Abstract
Studying architectural improvements for microprocessors and their memory hierarchies and evaluating the corresponding performance generally need trace driven simulation. Hardware is modeled, benchmark traces are generated and applied to the model and performance data is measured and analyzed. We present a hardware/software approach to collect perfect x86 execution traces using a commercial analyzer.