On Thursday 20 January 2005 15:44, Mike Jakubik wrote:
Wouldn't we want strict aliasing then?
Strict aliasing is a feature of the code being compiled, not the compiler.
The -fstrict-aliasing flag basically means this code follows the
strictest aliasing rules, so you can use extra
On Jan 18, 2005, at 7:48 PM, Kris Kennaway wrote:
Do yourself a favour and use -O2 -fno-strict-alias instead, because
some ports *are broken* with plain -O2.
I just tried this on a 5-STABLE dual opteron, and it complained about
unknown option. I had to specify the full option
On Thu, Jan 20, 2005 at 04:44:59PM -0500, Mike Jakubik wrote:
Paul Mather said:
On Thu, 2005-01-20 at 19:17 +, Chris wrote:
what does -fno-strict-alias do? I cannot find it in man gcc
It is a truncation of -fno-strict-aliasing. The flag does not appear
to be described in the gcc
On 18-Jan-2005 Chris wrote:
I have been compiling ports using -O2 since I started using FreeBSD
back in 2003 and only port that has had issues with this is lang/ezm3
in FreeBSD 5.2.1 it needed -O.
I fixed that at the beginning of November.
John
___
Yep its fixed now I know, I was saying its the only time I have ever
had a problem, the problem no longer exists now tho.
Chris
On Thu, 20 Jan 2005 08:56:33 -0800 (PST), John Polstra [EMAIL PROTECTED]
wrote:
On 18-Jan-2005 Chris wrote:
I have been compiling ports using -O2 since I started
what does -fno-strict-alias do? I cannot find it in man gcc
Chris
On Thu, 20 Jan 2005 18:54:05 +, Chris [EMAIL PROTECTED] wrote:
Yep its fixed now I know, I was saying its the only time I have ever
had a problem, the problem no longer exists now tho.
Chris
On Thu, 20 Jan 2005
On Thu, 2005-01-20 at 19:17 +, Chris wrote:
what does -fno-strict-alias do? I cannot find it in man gcc
It is a truncation of -fno-strict-aliasing. The flag does not appear
to be described in the gcc man page, but is documented in the gcc info
page (search for -fstrict-aliasing). To quote
Paul Mather said:
On Thu, 2005-01-20 at 19:17 +, Chris wrote:
what does -fno-strict-alias do? I cannot find it in man gcc
It is a truncation of -fno-strict-aliasing. The flag does not appear
to be described in the gcc man page, but is documented in the gcc info
page (search for
On Thu, 2005-01-20 at 16:44 -0500, Mike Jakubik wrote:
Paul Mather said:
On Thu, 2005-01-20 at 19:17 +, Chris wrote:
what does -fno-strict-alias do? I cannot find it in man gcc
It is a truncation of -fno-strict-aliasing. The flag does not appear
to be described in the gcc man
I have been compiling ports using -O2 since I started using FreeBSD
back in 2003 and only port that has had issues with this is lang/ezm3
in FreeBSD 5.2.1 it needed -O.
Chris
On Tue, 18 Jan 2005 08:46:24 +0100, Matthias Buelow [EMAIL PROTECTED] wrote:
Mike Tancsa wrote:
At 05:48 PM
On Tue, Jan 18, 2005 at 11:30:19PM +, Chris wrote:
I have been compiling ports using -O2 since I started using FreeBSD
back in 2003 and only port that has had issues with this is lang/ezm3
in FreeBSD 5.2.1 it needed -O.
Do yourself a favour and use -O2 -fno-strict-alias instead, because
On Tuesday 18 January 2005 01:46 am, Matthias Buelow wrote:
Then I'd leave it at that. The difference between gcc -O2 and -O is
marginal at best.
Sometimes.
I recently switched to -O2 -fno-strict-alias on my small Alpha machine and got
an increase of well over 100% in the OpenSSL
On Thu, 13.01.2005 at 19:38:27 -0500, Mike Tancsa wrote:
I'm using -Os -pipe ever since installing 5.x after the gcc update.
Haven't encountered any problem so far..
There are some differences between O2 and Os according to the man
page. Are you using that on your kernel and for buildworld ?
On Mon, Jan 17, 2005 at 08:38:44PM +0100, Ulrich Spoerlein wrote:
On Thu, 13.01.2005 at 19:38:27 -0500, Mike Tancsa wrote:
I'm using -Os -pipe ever since installing 5.x after the gcc update.
Haven't encountered any problem so far..
There are some differences between O2 and Os according to
At 05:48 PM 17/01/2005, Kris Kennaway wrote:
I let the box run over night building world 8 times with -j2 through -j5
and all worked just fine.
Ok, that should put hardware out of the question.
Sorry, no more ideas, besides getting rid of CPUTYPE as that is know to
cause problems.
Well,
Mike Tancsa wrote:
At 05:48 PM 17/01/2005, Kris Kennaway wrote:
Like I said before, switching back to -O makes the system stable once
again. I let it run continuous buildworlds without issue for 24hrs
along with burnP6, and memtest running in the background. No problem.
Then I'd leave it at
On Wed, 12.01.2005 at 17:38:56 -0500, Mike Tancsa wrote:
I did a buildworld /buildkernel with a RELENG_5 box using the following
flags
CPUTYPE=i686
KERNCONF=recycle
CFLAGS=-O2 -pipe
COPTFLAGS=-O2 -pipe
NO_MODULES=true# do not build modules with the kernel
MODULES_WITH_WORLD=true #
Ulrich Spoerlein wrote:
CPUTYPE=i686
KERNCONF=recycle
CFLAGS=-O2 -pipe
COPTFLAGS=-O2 -pipe
NO_MODULES=true# do not build modules with the kernel
MODULES_WITH_WORLD=true # do not build modules when building kernel
I'm using -Os -pipe ever since installing 5.x after the gcc update.
Haven't
At 05:51 PM 13/01/2005, Ulrich Spoerlein wrote:
On Wed, 12.01.2005 at 17:38:56 -0500, Mike Tancsa wrote:
I did a buildworld /buildkernel with a RELENG_5 box using the following
flags
CPUTYPE=i686
KERNCONF=recycle
CFLAGS=-O2 -pipe
COPTFLAGS=-O2 -pipe
NO_MODULES=true# do not build
I did a buildworld /buildkernel with a RELENG_5 box using the following
flags
CPUTYPE=i686
KERNCONF=recycle
CFLAGS=-O2 -pipe
COPTFLAGS=-O2 -pipe
NO_MODULES=true# do not build modules with the kernel
MODULES_WITH_WORLD=true # do not build modules when building kernel
After rebooting and trying
20 matches
Mail list logo
