Installing Tesla C2050 and C1060 on Centos

I have been waiting to get a NVIDIA Fermi GPU to prototype my algorithms, but I was waiting to get an optimized implementation on the Tesla C1060 and then scale up to the Fermi. I got a C2050 and set about installing it on my machine. (note: my host machine is a Lian-Li PC with an integrated NVIDIA nForce 980a/780a (8 cores). 

Installing the C2050 was easy, it has 2 power connections (6-pin and 8-pin). In most of the cases, connecting the 8-pin connector is more than enough. 

After installing the C2050, I had to boot the machine and change the BIOS setting to disable display from external GPUs. The C2050 has a display out and I don't intend to use it now. 

My deviceQuery output:

[vivekv@atstgpu release]$ more deviceQuery.txt
./deviceQuery Starting...
 CUDA Device Query (Runtime API) version (CUDART static linking)
There are 3 devices supporting CUDA
Device 0: "Tesla C2050"
  CUDA Driver Version:                           3.10
  CUDA Runtime Version:                          3.10
  CUDA Capability Major revision number:         2
  CUDA Capability Minor revision number:         0
  Total amount of global memory:                 3220897792 bytes
  Number of multiprocessors:                     14
  Number of cores:                               448
  Total amount of constant memory:               65536 bytes
  Total amount of shared memory per block:       49152 bytes
  Total number of registers available per block: 32768
  Warp size:                                     32
  Maximum number of threads per block:           1024
  Maximum sizes of each dimension of a block:    1024 x 1024 x 64
  Maximum sizes of each dimension of a grid:     65535 x 65535 x 1
  Maximum memory pitch:                          2147483647 bytes
  Texture alignment:                             512 bytes
  Clock rate:                                    1.15 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:                   Yes
  Device has ECC support enabled:                No
Device 1: "Tesla C1060"
  CUDA Driver Version:                           3.10
  CUDA Runtime Version:                          3.10
  CUDA Capability Major revision number:         1
  CUDA Capability Minor revision number:         3
  Total amount of global memory:                 4294770688 bytes
  Number of multiprocessors:                     30
  Number of cores:                               240
  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 2: "nForce 980a/780a SLI"
  CUDA Driver Version:                           3.10
  CUDA Runtime Version:                          3.10
  CUDA Capability Major revision number:         1
  CUDA Capability Minor revision number:         1
  Total amount of global memory:                 131399680 bytes
  Number of multiprocessors:                     1
  Number of cores:                               8
  Total amount of constant memory:               65536 bytes
  Total amount of shared memory per block:       16384 bytes
  Total number of registers available per block: 8192
  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:                                    1.20 GHz
  Concurrent copy and execution:                 No
  Run time limit on kernels:                     Yes
  Integrated:                                    Yes
  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
deviceQuery, CUDA Driver = CUDART, CUDA Driver Version = 3.10, CUDA Runtime Version = 3.10, NumDevs = 3, Device = Tesla C2050, Device = Tesla 
C1060

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

I hope to post the performance comparison of my kernels using the Tesla C2050 sometime soon.

More HPC tutorials

I want to quickly share this blogpost with lot of tutorials on MPI, OpenMP, CUDA and Graphics programming http://supercomputingblog.com


I like the tutorials here and rate it higher than Dobb’s Supercomputing for masses series.

CUDA and Computational Finance

Saw this on my Google Reader list and wanted to share it with everyone. This URL contains some videos/presentations on using CUDA for Computational Finance. (thanks to Argyn)

Lots more CUDA tutorials coming at SC09 !!!

Installing Boost C++ libraries

I know that installing Boost C++ libraries has nothing to do with GPGPU and multi-cores. Nevertheless, I have posted my procedure for installing Boost C++ libraries on a Mac.

I use Macports on my Mac and I find that it is the easiest way to install any open-source software on Mac OSX.

1. Search for boost packages

bash-3.2$ sudo port search boost

boost @1.40.0 (devel)

Collection of portable C++ source libraries

boost-build @2.0-m12 (devel)

Build system for large project software construction

boost-gil-numeric @1.0 (devel)

An algorithm extension to boost-gil.

boost-jam @3.1.17 (devel)

Boost.Jam (BJam) is a build tool based on FTJam

py26-pyplusplus @1.0.0 (python, devel)

Py++ is an framework for creating a code generator for Boost.Python library and ctypes package

Found 5 ports.

