Gshare

Gshare is a type of dynamic branch predictor used in computer processors to guess the direction of conditional branches. It belongs to the family of global history predictors, aiming to improve accuracy by exploiting correlations in branches that occur across the program, rather than relying solely on per-branch history.

The mechanism of a Gshare predictor centers on two main components: a global history register (GHR) and a pattern history table (PHT). The GHR keeps a short sequence of the outcomes of the most recent branches as bits. The PHT contains a number of saturating counters (typically 2-bit) that indicate whether a particular branch is predicted to be taken or not taken. To index the PHT, Gshare computes an address by taking the bitwise XOR of the branch’s program counter (PC) with the current value of the GHR. The counter stored at that PHT index determines the prediction: taken if the counter is in a taken state, not taken otherwise. After the actual branch outcome is known, the corresponding counter is updated toward the observed outcome, and the GHR is shifted to incorporate the latest outcome.

Gshare offers advantages over simple PC-based predictors by reducing aliasing and better capturing correlations among branches that share global context. Its performance depends on the size of the PHT and the length of the GHR; larger tables and longer histories can improve accuracy but incur higher hardware cost and potential diminishing returns. Limitations include potential aliasing within the PHT even with XOR indexing, and sensitivity to workload patterns that do not produce consistent global correlations. Gshare remains a foundational reference in the study of branch prediction and is often used in comparisons with more complex or hybrid predictors.