This occurred to me when looking at stencil computations in numpy. This python library compiles compute-intensive components, so I believe that it’s a valid example. Here making selections on an array is done with small offsets (0, 1, -1) a lot of times. Would a c++ compiler take into account that it doesn’t need a full integer/char/bool type and “pack” bits into bytes?
For example, could selecting every second element from 0 to 7, be represented as a single byte instead of 8 bytes:
0,1,0,1,0,1,0,1
instead of
0,0,0,0,0,0,0,0
0,0,0,0,0,0,0,1
0,0,0,0,0,0,0,0
0,0,0,0,0,0,0,1
0,0,0,0,0,0,0,0
0,0,0,0,0,0,0,1
0,0,0,0,0,0,0,0
0,0,0,0,0,0,0,1
