Ken Moffat
2018-03-28 00:49:44 UTC
Thought I'd mention this here, because with current gcc the use (or
omission!) of -march in CFLAGS isn't doing what I had expected.
Background: my buildscripts are on nfs, as are my sources, so in the
past I've found it convenient to bring up a new machine by building
what I hope will be a suitable LFS on an existing system, fixing up
the kernel, and then copying it over - usually followed by sorting
out changes to video drivers / kernel config / hardware monitoring.
The new box is my first Ryzen (nothing fancy, just an R3 1300X for
four real cores with good cache sizes), and I was hoping to use the
initial system to get to grips with whatever I may, or may not, need
to do to get nouveau running. Recently I've been using
-march=native in my CFLAGS (in the hope that compilation might be
less slow), but for this build I dropped that and made sure that gmp
used config.fsf.
It booted fine on the build machine (a Kaveri, model 15h), but when
I copied it over and tried to chroot from SystemRescueCD I got
'illegal instruction'. Eventually I dug out an OpenSuse disk and
tried that in recovery mode: same result, but it told me the problem
was in ld-2.27.so.
Googling around, I came upon old gentoo threads where -march=k8 was
recommended because Ryzens omit 4 instructions which were available
on Excavator CPUs. Return to start, do not pass go, do not collect
200 currency-units. So, I tried that. But the same result. That is
what really concerns me.
Then I got lucky - saved the kernel, modules, .config and tried
loading a backup from a different machine (an AMD Phenom - much
earlier x86_64, I think that is fam 10h) - with that I could chroot
from SystemRescueCD. Installed grub, fixed things up, booted. But
that wasn't a minimal system, so I've deferred looking at whatever
nouveau may, or may not, need for Xorg on this hardware (it's only a
GT710 - the bottom of the range, no fan and a vga output). It is now
getting towards the end of chapter 6 in a native build.
I'm really mentioning this in case anybody else gets a Ryzen and
tries the same approach.
ĸen
--
Before the universe began, there was a sound. It went: "One, two, ONE,
two, three, four" [...] The cataclysmic power chord that followed was
the creation of time and space and matter and it does Not Fade Away.
- wiki.lspace.org/mediawiki/Music_With_Rocks_In
--
http://lists.linuxfromscratch.org/listinfo/lfs-support
FAQ: http://www.linuxfromscratch.org/blfs/faq.html
Unsubscribe: See the above information page
Do not top post on this list.
A: Because it messes up the order in which people normally read text.
Q: Why is top-posting such a bad thing?
A: Top-posting.
Q: What is the most annoying thing in e-mail?
http://en.wikipe
omission!) of -march in CFLAGS isn't doing what I had expected.
Background: my buildscripts are on nfs, as are my sources, so in the
past I've found it convenient to bring up a new machine by building
what I hope will be a suitable LFS on an existing system, fixing up
the kernel, and then copying it over - usually followed by sorting
out changes to video drivers / kernel config / hardware monitoring.
The new box is my first Ryzen (nothing fancy, just an R3 1300X for
four real cores with good cache sizes), and I was hoping to use the
initial system to get to grips with whatever I may, or may not, need
to do to get nouveau running. Recently I've been using
-march=native in my CFLAGS (in the hope that compilation might be
less slow), but for this build I dropped that and made sure that gmp
used config.fsf.
It booted fine on the build machine (a Kaveri, model 15h), but when
I copied it over and tried to chroot from SystemRescueCD I got
'illegal instruction'. Eventually I dug out an OpenSuse disk and
tried that in recovery mode: same result, but it told me the problem
was in ld-2.27.so.
Googling around, I came upon old gentoo threads where -march=k8 was
recommended because Ryzens omit 4 instructions which were available
on Excavator CPUs. Return to start, do not pass go, do not collect
200 currency-units. So, I tried that. But the same result. That is
what really concerns me.
Then I got lucky - saved the kernel, modules, .config and tried
loading a backup from a different machine (an AMD Phenom - much
earlier x86_64, I think that is fam 10h) - with that I could chroot
from SystemRescueCD. Installed grub, fixed things up, booted. But
that wasn't a minimal system, so I've deferred looking at whatever
nouveau may, or may not, need for Xorg on this hardware (it's only a
GT710 - the bottom of the range, no fan and a vga output). It is now
getting towards the end of chapter 6 in a native build.
I'm really mentioning this in case anybody else gets a Ryzen and
tries the same approach.
ĸen
--
Before the universe began, there was a sound. It went: "One, two, ONE,
two, three, four" [...] The cataclysmic power chord that followed was
the creation of time and space and matter and it does Not Fade Away.
- wiki.lspace.org/mediawiki/Music_With_Rocks_In
--
http://lists.linuxfromscratch.org/listinfo/lfs-support
FAQ: http://www.linuxfromscratch.org/blfs/faq.html
Unsubscribe: See the above information page
Do not top post on this list.
A: Because it messes up the order in which people normally read text.
Q: Why is top-posting such a bad thing?
A: Top-posting.
Q: What is the most annoying thing in e-mail?
http://en.wikipe