A High-Level Synthesis Scheduling and Binding Heuristic for FPGA Fault Tolerance

Table 1

Benchmark summary.

Benchmark	Description	Number of nodes	Number of edges	Operation types
conv5x5	5x5 convolution kernel.	49	190	Add, Mult
conv9x9	9x9 convolution kernel.	161	932	Add, Mult
fft8	8-point radix-2 DIT FFT based on butterfly architecture.	32	116	Add, Sub, Mult
fft16	16-point radix-2 DIT FFT based on butterfly architecture.	88	648	Add, Sub, Mult
fftrad4	4-point radix-4 FFT based on dragonfly architecture. Efficiently decomposing complex operations into real operations.	40	232	Add, Sub, Mult
linsor	Single iteration kernel to solve a system of linear equations () in 5 variables, using successive-overrelaxation (SOR) method.	64	2657	Add, Sub, Mult, Div
linjacobi	Single iteration kernel to solve a system of linear equations in 5 variables, using Jacobi method.	60	230	Add, Sub, Mult, Div
lapsor	Single iteration kernel to solve Laplace’s equation (), using successive-overrelaxation (SOR) method.	7	16	Add, Sub, Mult
lapjacobi	Single iteration kernel to solve Laplace’s equation using Jacobi method.	5	9	Add, Sub, Mult
small0–7	8 randomly generated DFGs with 20 operations.	4–19	2–36	Add, Mult
medium0–4	5 randomly generated DFGs with 130 operations.	88–127	346–642	Add, Sub, Mult, Div