Slower execution under linux

Using Matlab 2012a under Windows 7..I wrote a code which includes one bottle nick at such line ....

K=zeros(Np,Np);

for i=1:ord

K=K+sparse(t(1:ord,:),repmat(t(i,:),ord,1),double(subs(Kv(:,i),Arg(Kv,1,1,6),Arg(Kv,1,2,6))),Np,Np);

end

%%%%

Note: Kv is symbolic matrix and Arg is a function that provides OLD and NEW and it uses a number of global variables.

Now, using Matlab 2013.. under ubuntu 12.04.. although the amount of memory consumed is much less but the execution time is extremely higher (80seconds became more than an hour)...why?

What am I missing??

[Jan]

What does this mean: "Arg provides OLD and NEW"?

Could you post a running piece of code, which reproduces the effect?

Answer 1

[Jan]

You compare R2012a/Win with R2013a/Linux. Then different timings are expected. Please post, what "extremely higher" means.

[Waleed Khedr]

it is an algorithm .... so in terms of one iteration:

R2012/win: 80 seconds

R2013/Linux: an hour untill now and not finished.

[Jan]

Does the profile reveal more details, e.g. if it is the repmat or the Arg() function?

The symbolic toolbox of the newer Matlab version might have substantially more power, with the drawback of needing much more time. The balance between power and run-time is always critical and known as a source for unexpected speed changes especially for symbolic maths.

 Example: Two trains 1 and 2 start a travel from city A to B, which have a distance of 160 km. Train 2 runs faster than 1 and when it reaches B it (immediately) returns until it meets train 1 on the railroads again, then it changes the direction again and so on. Train 1 has a speed of 40 km/h, train 2 75 km/h. Which is the traveled distance of train 2 when both trains reach the destination B?

A maths student can find the solution with some calculus in several minutes, a pupil of the elementary school in some seconds.

So do you find corresponding explanations in the release notes? I do neither have the symbolic toolbox nor 2013a.

[Waleed Khedr]

I made some tests actually ... and it seems to be "subs" command ....it goes very slow at some expressions ....Now the question is why "subs" goes slow ? is it something related to simplifications or assumptions?

Answer 2

[Asguard1]

Did anyone find a resolution to this problem? We have several simulations which must run perform sparse matrix manipulations under 2013a and under windows we are find a significantly faster performance than under linux. THe Linux is x86_64 quad socket AMD w/ 64 cores, running at 3Ghz. The Windows 7 is a quad core Xenon x86_64. The ODE is also faster under the windoze, than the linux as well.

[Waleed Khedr]

unlike my case, you are using two different machines. You need to check the multi-threading capabilities of each first. As far as I remember, this problem had something to do with the syntax of the sparse command itself in 2013a, you may need to check every piece of your command through Matlab's help documentary. Try your command with 2012a or with a newer version as well.