Quantcast

Got Homework?

Connect with other students for help. It's a free community.

  • across
    MIT Grad Student
    Online now
  • laura*
    Helped 1,000 students
    Online now
  • Hero
    College Math Guru
    Online now

Here's the question you clicked on:

55 members online
  • 0 replying
  • 0 viewing

agdgdgdgwngo

So I've got a simple C program that runs in 40 milliseconds on my x86 (1.6GHz Intel Atom). 40 milliseconds is not fast enough for me; I want it to happen in under 10 milliseconds. How do I optimize my C code? What are the sequence of steps that a programmer takes when optimizing code? How do I profile my program and find out what parts I need to refactor/ use a better algorithm, etc.

  • one year ago
  • one year ago

  • This Question is Closed
  1. agdgdgdgwngo
    Best Response
    You've already chosen the best response.
    Medals 1

    so far all i've done is use gcc -O2 . what else can i do to optimize my program to run in my calculator?

    • one year ago
  2. nick67
    Best Response
    You've already chosen the best response.
    Medals 1

    did you try with this ? -Ofast "Disregard strict standards compliance. -Ofast enables all -O3 optimizations. It also enables optimizations that are not valid for all standard compliant programs. It turns on -ffast-math and the Fortran-specific -fno-protect-parens and -fstack-arrays."

    • one year ago
  3. adhokshaj
    Best Response
    You've already chosen the best response.
    Medals 0

    Will you please provide the source code, so that I can have a look on it? You should not depend blindly on optimizations provided by compilers. Before optimizations, know when to optimize, what to optimize, and how to optimize. To answer first problem, if performance improvement is significant without TOOOOOOOOOO MUCH headach, go for it! To answer second problem, use a profiler. It will show you where most of the processing time is consumed in your program. Optimize that part first. For third one, you may use a better algorithm, or employ some *trikcy* fast solutions. It depends upon the case.

    • one year ago
  4. agdgdgdgwngo
    Best Response
    You've already chosen the best response.
    Medals 1

    I just tried -Ofast with no luck :( still giving me about 40 milliseconds. here is th e source: #include "prog.h" int main(int argc, char *argv[]) { switch (argc) { case 1: solve_from_stdin(); break; default: return -1; break; } return 0; }

    • one year ago
  5. adhokshaj
    Best Response
    You've already chosen the best response.
    Medals 0

    Please provide prog.h as well! In case you are working on some secret project, you may use profiler(s) or ask other project members.

    • one year ago
  6. nick67
    Best Response
    You've already chosen the best response.
    Medals 1

    which way are you measuring the time ?

    • one year ago
  7. agdgdgdgwngo
    Best Response
    You've already chosen the best response.
    Medals 1

    bash's builtin time command

    • one year ago
  8. nick67
    Best Response
    You've already chosen the best response.
    Medals 1

    did you try removing anything else but the pure main to have the minimum execution time ?

    • one year ago
  9. agdgdgdgwngo
    Best Response
    You've already chosen the best response.
    Medals 1

    this is what prog.h looks like #ifndef _PROG_H_ #define _PROG_H_ unsigned int a(int *, int *, const unsigned int, const unsigned int); unsigned int b(int *, int *, const unsigned int, const unsigned int, const unsigned int); unsigned int c(int *, const unsigned int); void solve_from_stdin(void); #endif /* ifndef _PROG_H_ */ what prfiler should i use?

    • one year ago
  10. adhokshaj
    Best Response
    You've already chosen the best response.
    Medals 0

    AMD APP Profiler is a free C/C++ Profiler

    • one year ago
  11. adhokshaj
    Best Response
    You've already chosen the best response.
    Medals 0

    Intel Parallel Studio also contains a profiler.

    • one year ago
  12. nick67
    Best Response
    You've already chosen the best response.
    Medals 1

    In the header file you just have declarations, not definitions, so nobody can understand what the code really does; but could you try to execute a main() without the function call and report the measured execution time ? #include "prog.h" int main(int argc, char *argv[]) { switch (argc) { case 1: // solve_from_stdin(); break; default: return -1; break; } return 0; }

    • one year ago
  13. agdgdgdgwngo
    Best Response
    You've already chosen the best response.
    Medals 1

    but those kits are both exclusive to Windows/visual studio :( without the solve_from_stdin(), time outputs 0.000 :-D so that one routine is taking 99.9% of cpu time :D

    • one year ago
  14. adhokshaj
    Best Response
    You've already chosen the best response.
    Medals 0

    Intell Parallel Studio is available for LINUX as well.

