Swapnajit Mittra

It all started when somebody (with little imagination, probably one of Brian Kernighan or Dennis Ritchie) wanted to execute a shell command from a C program. For example, if you want to list the contents of the current directory from your C program, it takes more than just the basic knowledge of C libraries. But, within a shell, all you need to do is just "ls -l". So why bother writing pages of code, if somehow you can invoke an "ls -l" and pass it to the shell !

So, the C system call system() was born and was placed in stdlib.h. If you call system("ls -l") from your C program, it will list the directory structure for you.

When Verilog was created and, later standardized, somehow this important utility was ignored. As a result, standard Verilog does not have a way to interact with the shell.

Both Verilog-XL and VCS provides a non-stadard system task $system()which does an equivalent job. You can invoke $system("ls -l") from your Verilog program to list the contents of the current directory.

Now, each command executed in a shell returns an exit status to the shell. Analyzing the exit status  is important for  analyzing and debugging the command (imagine the shell command to be a perl program or binary executable, rather than just "ls -l"). Since $system() is a system task and, not a system function, there is no way that a Verilog program can get that information.

This is where the following system call $my_system() can be useful. It reads in a shell command in form of a string, execute it and then return the exit status as an integer (32bit).

For sake of brevity, only the calltf function is shown here.

#include <stdlib.h>

#include <veriuser.h>

calltf_my_system() {

      tf_putp(0, system(tf_strgetp(1)));

}

Note, there is just one line in this function even though it does all these : it reads the input argument, executes it and then return the exit status.

Vineyard Research Inc.