specman e-language supports constraint random generation

specman e-language supports the concept of constrained random generation.

. Constraints are applied on struct members e.g Fields and methods.

· Constraints are Boolean equations

· Constraints are declarative statements

Specman e-language categories the constrained mechansim as follows :

1) Random Generation without any constraints parameters
2) Directed Random constrained Generation [ e.g from .. to ]
3) Directed constrained Generation

For constraining a Field use the following syntax :
keep Boolean-expression

Ex: struct packet_s {

Length : uint(bits:6);

Address : uint(bits:2);

// application of Directed Constraints Parameter for Length field

keep Length = = 10;

// application of Directed Random Constraints Parameter for Address field

keep Address in [0..2];
};




For constraining elements in a list (array):

keep for each (item) in list_name {

Boolean-expression; };


Ex: Struct packet_driver_s {

Packets : list of packet_s;

// Constraints list for the individual fields

keep for each (pkt) in packets {

pkt.len < 10; }; };



The syntax for implication constraints are given below:

keep Boolean-expr1 => Boolean-expr2;

Ex:

keep size = = SHORT => length < 10;

keep size = = LONG => length > 20;

Or, suppose if we want to constraint using lists:

Ex:

keep for each (pkt) in packet {

index = = 0 => pkt.addr = = 1 and pkt.kind = = GOOD;

index = = 1 => pkt.addr = = 2 and pkt.kind = = BAD; };


Weighted Constraints:

Weighted constraints are used on a specific application. It allows selection weight for value or range of values. In order to achieve a weighted constraint we need to apply a soft control on the distribution of generated values.

syntax: keep soft gen-item = = select {weight: value ;};

Ex: <’

keep soft length = = select {

20:14;

25:10;

10:20; };

‘>

specman e-language usage examples for extending methods

Here i am listing examples of extending methods using specman e-language concept.

Ex : extending using is_also

---> Filename packet.e
<’

struct packet_s {

my_method () is {

out (“I am here”);
};
};

‘>

-------> new file packet_extend.e

<’

import packet.e;

extend packet_s {

my_method () is also {

out("to learn e-language”);
};
};

‘>

when we execute the packet_extend.e we can see finally the output appearing as "I am here to learn e-language".


Ex : extend using is first

---> Filename packet.e

<’

struct packet_s {

my_method () is {

out (“I am here”);
};
};

'>

---> New file : packet_extend.e
<’

import packet_extend.e;

extend packet_s {

my_method () is first {

out (“to learn e-language”);
};
};

'>

when we execute the packet_extend.e we can see finally the output appearing as "to learn e-language i am here".


Ex: extend using is only

---> Filename packet.e

<’

struct packet_s {

my_method () is {

out (“I am here”); }; };

'>

---> New file packet_extend.e

<’

import packet.e;

extend transaction_s {

my_method () is only {

out (“to learn e-language”);
};
};

'>

when we execute the packet_extend.e we can see finally the output appearing as "to learn e-language ", this extend method operator overrules the previous contents of the same method defined.

How to extend methods in e-language

Specman e-language has the built in features to extend the predefined and
user-defined methods in order to have new contents.
The extension can be done in the same file where the original method is declared or in another new file.
Specman e-language supports the following features for extending the methods

a) is first => which means add before the existing e code
b) is also => which means add after the existing e code
c) is only => which means override/replace the existing e code

what are different output actions in e-language

e-language supports both formatted and non-formatted print statements, this is similar to any other high-level programming language.

print -> print data, useful at the command line and for struct printing.

Outf() -> it is a predefined method available in Specman e-language. This helps in printing the statements in a formatted manner.

Out() -> it is predefined method available in Specman e-language. This helps in printing the statements without formatted manner.

How to use methods using return values

The below code describes how to use methods using the return values
Ex: file_name : packet.e

<’

struct packet_s {

length: uint;

legal: bool;

legal_value (len: uint): bool is {

If (length >= len) then {

Result = TRUE;
};

If (result = FALSE) then {

legal = FALSE;
};
};

length_calc() is {

var length_incr : uint := 0;

if (length == 0) then {

length_incr = length_incr + 1;
};

// call the Legal_value method

legal_value(length_incr);
};

run() is also {

length_calc() ; // Need to call this method as it is calling other method

};
};

‘>

How to use methods using local struct fields

The below example shows how to use local struct fields for writing method using e-language.

