I don't have a contributor agreement on file, but I think its overkill
for this purpose, I've copied some diffs below, which gets it to run
up to the point where it runs out of processes on OSX. I just wanted
to see how libmicro worked using the machine I read email on, I'm not
doing serious benchmarking of OSX, but you may as well know whats
needed to make it more portable.
Adrian
As a normal user on OSX there is a 100 process limit set by the shell
which breaks the benchmark. As root with maxproc unlimited my machine
hung on the fork1000 test and was out of processes when I tried to run
something in another shell window. I gave up at this point.
Changes:
On OSX, there is no argument -k to arch so I changed it to uname -p
On OSX/G4 uname -m returns "Power Macintosh" and the script assumes a
single word so I changed it to uname -p which returns powerpc
Starting with Makefile.Linux I copied it to Makefile.Darwin and found
I needed to remove the bench tests for get/setcontext
In Makefile.com there is a -lrt link with librt, this is not needed
and causes an error on OSX
Diffs:
% diff Makefile.Linux Makefile.Darwin
40a41,42
getcontext \
setcontext \
% diff Makefile~ Makefile
72c72
< @mkdir -p bin-`uname -m`; cd bin-`uname -m`; MACH=`uname -m`
$(MAKE) -f ../Makefile.`uname -s` $@
---
@mkdir -p bin-`uname -p`; cd bin-`uname -p`; MACH=`uname -p` $(MAKE) -f
../Makefile.`uname -s` $@
% diff wrapper~ wrapper
37c37
< ARCH=`uname -m`
---
ARCH=`uname -p`
% diff Makefile.com~ Makefile.com
112c112
< $(CC) -o tattle $(CFLAGS) -I. ../tattle.c libmicro.a -lrt -lm
---
$(CC) -o tattle $(CFLAGS) -I. ../tattle.c libmicro.a -lm
On 4/9/07, Dan Price <[EMAIL PROTECTED]> wrote:
On Sun 08 Apr 2007 at 10:46AM, adrian cockcroft wrote:
> Thanks Dan, this is a generally useful test suite.
>
> I tried to get it working on my Mac OSX/G4 laptop, and found some
> portability issues. I have most of it working apart from the tests
> that involve fork. I have modified the makefile and wrapper slightly,
> let me know if you want the changes....
>
> Cheers
> Adrian
That would be great-- do you have a contributor agreement on file?
If so, post the diffs, and your agreement #, and someone will roll them in.
-dp
--
Daniel Price - Solaris Kernel Engineering - [EMAIL PROTECTED] - blogs.sun.com/dp
_______________________________________________
perf-discuss mailing list
perf-discuss@opensolaris.org
