Description of problem:
After a fresh install from the current "rawhide" tree, a look at the install.log reveals that package atlas-sse has been installed. The system is a Pentium 4 style one which supports SSE2 according to /proc/cpuinfo:
processor : 0
vendor_id : GenuineIntel
cpu family : 15
model : 2
model name : Intel(R) Celeron(R) CPU 2.00GHz
stepping : 9
cpu MHz : 2000.000
cache size : 128 KB
fdiv_bug : no
hlt_bug : no
f00f_bug : no
coma_bug : no
fpu : yes
fpu_exception : yes
cpuid level : 2
wp : yes
flags : fpu vme de pse tsc msr pae mce cx8 apic mtrr pge mca
cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss
ht tm pbe up pebs bts cid xtpr
bogomips : 3993.63
clflush size : 64
Version-Release number of selected component (if applicable):
Steps to Reproduce:
1. Install current Fedora development on Pentium 4 system.
Installed ATLAS package only supports SSE but not SSE2.
Installed ATLAS package is the best performing one thus atlas-sse2 when the CPU has the corresponding flag.
This might actually be an anaconda issue. In this case the bug report should be reassigned.
I don't understand how this can happen, can you provide more information about what is actually pulling in atlas during installation.
BTW, I also think this must be anaconda or yum issue rather than atlas packaging bug.
# yum install numpy
Loaded plugins: refresh-packagekit
Setting up Install Process
--> Running transaction check
---> Package numpy.i686 0:1.3.0-6.fc12 set to be updated
--> Processing Dependency: libatlas.so.3 for package: numpy-1.3.0-6.fc12.i686
--> Processing Dependency: libptcblas.so.3 for package: numpy-1.3.0-6.fc12.i686
--> Processing Dependency: liblapack.so.3 for package: numpy-1.3.0-6.fc12.i686
--> Processing Dependency: libptf77blas.so.3 for package: numpy-1.3.0-6.fc12.i686
--> Running transaction check
---> Package atlas-sse.i686 0:3.8.3-9.fc12 set to be updated
--> Finished Dependency Resolution
Package numpy is installed by default when using the live CD. It appears that the dependency on atlas gets established only indirectly through the shared libraries.
Looking at package names by alphabetical order, atlas-sse precedes atlas-sse2 and gets pulled in which understandable. I wonder whether packages atlas-sse2 or atlas-sse3 do get pulled in automatically on **any** system capable of taking advantage them.
Other systems of mine are a Pentium III-M system for which atlas-sse is the right one. A further Opteron based system simply has package atlas installed which is more or less ok because on x86_64 systems, SSE2 is enabled at the compiler level. However, atlas-sse3 would have to be pulled in by hand, too.
This raises the interesting question how to handle this package selection issue if it were possible at reasonable cost ..
Likewise, on a Pentium III-M system, package atlas instead of atlas-sse gets pulled in during install from the current "rawhide" tree. Now that there is a base package atlas for Fedora 11/12 i[5/6]86 (see bug 510498) is there agreement that the decision of choosing the most efficient subpackage atlas-sse, atlas-sse2, .. be up to the user?
In this case, I would close the bug as WONTFIX.
(In reply to comment #3)
> Likewise, on a Pentium III-M system, package atlas instead of atlas-sse gets
> pulled in during install from the current "rawhide" tree. Now that there is a
> base package atlas for Fedora 11/12 i[5/6]86 (see bug 510498) is there
> agreement that the decision of choosing the most efficient subpackage
> atlas-sse, atlas-sse2, .. be up to the user?
> In this case, I would close the bug as WONTFIX.
This bug appears to have been reported against 'rawhide' during the Fedora 12 development cycle.
Changing version to '12'.
More information and reason for this action is here: