Thread: Running a console program with maximum speed

  1. #1
    Registered User
    Join Date
    May 2008
    Posts
    10

    Running a console program with maximum speed

    Hello,

    I am programming a "agent based" computer simulation in gcc. It basically does vector operations and gives as output some text files. Its an console application. It is written for scientific purposes. I would like to find a way to running this program dedicating the complete (or like 80%) of my processing power to it. File is attached. (code is messy I am not a professional programmer)

    Some more details for the curious:It is basically a simulation of object oriented agent, which trade, produce and consume in a network. On to of that they change there trading partners, to get a competitive advantage

    Thank you very much

    Davoud

  2. #2
    and the hat of int overfl Salem's Avatar
    Join Date
    Aug 2001
    Location
    The edge of the known universe
    Posts
    39,660
    Well how long does it take to run as it is?

    I see a lot of console and file I/O, and that kind of stuff just doesn't use all the CPU time no matter how much you "optimise" it.
    If you dance barefoot on the broken glass of undefined behaviour, you've got to expect the occasional cut.
    If at first you don't succeed, try writing your phone number on the exam paper.

  3. #3
    C++まいる!Cをこわせ!
    Join Date
    Oct 2007
    Location
    Inside my computer
    Posts
    24,654
    As for the I/O, try using buffering approaches. Buffering can help a lot with file I/O. However, it would be platform specific.
    And that indentation could use some work too. I found it mostly confusing.
    Quote Originally Posted by Adak View Post
    io.h certainly IS included in some modern compilers. It is no longer part of the standard for C, but it is nevertheless, included in the very latest Pelles C versions.
    Quote Originally Posted by Salem View Post
    You mean it's included as a crutch to help ancient programmers limp along without them having to relearn too much.

    Outside of your DOS world, your header file is meaningless.

  4. #4
    Registered User
    Join Date
    May 2008
    Posts
    10
    Every round takes 20 to 30 second, so for 200 rounds, 100 minutes. Which is a problem as I have to try different configurations.

    I do not know what buffering is, could you give me a hint where to start?

    I do not think that it is only a problem of the i/o operations, they are only in the reporting. When I increased some thinking about the future of the agents, which increases the number of vector operations, but not I/O operations as thinking of the agents does not get reported processing time wen up drastically.

  5. #5
    C++まいる!Cをこわせ!
    Join Date
    Oct 2007
    Location
    Inside my computer
    Posts
    24,654
    I'm not familiar with Linux, sorry.
    Another good idea is to try a profiler to find the bottlenecks in your code. Then you can start optimizing.
    Quote Originally Posted by Adak View Post
    io.h certainly IS included in some modern compilers. It is no longer part of the standard for C, but it is nevertheless, included in the very latest Pelles C versions.
    Quote Originally Posted by Salem View Post
    You mean it's included as a crutch to help ancient programmers limp along without them having to relearn too much.

    Outside of your DOS world, your header file is meaningless.

  6. #6
    and the hat of int overfl Salem's Avatar
    Join Date
    Aug 2001
    Location
    The edge of the known universe
    Posts
    39,660
    > vector <int> concatinate(vector <int> a,vector <int>b)
    Look into returning the result by reference (in some way). But I've just noticed that you never call it anyway. For the sake of people reading the code, prune out all the dead wood.

    > void sort(vector<int> & mir,vector<float> vec)
    I was going to say this is horrible as well (look at std::sort in algorithm), but then I found this isn't used either.
    If you dance barefoot on the broken glass of undefined behaviour, you've got to expect the occasional cut.
    If at first you don't succeed, try writing your phone number on the exam paper.

  7. #7
    Registered User
    Join Date
    May 2008
    Posts
    10
    yes I purged the stuff.

  8. #8
    Kernel hacker
    Join Date
    Jul 2007
    Location
    Farncombe, Surrey, England
    Posts
    15,677
    Use either gprof or oprofile to find out where your code spends its time. That way, you are not spending time optimizing someones guess of what takes time.

    --
    Mats
    Compilers can produce warnings - make the compiler programmers happy: Use them!
    Please don't PM me for help - and no, I don't do help over instant messengers.

