If I am permitted to speak a little: There are many factors which can
dominate wrt. this.
1. Memory bandwidth. Transferring 64-bit numbers/pointers from RAM to
CPU or the other way requires up to twice as much as 32-bit ones. "Up
to twice", because this also depends on how well the actual transfers
fit in cache lines (e.g. how many cache lines the much-used structs or
objects require).
2. Big integer usage. If the 32-bit code does lots of 64-bit work,
that'll be slow.
3. RAM availability. If the code can benefit greatly from having more
than 2-3GB virtual memory available, 64-bit code obviously benefits and
32-bit code does not.
4. Optimisation on the CPU. If the designers spent the silicon on making
32-bit fast, or on making 64-bit fast, etc.
So depending on the workload, software, compiler, motherboard/CPU/RAM
combination and so on, either 32-bit or 64-bit may be faster.
Arnt
_______________________________________________
DBmail mailing list
[email protected]
https://mailman.fastxs.nl/mailman/listinfo/dbmail
