ZWOLF

Tutorial on using ZWOLF

Zwolf has two Intel Xeon X5670 processors @ 2.93GHz. As each processor has 6 six cores, zwolf is a 12 CPU shared memory parallel computer. In addition, zwolf has a NVIDIA GeForce GTX 480 which can be used for GPU computing.

Notes: In order to log onto zwolf, you must have credential with UC Davis, and have permissions to use the system from Prof. Bai.

Multicore Instructions:

Step 1: Log into zwolf.cs.ucdavis.edu using putty on windows or ssh on Mac or Linux.
Step 2: Supply your UCDavis Kerberos Username/Password to log in.
Step 3: Set the environment variables in order to use Intel's Compiler. To do this, execute the commands:
```
source /opt/intel/bin/iccvars.sh intel64
```
```
source /opt/intel/bin/ifortvars.sh intel64
```
Step 4: Create a new directory. Copy the examples omp_hello, omp_workshare1, omp_dgemm, omp_dgemm_fortran and the Makefile to this directory.
Step 5: Compile the examples with the command:
```
make openmp
```
Step 6: Remember to type ./ (dot slash) before the program name for running.
Step 7: By default, OpenMP programs run with as many as cores are available (12). You can change the number of threads via the command:
```
export OMP_NUM_THREADS=X
```
where X is any number of threads.

GPU Instructions:

Step 1: Log into zwolf.cs.ucdavis.edu using putty on windows or ssh on Mac or Linux.
Step 2: Supply your UCDavis Kerberos Username/Password to log in.
Step 3: Set the environment variables in order to use Intel's Compiler. To do this, execute the command:
```
source /opt/intel/bin/iccvars.sh intel64
```
Step 4: Create a new directory. Copy the examples gpu_dgemm, gpu_multd and the Makefile to this directory.
Step 5: Add the CUDA binary path to your environment. To do this, execute the command:
```
PATH=$PATH:/usr/local/cuda/bin
```
Step 6: Compile the examples with the command:
```
make gpu
```
Step 7: Remember to type ./ (dot slash) before the program name for running.

These are the instructions on getting started using OpenMP and CUDA on zwolf. You can use these files in other machines but check the path to CUDA and compiler flags in the Makefile.
If you don't want to type the environment setup commands each time you log in, just create a hidden file named .profile (dot profile) file in your home directory with all these commands.

If there are any questions, please email me andres AT cs.ucdavis.edu