-fomit-frame-pointer [was: Re: XFree86 host.def]
bdubbs at swbell.net
Thu Oct 3 21:17:03 PDT 2002
Greg Schafer wrote:
>On Fri, Oct 04, 2002 at 12:54:04AM +0200, Matthias Benkmann wrote:
>>Okay. I think the numbers speak for themselves.
>Excellent work MSB. I've been advocating the "-O3 is worse than -O2" line
>for a while now as well. But my tests weren't quite as scientific as yours,
>just some simple gzip decompression tests.
I agree. The analysis is supurb. I'd like to suggest another line of
thought, however. Most packages I've seen already use -O2. If we don't
supply any optimization, the packages usually default to -O2. On top of
that, IIRC, some packages may use CFLAGS internally, so if you add
anything, you effectively remove the package maintainer's default
optimizations. Also, in some cases, different programs or modules
within a package may use different optimizations. Setting a flag may
override problems that the developers already solved.
What I have been doing lately is not setting CFLAGS or CXXFLAGS at all.
In some cases, I may look to see if the -g flag is being used and take
action to remove it, but I generally stay pretty close to the package
Unsubscribe: send email to listar at linuxfromscratch.org
and put 'unsubscribe blfs-dev' in the subject header of the message
More information about the blfs-dev