#80 Configure error on AMD barcelona CPU

2018-02-22T13:19:33.790000Z

The cpu frequencies reported by the kernel differ as follows:

# cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_cur_freq 
1197864
 # cat /sys/devices/system/cpu/cpu0/cpufreq/cpuinfo_cur_freq 
2000000

https://askubuntu.com/questions/800970/cpu-scaling-how-can-current-frequency-be-out-of-min-max-range

Configure error on AMD barcelona CPU

2018-02-22T12:49:40.205000Z

Configure fails for me on a AMD barcelona CPU, with ATLAS 3.11.39 and 3.10.2.
The architecture is not detected.

Oddly it insists that cpu throttling is on, which it isn't. The clock rate is also not right. Related may be that in Linux 4.15/14 they changed the cpu frequency reporting in /proc/cpuinfo. https://www.phoronix.com/scan.php?page=news_item&px=cpuinfo-fix-for-freq

 # ../ATLAS/configure -v 2
gcc -I/mnt/data/tmp/portage/sci-libs/atlas-3.10.2/work/ATLAS_shared/../ATLAS//CONFIG/include  -g -w -c /mnt/data/tmp/portage/sci-libs/atlas-3.10.2/work/ATLAS_shared/../ATLAS//CONFIG/src/atlconf_misc.c
gcc -I/mnt/data/tmp/portage/sci-libs/atlas-3.10.2/work/ATLAS_shared/../ATLAS//CONFIG/include  -g -w -o xconfig /mnt/data/tmp/portage/sci-libs/atlas-3.10.2/work/ATLAS_shared/../ATLAS//CONFIG/src/config.c atlconf_misc.o 
/tmp/ccAqsBVd.o: In function `ATL_tmpnam':
/mnt/data/tmp/portage/sci-libs/atlas-3.10.2/work/ATLAS_shared/../ATLAS//CONFIG/include/atlas_sys.h:224: warning: the use of `tmpnam' is dangerous, better use `mkstemp'
./xconfig -d s /mnt/data/tmp/portage/sci-libs/atlas-3.10.2/work/ATLAS_shared/../ATLAS/ -d b /mnt/data/tmp/portage/sci-libs/atlas-3.10.2/work/ATLAS_shared  -v 2
cmnd=make IRun_OS args="-v 2 " | fgrep 'OS='
probe_OS.o: In function `ATL_tmpnam':
/mnt/data/tmp/portage/sci-libs/atlas-3.10.2/work/ATLAS//CONFIG/include/atlas_sys.h:224: warning: the use of `tmpnam' is dangerous, better use `mkstemp'

OS configured as Linux (1)
cmnd=make IRun_asm args="-v 2  -O 1" | fgrep 'ASM='
probe_asm.o: In function `ATL_tmpnam':
/mnt/data/tmp/portage/sci-libs/atlas-3.10.2/work/ATLAS//CONFIG/include/atlas_sys.h:224: warning: the use of `tmpnam' is dangerous, better use `mkstemp'
system(make IRun_GAS_x8632 args="-v 2" MYFLAGS="-DATL_OS_Linux" 2> /dev/null | fgrep SUCCESS)

ierr=256 in command='make IRun_GAS_x8632 args="-v 2" MYFLAGS="-DATL_OS_Linux" 2> /dev/null | fgrep SUCCESS'!

OUTPUT:
=======
system(make IRun_GAS_x8664 args="-v 2" MYFLAGS="-DATL_OS_Linux" 2> /dev/null | fgrep SUCCESS)

Assembly configured as GAS_x8664 (2)
cmnd=make IRun_vec args="-v 2  -O 1 -s 2" | fgrep 'VECFLAG='
probe_vec.o: In function `ATL_tmpnam':
/mnt/data/tmp/portage/sci-libs/atlas-3.10.2/work/ATLAS//CONFIG/include/atlas_sys.h:224: warning: the use of `tmpnam' is dangerous, better use `mkstemp'

ierr=256 in command='make IRun_VSX MYFLAGS='-DATL_OS_Linux -DATL_GAS_x8664' 2> /dev/null | fgrep SUCCESS'!

