y doesn't seem to make any difference. I have a couple of RHEL workstations that I have also tested. The first one has an ATI Radeon HD 5450 and didn't give any XRequest.146 errors but still didn't work:

Info) Detected 2 available CUDA accelerators:
Info) [0] Tesla C1060 30 SM_1.3 @ 1.30 GHz, 4.0GB RAM, OIO, ZCP
Info) [1] Tesla C1060 30 SM_1.3 @ 1.30 GHz, 4.0GB RAM, OIO, ZCP
Warning) Detected X11 'Composite' extension: if incorrect display occurs

Info) Features: STENCIL MSAA(8) MDE MTX NPOT PP PS

The OpenGL window displayed something similar to static. I tried installing the proprietary AMD drivers and now it causes the computer to crash and log off every time I try to run VMD remotely. Strangely enough, the local install still works fine:

Info) VMD for LINUXAMD64, version 1.9.1 (February 1, 2012)
Info) http://www.ks.uiuc.edu/Research/vmd/
Info) Email questions and bug reports to vmd_at_ks.uiuc.edu
Info) Please include this reference in published work using VMD:
Info) Humphrey, W., Dalke, A. and Schulten, K., `VMD - Visual
Info) Molecular Dynamics', J. Molec. Graphics 1996, 14.1, 33-38.
Info) -------------------------------------------------------------
Info) Multithreading available, 4 CPUs detected.
Info) Free system memory: 6445MB (82%)
Warning) Detected a mismatch between CUDA runtime and GPU driver
Warning) Check to make sure that GPU drivers are up to date.
Info) No CUDA accelerator devices available.
Warning) Detected X11 'Composite' extension: if incorrect display occurs
Warning) try disabling this optional X server feature.
Info) OpenGL renderer: ATI Radeon HD 5450
Info) Features: STENCIL MSAA(16) MDE CVA MTX NPOT PP PS GLSL(OVFGS)
Info) Full GLSL rendering mode is available.
Info) Textures: 2-D (16384x16384), 3-D (8192x8192x8192), Multitexture (8)

However, another RHEL workstation with proprietary nVidia drivers can remotely use VMD on the GPU cluster perfectly:

Info) Detected 2 available CUDA accelerators:
Info) [0] Tesla C1060 30 SM_1.3 @ 1.30 GHz, 4.0GB RAM, OIO, ZCP
Info) [1] Tesla C1060 30 SM_1.3 @ 1.30 GHz, 4.0GB RAM, OIO, ZCP
Warning) Detected X11 'Composite' extension: if incorrect display occurs
Warning) try disabling this optional X server feature.
Info) OpenGL renderer: Quadro FX 570/PCIe/SSE2
Info) Features: STENCIL MSAA(16) MDE MTX NPOT PP PS GLSL(OVFG)
Info) Full GLSL rendering mode is available.
Info) Textures: 2-D (8192x8192), 3-D (2048x2048x2048), Multitexture (4)

It is interesting that the 2-D textures numbers are lower here than what I see from Windows or the other linux machine. Instead they are the same as the those reported when running the locally installed version of VMD. I guess that is because of the limitations of the local graphics card. Could it be that reducing the textures makes remote visualisation more feasible? If so, is it possible to manually set this number lower than the limit of the graphics card? I've also found that MobaXTerm could not successfully run VMD remotely from either of the RHEL workstations. The only time MobaXTerm has succeeded was in the first example where the textures were very low. I've also tested VcXsvr as an alternative to MobaXTerm and didn't have any more luck there.

Does anyone have any other suggestions to explain this erratic pattern?

Thanks,

Cara