If you decide to use randomly generated data for benchmark in step 3, you could skip that step and run the following command with given vector size N, for example, 1024 and maximum number of iterations for the solver, for example, 100.
$ make run TARGET=hw PLATFORM=xilinx_u50_gen3x16_xdma_201920_3 N=1024 maxIter=100 deviceID=0
The following table lists the configurable parameters with the make command for the benchmark.
Parameter Name | Default Value | Notes |
---|---|---|
N | 1024 | Vector size, must be multiple of 16 |
maxIter | 100 | Maximum No. iterations for the solver <= 2000 |
tol | 1e-12 | Fault Tolerance |
deviceID | 0 | Alveo U50 Card ID |
condition_num | 128 | Conditioner number for matrix generated |