The Two Level Perceptron (TLP) predictor is a neural mechanism that effectively combines predicting whether an access will be off-chip with adaptive prefetch filtering at the first-level data cache (L1D).