180 likes | 278 Vues
Learn about the limits of ILP in instruction-level parallelism due to control and data flow dependencies. Explore the sources and uses of value prediction for improving accuracy and achieving speedups on PowerPC 620 and beyond. Discover the potential benefits and future developments in prediction models and parallel execution units.
E N D
Data value prediction Bas van der Tol
Limits to ILP Instruction Level Parallelism is limited by • Control flow • Data flow: true dependencies
Sources of predicatable data • Data redundancy • Error-checking • Program constants • Virtual function calls • Glue code • Call-subgraph identities • Register spill code
Penalties Penalties • Misprediction penalty • Structural hazards, both on correct and mispredictions
Improving Prediction Accuracy • Last Value Prediction • Stride Prediction • Finite Context Method Predictors (fcm)
Conclusions • Data value prediction increases performance by 5% on PowerPC 620 • A performance gain of 23% is possible Future Developments • More parallel execution units • Better prediction models