202406212114
Status: #idea
Tags: CMU Advanced Database Systems

CPU Overview

• CPU's organize instructions into pipeline stages
• Super-scalar CPU's have multiple pipelines per core
  • Out-of-order execution

CPU Problems

1. Dependencies
   • Dependent instructions can not be pushed into the same pipeline
2. Branch prediction
   • Misprediction requires a pipeline flush
   • Most common branching code in a DBMS is the filter operation during a sequential scan. But this is nearly impossible to predict correctly
   • Compiler hints exist, but they cant't talk to the CPU. It can make things worse

References