This is an OpenCL implementation of the Pagerank Algorithm. We use two ways to implement it, power method and edge-centric method.

You need to have an OpenCL supported environment to run this program. The table below lists the devices we used when we implement the program.

If you are using the OpenCL environment under a Linux OS, you can just run

make

to compile the program, then execute it by running

./pagerank ./dataset/hollins.txt 0

where "hollins.txt" is the input dataset and "0" is the mode selected.

If you are using the Altera SDK, you can run

make fpgapagerank

to compile the host program, and run

make edge_centric.aocx

or

make power.aocx

to compile the kernel you want. Then you should copy all the generated files (or simplely copy the whole directory) to the target Altera supported device (e.g. DE1-SoC Board), and program the FPGA like

aocl program /dev/acl0 edge_centric.aocx

Note: You should follow the instructions of your device to setup environment properly before this.

After you setup the environment properly, similarly run

./pagerank ./dataset/hollins.txt 2

to execute the program.

We already provide 3 sample dataset files and the program support 3 modes:

We test our program on both GPU (GeForce RTX 2080) and FPGA (DE1-SoC).