After logging into magic shell, execute make run all to run compile and run all the kernels.

H100, N = 2^30, 3300 GB/s max

Kernel	Bandwidth (GB/s)	% of Max Bandwidth	Implementation
two_pass_1	622.87	18.87%	Two-pass
two_pass_2	718.54	21.77%	Two-pass
two_pass_3	633.19	19.19%	Two-pass
two_pass_4	738.78	22.39%	Two-pass
one_pass_1	985.32	29.86%	One-pass
one_pass_2	1881.55	57.02%	One-pass
one_pass_3	1882.55	57.05%	One-pass
one_pass_4	3134.26	94.98%	One-pass
one_pass_5	2547.15	77.19%	One-pass

NVIDIA GeForce RTX 3060 Mobile, N = 2^29, 336 GB/s max

Kernel	Bandwidth (GB/s)	% of Max Bandwidth	Implementation
two_pass_1	108.23	32.21%	Two-pass
two_pass_2	160.18	47.67%	Two-pass
two_pass_3	107.27	31.93%	Two-pass
two_pass_4	160.25	47.69%	Two-pass
one_pass_1	167.45	49.84%	One-pass
one_pass_2	324.77	96.66%	One-pass
one_pass_3	324.77	96.66%	One-pass
one_pass_4	324.79	96.66%	One-pass
one_pass_5	324.42	96.55%	One-pass