search

JuliaInv / MUMPSjInv.jl

MUMPS interface for Julia
18 stars 15 forks source link

Julia results in benchmark #5

Closed KristofferC closed 9 years ago

KristofferC commented 9 years ago

Hello,

I was trying some of your benchmarks with my interface to Pardiso and I am getting much better results for backslash in Julia than you are:

Grid size Pardiso (sec) Julia (sec) Speedup
8^3 0.00305 0.00354 1.16009
16^3 0.04376 0.01373 0.31367
24^3 0.09810 0.06853 0.69857
32^3 0.25277 0.29828 1.18005
48^3 1.94173 2.45429 1.26397
64^3 9.91453 16.01175 1.61498

This was with this file: https://github.com/JuliaSparse/MUMPS.jl/blob/master/tests/benchmarkDivGrad.jl

I tried on both 0.3 and 0.4 with similar results.

I see that your benchmarks results are quite old, maybe it would be fair to rerun them?

lruthotto commented 9 years ago

Thanks Kristoffer, I just updated the results. They are more closely to the ones you have now.

KristofferC commented 9 years ago

Thanks.

Just a thought, it would be cool if there could be some uniformity in the interfaces to all the sparse solvers that you, me and others are making. Then you could make more generic code that works with all types of sparse solvers and users with access to different sparse solvers could still run the same code.

Oh well :)