I have written an UVM testbench that has 3 agents and am now in the process of writing a scoreboard/checker. I need to have a checker module for my SystemVerilog Assertions, but this checker module needs to be aware of register configuration that is done from the test (and can be random, decided during run_phase of the test).
I am unable to figure out how this would work? If I were to create a checker module for my assertions, and bind it at the top level (tb_top) to the dut, how does this checker module know my register configuration?
After reading some papers, I figured I could write my checker module as an interface, set it in tb_top. But this would give access to the variables in my interface to the UVCs. How does the interface access variables in the UVCs?
Any help is appreciated. I feel I am missing something key here as this has probably been done plenty of times before.
EDIT: Please don't tell me I have to implement some kind of API to set each individual register setting from my UVCs? I want to just get a handle to my reg_block (or any other config variable in my agents)
It seems that you want to pass information from tb_top to your UVC or vice versa. This information will be used by your assertion in tb_top, and shared by your UVC. My suggestion, you can either use uvm_resource_db or uvm_config_db.
I can think of two ways of achieving this communication.
First method is set the configuration in your tb_top, then your UVC grab this handle. From here on, you can communicate your register or whatever info you need for your assertion.
class my_tb_config extends uvm_object;
// ...
endclass
module tb_top;
my_tb_config tcfg;
initial begin
tcfg = new("tcfg");
uvm_config_db#(my_tb_config)::set(uvm_root::get(), "*", "my_tb_config", tcfg);
end
endmodule
// somewhere in your UVC
class my_uvc extends uvm_component;
my_tb_config tcfg;
function void build_phase(uvm_phase phase);
// now both tb_top and your UVC point to the same config object
void'(uvm_config_db#(my_tb_config)::get(this,"","my_tb_config", tcfg));
endfunction
endclass
Another method is the other way around. Pass your UVC configuration to your tb_top.
class my_other_uvc extends uvm_component;
my_tb_config tcfg;
function void build_phase(uvm_phase);
tcfg = new("tcfg");
uvm_resource_db#(my_tb_config)::set("*", "my_tb_config", tcfg);
endfunction
endclass
// somewhere in your tb_top
module tb_top;
my_tb_config tcfg;
initial begin
#1ps; // small delay, making sure resource is submitted
void'(uvm_resource_db#(my_tb_config)::read_by_name("*","my_tb_config",tcfg);
// Now both your tb_top and UVC share same object, so you can freely define your whatever communication between them
end
endmodule
I figured out a way to do this. Firstly, I realized that I had asked two separate questions:
1) My checker module needs to be aware of register configuration that is done from the test
I use cross module reference into my design to access my registers, and this provides me with up-to date register configuration as set by the test during the run-phase.
tb.sv
module tb;
dut my_dut( ... )
interface my_checker (
.input_registerA (tb.my_dut.my_sub_module.regA),
.input_registerB (tb.my_dut.my_sub_module.regB),
.input_registerC (tb.my_dut.my_other_sub_module.regC),
....
)
endmodule
my_checker.sv
interface my_checker (
input input_registerA,
input input_registerB,
input input_registerC,
....
);
// Here I can write properties/assertions that are register-aware
endinterface
2) How does the interface access variables in the UVCs?
This is a little trickier. I want to dynamically update my checker variables from the uvm_sequence or uvm_monitor etc.
I read this paper by Verilab that clearly describes the method to do this:
http://www.verilab.com/files/litterick_sva_encapsulation.pdf
In my checker module, I create a uvm_component. From this component, I now have access to the uvm_resource_db, through which I can exchange info with my UVM-testbench.
One thing to remember is that the uvm_component instantiated in the checker module is located at the top-level (uvm_root).
Related
I have a testbench in Vivado which has a hierarchy of IP--some custom IP and some Xilinx IP, like the Zynq Processing System. The Zynq Processing System also has an associated Verification IP library that has useful API for doing things like loading DDR.
I would like to write a task which leverages the Zynq Verification IP (and associated API) inside it. I can't figure out how I would implement this in my testbench? I am new to SV, and am guessing that I need to pass the zynq processing system object as an argument so I can access it's API inside my super-task.
