Hi Daniel,

Thanks for the reply. Are your suggesting mat[:,1] is making a copy of the 
entire matrix in the current version of Julia? I also tested the following 
code, for 10^4 times since each loop takes only a small amount of time:

@time for i in 1:10000;v=dtm[:,1]; w=dtm[:,2];end
# elapsed time: 0.011027003 seconds (17760000 bytes allocated)

-- Todd

On Monday, November 10, 2014 2:34:01 PM UTC+8, Daniel Høegh wrote:
>
> I think the reason why it is slow is due to indexing is making a copy of 
> your data and not passed by reference as in python. This behavior is going 
> to change in 0.4 where subarrays is introduced then the data is not copied.
>
> You could try it by.
> @time v=mat[:,1]; w=mat[:,2]
> This should take long time and allocate a lot of memory. 
>

Reply via email to