I just tracked down a problem down to a Sage ATLAS miscompile on my CPU; disabling SSE3 did the trick. How do I proceed to fix it permanently and submit a patch? My /proc/cpuinfo is below.
The strange thing is that /proc/cpuinfo doesn't contain SSE3, but still ATLAS didn't crash, in fact it would only a seldom time randomly produce wrong results (NaN in complex double matrix multiplication) when I had earlier produced a plot...that kind of thing...reproducable with 80% rate perhaps, not deterministic. Is there a chance SSE3 is buggy and disabled by the Linux kernel, or similar? The end-result is pretty nasty so if nobody knows an answer I should perhaps take it to the ATLAS list... The SAGE_FAT_BINARYfix did the trick. How do I proceed to fix it permanently and submit a patch? My /proc/cpuinfo is below. Also, "uname -p" returns "unknown" on my system, causing SAGE_FAT_BINARY to fail (so I had to hack the spkg). Would it be ok to replace this with (or fall back to) "uname -m"? $ uname -p unknown $ uname -m x86_64 $ uname -i unknown $ uname -a Linux corcaroli.uio.no 2.6.18-128.7.1.el5 #1 SMP Wed Aug 19 04:00:49 EDT 2009 x86_64 GNU/Linux /proc/cpuinfo: processor : 0 vendor_id : GenuineIntel cpu family : 15 model : 4 model name : Intel(R) Xeon(TM) CPU 3.20GHz stepping : 3 cpu MHz : 3200.218 cache size : 2048 KB physical id : 0 siblings : 2 core id : 0 cpu cores : 1 apicid : 0 fpu : yes fpu_exception : yes cpuid level : 5 wp : yes flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm syscall nx lm constant_tsc pni monitor ds_cpl cid cx16 xtpr bogomips : 6404.61 clflush size : 64 cache_alignment : 128 address sizes : 36 bits physical, 48 bits virtual power management: (same repeated for 3 more cores) Dag Sverre -- To post to this group, send an email to sage-devel@googlegroups.com To unsubscribe from this group, send an email to sage-devel+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/sage-devel URL: http://www.sagemath.org