Updated example of what I'm trying to do in my testbench. I realize this isn't proper SystemVerilog, it's just to demonstrate the functionality I'm trying to obtain. TOP is a module defined in some other .sv file that contains the definition of a task called T:
module tb();
TOP TIPTOP(), TIPITTYTOP();
task myTask(input TOP T);
begin
T.T;
end
endtask
initial begin
myTask(TIPTOP);
myTask(TIPITTYTOP);
end
endmodule
Another answer to the updated question
This can only be done if the module TOP is not a module but a different flavour of module, called an interface. There is a special kind of SystemVerilog variable called a virtual interface which is a variable that can store a reference to the instance of an interface. This is what you need here. So,
you need to make TOP an interface and
you need to add the keyword virtual to your task: task myTask(input virtual TOP T);
There are restrictions on an interface, however. (We are not quite using it for its normal purpose here.) The main one which might affect you is that you cannot instantiate a module inside an interface.
https://www.edaplayground.com/x/SM33
interface TOP;
task T;
$display("TOP.T");
endtask
endinterface
module tb();
TOP TIPTOP(), TIPITTYTOP();
task myTask(input virtual TOP T);
begin
T.T;
end
endtask
initial begin
myTask(TIPTOP);
myTask(TIPITTYTOP);
end
endmodule
You can call a task or function that is declared in another module. The code below has the following structure:
P
TOP ANOTHER_TOP
| |
BOT bot BOT bot
Package P and all modules have a task T declared in them. I can call all of the tasks from module TOP:
I can call the task in the package P using the score resolution operator, :::
P::T;
I can call the local task:
T;
I can call the task in the instance bot of BOT:
bot.T;
I can call the task in the other top-level module, ANOTHER_TOP:
ANOTHER_TOP.T;
I can call the task in the instance bot of BOT in the other top-level module:
ANOTHER_TOP.bot.T;
Notice how I have declared the various a tasks and modules "in the wrong order". This is OK, because Verilog takes 3 passes to compile and the relationships between the various tasks and modules are sorted out in passes 2 and 3. The package, however, has to be compiled first. This is because packages are a bit of an after-thought in the grand scheme of things.
https://www.edaplayground.com/x/KpJR
package P;
task T;
$display("P::T");
endtask
endpackage
module TOP;
initial
begin
T;
ANOTHER_TOP.T;
bot.T;
ANOTHER_TOP.bot.T;
P::T;
end
task T;
$display("TOP.T");
endtask
BOT bot ();
endmodule
module ANOTHER_TOP;
task T;
$display("ANOTHER_TOP.T");
endtask
BOT bot ();
endmodule
module BOT;
task T;
$display("BOT.T");
endtask
endmodule
Inside the build function of my env class trying to connect interfaces:
virtual my_if my_vif;
for (int i = 0; i<32; i++) begin
_agent[i]._vif = my_vif._if[i];
end
Inside my_if:
interface my_if();
if _if[32]();
endinterface :my_if
When running simulation I get this error:
Error-[MFNF] Member not found
my_env.sv, 229 "this.my_vif."
Could not find member '_if' in interface 'my_if', at "my_if.sv", 1.
_if is also an interface with the next signals:
interface if();
logic clk;
logic rstn;
logic [101:0] requests;
logic [63:0] dataOut;
endinterface :if
The thing is that everything worked fine when the interface _if was not wrapped under the my_if interface.
It looks like you are trying to instantiate some design element of type _if 32 times in the interface, using array instances.
array instances are related to generate blocks. As a result the only way to index them is to do it from another generate blocks. You cannot use a simple for loop with indexing to do it.
Generate blocks are a part of rtl design and in general have very limited applicability to the test bench, in particular with dynamic structs as an agent.
You need to re-think the way you want to implement it.
Is there a way to dump a memory content of a memory that is used in a class instead of using $writememh() system task? I want to check for memory content in the predictor class of a UVM environment used in a UVM scoreboard. I am using 2016 Synopsys VCS simulation vendor
E.g. say I have a predictor:
class my_predictor extends uvm_subscriber#(my_item);
`uvm_component utils(my_predictor)
logic [15:0] mem [512]; // Want to observe the change of this mem content
...
