Struct array wrapped in object, performance issues

I run simulations using several objects and need those objects to store results in one place. I decided to go with a struct array. I thought it would improve performance if I make the struct a property of a handle object. My reasoning was if I pass the struct array around to objects it would be copied every time and reduce performance. I thought if I make it a property in a handle object, it would not be copied if I pass the object around. I ran into some interesting performance issues though and would appreciate any input. I use R2011b.

Here is a simplified class definition for the data storage object with only the construction method and a couple of properties: Pstruct will be a struct array for storing data and PNum is just a double.

classdef anobj < handle
    properties
        PStruct
        PNum=1;
    end
    methods
        function obj = anobj()
        end
    end
end

Here is a script filling the structure in an object with 1’s which is pretty fast:

clear all
a = anobj(); % an object
b = anobj(); % another object
ntrials=10; niterations=1000;
a.PStruct(ntrials,niterations).field1=0; % 'initialize' the struct array
for t=1:ntrials
    tic;
    for i=1:niterations
        a.PStruct(t,i).field1=1; % store data
    end
    toc;
end

yielding:

    Elapsed time is 0.001008 seconds.
    Elapsed time is 0.000967 seconds.
    Elapsed time is 0.000972 seconds.
    Elapsed time is 0.001206 seconds.
    Elapsed time is 0.000992 seconds.
    Elapsed time is 0.000981 seconds.
    Elapsed time is 0.000975 seconds.
    Elapsed time is 0.001072 seconds.
    Elapsed time is 0.000951 seconds.
    Elapsed time is 0.000994 seconds.

When instead I use a property of another object (=1 as well), changing the line within the loops to:

          a.PStruct(t,i).field1=b.PNum; % store data

I get:

Elapsed time is 0.112418 seconds.
Elapsed time is 0.107359 seconds.
Elapsed time is 0.118347 seconds.
Elapsed time is 0.127111 seconds.
Elapsed time is 0.138606 seconds.
Elapsed time is 0.152675 seconds.
Elapsed time is 0.162610 seconds.
Elapsed time is 0.172921 seconds.
Elapsed time is 0.184254 seconds.
Elapsed time is 0.190802 seconds.

Not only performance is orders of magnitude slower, but also there is a very clear trend (verified more generally) of slowing down with each trial. I don’t get it. Furthermore, if I instead use a standalone uninitialized struct array which is not wrapped in an object (this line replaces the one within the loops):

          PStruct(t,i).field1=b.PNum; % store data

I get ok performance with no trends:

Elapsed time is 0.007143 seconds.
Elapsed time is 0.004208 seconds.
Elapsed time is 0.004312 seconds.
Elapsed time is 0.004382 seconds.
Elapsed time is 0.004302 seconds.
Elapsed time is 0.004545 seconds.
Elapsed time is 0.004499 seconds.
Elapsed time is 0.005840 seconds.
Elapsed time is 0.004210 seconds.
Elapsed time is 0.004177 seconds.

I like the convenience of data storage object, but it adds huge overhead to execution time due to some weird interaction between struct arrays and objects. Does anybody know what is happening and how to fix this? Thanks.

[per isakson]

Matlab release?

[matfan001]

R2011b

Answer 1

[Jan]

 My reasoning was if I pass the struct array around to objects it would be copied every time and reduce performance.

This assumption is wrong. You can enable a debug mode to see, if the data are copied or if more efficient shared data copies are created:

format debug    % [EDITED], not "feature"!

Matlab uses a copy-on-write strategy, which provides input data as shared data copy until the first modification of the data appears. E.g. for structs this is very efficient, when only a specific field is changed, while all others are kept.

[matfan001]

I guess 'feature' does not work for R2011b. I'll look more into that.

[per isakson]

See Undocumented feature() function. feature debug doesn't work in R2013a either.

Maybe this is useful

    feature(‘MemStats’), feature(‘DumpMem’), feature(‘ProcessMem’) – these
    are memory reports that are even recommended by official MathWorks tech
    notes (1,2), newsletter and technical solutions (1,2). Numerous 
    references to these features can be found online.

[matfan001]

Thanks for this info. I tried the feature function, but I not sure how to use the information it provides to improve the code.

[Jan]

How strange! I have edited this minutes after the posting to change "feature debug" to "format debug". Obviously I made something wrong.

Answer 2

[matfan001]

@Jan:

I tried some code which passes a struct array to an object to store data, but it seems this approach incurs a similar performance penalty.

Here is the modified class definition: no properties, only a method that adds some data to a passed structure and returns it back.

classdef anobj < handle
    methods
        function obj = anobj()
        end
        function data_struct = add_data(obj,data_struct,trial,iteration)
            data_struct(trial,iteration).field1 = 1;
        end
    end
end

Here is the script using the object:

clear all
a = anobj(); % an object
ntrials=10; niterations=1000;
PStruct(ntrials,niterations).field1=0; % 'initialize' the struct array
for t=1:ntrials
    tic;
    for i=1:niterations
        PStruct = a.add_data(PStruct,t,i); % store data
    end
    toc;
end

Here is the result:

Elapsed time is 0.060769 seconds.
Elapsed time is 0.063645 seconds.
Elapsed time is 0.070625 seconds.
Elapsed time is 0.077176 seconds.
Elapsed time is 0.084017 seconds.
Elapsed time is 0.092867 seconds.
Elapsed time is 0.099135 seconds.
Elapsed time is 0.108683 seconds.
Elapsed time is 0.119721 seconds.
Elapsed time is 0.128796 seconds.

There is that trend again. Copying of (parts of) the struct still takes time.