Assuming Methuselah doesn't work for you, you can still do that in about 200k of memory.

Only at the cost of poor cache locality. :)

Shouldn't it fit completely in L1 (or at worst, L2)? You need roughly 146,100 bytes: 120 years * 365.25 days/year * 4 bytes/day (uint32 as an accumulator) = 175,320 kb.

MacBooks have something like 192 kb shared + 128 kb/core, so that should be okay if not much else is going on....