As posted before, I am getting (consistent) test failures when running sage doctests on a core i7. I've been doing some more experiments and found some more interesting things about this.
I'm using two different boxes: core2: Intel(R) Core(TM)2 Quad CPU Q9550 @ 2.83GHz family=6 model=23 stepping=7 nehalem: Intel(R) Xeon(R) CPU E5520 @ 2.27GHz family=6 model=26 stepping=5 For the sake of this experiment I'm using the released sage-4.3 (doctest still fail on sage-4.3.alpha1 as reported before). A) sage-4.3 compiled on nehalem, ran on nehalem: --> doctest failures in symbolic/expression.pyx, numerical/optimize.py, interfaces/r.py (as reported earlier) B) sage-4.3 compiled on core2, ran on core2: --> all doctest pass C) sage-4.3 compiled on core2, ran on nehalem: --> all doctest pass D) sage-4.3 compiled on nehalem, ran on core2: --> many segfaults, the doctesting aborts in the end ----- Some conclusions: 1. the compilation on nehalem has some subtle issue which causes the doctest failures [actually the failures look all the same and related to numpy arrays, except the r failures, which may be due to a missing dependency since the system is a very bare bones debian lenny]. 2. whatever the failures are, they are triggered at compile time, since the code compiled on core2 works ok on the i7. 3. the sage-flags.txt are identical on both systems, namely "ssse3 mmx sse2 pni sse cmov". However, something seems to be using i7 features, causing segfaults when running the same binary on a core2. From /proc/cpuinfo, the flags are: nehalem: 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 pbe syscall nx rdtscp lm constant_tsc arch_perfmon pebs bts rep_good pni monitor ds_cpl vmx est tm2 ssse3 cx16 xtpr dca sse4_1 sse4_2 popcnt lahf_lm ida core2: 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 pbe syscall nx lm constant_tsc arch_perfmon pebs bts rep_good pni dtes64 monitor ds_cpl vmx smx est tm2 ssse3 cx16 xtpr pdcm sse4_1 lahf_lm tpr_shadow vnmi flexpriority Only suspicious flag I see is sse4_2, since both have sse4_1 (note that the old 65nm core2 don't have sse4_1 also). Best, Gonzalo
-- 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
