The demo application Makefile has a check target option which can be used to verify the output from the Binomial tree Kernel compared to CPU/Quantlib and the throughput.
For a 1 engine kernel with a tree height of 1024 we obtain a throughput of approximately 0.7K option calculations per second.
For a 4 engine kernel with a tree height of 1024 we obtain a throughput of approximately 2.7K option calculations per second.
Both these values are obtained when calculating 49 options (i.e. the stock and volatility test grid). The values are the same, whether European or American option prices are being calculated.