Ex : file_name : packet.e

<’

struct packet_s {

length: uint (bits; 6);

legal_value(): bool is {

if (length >= 64) then {

return (TRUE);
};
}; // end of method

// extend the pre-defined method to call the local method

run() is also {

// call the method declared

legal_value();

}; // end of run() method
};

‘>

Procedural Control statements in e-language

e-language is high-level programming language, it supports all Procedural flow control statements. The defined flow control statements are:

· If-else statement

· Case statement

· For loop

· While loop

· Repeat-Until loop

Ex: If statement execution

<’

struct packet_s {

Flag: bool;

Check_flag (): bool is {

If (flag = TRUE) {
Out (“flag is true”);
}
else {
Out (“flag is false”);
}; // end of if cond
}; // end of method

run() is also {

Check_flag(); // calling the method declared
};
}; // end of declared struct

‘>


Ex: Case Statement Execution

<’
struct packet_s {

Flag: bool;

Check_flag (): bool is {

Case flag {

TRUE: {out (“flag is true”)};

FALSE: {out (“flag is false”)};
}; // end of case statement
}; // end of method

run() is also {

Check_flag();
};

}; // end of struct

‘>

Ex: For loop condition execution
<’

struct packet_d {

! Pkt: Packet_s; // on-the fly generation

No_of_Pkt: uint (bits: 4);

Keep No_of_Pkt in [1.. 10];

Packet_gen () is {

For i from 0 to (no_of_Pkt-1) {

gen Pkt; }; // on_the_fly gen (gen is specman keyword)

}; // end of method


run() is also {

Packet_gen();
};
}; // end of struct

‘>

writing procedural code using e-language

Procedural Code forms one of the important features of the e-code.
Procedural Code is executed by struct / unit members called Methods (e.g similar to functions in c-language).

Ex: <’

struct packet_s {

Addr: uint (bits: 2);

my_address (addr: uint): bool is {

if (Addr == 0) {
result= TRUE;
}

else {
result= FALSE;
}; // end of if condition
}; // end of method declaration

// calling my_address method

run() is also {
my_address();
}; // end of run() method

}; // end of struct declaration

‘>

The above declared method has a return type of bool in nature and it cannot be executed until it is called.
In order to make it happen, the declared method is called in the pre-defined specman method run().
Since the run() is already a pre-defined method present in specman elite, we need to extend the method in order to call the my_address method.

Important Features of Methods:

i) Methods can contain actions (i.e. calculating parity, factorial...),

ii) Methods can Declare and use Local Variables,

iii) Methods can have Zero to Fourteen input arguments/parameters,

iv) Methods can optionally return a value.

v) Methods Optionally Consume time if declared to be a “Time-Consuming Methods (TCM)”, .

Types of Stimulus Generation in e-language

Stimulus needs to be generated and driven to the DUT in order verification process.

The Specman Elite e-language supports two types of stimulus generation :

i) Pre-Run Generation

ii) On-the-Fly Generation

Pre-Run Generates fields or struct once the Specman Test command is issued and sends them to the DUT with the help of BFM. The main advantage of this process is, it quickly shows what the generator will produce but consumes lot of memory in order to save all the generated values.
Pre-Run generation requires an understanding of the interface between struct definition and the Specman Elite Constraint Solver.

Ex: top file name : packet_top.e
<’

import packet.e;

extend sys{

Legal: bool;

Packet: packet_s;// Instantiate of struct

};

‘>



On-The–Fly Generation Generates items “on-the-fly” at the time they are driven into the DUT. On-the-Fly generation is achieved by using the “!” (Bang operator) Before the field, so that initially the field will be empty.
The main advantage of this mechansim is that it saves memory and the generation can be activated based on the DUT interaction.

