BUILD_NATIVE Vs Xeon E5 V2 Vs Xeon E5 V3

Home » Asterisk Users » BUILD_NATIVE Vs Xeon E5 V2 Vs Xeon E5 V3
Asterisk Users 2 Comments

hi,

i builded Asterisk 16.18 with BUILD_NATIVE on Xeon E5 v3 (Intel(R)
Xeon(R) CPU E5-2670 v3 @ 2.30GHz)

on machine when i want run Asterisk it hangs with Illegal instruction see below (Intel(R) Xeon(R) CPU E5-2660 v2 @ 2.20GHz)

is it only way to disable BUILD_NATIVE ?

or can be somehow specified to build for “something like xeon E5”?

OS is CentOS 7

stat(“/sys/fs/selinux”, {st_mode=S_IFDIR|0755, st_size=0, …}) = 0
brk(NULL)                               = 0x22aa000
brk(0x22cb000)                          = 0x22cb000
access(“/etc/selinux/config”, F_OK)     = 0
open(“/etc/pki/tls/legacy-settings”, O_RDONLY) = -1 ENOENT (No such file or directory)
access(“/etc/system-fips”, F_OK)        = -1 ENOENT (No such file or directory)
geteuid()                               = 0
stat(“/etc/sysconfig/64bit_strstr_via_64bit_strstr_sse2_unaligned”,
0x7ffcdace6c30) = -1 ENOENT (No such file or directory)
socket(AF_INET, SOCK_STREAM, IPPROTO_IP) = 3
ioctl(3, SIOCGIFCONF, {ifc_len=2 * sizeof(struct ifreq), ifc_buf=NULL}) = 0
ioctl(3, SIOCGIFCONF, {ifc_len=2 * sizeof(struct ifreq), ifc_buf=[{ifr_name=”lo”, ifr_addr={sa_family=AF_INET, sin_port=htons(0), sin_addr=inet_addr(“127.0.0.1″)}}, {ifr_name=”eth0”, ifr_addr={sa_family=AF_INET, sin_port=htons(0), sin_addr=inet_addr(“filtered”)}}]}) = 0
ioctl(3, SIOCGIFHWADDR, {ifr_name=”lo”, ifr_hwaddr={sa_family=ARPHRD_LOOPBACK, sa_data=00:00:00:00:00:00}}) = 0
ioctl(3, SIOCGIFHWADDR, {ifr_name=”eth0″, ifr_hwaddr={sa_family=ARPHRD_ETHER, sa_data=filtered}}) = 0
futex(0x910ba0, FUTEX_WAKE_PRIVATE, 2147483647) = 0
futex(0x910900, FUTEX_WAKE_PRIVATE, 2147483647) = 0
— SIGILL {si_signo=SIGILL, si_code=ILL_ILLOPN, si_addr=0x5a0cfe} —
+++ killed by SIGILL +++
Illegal instruction

Marek

2 thoughts on - BUILD_NATIVE Vs Xeon E5 V2 Vs Xeon E5 V3

  • Uh i forgot. the build machine is a VMware VM and asterisk build is inside a docker container (CentOS7) as RPM

    Dne 20/05/2021 v 11:57 marek napsal(a):

  • update

    my cpu family (hundred of VMs) is based on this (-march)

    corei7-avx core-avx2
    corei7

    note:

    you can check your march code like this

    gcc -c -Q -march=native –help=target |grep arch

    so it looks like i will stay with corei7  (anybody knows if avx is “big”
    performance save impact?)

    i’ll change build process to

    menuselect/menuselect –disable BUILD_NATIVE menuselect.makeopts

    then i need “force” CFLAGS=”-march=corei7″ in rpm build . any hints ?

    WARNING:

    there are changes in -march names through gcc versions

    https://stackoverflow.com/questions/943755/gcc-optimization-flags-for-xeon/25095818

    Dne 20/05/2021 v 12:31 marek napsal(a):