P-MapNet is a far-seeing map generator that integrates both SDMap and HDMap priors to enhance the performance of online HDMap generation. The paper presents P-MapNet, which focuses on incorporating map priors to improve model performance. Specifically, the method exploits priors from both SDMap and HDMap. On one hand, it extracts weakly aligned SDMap from OpenStreetMap and encodes it as an additional conditioning branch. Despite the misalignment challenge, the attention-based architecture adaptively attends to relevant SDMap skeletons and significantly improves performance. On the other hand, it uses a masked autoencoder to capture the prior distribution of HDMap, which can serve as a refinement module to mitigate occlusions and artifacts. The method is benchmarked on the nuScenes and Argoverse2 datasets. Through comprehensive experiments, it shows that the SDMap prior improves online map generation performance, using both rasterized and vectorized output representations. The HDMap prior improves map perceptual metrics by up to 6.34%. P-MapNet can be switched into different inference modes that cover different regions of the accuracy-efficiency trade-off landscape. P-MapNet is a far-seeing solution that brings larger improvements on longer ranges. The method achieves state-of-the-art results on public benchmarks and presents in-depth ablative analyses revealing the mechanism. The paper also discusses the implementation details of the SDMap and HDMap prior modules, including the use of a masked autoencoder for HDMap prior refinement and the integration of SDMap prior into end-to-end frameworks. The results show that P-MapNet outperforms existing methods in terms of mIoU and AP scores, and the method is effective in handling challenging scenarios such as occlusions and adverse weather. The paper also discusses the impact of inconsistencies between ground truth and SDMaps, and the effect of different downsampling factors on the performance of the model. Overall, P-MapNet demonstrates the effectiveness of incorporating SDMap and HDMap priors in improving the performance of online HDMap generation.P-MapNet is a far-seeing map generator that integrates both SDMap and HDMap priors to enhance the performance of online HDMap generation. The paper presents P-MapNet, which focuses on incorporating map priors to improve model performance. Specifically, the method exploits priors from both SDMap and HDMap. On one hand, it extracts weakly aligned SDMap from OpenStreetMap and encodes it as an additional conditioning branch. Despite the misalignment challenge, the attention-based architecture adaptively attends to relevant SDMap skeletons and significantly improves performance. On the other hand, it uses a masked autoencoder to capture the prior distribution of HDMap, which can serve as a refinement module to mitigate occlusions and artifacts. The method is benchmarked on the nuScenes and Argoverse2 datasets. Through comprehensive experiments, it shows that the SDMap prior improves online map generation performance, using both rasterized and vectorized output representations. The HDMap prior improves map perceptual metrics by up to 6.34%. P-MapNet can be switched into different inference modes that cover different regions of the accuracy-efficiency trade-off landscape. P-MapNet is a far-seeing solution that brings larger improvements on longer ranges. The method achieves state-of-the-art results on public benchmarks and presents in-depth ablative analyses revealing the mechanism. The paper also discusses the implementation details of the SDMap and HDMap prior modules, including the use of a masked autoencoder for HDMap prior refinement and the integration of SDMap prior into end-to-end frameworks. The results show that P-MapNet outperforms existing methods in terms of mIoU and AP scores, and the method is effective in handling challenging scenarios such as occlusions and adverse weather. The paper also discusses the impact of inconsistencies between ground truth and SDMaps, and the effect of different downsampling factors on the performance of the model. Overall, P-MapNet demonstrates the effectiveness of incorporating SDMap and HDMap priors in improving the performance of online HDMap generation.