**** 8 bits data size ****
threads per block ; occupancy
64 ; 0.5
128; 1024 0.5, rest 1.0
256; 1024 0.5, rest 1.0
288; 1024 0.562, rest 0.984 
320; 1024 0.625, rest 0.938
416; 1024 0.609, rest 0.812
448; 1024 0.656, rest 0.875
512; 1024 0.5, rest 1.000
400; 1024 0.609, rest 0.812
768; 0.750
832; 0.812
896; 0.875
1024; 1024 0.5, rest 1.0
=====================

K   ; BPT
1024; 320 256
2048; 320 288 256
4096; 128 64 256
8192; 128 320 64
16384; 256
32768; 128 256
100000; 128 256
200000; 128 256

**** 32 bits data size ****

K   ; BPT
1024; 64 128
2048; 64 128
4096; 64 128 256
8192; 320 64 128
16384; 256 128
32768; 768 384
100000; 1024 128 256
200000; 256 512

