Comment author: administrator

Float values in the bytecode format seem not to be portable between
x86 (or SPARC) and StrongARM.
For example, running a piece of bytecode for the following:
print_float 1.5
results in different outputs between x86 and StrongARM.

Right. The ARM has a non-standard format for double-precision floats,
which is neither big-endian nor little-endian. E.g. the IEEE double
"1.5" is represented as
00 00 00 00 00 00 f8 3f
on a little-endian machine,
3f f8 00 00 00 00 00 00
on a big-endian machine, and
00 00 f8 3f 00 00 00 00
on the ARM.

OCaml can handle the little-endian/big-endian conversion, but that's all.

• Xavier Leroy

Comment author: administrator

ARM has a non-standard float format (mixed-endian). Nothing we can do about it.