Ian you nailed it too - thanks for your advice about micro-benchmarks. I will fix them per your suggestion.
On Tue, Feb 28, 2017 at 7:41 AM, Shubha Ramani <shubharam...@gmail.com> wrote: > Dan thank you so much. Indeed you solved it and thanks for your advice > about the micro-benchmarks > too. > > Sincerely, > > Shubha > > On Mon, Feb 27, 2017 at 7:25 PM, Dan Kortschak < > dan.kortsc...@adelaide.edu.au> wrote: > >> When the python functions are actually called the comparison is more >> reasonable: >> >> ``` >> ~/concat $ go test -bench . concat_test.go >> BenchmarkUnicodeConcat-8 200000 10379 ns/op >> PASS >> ok command-line-arguments 2.193s >> ~/concat $ python concat_test.py >> time_taken = 8901.3030529 nsec >> >> ~/concat $ python3 concat_test.py >> time_taken = 7210.201049922034 nsec >> ``` >> >> This involved appending "; concat()" and "; mysetup()" to the stmt and >> setup parameters. >> >> On Mon, 2017-02-27 at 19:03 -0800, Ian Lance Taylor wrote: >> > On Mon, Feb 27, 2017 at 6:40 PM, Shubha Ramani <shubharam...@gmail.co >> > m> wrote: >> > > >> > > mysetup is being called here: >> > > >> > > times = timeit.Timer(stmt="from __main__ import >> > > concat",setup="gc.enable();from __main__ import >> > > mysetup").repeat(7,loops) >> > > >> > > Look at "from __main__ import mysetup" after gc.enable(); >> > > gc,enable() >> > > according to timeit rules needs to be called >> > > before the setup function. Are you saying Ian that I only import >> > > but never >> > > call mysetup() ? Perhaps you are right but I followed examples and >> > > this is >> > > what I thought was correct (maybe not). >> > I added print statements to the Python functions and they never >> > printed. I don't know whether that proves anything or not. >> > >> > Your micro-benchmarks are suspicious because they don't exist. It >> > would take very little for the compiler to simply discard all the >> > computation. I'm actually surprised that does not happen with the Go >> > compiler--seems like a minor bug. I wonder whether the Python >> > interpreter is doing that one way or another. >> > >> > You should always write a micro-benchmark to compute a real result >> > and >> > verify that the result is what you expect. Otherwise a clever >> > compiler will completely defeat the purpose of the benchmark. >> > >> > Ian >> > >> > > -- You received this message because you are subscribed to the Google Groups "golang-nuts" group. To unsubscribe from this group and stop receiving emails from it, send an email to golang-nuts+unsubscr...@googlegroups.com. For more options, visit https://groups.google.com/d/optout.