/* local logics, constructor, phase(s) and write function definitions */
...
endclass : my_predictor
I have tried the following:
task run_phase(uvm_phase phase);
forever begin
$vcdplusmemon(mem);
endtask : run_phase
But I was unable to observe it even when I load vcdplus.vpd. I then setup for simv in DVE and then ran. When I tried to view the waveform of my_predictor.mem, DVE crashed.
My motivation for doing this is to compare my predicted change of memory content with the actual memory content of the DUT I am testing via a constrained randomized test.
Is it a vendor tool feature limitation?
I was reading UVM cookbook and I got confused about virtual interface connection in between monitor, driver and their BFM. Does it mean there could be multiple driver or monitor, or this is independent of interfacing that does not know either its monitor or driver. Can anybody help?
The keyword virtual is re-used a number of times in SystemVerilog. The interface is virtual in the sense that its hierarchical path is set at runtime by passing it through a variable. All other connections in Verilog/SystemVerilog are fixed paths.
This does indeed allow you to have multiple instances of the same driver code connect to multiple interface instances. It also helps in block-to-system reuse so you can change the hierarchical path as the interface gets deeper into your system level.
Verilog was not created as a programming langue, more over, it was not suitable for object oriented programming. On the other hand, System verilog test bench language was created as an object oriented programming language.
One of the issues is to semantically connect HDL verilog with TB. All verilog HDL/RTL objects are statically compiled and cannot be manipulated dynamically (which is needed at TB). You cannot get pointers to modules, variables, ... (well except through some back-door PLI mechanism).
So, System verilog came up with the interface construct, which was intended as a connectivity object in RTL world. It is similar to a module in a sense, that it is a compile-time static object. But, SV also added a trick, which allows you to have a reference to an interface. The trick is called virtual interface.
From the point of view of a programmer, you can think of it as a reference, or a pointer to the static interface object. This gives you an ability to pass this reference to different TB class, or create multiple references tot he same interface.
Here is a schematic example:
class Transaction;
virtual dut_if trans; // <<< virtual interface
function new(virtual dut_if t);
trans = t; // <<<< assign it to trans
endfunction // new
endclass // Transaction
// definition of the interface
interface dut_if import trans_pkg::*;
(input trans_t trans);
endinterface
// instantiate the interface in module 'rtl'
bind rtl dut_if dut_if(data);
program tb;
// pass it to the constructor as a virtual interface pointer.
Transaction trans1 = new (rtl.dut_if);
Transaction trans2 = new (rtl.dut_if);
endprogram // tb
I want to write some tasks in a package, and then import that package in some files that use those tasks.
One of these tasks toggles a reset signal. The task is reset_board and the code is as follows:
package tb_pkg;
task reset_board(output logic rst);
rst <= 1'b0;
#1;
rst <= 1'b1;
#1;
rst <= 1'b0;
#1;
endtask
endpackage
However, if I understand this correctly, outputs are only assigned at the end of execution, so in this case, the rst signal will just get set to 0 at the end of the task's execution, which is obviously not what I want.
If this task were declared locally in the module in which it is used, I could refer to the rst signal directly (since it is declared in the module). However, this would not allow me to put the task in a separate package. I could put the task in a file and then `include it in the module, but I'm trying to avoid the nasty complications that come with the way SystemVerilog handles includes (long-story-short, it doesn't work the way C does).
So, is there any way that the task can drive an output with different values across the duration of its execution without it having to refer to a global variable?
A quick solution is to use a ref that passes the task argument by reference instead of an output argument that is copied after returning from the task.
task reset_board(ref logic rst);
There are a few drawbacks of doing it this way. You can only pass variables of matching types by reference, so when you call reset_board(*signal*), signal cannot be a wire. Another problem is you cannot use an NBA <= to assign a variable passed by reference, you must use a blocking assignment =. This is because you are allowed to pass automatic variables by reference to a task, but automatic variable are not allowed to be assigned by NBAs. There is no way for the task to check the storage type of the argument passed to it.
Standard methodologies like the UVM recommend using virtual interfaces or abstract classes to create these kinds of connections from the testbench to the DUT. See my DVCon paper for more information.