Ex: top file name : packet_top.e
<’

import packet.e;

extend sys {

! Cur_packet: packet_s;

! Packet_count: uint;

};

‘>

** sys is a Specman Elite pre-defined struct

extend feature in e-language

e-language supports extensibility feature in order to add new elements in the declared struct or unit. This feature is very useful in writing Top level Testcases using the constraint parameters.
To understand more, we will look at a small example listed below :

Ex: The e-description for the Packet structure

file name : packet.e

<’ struct packet_s {

Addr: uint (bits: 2);

Len: uint (bits: 6);

Data: list of byte; };

‘>

Extension file name : packet_ext.e

<’ import packet.e;

extend packet_s {

Legal: bool;

Parity: byte; };

‘>

The Legal and Parity fields will be added into the packet structure based on the order of generation if the packet_ext.e is executed.

import, extend are the pre-defined keywords of Specman Elite e-language.

Creation of Stimulus using e-language

In order to start verification process, we need to generate the input stimulus defined by protocol for the given DUT (Device under Test) either using a random approach or using a constrained random. To generate the required information we need to declare the struct followed by the corresponding fields. Each field should be of Specific type either predefined or user-defined (e.g enumerated types). To generate the defined fields we make use of Specman elite generator to generate each field with a random value unless it is not constrained. Let’s consider a Network packet having data, length and address, which are to be generated randomly.

Ex : Let's look at the Network Packet Structure definition
file name : packet.e

<’
struct packet_s {

Addr: uint (bits: 2);
Len: uint (bits: 6);
Data: list of byte;

};
‘>


once the struct is defined, it needs to be Instantiated at the specman top level for generation.

Ex : file name top.e

<'
import Packet.e ;

extend sys {

packet : packet_s; // Instantiate the packet struct

};

‘>

How to write an e-program block

The e program block is indicated by the keyword struct.

The Program Block contains the following format.

struct struct_name {

--- Field declaration;

--- Methods with /without return type;

};

The main e-Program block is where:

· Fields are declared inside the struct_name

· Variables are declared inside the methods

· Executable statements are carried out

· Calls to methods with arguments are made

Basic syntax of e-language

Specman e-language is case sensitive in nature for both the predefined syntax keywords and user-defined names.
The name my_struct_s and my_STRUCT_S are both different structs defined in the e-language.

struct my_struct_s {
--- Action to be performed
}

struct my_STRUCT_S {
---- Action to be performed
}

Here struct is a specman e-language keyword.

The important aspect of the e-language is that the e-code written should be within the delimiters (e.g <' '>) and the code written outside the delimiters are considered as comments. In order to have single inline comments within the delimiters we can have both the VHDL / Verilog format ( e.g -- or // ) .

Ex :
The below code shows a sample e-language code structure
<'
struct my_packet_s {
// Field Declaration ( single line comment)
Addr : byte;
Data : byte;
Len : uint (bits:6);
method_name() is {
--- Action required
}
}; // end of struct my_packet_s

'>

Know more about what e-language supports

e-language can be distinguished as Programming Language ( like any other programming languages e.g C, C++) and as well Simulation Language ( Simulation related concept).
e-language supports the lexical elements and also some of the basic Programming Concepts (e.g data types).
It supports the following sections.
1) Lexical Elements
2) Data types and Variable Declaration
3) Arrays ( one-dimensional)
4) Enumerated Types
5) Operators
6) Variable assignment

Introduction to specman e-language

Specman e verification language is an HVL (Hardware Verification Language) Language and an IEEE standard language used by most of the verification engineers for Block/SOC Verification. It is one of most advanced language in terms of methodology and technology for building the Verification environment.
Specman e-language has a unique concept called Aspect-oriented Programming, that enables easier reuse and provide "extendibility" features for derivative environments and follows the eRM Methodology.
According to the eRM (e Reusable Methodology) the Verification environment is designed to have a Stimulus Generator, Driver, BFM (Bus Functional Module) and to intergrate all these elements it has an Agent which could be Transmitter (Tx) or Receiver (RX).
The Top Verification Environment categories into Active and Passive Elements. The Active Elements consists of Stimulus Generator, Driver and BFM whereas the Passive Elements Consists of Monitors (e.g Checkers, Signal Map and Functional Coverage). These Elements can be Enabled / Disabled by built in switch depending on the requirement. The Functional Coverage is one of the important element in the Specman e-verification which gives the verification engineer to cover all the important functionality in terms of coverage elements.