OUTPUT:
=======

ierr=256 in command='make IRun_AltiVec MYFLAGS='-DATL_OS_Linux -DATL_GAS_x8664' 2> /dev/null | fgrep SUCCESS'!

OUTPUT:
=======

ierr=256 in command='make IRun_AVXMAC MYFLAGS='-DATL_OS_Linux -DATL_GAS_x8664' 2> /dev/null | fgrep SUCCESS'!

OUTPUT:
=======

ierr=256 in command='make IRun_AVXFMA4 MYFLAGS='-DATL_OS_Linux -DATL_GAS_x8664' 2> /dev/null | fgrep SUCCESS'!

OUTPUT:
=======

ierr=256 in command='make IRun_3DNow MYFLAGS='-DATL_OS_Linux -DATL_GAS_x8664' 2> /dev/null | fgrep SUCCESS'!

OUTPUT:
=======

ierr=256 in command='make IRun_NEON MYFLAGS='-DATL_OS_Linux -DATL_GAS_x8664' 2> /dev/null | fgrep SUCCESS'!

OUTPUT:
=======

Vector ISA Extension configured as  AVX (5,480)
cmnd=make IRun_arch args="-v 2  -O 1 -s 2 -a" | fgrep 'MACHTYPE='
probe_arch.o: In function `ATL_tmpnam':
/mnt/data/tmp/portage/sci-libs/atlas-3.10.2/work/ATLAS//CONFIG/include/atlas_sys.h:224: warning: the use of `tmpnam' is dangerous, better use `mkstemp'
ERROR: enum fam=0, chip=32526, model=48, mach=0
make[3]: *** [Makefile:106: atlas_run] Error 100
make[2]: *** [Makefile:404: IRunArchInfo_x86] Error 2

Architecture configured as  UNKNOWNx86 (39)
cmnd=make IRun_arch args="-v 2  -O 1 -s 2 -m" | fgrep 'CPU MHZ='
ERROR: enum fam=0, chip=32569, model=48, mach=0
make[3]: *** [Makefile:106: atlas_run] Error 100
make[2]: *** [Makefile:404: IRunArchInfo_x86] Error 2

Clock rate configured as 1198Mhz
cmnd=make IRun_arch args="-v 2  -O 1 -s 2 -n" | fgrep 'NCPU='
ERROR: enum fam=0, chip=32522, model=48, mach=0
make[3]: *** [Makefile:106: atlas_run] Error 100
make[2]: *** [Makefile:404: IRunArchInfo_x86] Error 2

Maximum number of threads configured as  4
cmnd=make IRun_pmake args="-v 2 -O 1 -t 4" | fgrep 'PMAKE='
probe_pmake.o: In function `ATL_tmpnam':
/mnt/data/tmp/portage/sci-libs/atlas-3.10.2/work/ATLAS_shared/../ATLAS//CONFIG/include/atlas_sys.h:224: warning: the use of `tmpnam' is dangerous, better use `mkstemp'
Parallel make command configured as '$(MAKE) -j 4'
cmnd=make IRun_arch args="-v 2  -O 1 -s 2 -t" | fgrep 'CPU THROTTLE='
ERROR: enum fam=0, chip=32695, model=48, mach=0
make[3]: *** [Makefile:106: atlas_run] Error 100
make[2]: *** [Makefile:404: IRunArchInfo_x86] Error 2
CPU Throttling apparently enabled!
It appears you have cpu throttling enabled, which makes timings
unreliable and an ATLAS install nonsensical.  Aborting.
See ATLAS/INSTALL.txt for further information
xconfig exited with 1

I also had an openblas detection issue which was then fixed: https://github.com/xianyi/OpenBLAS/issues/1139

System information:

