I closed the issue (https://github.com/JuliaLang/julia/issues/10039) since a complete rebuild on the i7 appears to have put the i7 ahead of the i5 by a factor of 10. A full rebuild of 0.3.4 and 0.4 on the i7 now puts the two versions within the same order of magnitude, so that seems to have been resolved as well.
On Wednesday, February 4, 2015 at 8:10:22 PM UTC-8, Viral Shah wrote: > > Thanks for filing the issue. For future reference, could you post the > issue link here too? This is exactly why we have peakflops(), to detect all > these various idiosyncracies. There should ideally be no difference, since, > IIRC we are using the same OpenBLAS version in 0.3 and 0.4. > > -viral > > On Tuesday, February 3, 2015 at 9:05:41 AM UTC+5:30, Seth wrote: >> >> I rebuilt from scratch and am now getting >> >> julia> peakflops() >> 7.735382541885574e10 >> >> Which is about 30% slower than in 0.3.4 but much better than before. >> >
