ECE 572 AVDVANCED LOGIC SYNTHESIS PROJECTS. FALL 2006

PROJECT 1.


CUBE CALCULUS TOOLKIT IN MATLAB FOR DYDACTIC APPLICATIONS.
Students: Nekkalapu Satyanarayana, Jothi Komalan, Thankappan Achary Retnamma Renjith, Kusugal Vishwanath Arun.
Subject: The subject of this project is to design a simple to use toolkit in Matlab for classical binary cube calculus. The functions should be displayed as Karnaugh Maps.
Materials:
Deliverables:
Students should demonstrate a working program that will use all logic operations on arrays of cubes and prime implicant generation using them. Demonstrate operating code in class on practical examples. Explain how it works. Write user manual and detailed documentation. The user manual should be oriented towards future students of ECE 572 class. It should show examples of applications in logic synthesis that the toolkit can be used for.



PROJECT 2.

P
RESEARCH PROJECT: MMD FOR DON'CARES.
Students: Metzger Natalie, Kumar Manjith, Iyer Balachandar, Wang Ying
Subject: Modify the original MMD algorithm for synthesis of reversible functions to be able to synthesize reversible functions with don't cares. Take care to keep the function being reversible when you replace dashes on outputs with cares 0 and 1. For instance if you change output vector 000- to 0000 you should change another output vector 000- down the truth table to 0001. Try to find a smart strategy for MMD, better than it uses now, how the Toffoli gates are created and in what order. The algorithm may be slower than MMD, but it should find correct and hopefully better solutions.
Materials:
Deliverables:
Students should demonstrate working MMD program to demonstrate using it with various percentages of don't cares, from very low percentag to very high. Data structure in MMD uses only zeros and ones for function. You will have to use the third symbol for don't cares. You will have to verify that your generated functions are always reversible, even for a very high percentage of don't cares. You can use MMD software for this. Repeating your program with different assignments of don't cares should lead to better results. You can test probabilistic and backtracking strategies of search.

PROJECT 3.


RESEARCH PROJECT: MMD WITH IMPLICIT REPRESENTATION.
Students: Armagost Jesse, Hyunh Noda, Lafond Kenneth
Subject: MMD is limited by its representation of reversible function as a truth table. We will improve the MMD by using some better representation. Consider using (A) cube calculus, (B) BDDs, (D) quantum decision diagrams of Miller and Thornton.
Do not change MMD algorithm, just find a way of using implicit representation of the truth table to generate only one line of truth table at a time, to generate a new product for Toffoli gates in selected columns in which input and output values in the truth table differ. Your method should also check if the input - output line for given input line is in good (Natural) MMD order. (MMD makes transformations to have all lines from top of good order).
Materials:
Deliverables:
You should create a program that will allow to solve the same problems as MMD and also problems of much higher size, even if your program will be slower on large examples. Functions of 19 variables should be minimum.

PROJECT 4.


RESEARCH PROJECT: NEW IDEAS IN A MMD-LIKE ALGORITHM FOR REVERSIBLE LOGIC MMD WITH IMPLICIT REPRESENTATION.
Students: Ibrahim Othman
Subject:
The task of this project is to investigate generalizations of MMD in the following directions: (a) using gates that are more complicated than Toffoli, (b) converting from non-reversible to reversible with minimum number of bits. (c) synthesis with only some output bits in the reversible functions carrying information and other output bits treated as garbage (they are needed for reversibility, but their values are irrelevant).
Materials:
Deliverables:
Depending on complexity of tasks, all or some of the above points should be programmed and demonstrated in class. All theory is basically done, but the student can improve it or find some new ways of solving problems.

PROJECT 5.


RESEARCH PROJECT: COMPARISON OF COSTS OF STANDARD CANONICAL AND/EXOR FORMS AND NEW LINEARLY INDEPENDENT FORMS.
Students: Keshavamurthy Savya Saachi, Bali.
Subject:
The students are familiar with Fixed Polarity Reed Muller Forms. The set of coefficients in them can be calculated by multiplying the matrix of the transform by the binary vector of minterms. By going through all such matrices and finding the vector of coefficients with the smallest number of non-zero coefficients the best circuit is found. As we remember, FPRM forms have standard trivial functions that are products of variables. It was shown that many forms exist where the standard trivial functions use other functions than AND. The project is to compare the number of terms for old and new expansions (matrices).
Materials:
Deliverables:
Students should demonstrate software that will create transform matrices of new transforms. Next the comparison of spectral coefficient costs for a number of benchmark functions should be done. Good report is expected in addition to software and experimental results. Programming is not difficult and can be done in Matlab.