| Summary: | [4.5] Core i5 misdetected as Atom w/ -march=native (was: illegal instruction in xulrunner) | ||
|---|---|---|---|
| Product: | Gentoo Linux | Reporter: | Marc-Antoine Perennou <Marc-Antoine> |
| Component: | Current packages | Assignee: | Gentoo Toolchain Maintainers <toolchain> |
| Status: | RESOLVED FIXED | ||
| Severity: | normal | CC: | canarauc |
| Priority: | High | Keywords: | Inclusion |
| Version: | unspecified | ||
| Hardware: | AMD64 | ||
| OS: | Linux | ||
| URL: | http://gcc.gnu.org/PR44046 | ||
| Whiteboard: | |||
| Package list: | Runtime testing required: | --- | |
| Attachments: |
build log
paludis --info cpuinfo gcc45-PR44046-core2.patch |
||
|
Description
Marc-Antoine Perennou
2010-04-27 17:19:09 UTC
Created attachment 229417 [details]
build log
Created attachment 229419 [details]
paludis --info
with CFLAGS="-march=core2 -O2 -pipe" it works well, I'll investigate to find which of the CFLAGS implied by -march=native with a core i5 causes the failure ... Ok it's definitely a gcc 4.5 failure. -march=native is equivalent to -march=atom -msse4.2 -maes -mpclmul -mpopcnt -mcx16 -msahf -mtune=core2 replacing -march=atom (which appeared in gcc 4.5 afaik) by -march=core2 solves the problem. so all the other flags work fine with -march=core2 ? yep, built it fine changing only atom to core2 after copy/pasting this list (I even changed -msse4.2 to -msse4 to enable both .1 and .2 and it still works) scheduling differences sholudnt matter. docs dont say it, but source indicates atom provides -mmovbe over core2. try using that flag with -march=core2 and see if you get the same result. CFLAGS="-march=atom -msse4 -maes -mmovbe -mpclmul -mpopcnt -mcx16 -msahf -mtune=core2 -O2 -pipe" => failure CFLAGS="-march=core2 -msse4 -maes -mmovbe -mpclmul -mpopcnt -mcx16 -msahf -mtune=core2 -O2 -pipe" => success I finally switched back to gcc 4.4.3 for now, and I can confirm that -march=native gives me a -march=core2, not a -march=atom with it, and compilation works fine, so it's a 100% gcc 4.5 bug for sure xulrunner strips flags (even w/ USE=custom-optimization), so adding -mmovbe won't work. i just added a few 4.5 flags to the exception list so -mno-movbe will be allowed now. -march=atom -mno-movbe fixes the error. Your cpu is a Core i5 M 520, correct? As far as I can see they don't have movbe - it was introduced with the Atom series. So I would say that -march=native should instead be -march=core2 + a bunch of -m flags for that processor.
Could you provide the contents of /proc/cpuinfo and the output of
$ echo "int main() { return 0; };" > test.c
$ gcc-4.5.0 -march=native -v test.c -o test
Created attachment 230765 [details]
cpuinfo
yep, it's a core i5 M 520
Lou ~ # echo "int main() { return 0; };" > test.c [10:35:26 - 08/05/2010]
Lou ~ # gcc-4.5.0 -march=native -v test.c -o test [10:35:33 - 08/05/2010]
Utilisation des specs internes.
COLLECT_GCC=gcc-4.5.0
COLLECT_LTO_WRAPPER=/usr/libexec/gcc/x86_64-pc-linux-gnu/4.5.0/lto-wrapper
Target: x86_64-pc-linux-gnu
Configuré avec: /var/tmp/paludis/sys-devel-gcc-4.5.0/work/gcc-4.5.0/configure --prefix=/usr --bindir=/usr/x86_64-pc-linux-gnu/gcc-bin/4.5.0 --includedir=/usr/lib/gcc/x86_64-pc-linux-gnu/4.5.0/include --datadir=/usr/share/gcc-data/x86_64-pc-linux-gnu/4.5.0 --mandir=/usr/share/gcc-data/x86_64-pc-linux-gnu/4.5.0/man --infodir=/usr/share/gcc-data/x86_64-pc-linux-gnu/4.5.0/info --with-gxx-include-dir=/usr/lib/gcc/x86_64-pc-linux-gnu/4.5.0/include/g++-v4 --host=x86_64-pc-linux-gnu --build=x86_64-pc-linux-gnu --disable-altivec --disable-fixed-point --with-ppl --with-cloog --disable-lto --enable-nls --without-included-gettext --with-system-zlib --disable-checking --disable-werror --enable-secureplt --enable-multilib --enable-libmudflap --disable-libssp --enable-libgomp --enable-cld --with-python-dir=/share/gcc-data/x86_64-pc-linux-gnu/4.5.0/python --enable-java-awt=gtk --enable-languages=c,c++,java,fortran --enable-shared --enable-threads=posix --enable-__cxa_atexit --enable-clocale=gnu --with-bugurl=http://bugs.gentoo.org/ --with-pkgversion='Gentoo 4.5.0 p1.0'
Modèle de thread: posix
gcc version 4.5.0 (Gentoo 4.5.0 p1.0)
COLLECT_GCC_OPTIONS= '-v' '-o' 'test'
/usr/libexec/gcc/x86_64-pc-linux-gnu/4.5.0/cc1 -quiet -v test.c -D_FORTIFY_SOURCE=2 -march=atom -mcx16 -msahf -maes -mpclmul -mpopcnt -msse4.2 --param l1-cache-size=32 --param l1-cache-line-size=64 --param l2-cache-size=3072 -mtune=core2 -quiet -dumpbase test.c -auxbase test -version -o /tmp/ccmAagyo.s
GNU C (Gentoo 4.5.0 p1.0) version 4.5.0 (x86_64-pc-linux-gnu)
compiled by GNU C version 4.5.0, GMP version 4.3.2, MPFR version 2.4.2-p3, MPC version 0.8.1
heuristiques GGC: --param ggc-min-expand=100 --param ggc-min-heapsize=131072
le répertoire « /usr/local/include » est ignoré car inexistant
le répertoire « /usr/lib/gcc/x86_64-pc-linux-gnu/4.5.0/../../../../x86_64-pc-linux-gnu/include » est ignoré car inexistant
la recherche pour #include "..." débute ici :
la recherche pour #include <...> débute ici:
/usr/lib/gcc/x86_64-pc-linux-gnu/4.5.0/include
/usr/lib/gcc/x86_64-pc-linux-gnu/4.5.0/include-fixed
/usr/include
Fin de la liste de recherche.