# uname -a
Linux hostname 4.15.2-gentoo #18 SMP Sat Feb 17 01:29:36 -03 2018 x86_64 AMD A8-6410 APU with AMD Radeon R5 Graphics AuthenticAMD GNU/Linux
# cat /sys/devices/system/cpu/cpufreq/policy?/scaling_governor 
performance
performance
performance
performance
# cat /proc/cpuinfo 
processor   : 0
vendor_id   : AuthenticAMD
cpu family  : 22
model       : 48
model name  : AMD A8-6410 APU with AMD Radeon R5 Graphics
stepping    : 1
microcode   : 0x7030104
cpu MHz     : 1197.843
cache size  : 2048 KB
physical id : 0
siblings    : 4
core id     : 0
cpu cores   : 4
apicid      : 0
initial apicid  : 0
fpu     : yes
fpu_exception   : yes
cpuid level : 13
wp      : yes
flags       : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc rep_good acc_power nopl nonstop_tsc cpuid extd_apicid aperfmperf pni pclmulqdq monitor ssse3 cx16 sse4_1 sse4_2 movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt topoext perfctr_nb bpext ptsc perfctr_llc cpb hw_pstate vmmcall bmi1 xsaveopt arat npt lbrv svm_lock nrip_save tsc_scale flushbyasid decodeassists pausefilter pfthreshold overflow_recov
bugs        : fxsave_leak sysret_ss_attrs null_seg spectre_v1 spectre_v2
bogomips    : 3992.86
TLB size    : 1024 4K pages
clflush size    : 64
cache_alignment : 64
address sizes   : 40 bits physical, 48 bits virtual
power management: ts ttp tm 100mhzsteps hwpstate cpb acc_power [13]

processor   : 1
vendor_id   : AuthenticAMD
cpu family  : 22
model       : 48
model name  : AMD A8-6410 APU with AMD Radeon R5 Graphics
stepping    : 1
microcode   : 0x7030104
cpu MHz     : 1197.859
cache size  : 2048 KB
physical id : 0
siblings    : 4
core id     : 1
cpu cores   : 4
apicid      : 1
initial apicid  : 1
fpu     : yes
fpu_exception   : yes
cpuid level : 13
wp      : yes
flags       : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc rep_good acc_power nopl nonstop_tsc cpuid extd_apicid aperfmperf pni pclmulqdq monitor ssse3 cx16 sse4_1 sse4_2 movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt topoext perfctr_nb bpext ptsc perfctr_llc cpb hw_pstate vmmcall bmi1 xsaveopt arat npt lbrv svm_lock nrip_save tsc_scale flushbyasid decodeassists pausefilter pfthreshold overflow_recov
bugs        : fxsave_leak sysret_ss_attrs null_seg spectre_v1 spectre_v2
bogomips    : 3992.86
TLB size    : 1024 4K pages
clflush size    : 64
cache_alignment : 64
address sizes   : 40 bits physical, 48 bits virtual
power management: ts ttp tm 100mhzsteps hwpstate cpb acc_power [13]

processor   : 2
vendor_id   : AuthenticAMD
cpu family  : 22
model       : 48
model name  : AMD A8-6410 APU with AMD Radeon R5 Graphics
stepping    : 1
microcode   : 0x7030104
cpu MHz     : 1197.838
cache size  : 2048 KB
physical id : 0
siblings    : 4
core id     : 2
cpu cores   : 4
apicid      : 2
initial apicid  : 2
fpu     : yes
fpu_exception   : yes
cpuid level : 13
wp      : yes
flags       : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc rep_good acc_power nopl nonstop_tsc cpuid extd_apicid aperfmperf pni pclmulqdq monitor ssse3 cx16 sse4_1 sse4_2 movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt topoext perfctr_nb bpext ptsc perfctr_llc cpb hw_pstate vmmcall bmi1 xsaveopt arat npt lbrv svm_lock nrip_save tsc_scale flushbyasid decodeassists pausefilter pfthreshold overflow_recov
bugs        : fxsave_leak sysret_ss_attrs null_seg spectre_v1 spectre_v2
bogomips    : 3992.86
TLB size    : 1024 4K pages
clflush size    : 64
cache_alignment : 64
address sizes   : 40 bits physical, 48 bits virtual
power management: ts ttp tm 100mhzsteps hwpstate cpb acc_power [13]