  9. #9
    Kernel hacker
    Join Date
    Jul 2007
    Location
    Farncombe, Surrey, England
    Posts
    15,677
    One thought that strikes me after running gprof on the code is that you seem to maintain multiple deque's with the same data. E.g.
    Code:
    	    while ((_Price_List<Price_List[Auction_Color].end()) && (_Price_Response<Price_Response.end())){
    		if (*_Price_List<*_Price_Response) {                //  This is verz cricial
    		    Price_Help[Auction_Color].push_back(*_Price_List);
    		    Name_Help [Auction_Color].push_back(*_Name_List);
    		    _Price_List++;
    		    _Name_List++;
    		} else {
    		    Price_Help[Auction_Color].push_back(*_Price_Response);
    		    Name_Help [Auction_Color].push_back(friends[i]);
    		    _Price_Response++;
    		}
    	    }
    	    while (_Price_List<Price_List[Auction_Color].end()) {
    		Price_Help[Auction_Color].push_back(*_Price_List);
    		Name_Help[Auction_Color].push_back(*_Name_List);
    		_Price_List++;
    		_Name_List++;
    	    }
    	    while (_Price_Response<Price_Response.end()) {
    		Price_Help[Auction_Color].push_back(*_Price_Response);
    		Name_Help[Auction_Color].push_back(friends[i]);
    		_Price_Response++;
    	    }
    I think it would create a bit less overhead if you had a struct/class containing the price and the name, and store that in a single deque. This will reduce the overhead of push_back() and iterator updates.

    Here's the first five functions of gprof-output (running only 5 iterations):
    Code:
    Each sample counts as 0.01 seconds.
      %   cumulative   self              self     total           
     time   seconds   seconds    calls   s/call   s/call  name    
     10.88      5.86     5.86 117668390     0.00     0.00  cAgent::_Production(int*, float&)
      9.23     10.82     4.97    48064     0.00     0.00  cAgent::Ask_Prices()
      5.96     14.04     3.21 58874421     0.00     0.00  cAgent::MPX(int, int*, int)
      5.30     16.89     2.85 387271859     0.00     0.00  std::_Deque_iterator<float, float&, float*>::_Deque_iterator(std::_Deque_iterator<float, float&, float*> const&)
      5.17     19.66     2.78 470673560     0.00     0.00  powf
    Since _production is one of the top functions, and it calls pow() which is number 5 on the list, would it be possible to pre-calculate this calculaiton of pow():
    Code:
    void cAgent::_Production(int *Colors, float &y)
    {
        int i;
        y=productivity;
        for (i = 0; i < Number_of_Different_Colors; i++) {
    	y=y*pow(Colors[i],p[i]);
    	Colors[i]=1;
        }
    }
    You may gain some speed in thise type of thing:
    Code:
        for (i = 0; i < Number_of_Different_Colors; i++){
    	lColors[i]=i_Colors[i];
    	iColors[i]=i_Colors[i];
        }
    by using memcpy:
    Code:
        memcpy(lColors, i_Colors, sizeof(lColors));
        memcpy(iColors, i_Colors, sizeof(iColors));

    Edit: I ran the last change, and MPX() went from being third on the list to about tenth place in time taken. There may be similar cases in other places - I only scanned the code briefly.

    --
    Mats
    Last edited by matsp; 05-26-2008 at 05:13 PM.
    Compilers can produce warnings - make the compiler programmers happy: Use them!
    Please don't PM me for help - and no, I don't do help over instant messengers.

  10. #10
    Kernel hacker
    Join Date
    Jul 2007
    Location
    Farncombe, Surrey, England
    Posts
    15,677
    Code:
        int   lColors[Number_of_Different_Colors];
        for (i = 0; i < Number_of_Different_Colors; i++)  
    	lColors[i]=Colors[i];
    at the end of "Buy" seems to be completely unnecessary.

    --
    Mats
    Compilers can produce warnings - make the compiler programmers happy: Use them!
    Please don't PM me for help - and no, I don't do help over instant messengers.

