Xilinx FPGA Design Tools for Linux

Now that we've selected HDL, the next step is to let the tool know with which kind of FPGA we're going to design. For illustration, we use the most advanced FPGA devices, the Virtex-II Pro family. This is shown in Figure 4, where the xc2vp7 device is selected, along with an fg456 package. This device is one of the smaller Virtex-II Pro parts, but it still has numerous resources. Here, we use only a small part of the device.

The menu window in Figure 4 also allows other choices. You can select the part speed grade, and here we accept the default of -6. The Project Navigator can be used to organize your entire project flow. For instance, you can perform both behavioral simulation and functional simulation. The first type of simulation is a check that you have a logically correct design—that the design does what it's supposed to do. The second type of simulation is post-FPGA implementation, used for design verification of a completed chip. The device selection screen in Figure 4 also includes other options, such as the particular simulator you want to use for HDL simulation and what language to use for simulating an implemented FPGA. This might include, respectively, some industry-standard tools provided by Xilinx partners, or another simulator, and Verilog or VHDL.

Figure 4. Setting FPGA Device Type

Next, we select a new source file for the design and give the design a filename. In this process, we tell the tool what kind of CAD document we're creating; in this case, it's a Verilog module. We enter a filename of mpy16.v, with the .v being the standard filename suffix for Verilog. It is customary, but not required to make the filename for the top-level module the same as the module, or a name like toplevel.v.

Several other kinds of documents can be entered for the tool and added to the project. We don't have time to examine all of these capabilities, which include alternative entry modes (schematics) and the inclusion of standard and custom HDL libraries made by the user.

To define this (first) Verilog source for the design, the Design Manager offers some help. For the top-level module mpy16, we fill out a module port table using a tabular entry tool (Figure 5). Here, we define the wires that enter and exit the top-level module, and these will end up as the external I/O pins on the FPGA. The names of the ports entered are p, x, y and clk.

We specify p as a 32-bit-wide output and the main inputs, x and y, as 16-bits wide. Because this multiplier will be pipelined, we also include a port named clk, which will provide the synchronous timing source for the multiplier. The port clk is only a single wire, or net, so we leave out content for MSB or LSB in the table. This means clk will be a scalar. In Verilog, vectors are groups of wires or nets, and these are zero-base indexed.

Figure 5. Setting Up the External I/O Pins

After completing tabular entry for the top-level module, we obtain a summary dialog. Then with the project set up, the Project Navigator brings up all the tools and the initial outline for our Verilog module. This is shown in Figure 6, with the skeleton source code in the upper-right corner. An editor is supplied with ISE 6.1i, and you also can import HDL source code created with the Linux editor of your choice.

Figure 6. Project, Tools and Skeleton Source

module mpy16(p,x,y,clk);
        output [31:0] p;
        input [15:0] x;
        input [15:0] y;
        input clk;

        // inferable storage via synthesis
        reg [31:0] p;
        reg [15:0] xq;
        reg [15:0] yq;

        // 16x16 unsigned multiplier specified
        // behaviorally
        always @(posedge clk)
                begin
                        xq <= x;
                        yq <= y;
                        p <= xq * yq;
                end
endmodule // mpy16

Listing 1 is the Verilog source code for a 16-bit pipelined multiplier. This code is done in a behavioral style, and we're going to allow Xilinx Synthesis Technology (XST) to figure how to implement what we mean by the code. Today, synthesis is very powerful, and we simply can infer the multiplier hardware, without having to specify its logic design in detail.