Memory bandwidth and latency is a major performance bottleneck these days. Usually the CPU spends a lot of time starving for data instead of executing stuff.
x32 ABI is useful because:
There are applications that don't need to map more than 4GB of memory, and could get a performance boost from these benefits.
x32 ABI is useful because:
- As has been said, you can transfer double the pointers at a time (compared to x86_64)
- You can fit more data in CPU cache since you now save 4 bytes for every pointer (very good to avoid trashing data between RAM and CPU)
- You have a lot more (bigger) registers compared to x86
- Probably more I can't think of right now
There are applications that don't need to map more than 4GB of memory, and could get a performance boost from these benefits.
Comment