Is PCA a valid option and if so, how would I even apply it to my datasets?

I have 2 datasets, one for training and one for testing, how if possible with the datasets would I apply PCA and and then evaluate my model with lets say decision tree, with "my-train-data" as training the model and "my-test-data" for testing the model? i know how to do it without pca, but cant seem to get pca to work