Jacket GBENCH (v1.1)
Jacket GBENCH allows users to gauge the GPU performance of their computer relative to equivalent benchmarks obtained from a variety of other computers, including the CPU of the same computer. Benchmarks include six different tasks, common to the technical computing community:
| 1. | LU: | LU decomposition of 1024 x 1024 matrix |
| 2. | FFT: | Fast Fourier Transform of a 2^20 x 1 vector |
| 3. | BLAS: | Matrix multiplication of two 1024x1024 matrices |
| 4. | 3D Conv: | Convolution of 64x64x64 array with 3x3x3 kernel |
| 5. | FOR/GFOR: | Matrix-vector multiplication of 1024x1024x32 array |
| 6. | Equations: | Solution of a system of 1024 equations |
GBENCH is a practical application benchmark measured in real seconds and is not meant to be a scientific or theoretical benchmark measured in GFLOPs. Also note that for fairness, arithmetic precisions (e.g. double, single) have been matched on the CPU and GPU. Finally, the data sizes used in these computations are large enough to exploit data parallelism (e.g. no scalar arithmetic was attempted). This benchmark assumes a data parallel problem.
Note: GBENCH v1.1 was built using Jacket v1.2.1 and MATLAB R2009b. GBENCH requires CUDA 2.3 (driver and toolkit).
Instructions to Run
GBENCH can be run in two ways: 1) by means of executables 2) from sourceRunning from Executables:
* Requires MATLAB Compiler Runtime (MCR)
1. Extract the given zip file.
2. Install Matlab Compile Runtime using the installer given.
On Windows systems only, append the directory [MCR_ROOT_LOCATION]\v711 to the environment variable PATH.
3. - On Windows Systems run the executable file gbench
- On Linux Systems, do ./run_gbench.sh [MCR_ROOT_LOCATION]/v711
Running from Source:
GBENCH uses the syntax [gpu_times, cpu_times] = gbench(N)
To run gbench from source in MATLAB, add path to the engine directory in Jacket, and run any of the following commands:
* gbench
* gbench(N) % to run the tests N times
GBENCH returns two arrays (each Nx6 in size) with the first containing the results of each of the N runs on the GPU and the second containing the CPU equivalent. In addition, two charts are displayed - the first chart is a bar graph representation of the performance on the given computer (shown in distinct colors) relative to other computers. The second is a full report of the exact times taken to complete the benchmark tests.
Download
Executables:
Linux 64-bit (307 MB)
Linux 32-bit (313 MB)
Windows 64-bit (208 MB)
Windows 32-bit (204 MB)
Source:
GBench Source Package (6 KB)
(Note: Needs MATLAB & Jacket to run)
Sample Results
« Back to Home Page