    • one year ago
  15. agdgdgdgwngo
    Best Response
    You've already chosen the best response.
    Medals 1

    that's awesome I'll check my package manager then

    • one year ago
  16. agdgdgdgwngo
    Best Response
    You've already chosen the best response.
    Medals 1

    alright I have intel parallel studio xe in my package manager

    • one year ago
  17. nick67
    Best Response
    You've already chosen the best response.
    Medals 1

    ok, now reactivate the call to the function but eliminate any action in the function body, then go on reactivating part of the code in the function body until you can understand which part is taking more execution time

    • one year ago
  18. nick67
    Best Response
    You've already chosen the best response.
    Medals 1

    old times profiling style :-)

    • one year ago
  19. agdgdgdgwngo
    Best Response
    You've already chosen the best response.
    Medals 1

    okay I deactivated procedure a() and I'm also getting 0.000 from bash

    • one year ago
  20. agdgdgdgwngo
    Best Response
    You've already chosen the best response.
    Medals 1

    but procedure a() calls procedure b()

    • one year ago
  21. nick67
    Best Response
    You've already chosen the best response.
    Medals 1

    then reactivate procedure a() but not procedure b()

    • one year ago
  22. agdgdgdgwngo
    Best Response
    You've already chosen the best response.
    Medals 1

    that decreased my time from 40 to 22 ms

    • one year ago
  23. nick67
    Best Response
    You've already chosen the best response.
    Medals 1

    does procedure a() do anything else apart calling procedure b() ?

    • one year ago
  24. agdgdgdgwngo
    Best Response
    You've already chosen the best response.
    Medals 1

    procedure a also calls itself before calling procedure b

    • one year ago
  25. nick67
    Best Response
    You've already chosen the best response.
    Medals 1

    it sounds strange, it should loop forever, unless there is some kind of counter to avoid it

    • one year ago
  26. agdgdgdgwngo
    Best Response
    You've already chosen the best response.
    Medals 1

    yeah at the start it tests for the value of (const int $1 + const int $2) / 2

    • one year ago
  27. nick67
    Best Response
    You've already chosen the best response.
    Medals 1

    is the function solve_from_stdin() recursive ?

    • one year ago
  28. agdgdgdgwngo
    Best Response
    You've already chosen the best response.
    Medals 1

    here's procedure b() void merge_int(int* left, unsigned int len_left, int* right, unsigned int len_right, int* end) { unsigned int i, j, k; for (i = j = k = 0; i < len_left && j < len_right; ++k) { if (left[i] < right[j]) { end[k] = left[i]; ++i; } else { end[k] = right[j]; ++j; } } for (; i < len_left; ++i, ++k) { end[k] = left[i]; } for (; j < len_right; ++j, ++k) { end[k] = right[j]; } }

    • one year ago
  29. nick67
    Best Response
    You've already chosen the best response.
    Medals 1

    is the b() procedure called just once in the a() procedure ?

    • one year ago
  30. agdgdgdgwngo
    Best Response
    You've already chosen the best response.
    Medals 1

    right just once

    • one year ago
  31. agdgdgdgwngo
    Best Response
    You've already chosen the best response.
    Medals 1

    but since a() calls itself recursively, it ends up calling b quite a lot of times :)

    • one year ago
  32. nick67
    Best Response
    You've already chosen the best response.
    Medals 1

    if so you could try to directly write the b() procedure's content into the a() procedure body, so that you save one function call time (context saving time into the stack)

    • one year ago
  33. agdgdgdgwngo
    Best Response
    You've already chosen the best response.
    Medals 1

    I got a seg fault :(

    • one year ago
  34. nick67
    Best Response
    You've already chosen the best response.
    Medals 1

    did you use correct types for left, right, end variables ?

    • one year ago
  35. agdgdgdgwngo
    Best Response
    You've already chosen the best response.
    Medals 1

    alright I'm going to refurbish my code and use a different data structure and see how it goes...

    • one year ago
    • Attachments:

See more questions >>>

Your question is ready. Sign up for free to start getting answers.

spraguer (Moderator)
5 → View Detailed Profile

is replying to Can someone tell me what button the professor is hitting...

23

  • Teamwork 19 Teammate
  • Problem Solving 19 Hero
  • You have blocked this person.
  • ✔ You're a fan Checking fan status...

Thanks for being so helpful in mathematics. If you are getting quality help, make sure you spread the word about OpenStudy.

This is the testimonial you wrote.
You haven't written a testimonial for Owlfred.