Method to optimize net lists using simultaneous placement and logic optimization

[0002] Invention relates to design of integrated circuits and more particularly to a method for optimizing net lists using simultaneous placement optimization, logic function optimization and net buffering.

[0003] Prior art methods use a mathematical expression to provide an analytical solution for simultaneous placement of logic functions and logic optimization. The solution of such a mathematical expression would lie in the continuous domain. Furthermore, mapping this solution into a valid placement and a valid implementation of each logic function can lead to loss of optimality.

[0004] Therefore, there is a need for a method that will overcome the deficiencies found in the prior art.

[0005] Invention discloses a method for the design and fabrication of integrated circuits to optimize signal timing and circuit performance.

[0006] The method simultaneously obtains a placement of logic functions, mapping of logic functions onto library elements and buffering of nets connecting the logic functions to optimize signal timing. Logic functions may be moved between bins and or may be optimized to obtain the best cost. A best buffering solution is also computed for each logic function's location and size.

[0007] FIG. 1 shows logic functions placed in bins.

[0008] FIG. 2 shows a critical region of an integrated circuit.

[0009] FIG. 3 shows a list of possible moves for a logic function.

[0010] FIG. 4 shows a list of best moves for each logic function.

[0011] FIG. 5 shows a physical move of a logic function.

[0012] FIG. 6 shows a graph plotting the best cost point for a logic function.

[0013] FIG. 7 shows a flow diagram of the method.

[0014] Preferred embodiment is a method to simultaneously obtain a placement of logic functions wherein a logic function is an electrical circuit including logic gates and memory devices, mapping of logic functions onto library elements and buffering of nets to optimize timing in integrated circuits. Preferred embodiment is based on the Fidducia-Mattheyses optimization algorithm. Other optimization algorithms that are known in the art, such as simulated annealing and genetic algorithms, for example, could be used without deviating from the invention.

[0015] Method 100 consists of the following steps as shown in FIG. 7:

[0016] 1. Compute and store timing for the design 101. A preliminary circuit is created to accomplish desired function of an integrated circuit. Timing information for each of the logic functions in the circuit is computed and stored to provide a base-line for subsequent attempts to reduce signal timing. FIG. 1 shows a possible initial configuration for the design. Possible locations are given by the center of the bins where each bin contains logic functions.

[0017] 2. Initialize move stack with current timing information for the design 102.

[0018] 3. Initialize previously moved set 103. Set contains a historical list of previously moved logic functions.

[0019] 4. Extract a critical region of the integrated circuit 104. Critical region contains a subset of logic functions that do not satisfy a desired timing specification for an integrated circuit. Timing specification of an integrated circuit states when to expect signals at the output of a logic function and when signals arrive at its input. Slack is the difference in time between when a signal is required at the output of a logic function and the time when it actually arrives at the output. Negative slack signifies an undesirable delay in signal propagation through a logic function. Each logic function that has a negative slack is classified as critical. FIG. 2 shows a critical region of an integrated circuit with logic functions and connecting nets.

[0020] 5. Compute cost of moving each logic function 105. By excluding logic functions listed in the previously moved set, a simultaneous operation is performed on each logic function to bring it within the required timing specification. The following simultaneous operation is performed on each critical logic function:

[0021] a) Physically relocate each logic function to improve slack by making moves that optimize the distribution of electrical load along the signal path. The locations of the logic functions are chosen from a set of valid locations such as within a bin, for example. Bins are obtained by dividing the area of the integrated circuit into equal sized regions.

[0022] b) Perform a logic optimization move by changing the mapping of a logic function from one library element to another. A logic optimization move will pick a sized function from the library to achieve optimum timing. A logic function can be implemented by an equivalent plurality of circuit configurations having a plurality of physical sizes. A circuit is implemented by a plurality of semiconductor gates. The library contains elements of equivalent circuit configurations for each logical function.

[0023] c) Determine best buffering solution on nets connected to each logic function. Repeaters are inserted at points along a net where needed to achieve best cost solution. FIG. 3 and FIG. 4 show a set of possible moves in an implementation of the method. The example shows logic functions consisting of individual cells and contemplated moves of the cells to another bin location, changing its function size or both sizing and moving the cells simultaneously.

[0024] 6. Replace critical logic function with logic function computed and relocate it to have the best cost solution 106. FIG. 5 shows an example of a possible replacement. The cell is relocated to a different section of the chip. A larger sized implementation of its logic function has been selected from the library.

[0025] 7. Update the timing information for design 107. The original location and original physical implementation of each moved logic function along with the new timing is pushed onto the move stack.

[0026] 8. Store the moved logic function in the previously moved set 108.

[0027] 9. Repeat steps 4 through 7 until all logic functions in the extracted critical region are stored in the previously moved set 109.

[0028] 10. Sequentially sort through the moves in the move stack to find the best timing during the sequence of moves and note its location in the move stack 110. FIG. 6 shows a graph of the cost of each move in a sequence where the best cost point is shown.

[0029] 11. Revert all moves made after best timing is found using the move stack and the location of the best timing found in step 10110.

[0030] Foregoing described embodiments of the invention are provided as illustrations and descriptions. They are not intended to limit the invention to precise form described.

[0031] In particular, Applicants contemplate that functional implementation of invention described herein may be implemented equivalently in hardware, software, firmware, and/or other available functional components or building blocks.

[0032] Other variations and embodiments are possible in light of above teachings, and it is thus intended that the scope of invention not be limited by this Detailed Description, but rather by Claims following.