  11. #11
    Kernel hacker
    Join Date
    Jul 2007
    Location
    Farncombe, Surrey, England
    Posts
    15,677
    And I can confirm that using a single struct, like this:
    Code:
    struct pricename
    {
        float price;
        int name;
    };
    ...
        deque <pricename> Price_List[Number_of_Different_Colors];
    ...
    and removing all references to deque<int> Name_List[] reduces the runtime by approximately 25&#37; (Went from around 52 seconds to around 39 seconds)

    There are likely other possible gains. The biggest hit is still _Production(), which uses up about 13% of the total runtime. But add to that about 25% of the calls to powf, so it's more like 16% of the total time.

    Just iterating through the deque using the pricename iterator takes about 5% of the runtime.

    --
    Mats
    Compilers can produce warnings - make the compiler programmers happy: Use them!
    Please don't PM me for help - and no, I don't do help over instant messengers.

  12. #12
    Registered User
    Join Date
    May 2008
    Posts
    10
    mats,

    thank you for all the effort you put!

    I have put the memcp.

    The structure is a bit difficult, As you have tried it out already could you send me the code?
    My mail [email protected]

    As you seam to be interested in this program, I attach an auxilliary porgram, which converts the output, to .dot files, which than can be converted using http://www.graphviz.org/ to JPG. Gladly I have no speed issues with that one. If you are interested I can send you the paper, for this model

    Davoud

  13. #13
    Cat without Hat CornedBee's Avatar
    Join Date
    Apr 2003
    Posts
    8,895
    You do realize that names starting with two underscores or an underscore and a capital are reserved for the compiler, right?
    All the buzzt!
    CornedBee

    "There is not now, nor has there ever been, nor will there ever be, any programming language in which it is the least bit difficult to write bad code."
    - Flon's Law

  14. #14
    Registered User
    Join Date
    May 2008
    Posts
    10
    I did not realize that underscores are reserved. Does it lead to trouble or is it just a question od style. I am not really a programmer, I just write a msters thesis

    Davoud

    ATTACHMENT: JPG - generator

  15. #15
    Kernel hacker
    Join Date
    Jul 2007
    Location
    Farncombe, Surrey, England
    Posts
    15,677
    Quote Originally Posted by arno-nyme View Post
    mats,

    thank you for all the effort you put!

    I have put the memcp.

    The structure is a bit difficult, As you have tried it out already could you send me the code?
    My mail [email protected]

    As you seam to be interested in this program, I attach an auxilliary porgram, which converts the output, to .dot files, which than can be converted using http://www.graphviz.org/ to JPG. Gladly I have no speed issues with that one. If you are interested I can send you the paper, for this model

    Davoud
    I think it's worth you doing the structure change, as it's actually your work. I did it quickly, I may have done it wrong (I don't think so), and it's on my home machine which I'm not at right now.

    One thought: Since pow() is so frequently used, perhaps switching to -ffast-math and using basic functionality instead of the complexities of "pow" (it does all sorts of quite complicated stuff).

    I wrote this:
    Code:
    static inline float mypowf(float x, float y)
    {
        return exp(log(x) * y);
    }
    In my little test-app, that almost halves the time used by pow (which is essentially all that the app does, so halving the time of the whole test-code).

    --
    Mats
    Compilers can produce warnings - make the compiler programmers happy: Use them!
    Please don't PM me for help - and no, I don't do help over instant messengers.

Popular pages Recent additions subscribe to a feed

Similar Threads

  1. Running a program
    By dingolay in forum C Programming
    Replies: 4
    Last Post: 10-01-2006, 09:03 PM
  2. Running a console application
    By maxorator in forum C++ Programming
    Replies: 4
    Last Post: 10-03-2005, 04:23 AM
  3. Replies: 3
    Last Post: 09-05-2005, 08:57 AM
  4. Get the PID of a console program
    By BianConiglio in forum Windows Programming
    Replies: 6
    Last Post: 05-24-2004, 05:24 AM
  5. C++ console program using Ms Visual Studio .NET
    By matheo917 in forum C++ Programming
    Replies: 4
    Last Post: 09-05-2002, 11:16 PM

Tags for this Thread