So what are the advantages of AArch32? Given than the total ram is (I believe) 1G, there is no use to a 64 bit pointer. ILP64 or LP64 code would just cause unnecessary code bloat and cache pressure over ILP32. Is there something in AArch32 that is compelling enough to go to ArmV8 over ArmV7?
A register file that's twice as large (NEON really suffers without the extra registers), and atomic operations that aren't dog slow. Doing anything other than scalar sequential programming on ARMv7 is pretty painful coming from x86.
Yes, OK, that makes sense, at least for a chip that isn't doing register renaming the the back end, which takes a lot of pressure off the compiler's register allocator.
