Analysis Of Branch Prediction Buffer

1477 Words6 Pages
e. Correlating Branch Predictor: The 2-bit BHT uses only the recent behavior of a single branch to predict the future behavior of that branch. BHT contains prediction regarding the next branch – whether it will be a taken (T) or not-taken (N), but does not supply the target PC value to initiate fetch of next instruction into the pipeline. Branch predictors that use knowledge about the behavior of other branches to make prediction are called Correlating Predictors or 2-Level Predictors, where: 1st Level: Stores the branch history in Branch History Table (BHT) or Branch Prediction Buffer (BPB). 2nd Level: Stores prediction pattern in the Pattern History Table (PHT) or Branch Target Buffer (BTB). Traditionally, Correlating Predictors…show more content…
Differing by the way the 1st level branch history information is maintained in BHT, i.e., global (G) or on per-address (P) basis, and the way the 2nd level PHTs are associated with the BHT, i.e., global (g) or on per-address basis (p), Yeh and Patt [18] have presented three variations of the Two-Level Adaptive Branch Prediction schemes. These schemes are identified as GAg, PAg and PAp, embedded A being signifying ‘Adaptive’ and GAp being the Correlating Branch Predictor. Considering the addresses that contains branch instructions partitioned into sets (represented by S in the 1st level and by s in the 2nd level), the Two-Level Adaptive Branch Prediction scheme yiels nine possible variations, as listed below in the Table-1: Table-1: Varieties of Two-Level Adaptive Branch Predictor Scheme Description GAg Global Adaptive Branch Prediction using one global pattern history table. GAs Global Adaptive Branch Prediction using per-set pattern history tables. GAp Global Adaptive Branch Prediction using per-address pattern history tables. PAg Per-address Adaptive Branch Prediction using one global pattern history table. PAs Per-address Adaptive
Open Document