You can denoise your raw sequence data using DADA2 or Deblur in QIIME2. Depending on the tool you use, you may or may not need to denoise your raw data on a per-run basis.
I guess you mean the samples were sequenced in 2 separate runs? If so, you need to denoise your raw sequence data for each run separately when using the DADA2. After that, you merge your feature table and representative features from different sequencing runs. Check the QIIME2 FMT tutorial for more details on how to merge feature table and representative sequences.
If you use Deblur for the sequence denoising, you can run the algorithm for all the samples at the same time to get the feature table and representative features.
You just need to make one mapping file (metadata) for all the samples as the mapping file is not required for the sequence denoising.