The serial/parallel converter shown in most OFDM transmitter block diagrams is required in a hardware implementation but not very meaningful for software implementation. Because in software there is no IFFT block that actually works with parallel inputs, it's all serial processing in reality.
Let $S(k)$ be your input stream of complex numbers and $N$ be the number of subcarriers, i.e. the IFFT size. Then you have to calculate the IFFT of every $N$ samples in $S(k)$, i.e. of the vectors $[S(iN),\ldots,S((i+1)N - 1)]$, with symbol index $i=0,1,2,\ldots$. The result is the OFDM modulated time domain signal. If I understand your example correctly you have to apply the IFFT to both $S_1$ and $S_2$, yes.
The output of the IFFT operation is also complex, in general. Unless you use two audio channels followed by an Inphase/quadrature mixer that shifts the signal to a carrier frequency the vectors defined above have to exhibit a complex conjugate symmetry in order to obtain a real-valued time domain signal. (So-called discrete multitone (DMT))
The allocation of data to subcarriers happens through the IFFT algorithm you're using. The FFTW library, for example, allocates $S(0)$ to frequency zero and $S(N-1)$ to "frequency" $-N/2$. What physical frequency this corresponds to in turn depends on the sampling frequency of the digital-to-analog converter you're using. Orthogonality is inherent to the (I)FFT algorithm, you don't have to apply any further processing to obtain it.
