FPGA module minimization
Dimitri Kagaris, Spyros Tragoudas
Abstract
Dimitri Kagaris, Spyros Tragoudas
Abstract
We examine the problem of minimizing the number of modules in an FPGA with combinational and sequential modules (like the C-modules and S-modules of the ACT2 and ACTS architectures). The constraint is that a combinational module can be combined with one flip-flop in a single sequential module, only if the combinational module drives no other combinational modules. We show that the problem of rearranging the flip-flops by retiming so as to satisfy prescribed individual bounds on the number of combinational and sequential modules is NP-complete. However for the problem of rearranging the flip-flops by retiming so as to minimize the total number of combinational and sequential modules, we present a quadratic-time algorithm. The algorithm uses a minimum-cost flow formulation and offers a significant time improvement over a previous approach that used a general linear program.
OpenAlex reports 1 citations for this work. Citation counts describe recorded attention and do not establish research quality.
A contribution statement is not available in the OpenAlex record.
Method details are not available in the OpenAlex metadata.
Findings are not separately available in the OpenAlex metadata.
Limitations are not available in the OpenAlex metadata.
Application details are not available in the OpenAlex metadata.
We examine the problem of minimizing the number of modules in an FPGA with combinational and sequential modules (like the C-modules and S-modules of the ACT2 and ACTS architectures). The constraint is that a combinational module can be combined with one flip-flop in a single sequential module, only if the combinational module drives no other combinational modules. We show that the problem of rearranging the flip-flops by retiming so as to satisfy prescribed individual bounds on the number of combinational and sequential modules is NP-complete. However for the problem of rearranging the flip-flops by retiming so as to minimize the total number of combinational and sequential modules, we present a quadratic-time algorithm. The algorithm uses a minimum-cost flow formulation and offers a significant time improvement over a previous approach that used a general linear program.
Key concepts: Retiming, Combinational logic, Sequential logic, Computer science, Field-programmable gate array, Minification, Parallel computing, FLOPS