A new technique to acquire pseudo-noise (PN) sequences has been recently proposed in [1] and [2]. It is based on the paradigm of iterative Message Passing (iMP) to be run on loopy graph. This technique approximates the maximum-likelihood (ML) estimator, providing a sub-optimal algorithm that searches all possible code phases in parallel, at low complexity and fast acquisition time. This work is addressed to the design of the architecture of an iMP detector, following the implementation methodologies typical of standard lowdensity parity-check (LDPC) decoders, and demonstrates its benefits in terms of acquisition time and complexity, compared with standard techniques.