Re: Bounds checking, Optimization techniques and undefined behavior

>[Are you referring to the old BOUND instruction or the newer MPX feature. -John]


Mostly I was referring to the new(ish, Skylake and later) MPX stuff.


The larger point is that the architecture had at least some ability to
check bounds in hardware for a very long time: BOUND existed at least
from i286 onward (if not before). 64-bit mode dropped it, but the
need for *something* to do the same job was deemed great enough that
the MPX extensions were introduced.


The MPX approach is more flexible and faster than BOUND was ... but it
still is far too slow for heavy use. IMO it just rehashes many of the
old problems that existed with segments: programmer visible registers,
too few of them, too complicated setup of data structures in memory,
etc.


YMMV,
George
[Bounds checking in the x86 is pretty old. It wasn't in the 8086/88 but
it was in the 286. -John]

Return to the comp.compilers page.
Search the comp.compilers archives again.