What is Branch Prediction
The branch prediction unit, as the name suggests, needs to perform a basic task—branch prediction. Before delving into the branch prediction unit, it is necessary to understand what branch prediction is.
Basic Design of Shanshan Branch Prediction Unit (BPU)
This document introduces the basic design principles of the Shanshan branch prediction unit. By reading this document, you can understand the approximate workflow of the Shanshan BPU without needing to know specific signal names and code details.
Categories:
In processor design, a well-designed branch predictor (BPU) is a key component for improving processor performance. It is responsible for guiding the processor’s fetch, determining where the next instruction should be fetched and executed. The BPU is the starting point of an instruction’s lifecycle, so exploring a high-performance processor from the BPU is a good starting point.
This is also true for Shanshan, a high-performance processor with out-of-order six-issue execution, which naturally requires a branch prediction unit with high accuracy and efficiency. The design of a branch prediction unit often needs to consider many factors, such as timing, complexity of structure, silicon area occupation, prediction accuracy, and speed of recovery from prediction errors. The branch prediction unit of the Shanshan processor has achieved a good balance among these factors through many clever designs, giving it high branch prediction efficiency and accuracy, providing a basic guarantee for the supply of instructions to the backend.
In this section, we will introduce the basic design of the Shanshan prediction unit. By reading this section, you can learn the following:
- Basic concepts of branch prediction
- Basic prediction unit of the Shanshan branch prediction unit - branch prediction block
- External interfaces of the Shanshan branch prediction unit
- Basic structure of the Shanshan branch prediction unit
- Basic timing of the Shanshan branch prediction unit
Anyone participating in the BPU crowdsourcing verification work should read this section first to have a basic understanding of the Shanshan branch prediction unit.
Basic of the Xiangshan Branch Prediction Unit (BPU)
This section introduces the basic ideas and working principles of the Xiangshan Branch Prediction Unit (BPU), including the use of branch prediction block concepts, multiple predictors, multiple pipeline structures, and the role of the Fetch Target Queue (FTQ), explaining the main interfaces of BPU for external interaction.
Introduction to the Xiangshan Branch Prediction Unit Structure
This section introduces the structure of the Xiangshan Branch Prediction Unit (BPU), including the integration of multiple predictors and multiple pipeline schemes, as well as the organization structure and interface design of internal sub-predictors, demonstrating how the BPU interacts with the Composer, and explaining the connection methods between sub-predictors.
Introduction to the Timing of Xiangshan Branch Prediction Unit
The timing design of the three-stage pipeline is the essence of the Xiangshan BPU. This section will introduce how the prediction result redirection signal is generated, how a new PC is generated based on the prediction result, and how the prediction results of the three channels are handled.