View previous topic :: View next topic |
Author |
Message |
Ruairi Guest
|
Posted: Fri Aug 16, 2002 7:52 pm Post subject: gcc 3.2-pre -march=pentium4 -msse2 |
|
|
If I try to compile mozilla or X-4.2 with the above optimizations I get errors complaining about movd operands being incorrect... Any ideas ???
Ruairi |
|
Back to top |
|
|
rac Bodhisattva
Joined: 30 May 2002 Posts: 6553 Location: Japanifornia
|
Posted: Fri Aug 16, 2002 8:13 pm Post subject: |
|
|
From Installing Gentoo to GCC3, because it involves a gcc 3.x problem with something after completion of the install guide. _________________ For every higher wall, there is a taller ladder |
|
Back to top |
|
|
Locke n00b
Joined: 02 Aug 2002 Posts: 50 Location: Baton Rouge, LA, US
|
Posted: Fri Aug 16, 2002 9:06 pm Post subject: |
|
|
GCC 3.1.1-r1 has trouble compiling a few packages with -msse2 because certain ebuilds (XFree for example) knock my -march= setting down from pentium4 to pentium3, thus the invalid asm instructions. Why is this? _________________ ---BEGIN GEEK CODE BLOCK---
GH/MU d+ s+:+>+: a--- C++ UL+++>++++ P>++ L+++(++) E- W++(-) N+ o?
K+ w !O M-- !V PS+@ PE Y+ PGP t 5 X- R-@ tv- b+++ DI+ D G e>++++ h- r y
---END GEEK CODE BLOCK--- |
|
Back to top |
|
|
rac Bodhisattva
Joined: 30 May 2002 Posts: 6553 Location: Japanifornia
|
Posted: Fri Aug 16, 2002 9:16 pm Post subject: |
|
|
Locke wrote: | GCC 3.1.1-r1 has trouble compiling a few packages with -msse2 because certain ebuilds (XFree for example) knock my -march= setting down from pentium4 to pentium3, thus the invalid asm instructions. Why is this? |
There were bugs with pentium4 optimizations in GCC 3.1. IIRC, the original reason there was an ebuild of a prerelease of 3.1.1 was because the ebuild author had a P4 and was hacked off. I am guessing that the xfree ebuild tried to work around the P4 problems, and maybe didn't distinguish between 3.1 and 3.1.1. Which xfree ebuild version are you working on? _________________ For every higher wall, there is a taller ladder |
|
Back to top |
|
|
Locke n00b
Joined: 02 Aug 2002 Posts: 50 Location: Baton Rouge, LA, US
|
Posted: Fri Aug 16, 2002 9:35 pm Post subject: |
|
|
4.2.0-r12. That does make a bit of sense. Thanks for the info _________________ ---BEGIN GEEK CODE BLOCK---
GH/MU d+ s+:+>+: a--- C++ UL+++>++++ P>++ L+++(++) E- W++(-) N+ o?
K+ w !O M-- !V PS+@ PE Y+ PGP t 5 X- R-@ tv- b+++ DI+ D G e>++++ h- r y
---END GEEK CODE BLOCK--- |
|
Back to top |
|
|
rac Bodhisattva
Joined: 30 May 2002 Posts: 6553 Location: Japanifornia
|
Posted: Fri Aug 16, 2002 10:01 pm Post subject: |
|
|
OK, that's the version I was looking at. I think this is the culprit: Code: | if [ "`gcc -dumpversion`" != "2.95.3" ]
then
# should fix bug #4189. gcc-3.x have problems with -march=pentium4
echo "#define OptimizedCDebugFlags ${CFLAGS/pentium4/pentium4 -mno-mmx} -fno-merge-constants" >> \
config/cf/host.def
else
echo "#define OptimizedCDebugFlags ${CFLAGS}" >> config/cf/host.def
fi |
You could try modifying your copy of the ebuild to make the test always fail by replacing the "if' line with something like:
...and see if it does what you want it to do. _________________ For every higher wall, there is a taller ladder |
|
Back to top |
|
|
Ard Righ Guru
Joined: 24 Jun 2002 Posts: 337 Location: Wellington, NZ
|
Posted: Sun Aug 18, 2002 8:27 am Post subject: |
|
|
Hmm, I just emerged Gnome on a new 1.4/3.2 system using -march=pentium4 -mmmx -msse -msse2 -mfpmath=sse -O3 -pipe and I didn't get any errors at all. |
|
Back to top |
|
|
|