alloca() portability

[Dan Bernstein]

In article <s64421.658490039 at zeus> s64421 at zeus.usq.EDU.AU (house ron) writes:
> given that all sane computers nowadays have a good
> efficient stack.

Oh? I regularly use a supermini (not a slow little Sun---this thing can
run up to twice as fast as a Convex on non-vector codes) that doesn't
have *any* stack. *None*. It also takes two instructions to load from
memory. It also has separate integer and floating-point instruction
streams, each dealing with thirty-two registers.

As you can imagine, the compiler does a lot of work.

As does the instruction scheduler.

Do you have some religious objection to machines with simulated stacks,
or do you have some rational reason that they're not ``sane''?

---Dan