This utility has two operating modes:

1. Equivalence Checking (enabled with the flag --check): designed to check the equivalence of two given circuits (in Verilog) based on the provided error threshold; if they are not equivalend, it will trigger an "ET beached!!!" message, otherwise, a "TEST -> PASS" is shown.
2. Difference Evaluation (enabled with the flag --evaluate): designed to evaluate the error of a given circuit (in Verilog). The output is "error=[errorValue]"

Note: the error criteria can be computed based on different metrics such as Worst-Absolute Error, Mean Error Distance (MED), Mean-Squared Error Distance (MSED, Error Rate (ER), and Mean-Relative Error Distance (MRED)

• Install the following tools:

1. Linux
2. Yosys: link (https://github.com/YosysHQ/yosys)
3. Icarus Verilog: link (https://github.com/steveicarus/iverilog)

Note: add the binaries of 2 and 3 to your PATH

• ./input/: contains the input circuits to the tool.
• ./input/exact/: contains the exact circuits of our benchmark suite
• ./input/test/: several approximate circuits generated by MECALS as a simple test
• ./report/: in case of a breach to ET, a .txt file is dumped into this directory that includes the input combinations for which a breach of ET happend.
• ./temp: all the intermediary files such as yosys, iverilog, and others are stored here.

To run the checker, use the following syntax:

• --input, -i: Specify the path to the circuits to be checked. This option can be used multiple times to input multiple circuits.
• --input_port_orders, -ipo: Define the order of the input ports. This can be one of the predefined order types. The default is ['1', '1'].
• --output_port_orders, -opo: Define the order of the output ports similar to the input port orders. The default is ['1', '1'].
• '--error_threshold', -et: Set the error threshold, which is an integer value. The default is -1.

verifier.py accepts the following command-line arguments to control its behavior and specify the parameters for the equivalence checking process:

• --input, -i:

  • Description: Specify the path(s) to the circuit file(s) that you want to check for equivalence. You can use this option multiple times if you need to input multiple circuit files.
  • Usage Example: -i path/to/circuit1.v -i path/to/circuit2.v

• --input_port_orders, -ipo:

  • Description: Define the order of the input ports. This determines how the input ports of the circuits are arranged and compared during the checking process.
  • Choices: 11, 12, 21, 22
  • Default: ['1', '1']
  • Usage Example: -ipo 21

• --output_port_orders, -opo:

  • Description: Define the order of the output ports, similar to the input port orders. This affects how the output ports are arranged and compared.
  • Choices: 11, 12, 21, 22
  • Default: ['1', '1']
  • Usage Example: -opo 12

• --error_threshold, -et:

  • Description: Set the error threshold, a float/int value that defines the acceptable error limit for the equivalence checking.
  • Default: -1
  • Usage Example: -et 2.5

• --metric_type, -t:

  • Description: Specify the metric type to use for evaluating the circuits' equivalence.
  • Choices: wae (Worst Average Error), med (Mean Error Distance), msed (Mean Squared Error Distance), er (Error Rate), mred (Maximum Relative Error Distance)
  • Default: wae
  • Usage Example: -t msed

• --check:

  • Description: Checks whether the two input Verilog circuits are equivalent within a specified error threshold. This mode is used to verify that the differences between the circuits fall within an acceptable range defined by the --error_threshold parameter.
  • Usage Example: python3 checker.py -i path/to/circuit1.v -i path/to/circuit2.v -et 2 -t wae --check

• --evaluate:

  • Description: Computes and reports the error between the two input circuits using the specified error metric. Available metrics are wae (Weighted Average Error), med (Median Error), msed (Mean Squared Error), mred (Maximum Relative Error Difference), and er (Error Rate). This mode is used for detailed analysis of how and where the circuits differ.
  • Usage Example: python3 checker.py -i path/to/circuit1.v -i path/to/circuit2.v -et 0.5 -t med --evaluate

Assuming that an n-bit binary number X is represented as X = x_n-1...x₀, Port orders define how the circuit ports are arranged in the checking process:

• Input Order Types:

  • 1: [n:0]a, [n:0]b => circuit(a₀, a₁, ..., a_n, b₀, b₁, ..., b_n, [outputs])
  • 2: [n:0]a, [n:0]b => circuit(a_n, a_n-1, ..., a₀, b_n, b_n-1, ..., b₀, [outputs])

• Output Order Types:

  • 1: [n:0]y, [n:0]z => circuit([inputs], y₀, y₁, ..., y_n, z₀, z₁, ..., z_n)
  • 2: [n:0]y, [n:0]z => circuit([inputs], y_n, y_n-1, ..., y₀, z_n, z_n-1, ..., z₀)

The order types are specified using numbers, where '1' and '2' define the specific arrangement of circuit pins.

• Note: all exact files follow the order number 1 for both inputs and outputs

Here's an example command to run the verifier with specific input and output port orders:

• In this example, the approximate circuit is located in ./input/test/ directory and its name is adder_i4_o3_wce1 (that supposed to have a WAE of 1). The exact circuit is located in ./input/exact/ directory and its name is adder_i4_o3.v.
• The -i argument is used twice to specify two circuit files: adder_i4_o3_wce1.v and adder_i4_o3.v.
• The -ipo and -opo arguments set the input and output port orders respectively. Here, an -ipo 21 (or -opo 21) means that the input order (output order) type of the first circuit is 2 and the second file is 1.
• The error threshold set is 1.

$ python3 checker_blahBlah.py -i  input/test/adder_i4_o3_wce1.v -i input/exact/adder_i4_o3.v -ipo 21 -opo 21 -et 1 -t wae --check

Upon running this command, you should get the following output:

args = An object of class Arguments:
self.circuits = ['input/test/adder_i4_o3_wce2.v', 'input/exact/adder_i4_o3.v']
self.input_port_orders = ['1', '2']
self.output_port_orders = ['1', '2']
self.metric_type = <function wae at 0x7fc0add399d0>
self.check = True
self.evaluate = False
self.et = 1.0

Exhaustive Simulation
Error Breached!!!

This means that at least, given this specific order, this circuit does not fulfill the error threshold it claims to have.

In this example, we will compute the error between two Verilog circuits using the Mean Squared Error (MSE) metric.

• Assume the approximate circuit is located in ./input/test/ directory and named adder_i4_o3_wce2 (which has a different error characteristic).
• The exact circuit is in the ./input/exact/ directory, named adder_i4_o3.v.
• We use the -i argument twice to specify the two circuit files: adder_i4_o3_wce2.v and adder_i4_o3.v.
• The -ipo and -opo arguments set the input and output port orders respectively. In this case, -ipo 12 and -opo 12 set the port orders for both circuits respectively.
• We choose the Mean Squared Error (msed) as the metric type to evaluate the error between the circuits.

Here's the command to run the verifier in evaluate mode:

$ python3 checker_blahBlah.py -i input/test/adder_i4_o3_wce2.v -i input/exact/adder_i4_o3.v -ipo 12 -opo 12 -t msed --evaluate

Upon running this command, the output might look like:

args = An object of class Arguments:
self.circuits = ['input/test/adder_i4_o3_wce2.v', 'input/exact/adder_i4_o3.v']
self.input_port_orders = ['2', '1']
self.output_port_orders = ['2', '1']
self.metric_type = <function med at 0x7f8890272a60>
self.check = False
self.evaluate = True
self.et = -1

Exhaustive Simulation
error = 0.75

This output indicates the Mean Error Distance between the two specified circuits is