I hope another set of benchmarks isn't overkill for the list. I see different results comparing 2.1 with 2.0 (both -O3) using pybench 0.6. The interesting differences I see in this benchmark that I didn't see in MAL's are: DictCreation +15.87% SeoncdImport +20.29% Other curious differences, which show up in both benchmarks, include: SpecialClassAttribute +17.91% (private variables) SpecialInstanceAttribute +15.34% (__methods__) Jeremy PYBENCH 0.6 Benchmark: py21 (rounds=10, warp=20) Tests: per run per oper. diff * ------------------------------------------------------------------------ BuiltinFunctionCalls: 305.05 ms 2.39 us +4.77% BuiltinMethodLookup: 319.65 ms 0.61 us +2.55% ConcatStrings: 383.70 ms 2.56 us +1.27% CreateInstances: 463.85 ms 11.04 us +1.96% CreateStringsWithConcat: 381.20 ms 1.91 us +2.39% DictCreation: 508.85 ms 3.39 us +15.87% ForLoops: 577.60 ms 57.76 us +5.65% IfThenElse: 443.70 ms 0.66 us +1.02% ListSlicing: 207.50 ms 59.29 us -4.18% NestedForLoops: 315.75 ms 0.90 us +3.54% NormalClassAttribute: 379.80 ms 0.63 us +7.39% NormalInstanceAttribute: 385.45 ms 0.64 us +8.04% PythonFunctionCalls: 400.00 ms 2.42 us +13.62% PythonMethodCalls: 306.25 ms 4.08 us +5.13% Recursion: 337.25 ms 26.98 us +19.00% SecondImport: 301.20 ms 12.05 us +20.29% SecondPackageImport: 298.20 ms 11.93 us +18.15% SecondSubmoduleImport: 339.15 ms 13.57 us +11.40% SimpleComplexArithmetic: 392.70 ms 1.79 us -10.52% SimpleDictManipulation: 350.40 ms 1.17 us +3.87% SimpleFloatArithmetic: 300.75 ms 0.55 us +2.04% SimpleIntFloatArithmetic: 347.95 ms 0.53 us +9.01% SimpleIntegerArithmetic: 356.40 ms 0.54 us +12.01% SimpleListManipulation: 351.85 ms 1.30 us +11.33% SimpleLongArithmetic: 309.00 ms 1.87 us -5.81% SmallLists: 584.25 ms 2.29 us +10.20% SmallTuples: 442.00 ms 1.84 us +10.33% SpecialClassAttribute: 406.50 ms 0.68 us +17.91% SpecialInstanceAttribute: 557.40 ms 0.93 us +15.34% StringSlicing: 336.45 ms 1.92 us +9.56% TryExcept: 650.60 ms 0.43 us +1.40% TryRaiseExcept: 345.95 ms 23.06 us +2.70% TupleSlicing: 266.35 ms 2.54 us +4.70% ------------------------------------------------------------------------ Average round time: 14413.00 ms +7.07% *) measured against: py20 (rounds=10, warp=20)