Skip to contentSkip to conversation

Add Intel Core 2 machine support (prescott and nocona)

DISCUSSION Leon Woestenberg 3 participants 6 messages Aug 20, 2007 – Aug 22, 2007

lore
Share & tools
lore
#0 lore raw

Hello,

I would like to add support for Intel's Core 2 processors, in both 32 and 64 bit mode.

I am wondering what namespace to use for the PACKAGE ARCHs?

Given all the options[*], I think "x86-prescott" and "x86_64-nocona" are suitable:

x86_64-prescott= x86_32-core = x86_32-core2
x86_64-nocona = x86_64-core2

The naming is a bit to long to suite my taste, maybe prescott and nocona are enough?

I would like some input on this before I commit anything and pollute the namespace.

Regards,

Signature · 4 lines
Leon

[*]
http://gentoo-wiki.com/Safe_Cflags#Intel_Core_2_Duo.2FQuad_.2F_Xeon_51xx.2F53xx
#1 lore raw

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Quoted 13 lines · Leon Woestenberg

Hello,

I would like to add support for Intel's Core 2 processors, in both 32 and 64 bit mode.

I am wondering what namespace to use for the PACKAGE ARCHs?

Given all the options[*], I think "x86-prescott" and "x86_64-nocona" are suitable:

x86_64-prescott= x86_32-core = x86_32-core2
x86_64-nocona = x86_64-core2

Quoted 7 lines

The naming is a bit to long to suite my taste, maybe prescott and nocona are enough?

I would like some input on this before I commit anything and pollute the namespace.

Regards,

I would suggest just add conf/machine/core2.conf and another conf/machine/tune-core2.inc file to have the special gcc flags. I would not add it as a separate arch but stick with x86. To me it sounds more confusing to have x86 and core separated out into two different basic arches.

Just my 2 cents.

- -Khem

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.7 (Darwin)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFGygM3HnJKy6V6em4RAnk4AJ0Xw4JlBqos/8xA60x59GkAbOYvhwCffGm+ AMhMc3UVpCmd1qTCmlVT570=
=ES0x
-----END PGP SIGNATURE-----

#2 lore raw

Hello Khem,

thanks for the feedback.

Quoted 6 lines · Khem Raj

I would suggest just add conf/machine/core2.conf and another conf/machine/tune-core2.inc file to have the special gcc flags. I would not add it as a separate arch but stick with x86. To me it sounds more confusing to have x86 and core separated out into two different basic arches.

My current proposal is:

conf/machine/x86_32-prescott.conf
conf/machine/include/x86_32-prescott.conf
-arch=prescott
TARGET_ARCH=i686

conf/machine/x86_64-nocona.conf
conf/machine/include/x86_64-nocona.conf
-arch=nocona
TARGET_ARCH=x86_64

starting with GCC 4.3:
conf/machine/x86_64-core2.conf
conf/machine/include/x86_64-core2.conf
-arch=core2
TARGET_ARCH=x86_64

The remaining issue is that the TARGET_ARCH suggests that the resulting packages run on any i686 or x86_64 which is not necessarely true due to the instruction set being used.

Although we could create more generic machines, my direct interest is in (live) media encoding and some packages will use SSE3 and SSSE3.

Regards,

Leon.

#3 lore raw
Quoted 12 lines · Leon Woestenberg

My current proposal is:

conf/machine/x86_32-prescott.conf
conf/machine/include/x86_32-prescott.conf
-arch=prescott
TARGET_ARCH=i686

conf/machine/x86_64-nocona.conf
conf/machine/include/x86_64-nocona.conf
-arch=nocona
TARGET_ARCH=x86_64

Quoted 5 lines

starting with GCC 4.3:
conf/machine/x86_64-core2.conf
conf/machine/include/x86_64-core2.conf
-arch=core2
TARGET_ARCH=x86_64

Please don't use '_' in the names. It may work but it has the potential to mess up bitbake's override handling and would be better avoided. I know people have had issues with "x86_64" already due to this.

Quoted 6 lines

The remaining issue is that the TARGET_ARCH suggests that the resulting packages run on any i686 or x86_64 which is not necessarely true due to the instruction set being used.

Although we could create more generic machines, my direct interest is in (live) media encoding and some packages will use SSE3 and SSSE3.

When using optimised output like that you really do need to use a different TARGET_ARCH since as you say, the packages will not run on standard i686/x86_64...

Cheers,

Richard

#4 lore raw

Hello Richard,

thanks for your feedback!

Quoted 9 lines · Richard Purdie
Quoted 8 lines · Leon Woestenberg

My current proposal is:

...
conf/machine/x86_64-nocona.conf
conf/machine/include/x86_64-nocona.conf
-arch=nocona
TARGET_ARCH=x86_64

Quoted 10 lines
Quoted 5 lines

starting with GCC 4.3:
conf/machine/x86_64-core2.conf
conf/machine/include/x86_64-core2.conf
-arch=core2
TARGET_ARCH=x86_64

Please don't use '_' in the names. It may work but it has the potential to mess up bitbake's override handling and would be better avoided. I know people have had issues with "x86_64" already due to this.

What are the constraints TARGET_ARCH should meet?

Does it have anything to do with the GNU machine tuple?

Doesn't TARGET_ARCH propogated into several package's configure/Makefile's etc?

For example:

linux-libc-headers:

do_configure () {
    case ${TARGET_ARCH} in
        alpha*)   ARCH=alpha ;;
        arm*)     ARCH=arm ;;
        cris*)    ARCH=cris ;;
        hppa*)    ARCH=parisc ;;
        i*86*)    ARCH=i386 ;;
        ia64*)    ARCH=ia64 ;;
        mips*)    ARCH=mips ;;
        m68k*)    ARCH=m68k ;;
        powerpc*) ARCH=ppc ;;
        s390*)    ARCH=s390 ;;
        sh*)      ARCH=sh ;;
        sparc64*) ARCH=sparc64 ;;
        sparc*)   ARCH=sparc ;;
        x86_64*)  ARCH=x86_64 ;;
    esac
    if test !  -e include/asm-$ARCH; then
        oefatal unable to create asm symlink in kernel headers
    fi
Quoted 4 lines

When using optimised output like that you really do need to use a different TARGET_ARCH since as you say, the packages will not run on standard i686/x86_64...

That's why I suggested "x86_64-nocona".

I could only think of using "amd64" instead of "x86_64" but most distributions are favouring the last one over the first one recently.

Regards,

Leon.

#5 lore raw

Hello Richard,

Quoted 6 lines · Leon Woestenberg
Quoted 4 lines

When using optimised output like that you really do need to use a different TARGET_ARCH since as you say, the packages will not run on standard i686/x86_64...

That's why I suggested "x86_64-nocona".

Whoops, I didn't actually suggest x86_64-nocona as TARGET_ARCH.

BTW, I don't understand how the *value* of a variable may not contain underscores.

Do I misunderstand when I think bitbake overrides apply to variable names only?

Regards,

Leon