13.07.2015 Views

Intel® 64 and IA-32 Architectures Optimization Reference Manual

Intel® 64 and IA-32 Architectures Optimization Reference Manual

Intel® 64 and IA-32 Architectures Optimization Reference Manual

SHOW MORE
SHOW LESS
  • No tags were found...

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

USING PERFORMANCE MONITORING EVENTSFigure B-2. Performance Events Drill-Down <strong>and</strong> Software Tuning Feedback LoopTypically, the logic in performance monitoring hardware measures microarchitecturalconditions that varies across different counting domains, ranging from cycles, microops,address references, instances, etc. The drill-down methodology attempts toprovide an intuitive, cycle-based view across different phases by making suitableapproximations that are described below:• Total cycle measurement — This is the start to finish view of total number ofcycle to complete the workload of interest. In typical performance tuningsituations, the metric Total_cycles can be measured by the eventCPU_CLK_UNHALTED.CORE. See Appendix A, “Performance Monitoring Events,”of Intel® <strong>64</strong> <strong>and</strong> <strong>IA</strong>-<strong>32</strong> <strong>Architectures</strong> Software Developer’s <strong>Manual</strong>, Volume 3B).• Cycle composition at issue port — The reservation station (RS) dispatchesmicro-ops for execution so that the program can make forward progress. Hencethe metric Total_cycles can be decomposed as consisting of two exclusivecomponents: Cycles_not_issuing_uops representing cycles that the RS is notB-46

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!