2. Install

bash-3.2$ sudo port install boost-jam  

---> Fetching boost-jam

---> Verifying checksum for boost-jam

---> Extracting boost-jam

---> Configuring boost-jam

---> Building boost-jam with target all

---> Staging boost-jam into destroot

---> Installing boost-jam

bash-3.2$ sudo port install boost

---> Fetching boost

---> Verifying checksum for boost

---> Extracting boost

---> Configuring boost

---> Building boost with target all

---> Staging boost into destroot

---> Installing boost

bash-3.2$ sudo port install boost-build  

---> Fetching boost-build

---> Verifying checksum for boost-build

---> Extracting boost-build

---> Configuring boost-build

---> Building boost-build with target all

---> Staging boost-build into destroot

---> Installing boost-build

3. Here's my example program:

#include <iostream>

#include <boost/any.hpp>

using namespace std;

int main()

{

boost::any

a(5);

a = 7.67;

std:cout<<boost::any_cast<double>(a)<<std::endl;

}

Now when I tried compiling my example program, I got a lot of errors such as:

example3.cpp:11:25: error: boost/any.hpp: No such file or directory

example3.cpp: In function ‘int main()’:

example3.cpp:18: error: ‘boost’ has not been declared

example3.cpp:18: error: ‘any’ was not declared in this scope

example3.cpp:18: error: expected `;' before ‘a’

example3.cpp:19: error: ‘a’ was not declared in this scope

example3.cpp:20: error: ‘boost’ has not been declared

example3.cpp:20: error: ‘any_cast’ was not declared in this scope

example3.cpp:20: error: expected primary-expression before ‘double’

example3.cpp:20: error: expected `;' before ‘double’

So I googled around and came to this link and tried compiling using the full path:

$ g++ -I /opt/local/var/macports/software/boost/1.40.0_1/opt/local/include/ example3.cpp -o example3_new

and that worked...!!!!

I also did this as a shortcut for compiling my programs that need the Boost libraries:

$ export BOOST=/opt/local/var/macports/software/boost/1.40.0_1/opt/local/include/

$ echo $BOOST

/opt/local/var/macports/software/boost/1.40.0_1/opt/local/include/

$ g++ -I $BOOST example3.cpp -o example3_new

$./example3_new

7.67

Hope this helps :)

OpenCL drivers available from Nvidia

OpenCL drivers are available for download from Nvidia.

You can download the SDK, Best Practices guide from here: Nvidia OpenCL download

Waiting to try some sample code examples using OpenCL and evaluate how it differs from CUDA :)

SAAHPC presentation

I recently presented my work at the SAAHPC conference, held at NCSA Urbana, IL. The Keynote talk was by Pradeep Dubey on Massive Data Computing using Intel Larrabee. Excellent overview on why data transfer and management is the challenge in today's computing. I liked the Connected Computing factor - Content, Connect and Compute. I have found out from real-time problems that having the fastest computation platform is not enough, it's even more important to sustain the streaming bandwidth of the data into the platform. Obviously, once the data is inside the memory, computation is fast. But the real bottleneck is in importing and offloading the data and trying to streamline and synchronize the data (some of my PhD dissertation grumble).

I liked the talk by Michael Garland on GPU Computing using CUDA. This was informative in terms of his insight on the Thrust template. Thrust is open source and is hosted on Google Code. I have to start using this for my next CUDA project.

Finally, here's a link to my presentation on "Accelerating Particle Image Velocimetry using Hybrid Architectures".

Installing CUDA 2.3

Installing CUDA 2.3 is pretty easy and straightforward. However, on my Mac, the CUDA SDK examples are now in

/Developer/GPU Computing

In addition to the drivers for Leopard, there's a separate version for Snow Leopard :) New documentation includes the CUDA Best Practices Guide. This is the correct link, thanks to the NVIDIA forums post. The CUDA Resource page does not point you to the correct link.

Have a good time accelerating your apps using CUDA :)