[QST][Cute-DSL] How to return more than 1 value from llvm.inline_asm

I'm using llvm.inline_asm in cute-dsl and currently it can only return 1 value, like this:

@dsl_user_op
def rsqrt(a: float | Float32, *, loc=None, ip=None) -> Float32:
    return Float32(
        llvm.inline_asm(
            T.f32(),
            [Float32(a).ir_value(loc=loc, ip=ip)],
            "rsqrt.approx.ftz.f32 $0, $1;",
            "=f,f",
            has_side_effects=False,
            is_align_stack=False,
            asm_dialect=llvm.AsmDialect.AD_ATT,
        )
    )

I'm looking for ways to return more than 1 value. I've tried using T.vector(2, T.f32()) as the output type and "=f,=f,f" as the constraint but that didn't work.

Start with the llvm.inline_asm call and the rsqrt example in the issue, then trace how Cute-DSL represents its output type and constraints. No implementation files or tests are named, so first identify the relevant inline-assembly entry point and existing tests. Done would require a confirmed, tested way to represent more than one returned value, or a clearly scoped implementation plan.