Re: [PATCH] libpthread NOLOAD removal

[Aleksej Saushev <asau%inbox.ru@localhost>]

  Hello,

manu%netbsd.org@localhost (Emmanuel Dreyfus) writes:

> Here is my llatest patch for libpthread NOLOAD removal. It implements
> minimal mutex, condvar and rwlock operations in libc thread stubs, so
> that theses objects can be handled by both libc and libpthread after
> libpthread has been loaded by dlopen.
>
> http://ftp.espci.fr/shadow/manu/pthread5.patch
>
> Performance hit has been evaluated: it is not significant for running
> NetBSD's build.sh. On a billion calls to gmtime(), I get a 1%
> performance hit.

I'm sorry to intervene, but am I the only person who thinks that running
build.sh isn't correct way to test performance hit here?

build.sh generates "visible" pressure on I/O subsystem that may trump
all effects of mutex introduction. I hope we do agree that NetBSD is not
meant to run build.sh exclusively.


-- 
HE CE3OH...