On 30/06/2019 21:02, Richard Kettlewell wrote:
> For both Intel and ARM the 64-bit instruction sets double the number of
> architectural GPRs. Particularly on Intel where the 32-bit situation is
> pretty poor, that’s a big improvement for anyone writing in assembler.

Don't expect general performance gains with 64bit ARM. It's only because
x86 is such an atrocious outdated mess, that the AMD64 instruction set 
shows significant gains. On every other architecture supporting both 32 
bit and 64 bit code (Power, Sparc, MIPS) the 64 bit code is usually 
slower than the 32 bit on the same chip, due to the increased demand on 
the data cache from 64 bit pointers.

With the current state of ARMv8 support on GCC, CPU and memory bound 
programs vary from a barely noticeable few percent drop, to 50% slower. 
So it only makes sense to move from running 32 bit to 64 bit code when 
moving to a chip with which supports 4GB or more of memory, and a much 
larger data cache.

---druck

--- SoupGate-Win32 v1.05