Is it possible to force 16/32-bit stack alignment without using the optimizer?

Question

The compiler emits code targeted at the classic Pentium architecture for the -m32 memory model.  I'm running into problems mixing Sun Studio compiled code with code built with other compilers because the other compiler builds under the assumption that the stack is 16-byte aligned.
The only way I've found to force Sun Studio to comply with that restriction is with -xarch={sse2a,sse3,...}, but this causes the code to pass through the optimizer.  As noted in the documentation, if you want to avoid optimizations you must remove all flags that imply optimizations -- that is to say, there's no way to disable optimizations once enabled.  This should not, however, be treated as an optimization because it's an ABI requirement.
I've scoured the documentation, spent many hours googling, digging through forums, and asking questions.
The best I've come up with is the -xarch option which is sub-optimal because it has side effects.  I tried -xchip=pentium4 (this is what my other compilers have set as their default target), but the generated code doesn't force 16-byte stack alignment.
Is there a way to force the compiler to emit code conforming to a different ABI without using the optimizer?
-Brian

Answer 1

Thank you for your response.
I hope you won't mind my asking: do you have a way to prove that it's not possible to force 16-byte alignment without using the optimizer?  I ask because your username / profile don't give the impression you work for Oracle, so while I think you're probably right it's at least possible that we're both mistaken.  I haven't been able to find any documentation on either stack alignment or altering the targeted ABI short of using the -xarch flag, and even there the details are fairly sketchy.
-Brian