Lab: Pynq Memory Mapped IO (s_axilite)

This lab describes how to use Pynq to develop an application on the Zynq SoC. The application performs a simple hardware accelerated function on the programmable logic. We first create the IP core that performs the function \(f(x) = 2x\) using high level synthesis. We synthesize it to the programmable logic using the Vivado tools. Using the PYNQ infrastructure, we talk to the IP core from ARM processor using memory mapped I/O. We develop a Pynq notebook that sends data to the IP core, executes the core, and receives the computed results.

0) Vivado Design suite installation

Check here for installation and UCSD license server guide, though you probably would not need the license server for the projects in this course.

We need Vitis HLS for C synthesis and generating the RTL, and Vivado for FPGA prototyping and generating the bitstream. Vitis is primarily for firmware design, thus not required.

Select Standard edition.

To save disk space, you do not need to install the simulation dependencies for all supported devices since we are only using PYNQ-Z2.

1) Vitis HLS: C/C++ to RTL

In this section, you will write your code in C/C++, test it, and convert it to RTL using Vitis HLS.

1.1) Write your code

Open Vitis HLS, create a new project, and name it pynq_mul

Set top function name to mul_test

Do not add any files to your project and proceed to part selection and select xc7z020clg400-1

In Explorer section, right click on Source, select New file and create mul_test.cpp. Complete the body of mul_test.cpp as following:

void mul_test(int* out, int in){
      *out = 2*in;
}

Create a test bench file by right clicking on Test Bench in Explorer section and create a new file named mul_tb.cpp. Complete the body of this file as following:

#include <iostream>

using namespace std;

void mul_test(int* out, int in);

int main(int argc, char *argv[]){
      int x=5;
      int y;
      mul_test(&y, x);
      if(y!=2*x){
              cout << "Test Failed: output(" << y << ") is not equal to 2x" << x << endl;
      }else{
              cout << "Test Passed" << endl;
      }
      return 0;
}

1.2) Test your code

Run C simulation and make sure your code passes your test bench.

1.3) Set port types

You can set the ports with GUI. Make sure that mul_test.cpp is open. Open Directive. In the directive window, right click on the the testbench module itself (in our case, mul_test), and select Modify Directive (or Insert Directive). In the drop down, select INTERFACE and set all the ports to s_axilite by right clicking on available options. Repeat the same for all its ports (right click on out and in). In this example, you should configure the ports for 3 times.

Carefully check your directive tab with the picture below after setting your ports. Make sure that return, out and in ports are all set to s_axilite.

An alternate (and better) way of setting the ports is by making HLS pragma declerations in mul_test.cpp. Simply add 3 lines of commands, and you don’t have to go through the steps with GUI above.

void mul_test(int* out, int in){
      #pragma HLS INTERFACE mode=s_axilite port=return
      #pragma HLS INTERFACE mode=s_axilite port=in
      #pragma HLS INTERFACE mode=s_axilite port=out
      *out = 2*in;
}

1.4) Synthesis and export your design

Run C Synthesis. After the synthesis is complete, you should see the following synthesis summary report-

and after finished, click on export RTL and export your design.

At this point, you can exit and close Vitis HLS.

2) Vivado: RTL to bitstream

In this section, you will import your RTL code to Vivado and generate a bitstream.

2.1) Create a new project

Open Vivado and create a new project and Name your project as pynq_mul

Select RTL Project and check Do not specify sources at this time

Set default part to xc7z020clg400-1

Under IP Integrator, click on Create Block Design

2.2) Import your design

Under Project Manager, click on IP Catalog. Right click inside the newly open ‘IP Catalog’ tab and select Add Repository. In the open window navigate to your Vivado HLS project folder and select <path_to_vivado_hls_folder>/solution1/impl/ip

You can see Mul_test under IP Catalog.

Click Open block design, then click +, add Mul_test IP block into our block design.

2.3) Add connections

In the same window, search for “zynq” and add ZYNQ7 Processing System to your block design.

Your diagram should look like the following:

On top of Diagram window, first click and complete Run Block Automation and then Run Connection Automation with default settings. Your diagram should change and show connections and a couple of extra IPs:

2.4) Generate bitstream

In Sources, right click on design_1 and select Create HDL Wrapper

Under Program and Debug, click on Generate Bitstream and follow instructions to complete synthesis, implementation and bitstream generation.

2.5) Bitstream, .hwh, and addresses

Before closing Vivado, we need to note our IP and its ports addresses.

Under Sources, open mul_test_control_s_axi.v (the exact name may vary across different versions of Vivado), scroll down and note addresses for in and out ports. We need these addresses for our host program.

In the example below for the streamMul, the addresses to pay attention to are 0x00 (control bus ap_ctrl), 0x10 (output), and 0x20 (input). These are the addresses you will need to use to write data to the fabric from the ARM core, start the fabric to run your design and generate your outputs, and then read your outputs from the fabric into the ARM core on the Pynq board.

Under Address Editor note IP’s address

3) PYNQ board and Host program

Download an appropriate image for your board from here and write it to your micro SD Card (PYNQ-Z2 setup instructions). It is recommended to use a dedicated software, e.g. Win32 Disk Imager to burn the .img file to the SD card instead of a simple copy-paste operation, if you find the board cannot booted properly. If you set the jumpers correctly, the boot process should not take more than a few minutes.

Use the ethernet cable to connect the board to your machine, and set the static address as stated in the PYNQ-Z2 tutorial. Connect the JTAG port on the board with your machine using MicroUSB line, and use serial communication software (like PuTTY or Serial Port Unility) to access the command-line-tools (The picture below is for demo only, you don’t need to run those commands). This is especially useful for fixing Linux-related issues of the board.

You can access jupyter notebook through the ipv4 address via a web browser. Create a new folder and notebook. Upload design_1_wrapper.bit from vivado_project_path/pynq_mul.runs/impl1 and copy design_1.hwh from vivado_project_path/pynq_mul.gen/sources_1/bd/design_1/hw_handoff to the folder you just created in Jupyter.

Make sure the .bit file and the .hwh file have the same name. In this case, we name them “design_1_wrapper.bit” and “design_1_wrapper.hwh”.

In the notebook, run the following code to test your IP

from pynq import Overlay
from pynq import MMIO

ol = Overlay("./design_1_wrapper.bit") # designate a bitstream to be flashed to the FPGA
ol.download() # flash the FPGA

mul_ip = MMIO(0x40000000, 0x10000) # (IP_BASE_ADDRESS, ADDRESS_RANGE), told to us in Vivado
inp = 5 # number we want to double

mul_ip.write(0x20, inp) # write input value to input address in fabric
print("input:", mul_ip.read(0x20)) # confirm that our value was written correctly to the fabric
mul_ip.write(0x00, 1) # set ap_start to 1 which initiates the process we wrote to the fabric
print("output:", mul_ip.read(0x10)) # read corresponding output value from the output address of the fabric