GNU C (Gentoo 4.5.0 p1.0) version 4.5.0 (x86_64-pc-linux-gnu)
compiled by GNU C version 4.5.0, GMP version 4.3.2, MPFR version 2.4.2-p3, MPC version 0.8.1
heuristiques GGC: --param ggc-min-expand=100 --param ggc-min-heapsize=131072
Compiler executable checksum: 12095ba8527fd11f187b76cec8ccc6c5
COLLECT_GCC_OPTIONS= '-v' '-o' 'test'
/usr/lib/gcc/x86_64-pc-linux-gnu/4.5.0/../../../../x86_64-pc-linux-gnu/bin/as -V -Qy --64 -o /tmp/ccJvSQFc.o /tmp/ccmAagyo.s
GNU assembler version 2.20.1 (x86_64-pc-linux-gnu) using BFD version (GNU Binutils) 2.20.1.20100303
COMPILER_PATH=/usr/libexec/gcc/x86_64-pc-linux-gnu/4.5.0/:/usr/libexec/gcc/x86_64-pc-linux-gnu/4.5.0/:/usr/libexec/gcc/x86_64-pc-linux-gnu/:/usr/lib/gcc/x86_64-pc-linux-gnu/4.5.0/:/usr/lib/gcc/x86_64-pc-linux-gnu/:/usr/lib/gcc/x86_64-pc-linux-gnu/4.5.0/../../../../x86_64-pc-linux-gnu/bin/
LIBRARY_PATH=/usr/lib/gcc/x86_64-pc-linux-gnu/4.5.0/:/usr/lib/gcc/x86_64-pc-linux-gnu/4.5.0/../../../../lib64/:/lib/../lib64/:/usr/lib/../lib64/:/usr/lib/gcc/x86_64-pc-linux-gnu/4.5.0/../../../../x86_64-pc-linux-gnu/lib/:/usr/lib/gcc/x86_64-pc-linux-gnu/4.5.0/../../../:/lib/:/usr/lib/
COLLECT_GCC_OPTIONS= '-v' '-o' 'test'
/usr/libexec/gcc/x86_64-pc-linux-gnu/4.5.0/collect2 --eh-frame-hdr -m elf_x86_64 -dynamic-linker /lib64/ld-linux-x86-64.so.2 -o test /usr/lib/gcc/x86_64-pc-linux-gnu/4.5.0/../../../../lib64/crt1.o /usr/lib/gcc/x86_64-pc-linux-gnu/4.5.0/../../../../lib64/crti.o /usr/lib/gcc/x86_64-pc-linux-gnu/4.5.0/crtbegin.o -L/usr/lib/gcc/x86_64-pc-linux-gnu/4.5.0 -L/usr/lib/gcc/x86_64-pc-linux-gnu/4.5.0/../../../../lib64 -L/lib/../lib64 -L/usr/lib/../lib64 -L/usr/lib/gcc/x86_64-pc-linux-gnu/4.5.0/../../../../x86_64-pc-linux-gnu/lib -L/usr/lib/gcc/x86_64-pc-linux-gnu/4.5.0/../../.. /tmp/ccJvSQFc.o -lgcc --as-needed -lgcc_s --no-as-needed -lc -lgcc --as-needed -lgcc_s --no-as-needed /usr/lib/gcc/x86_64-pc-linux-gnu/4.5.0/crtend.o /usr/lib/gcc/x86_64-pc-linux-gnu/4.5.0/../../../../lib64/crtn.0
Created attachment 230885 [details, diff]
gcc45-PR44046-core2.patch
From patch: + case 0x1a: + case 0x1e: + case 0x1f: + case 0x2e: + /* FIXME: Optimize for Nehalem. */ + cpu = "core2"; + break; I presume this fix misdetection for Core i7. Am I right ? I assume it will. If you'd like to test it there's a copy in my personal overlay with all the patches we have so far. I applied the patch manually and it's worked. Are there any method as --param l2-cache-size=8192 to be used for Nehalem architecture ? By default 256kb is used when reverting to core2 from atom:--param l2-cache-size=256 Or, perhaps, am I doing something wrong ? Added to cvs. Will be included in the next patchset. Released in 1.2. |
