Novel Dynamic Partial Reconfiguration Implementation of K-Means Clustering on FPGAs: Comparative Results with GPPs and GPUs

Table 3

Execution result of K-means in GPP, FPGA, and GPU, for single dimension data.

Clusters	GPP avg. time per iteration (sec.) [12]	GPP avg. time for complete execution (sec.) [12]	GPU avg. time per iteration (sec.) [12]	GPU avg. time for complete execution (sec.) [12]	FPGA per iteration (sec.)	FPGA complete execution (sec.)
0.4 MPx
16	0.269	4.314	0.021	0.443	0.0028	0.0392
32	0.516	7.637	0.020	0.421	0.0028	0.042
64	1.004	12.78	0.023	0.508	0.0028	0.0454
6 MPx
16	4.279	67.07	0.256	5.176	0.0425	0.723
32	8.144	110.7	0.247	4.439	0.0425	0.638
64	15.86	208.2	0.270	5.220	0.0425	0.723