Constraint

Question

Hi, I run this code in QuestaSim but it has an error. Can anyone help me why this randomization failed?

class Packet;
    rand bit p[2048];
    int sum=0;
    function int total(bit p[2048]);
        for (int i=0;i<2048;i++)
            sum = sum + p[i];
    endfunction
    constraint c_stim{
        total(p)==256;
    }
endclass:Packet

Answer 1

dave_59

Forum Moderator

10890 posts

November 03, 2022 at 10:40 am

In reply to Moein75:

There are several problems with this code.

Calling user defined functions in a constraint break the solution space in that the random variables as inputs to the function must choose their values first. Then the return value of the function becomes a non-random state variable. The constraint solver is not required to keep trying different inputs to make sure its return value satisfies the remaining constraint (==256).

Also functions used in constraint must be "pure" they cannot reference or modify variables outside their scope. Imagine the behavior if there were multiple calls to your function total.

And finally, the LRM does not allow references to random variables that are aggregate arrays without using an iterative construct like foreach or sum reduction method that can be unrolled into a set of boolean equations.

— Dave Rich, Verification Architect, Siemens EDA

Answer 2

Shubhabrata

Forum Access

102 posts

November 03, 2022 at 11:14 am

In reply to Moein75:

Hi, Observed your code from top to bottom. There are loads of things to say and there are few speculations of mine.
1)You have called a user-defined method which will be acting after the randomization method and will return a value. That's why you might get the right result but most of the time, you will get the wrong one.
2)It's the dynamic array you are using. So in order to get the proper answer you have to use system defined method (array_manipulation). You gotta use p.sum().
3)Now return type of the sum will be depending upon the variable type you are dealing with. Here variable type is a bit. So you will get either 1 or 0.
4)Now finally, to get rid of all of these hindrances, you have to use the int variable. I tried to do so in the following code.

class Packet;
  rand int p[20];
 constraint values{
    foreach(p[i])
      p[i] inside {[0:1]};
  }
    constraint c_stim{
      p.sum()==13;
    }
endclass:Packet
 
module test;
 
  Packet P=new;
  initial begin
    assert(P.randomize)
    $display("%p",P.p);
  end
 
endmodule
 
 
// Another way
class Packet;
  rand bit[3:0][7:0] p[20];
  rand bit[31:0]p1[20];
 //rand int unsigned p1[20]; // by-defualt signed
 constraint values{
    foreach(p[i])
      p[i] inside {[0:1024]};
  }
    constraint c_stim{
      p.sum()>1352;
    }
  constraint eq{p[5] == p1[5];p[5]<=105 && p[5] >=99;} //extra ; some observation ;
 
endclass:Packet
 
module test;
 
  Packet P=new;
  initial begin
    assert(P.randomize)
    $display("%p",P.p);
    $display("%0p",P.p1);
  end
 
endmodule

Now go through it gingerly.
I have one tiny little problem here. During declaring the array, instead of using rand int along with constraint values, when I was using rand unsigned int along with no such values constraint, I was getting garbage value. WHY?

Answer 3

Moein75

Full Access

15 posts

November 04, 2022 at 9:05 am

Good clarification, thank you

Answer 4

dave_59

Forum Moderator

10890 posts

November 04, 2022 at 9:21 am

In reply to Shubhabrata:

If you remove the constraint values, the constraint solver is giving you back values that satisfy the other active constraints. You need to plug in the "garbage values" into the constraint equation and see that it does in fact satisfy those constraints.

— Dave Rich, Verification Architect, Siemens EDA

Answer 5

Shubhabrata

Forum Access

102 posts

November 05, 2022 at 12:14 pm

In reply to dave_59:

Hello sir, I have clarity of that. But that will happen in case of signed int (rand int p[20]) . In case of unsigned int declaration (rand int unsigned p[20]) , it's giving garbage values which is not satisfying the constraint (2nd one) .
I should have got series of 0s and 1s.

Answer 6

dave_59

Forum Moderator

10890 posts

November 05, 2022 at 2:51 pm

In reply to Shubhabrata:
Show me exactly the code you ran and the "garbage" values you got.

— Dave Rich, Verification Architect, Siemens EDA

Answer 7

Shubhabrata

Forum Access

102 posts

November 06, 2022 at 1:11 am

In reply to dave_59:

class Packet;
  rand int unsigned p[20];
  constraint c_stim{
      p.sum()==13;
    }
endclass:Packet
 
module test;
 
  Packet P=new;
  initial begin
    assert(P.randomize)
    $display("%p",P.p);
  end
 
endmodule

The result , I am getting -

vsim -voptargs=+acc=npr
# run -all
# '{257152500, 3154634279, 1845763276, 3696871621, 631626809, 2108533005, 2922957038, 2580893384, 1715310603, 828324944, 3022238876, 3785499950, 2835521728, 3653584145, 2354463875, 2745584841, 1287618582, 201913747, 2622534294, 698645476}
# exit

As I am using unsigned , I was supposed to have 1s and 0s . Instead , I am these values.

Answer 8

dave_59

Forum Moderator

10890 posts

November 06, 2022 at 7:45 am

In reply to Shubhabrata:

module top;
 
  int unsigned p[20] =
 
'{257152500, 3154634279, 1845763276, 3696871621, 631626809, 2108533005, 2922957038, 2580893384, 1715310603, 828324944, 3022238876, 3785499950, 2835521728, 3653584145, 2354463875, 2745584841, 1287618582, 201913747, 2622534294, 698645476};
 
  initial begin
    $display("p.sum =                    %16d",p.sum);
    $display("p.sum =                  0x%16h",p.sum);
    $display("p.sum with (64'(item)) = 0x%16h",p.sum() with (64'(item)));
  end
endmodule

— Dave Rich, Verification Architect, Siemens EDA

Answer 9

Shubhabrata

Forum Access

102 posts

November 07, 2022 at 1:02 am

In reply to dave_59:

Ok, Got it. Numbers are made of 32 bits, but the summation consists of more than 32 bits. As we are taking int variable that's why we'll be getting the sum result up to the 32nd bit. And the answer is 13. Hope I got it correct.

Answer 10

UVM_geek

Full Access

18 posts

April 30, 2023 at 6:28 am

Hi Dave,

https://www.edaplayground.com/x/jwjp

Here I am getting unexpected result
p.sum with (64'(item)) = 0x0000000b0000000d
may I know why?

Thanks !!!

Answer 11

dave_59

Forum Moderator

10890 posts

April 30, 2023 at 7:13 am

In reply to UVM_geek:

Shubhabrata wrote:

Ok, Got it. Numbers are made of 32 bits, but the summation consists of more than 32 bits. As we are taking int variable that's why we'll be getting the sum result up to the 32nd bit. And the answer is 13. Hope I got it correct.

That is your answer.

— Dave Rich, Verification Architect, Siemens EDA

Languages & Standards

Methodologies

Techniques & Tools

Universal Verification Methodology (UVM)

Featured Courses

Additional Courses

Formal-Based Techniques

Analog/Mixed Signal

UVM Forum

SystemVerilog Forum

Coverage Forum

Additional Forums

Implementation Patterns

Specification Patterns

Pattern Resources

UVM Cookbook

Coding Guidelines & Deployment

Coverage Cookbook

Featured & On-Demand

On-Demand Library

Recording Archive

Conferences & WRG

Siemens EDA Learning Center

Blog & News

Verification Horizons Publication

About Us

Training

Portfolio

Explore

Siemens

Contact