search

JuliaLang / Microbenchmarks

Microbenchmarks comparing the Julia Programming language with other languages
https://julialang.org/benchmarks/
Other
88 stars 48 forks source link

update `hex_string` in parse_int fortran benchmark #68

Closed oscardssmith closed 2 years ago

oscardssmith commented 2 years ago

changes based on https://github.com/JuliaLang/Microbenchmarks/issues/67

Closes #67

oscardssmith commented 2 years ago

@xecej4 any idea why this is failing the tests?

acxz commented 2 years ago

I believe this is failing due to the newer julia version. I need to redo some of the julia benchmarks with newer julia functionality. Particularly trying to get rid of this compat file: https://github.com/JuliaLang/Microbenchmarks/blob/master/perfutil.jl

xecej4 commented 2 years ago

Sorry, I have no idea what the testing procedure is or what tests are run. I am not familiar with Github (other than just downloading files from it). I did compile and run perf.f90 after making my suggested changes, using several Fortran compilers on my Windows PC and on a cloud Linux server, and the program ran normally, with less time used in the parse_string tests.

Here are the console outputs from perf.f90, on a Windows 11 PC with an i7-10710U, using Intel Fortran 2022 64-bit.

Previous Github version:

fortran,recursion_fibonacci,.015000 fortran,parse_integers,.140000 fortran,print_to_file,31.000000 fortran,userfunc_mandelbrot,.046000 fortran,recursion_quicksort,.297000 fortran,iteration_pi_sum,.000000 fortran,matrix_statistics,.000000 fortran,matrix_multiply,15.000000 Elapsed time: 3.006 s

With my changes applied: fortran,recursion_fibonacci,.015000 fortran,parse_integers,.031000 fortran,print_to_file,31.000000 fortran,userfunc_mandelbrot,.046000 fortran,recursion_quicksort,.312000 fortran,iteration_pi_sum,.000000 fortran,matrix_statistics,.000000 fortran,matrix_multiply,16.000000 Elapsed time: 2.512 s

@xecej4https://github.com/xecej4 any idea why this is failing the tests?

— Reply to this email directly, view it on GitHubhttps://github.com/JuliaLang/Microbenchmarks/pull/68#issuecomment-1152599360, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AQNQ2RLCKLAOKXYBWTO6TK3VON57LANCNFSM5X5NBABA. You are receiving this because you were mentioned.Message ID: @.***>

acxz commented 2 years ago

@oscardssmith sorry that it took me a while to get around to this, but if you could rebase this PR off the latest master, it should resolve the CI errors we are seeing.

@xecej4 thanks for the analysis on your local machine we should be able to confirm from the CI benchmarks as well. You can check the CI output yourself by clicking on the Details link. It will have a bunch of build/run output as well as the timings.

acxz commented 2 years ago

CI passes and shows similar improvement as displayed by local results of @xecej4

Given that the benchmark in question (parse_int) does not focus on how to write out the string in hex, this is a valid change to the benchmark and merging this in.

Thx for the contrib @xecej4 and for the help in pushing this along @oscardssmith !