[Gc] Re: conclusions (benchmarking, compilation, failures)

Christophe Meessen meessen at cppm.in2p3.fr
Sat Apr 28 00:50:27 PDT 2007

If you want to do benchmarking, please disable the debug option.
remove $(cdebug) from .c.obj: and .cpp.obj: to do so.
A specific make file would be preferable.

Building the cord\de.exe failed for me too, but the gc.lib was
successfuly compiled.

>From a quick inspection it looks like the $(cvars) should be removed in
the line
cord\de_win.res: cord\de_win.rc cord\de_win.h cord\de_cmds.h
	$(rc) $(rcvars) -r -fo cord\de_win.res $(cvars) cord\de_win.rc

These are variables for c. It should use recvars instead which is
already there.

It might be that the DLL version you tried was compiled with DUBUG mode.
This could explain why the malloc was so slow.

The NT_THREAD_MAKEFILE is much more complext than the one building the
static files. So I don't know what to change to remove DEBUG option.

Make sure your main program is also compiled in no debug mode. Don't
know what bench program you use. Remember thus to remove any debug flags
in the linking options too.

Achilleas Margaritis a e'crit :
> I've tried with NT_STATIC_THREADS_MAKEFILE, but I got the following error:
>         Rc -DWIN32 -D_WIN32 -DWINVER=0x0500 -DDEBUG -D_DEBUG  -r -fo
> cord\de_win
> .res   -D_MT -MTd cord\de_win.rc
> fatal error RC1106: invalid option: -Td
> NMAKE : fatal error U1077: '"C:\Program Files\Microsoft Visual Studio
> 8\VC\BIN\R
> c.EXE"' : return code '0x1'
> Stop.
> It seems the only makefile that works for me is the NT_THREADS_MAKEFILE.
> O/H Christophe Meessen έγραψε:
>> Hello,
>> could you do the same bench test using the static version and see how
>> it affects numbers ?
> _______________________________________________
> Gc mailing list
> Gc at linux.hpl.hp.com
> https://www.hpl.hp.com/hosted/linux/mail-archives/gc/

More information about the Gc mailing list