[Haskell-cafe] GHC profiling overhead

[Carter Schonwald]

its worth nothing that if you build the application / libraries / ghc  with
-g or g2 or the like, you can do low over head statistical profiling using
DWARF metadata related tools, though the source code mapping isn't as nice
as what cost center time / allocation profilng can accomplish.

On Tue, Feb 25, 2020 at 12:59 PM Jaro Reinders <jaro.reinders at gmail.com>
wrote:

> Dear Dusan,
>
> (Automatic) profiling can prevent some very important optimizations such as
> inlining. I would recommend this blog post:
> https://www.tweag.io/posts/2020-01-30-haskell-profiling.html.
>
> Kind regards,
>
> Jaro
>
> On 25-02-2020 18:18, Kolář Dušan wrote:
> > Dear Café,
> >
> > I'm quite curious about overhead of the profiler...
> > I've compiled my program with --make -threaded -rtsopts -O2 -prof
> options.
> > It was started with +RTS -N5 -p
> >
> > Usually, the program evaluates like 6-7 hours of real time. What is the
> > overhead of the profiling? After 8 hours of working, the program has not
> > finished yet. I've looked over the Internet and it seems like the
> overhead may
> > be 25 times slowdown and even worse. Is it true? If yes, it's not worth
> waiting
> > then :-)
> >
> > Best regards,
> > Dusan
> > _______________________________________________
> > Haskell-Cafe mailing list
> > To (un)subscribe, modify options or view archives go to:
> > http://mail.haskell.org/cgi-bin/mailman/listinfo/haskell-cafe
> > Only members subscribed via the mailman list are allowed to post.
> _______________________________________________
> Haskell-Cafe mailing list
> To (un)subscribe, modify options or view archives go to:
> http://mail.haskell.org/cgi-bin/mailman/listinfo/haskell-cafe
> Only members subscribed via the mailman list are allowed to post.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.haskell.org/pipermail/haskell-cafe/attachments/20200225/75b21cb3/attachment.html>