TeslaGpuKMBnote1

Accessing the Tesla host machine

If you have an account on the host for the Tesla card, you should be able to SSH to red-tomatoes from your ECS workstation.

You have a home directory which should be considered simply a "landing point", which is:

/home/username

DO NOT use this for storing files at present.

You have the same shell as you have on the ECS/MSOR systems.

Your ECS/MSOR filestore is not currently accesible but you can simply scp or rsync files across

There is another area of filestore owned by you on the path

/local/scratch/username

which can also be accessed by using the environmental variable

$LOCSCR

This is the area you should be looking to work in at present.

The $PATH and $LD_LIBRARY_PATH environmental variables are automatically set to contain the CUDA-specific stuff when you login.

Experimenting with the SDK

A copy of the unpacked SDK including examples can be found here

/opt/NVIDIA_GPU_Computing_SDK-3.2

and so source codes, Makefiles, etc can be studied from there, however, the Guide recommends that you unpack your own copy if you wish to build examples within the tree, and the way I suggest you do that is as follows

This first bit is important, change directory to your local scratch area:

cd $LOCSCR

then run the following command which accesses the original downloaded files and other related stuff living in my scratch area:

sh /local/scratch/kevin/CUDA_CD/gpucomputingsdk_3.2_linux.run

This will suggest you create a directory:

~/NVIDIA_GPU_Computing_SDK

DO NOT, repeat DO NOT accept the deafult at this time, enter

/local/scratch/USERNAME/NVIDIA_GPU_Computing_SDK

and the whole of the SDK will unpack for you.

You should thus see the following, assuming you:

1. Replace USERNAME with your username
2. Accept the default for the CUDA install path

Verifying archive integrity... All good.
Uncompressing NVIDIA GPU Computing SDK................
...
Enter install path (default ~/NVIDIA_GPU_Computing_SDK): /local/scratch/USERNAME/NVIDIA_GPU_Computing_SDK

Located CUDA at /opt/cuda
If this is correct, choose the default below.
If it is not correct, enter the correct path to CUDA

Enter CUDA install path (default /opt/cuda):

   ... lots files being upacked ...

========================================

Configuring SDK Makefile (/local/scratch/USERNAME/NVIDIA_GPU_Computing_SDK/C/common/common.mk)...

========================================

 * Please make sure your PATH includes /opt/cuda/bin
 * Please make sure your LD_LIBRARY_PATH includes /opt/cuda/lib

 * To uninstall the NVIDIA GPU Computing SDK, please delete /local/scratch/USERNAME/NVIDIA_GPU_Computing_SDK
 * Installation Complete

Building SDK examples

If you then follow the instructions from the Guide available at the CUDA web site and make all the examples, you will end up using nearly 800 MEGABYTES of space. Not Good.

However, building a single example is not as simple as simply changing into an example directory and running make.

The examples use some support libraries that are themselves part of the SDK, so you have to build those first.

So I'd suggest doing the following to build, for example, the deviceQuery example, but note

1. By default the SDK make invocations run silently, so if you want to see what is happening, use

  make verbose=1 ...rest of your arguments...

so,

  cd  /local/scratch/USERNAME/NVIDIA_GPU_Computing_SDK/C/
  make verbose=1 -C common
  make verbose=1 -C common -f Makefile_paramgl
  make verbose=1 -C common -f Makefile_rendercheckgl
  make verbose=1 -C ../shared

Those commands create these supporting libraries:

  /local/scratch/USERNAME/NVIDIA_GPU_Computing_SDK/C/lib/libcutil_x86_64.a
  /local/scratch/USERNAME/NVIDIA_GPU_Computing_SDK/C/lib/libparamgl_x86_64.a
  /local/scratch/USERNAME/NVIDIA_GPU_Computing_SDK/C/lib/librendercheckgl_x86_64.a

  /local/scratch/USERNAME/NVIDIA_GPU_Computing_SDK/shared/lib/libshrutil_x86_64.a
  /local/scratch/USERNAME/NVIDIA_GPU_Computing_SDK/shared/lib/linux/libGLEW.a
  /local/scratch/USERNAME/NVIDIA_GPU_Computing_SDK/shared/lib/linux/libGLEW_x86_64.a

You can then build the example as follows:

 cd  /local/scratch/USERNAME/NVIDIA_GPU_Computing_SDK/C/
 make verbose=1 -C src/deviceQuery

which creates

/local/scratch/USERNAME/NVIDIA_GPU_Computing_SDK/C/bin/linux/release/deviceQuery

You can then run it as:

/local/scratch/USERNAME/NVIDIA_GPU_Computing_SDK/C/bin/linux/release/deviceQuery

and you will see something similar to the following.

$ /local/scratch/kevin/NVIDIA_GPU_Computing_SDK/C/bin/linux/release/deviceQuery
./deviceQuery Starting...

 CUDA Device Query (Runtime API) version (CUDART static linking)

There is 1 device supporting CUDA

Device 0: "Tesla T10 Processor"
  CUDA Driver Version:                           3.20
  CUDA Runtime Version:                          3.20
  CUDA Capability Major/Minor version number:    1.3
  Total amount of global memory:                 4294770688 bytes
  Multiprocessors x Cores/MP = Cores:            30 (MP) x 8 (Cores/MP) = 240 (Cores)
  Total amount of constant memory:               65536 bytes
  Total amount of shared memory per block:       16384 bytes
  Total number of registers available per block: 16384
  Warp size:                                     32
  Maximum number of threads per block:           512
  Maximum sizes of each dimension of a block:    512 x 512 x 64
  Maximum sizes of each dimension of a grid:     65535 x 65535 x 1
  Maximum memory pitch:                          2147483647 bytes
  Texture alignment:                             256 bytes
  Clock rate:                                    1.30 GHz
  Concurrent copy and execution:                 Yes
  Run time limit on kernels:                     No
  Integrated:                                    No
  Support host page-locked memory mapping:       Yes
  Compute mode:                                  Default (multiple host threads can use this device simultaneously)
  Concurrent kernel execution:                   No
  Device has ECC support enabled:                No
  Device is using TCC driver mode:               No

deviceQuery, CUDA Driver = CUDART, CUDA Driver Version = 3.20, CUDA Runtime Version = 3.20, NumDevs = 1, Device = Tesla T10 Processor


PASSED

Press <Enter> to Quit...
-----------------------------------------------------------
$

Building your own programs

It should be possible to use the SDK examples and directory structure to create your own external build environment.