Passive radar exploits an external illuminator signal to detect targets. It has the advantages of silence, anti-interference, and counter-stealth ability. In most cases, direct and multipath clutters should be suppressed first. Then coherent detection can be made by performing a cross-ambiguity function of the remaining target echoes and the reference signal. However, under a wide-band signal, a long-integration time, or multi-beam circumstances, a large number of computations and amount of memory is required for normal processing. This paper expresses the mathematical relationships of clutter suppression algorithms based on the Minimum Mean Square Error (MMSE) principle and coherent detection algorithms based on the cross-ambiguity function. Herein, a joint-optimize and processing method is presented. This method reduces the number of computations and amount of memory required, is easy to implement on GPU devices such as CUDA, and will be useful for engineering applications. Its high-efficiency and real-time properties are validated in the experimental results.