processor   : 3
vendor_id   : AuthenticAMD
cpu family  : 22
model       : 48
model name  : AMD A8-6410 APU with AMD Radeon R5 Graphics
stepping    : 1
microcode   : 0x7030104
cpu MHz     : 1197.879
cache size  : 2048 KB
physical id : 0
siblings    : 4
core id     : 3
cpu cores   : 4
apicid      : 3
initial apicid  : 3
fpu     : yes
fpu_exception   : yes
cpuid level : 13
wp      : yes
flags       : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc rep_good acc_power nopl nonstop_tsc cpuid extd_apicid aperfmperf pni pclmulqdq monitor ssse3 cx16 sse4_1 sse4_2 movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt topoext perfctr_nb bpext ptsc perfctr_llc cpb hw_pstate vmmcall bmi1 xsaveopt arat npt lbrv svm_lock nrip_save tsc_scale flushbyasid decodeassists pausefilter pfthreshold overflow_recov
bugs        : fxsave_leak sysret_ss_attrs null_seg spectre_v1 spectre_v2
bogomips    : 3992.86
TLB size    : 1024 4K pages
clflush size    : 64
cache_alignment : 64
address sizes   : 40 bits physical, 48 bits virtual
power management: ts ttp tm 100mhzsteps hwpstate cpb acc_power [13]

#19 Patch allow support of FreeBSD Kernel in Debian

2017-08-28T15:35:43.790000Z

Thanks, I confirm that the patch fixes the build on GNU/kFreeBSD (without breaking GNU/Hurd).

#79 Fix endianness detection on 64-bit PowerPC

2017-08-25T12:57:51.534000Z

status: open --> closed-accepted
assigned_to: R. Clint Whaley

#78 Fix typos (preceeding -> preceding)

2017-08-25T12:56:05.864000Z

Thanks. Note that I'm not particularly interested in hunting spelling mistakes, but that one was dectected by lintian (the Debian package checker), so I felt compelled to fix it!

#19 Patch allow support of FreeBSD Kernel in Debian

2017-08-25T12:51:58.437000Z

status: open --> open-accepted
Group: --> developer

#19 Patch allow support of FreeBSD Kernel in Debian

2017-08-25T12:51:40.509000Z

The amusing part is that a patch for accepting HURD as Linux by Sylvestre was what broke it, if you read the comments of the if :)

If you look on git, I have taken your suggestion to move the GNU == HURD == Linux assumption to dead last.

Can you confirm that the attached probe_OS.c (extracted from git basefiles) fixes the problem?

#78 Fix typos (preceeding -> preceding)

2017-08-25T12:05:39.052000Z

status: open --> closed-fixed
assigned_to: R. Clint Whaley

Fix endianness detection on 64-bit PowerPC

2017-08-16T19:40:42.645000Z

ATLAS incorrectly detects big-endian 64-bit PowerPC platforms as being little endian.

The reason is that it tests whether the macro ORDER_LITTLE_ENDIAN is defined in CONFIG/src/SpewMakeInc.c. But this macro is defined even on big-endian systems. The right tests is to check whether BYTE_ORDER == ORDER_LITTLE_ENDIAN (at least with GCC).

I attach a patch for ATLAS 3.10.3 that is applied to the Debian package. It is straightforward to adapt it for ATLAS 3.11 branch, where two instances of the same test need to be fixed.

Fix endianness detection on 64-bit PowerPC

2017-08-16T19:40:42.645000Z

Ticket 79 has been modified: Fix endianness detection on 64-bit PowerPC
Edited By: R. Clint Whaley (rwhaley)
Status updated: u'open' => u'closed-accepted'
Owner updated: None => u'rwhaley'

Recent changes to patches

#80 Configure error on AMD barcelona CPU

Configure error on AMD barcelona CPU

#19 Patch allow support of FreeBSD Kernel in Debian

#79 Fix endianness detection on 64-bit PowerPC

#78 Fix typos (preceeding -> preceding)

#19 Patch allow support of FreeBSD Kernel in Debian

#19 Patch allow support of FreeBSD Kernel in Debian

#78 Fix typos (preceeding -> preceding)

Fix endianness detection on 64-bit PowerPC

Fix endianness detection on 64-bit PowerPC