Interesting question is what was being called that was taking 30 seconds then, and why no warning? Was it a fallback to a slow system blas/lapack, or a fallback to generic naive LU in Julia?
On Saturday, August 23, 2014 6:08:16 AM UTC-7, Milan Bouchet-Valat wrote: > > Le samedi 23 août 2014 à 08:55 -0400, Stefan Karpinski a écrit : > > There's a reasonable chance that Debian/Ubuntu has a pessimized > > OpenBLAS for the sake of portability. > It's also possible that the Debian package (these are the nighlties, > right?) should depend on (or recommend) openblas-devel, and not only > openblas, since Julia is currently not able to dlopen() a library if > there's only libopenblas.so.0, and no libopenblas.so (#6742). I know I > initially forgot to add this dependency in my RPM package. > > > Regards > > > On Sat, Aug 23, 2014 at 3:18 AM, Don MacMillen > > <[email protected] <javascript:>> wrote: > > sudo apt-get install libopenblas-dev > > > > > > julia> @time begin a = rand(5000,5000); b = rand(5000); x =a > > \b end; > > elapsed time: 2.517259033 seconds (400100776 bytes allocated, > > 1.44% gc time) > > > > > > Thanks John. > > > > On Friday, August 22, 2014 11:11:09 PM UTC-7, John Myles White > > wrote: > > Perhaps you have a defective BLAS installation for > > 0.3? > > > > > > — John > > > > > > On Aug 22, 2014, at 11:08 PM, Don MacMillen > > <[email protected]> wrote: > > > > > Hmmm, thx for the experiment. To clarify, I assumed > > > that the normal behavior would have been closer > > > to the 4.0 time, but I don't have anything earlier. > > > The 0.3 was just updated using apt-get on Ubuntu > > > 14.04 > > > (VM running on windows). 0.4 was of course cloned > > > and compiled. Just scraped and re-installed 0.3 > > > with apt-get, and I still see the same behavior. > > > (This time remembering to discard initial compile > > > time) > > > > > > > > > julia> @time begin a = rand(5000,5000); b = > > > rand(5000); x =a\b end; > > > elapsed time: 30.509950844 seconds (400100776 bytes > > > allocated, 0.11% gc time) > > > > > > > > > It's mystery to me. If I learn anything more, will > > > let you know. > > > Thanks again. > > > > > > > > > Don > > > > > > > > > > > > > > > On Friday, August 22, 2014 9:54:22 PM UTC-7, John > > > Myles White wrote: > > > I don’t see this behavior at all on my > > > system. > > > > > > After discarding an intial compilation step, > > > here’s what I get: > > > > > > 0.3 — elapsed time: 3.013803287 seconds > > > (400120776 bytes allocated, 1.77% gc time) > > > 0.4 — elapsed time: 2.920384195 seconds > > > (400120776 bytes allocated, 1.89% gc time) > > > > > > Also to clarify: do you mean to refer to 0.3 > > > as a regression relative 0.4? > > > > > > — John > > > > > > On Aug 22, 2014, at 9:50 PM, Don MacMillen > > > <[email protected]> wrote: > > > > > > > Is anyone else seeing the following? If > > > not, what could I have done to my env to > > > trigger it? > > > > > > > > Thx. > > > > > > > > julia> VERSION > > > > v"0.3.0" > > > > > > > > julia> @time begin a = rand(5000,5000); b > > > = rand(5000); x =a\b end; > > > > elapsed time: 31.413347385 seconds > > > (440084348 bytes allocated, 0.12% gc time) > > > > > > > > julia> > > > > > > > > ---------------- > > > > > > > > julia> VERSION > > > > v"0.4.0-dev+308" > > > > > > > > julia> @time begin a = rand(5000,5000); b > > > = rand(5000); x =a\b end; > > > > elapsed time: 1.686715561 seconds > > > (431769824 bytes allocated, 0.87% gc time) > > > > > > > > julia> > > > > > > > > > > > > > > > > >
