Optimizing existing large codebase Measure Mem threads low Software instrumentation in practice o give precise measurements of where you spend instructions including many details oreproducible,so your can compare stuff but not always realistic no real timing,only instructions count memory caching is only simulated,often far from real case no clue on low level efficiency (CPI in particular) and gives no clue on hardware /OS behavior o Main tool available:valgrind family ef tools bottlenecks 11/62 S.Ponce-CERNOptimizing existing large codebase 11 / 62 S. Ponce - CERN Measure Modernize Mem threads low level c/c perf tools bottlenecks Software instrumentation in practice give precise measurements of where you spend instructions including many details reproducible, so your can compare stuff but not always realistic no real timing, only instructions count memory caching is only simulated, often far from real case no clue on low level efficiency (CPI in particular) and gives no clue on hardware / OS behavior Main tool available : valgrind family