Re: Optimisation / parallelisation Problems

[somealias]

> I think your speed-up would be closer to linear if you don't try to use
hyperthreading.

I get a speed-up of 7.27x and 4.89x with 8 cores (16 cores = 8.88x and
6.44x) for Dataset 1 and 2 respectively. There is a somewhat linear speed-up
on 8-cores for the smaller dataset but not for the larger dataset.

> I think your best option would be to run both jobs at the same time, each
> on 4 cores.

I will give this a try and see how it goes once my work computer is
available. I started an experiment on a much larger dataset (500k+ examples)
after my previous post.

Thanks again!


--
View this message in context: 
http://octave.1599824.n4.nabble.com/Optimisation-parallelisation-Problems-tp4572922p4589440.html
Sent from the Octave - General mailing list archive at Nabble.com.