Beefy Boxes and Bandwidth Generously Provided by pair Networks RobOMonk
"be consistent"
 
PerlMonks  

Re^2: The horribly slow Julia set viewer with the extremely inefficient plotting method

by kikuchiyo (Monk)
on Dec 17, 2009 at 15:11 UTC ( #813194=note: print w/ replies, xml ) Need Help??


in reply to Re: The horribly slow Julia set viewer with the extremely inefficient plotting method
in thread The horribly slow Julia set viewer with the extremely inefficient plotting method

Oh, no need to be so serious about it. I wrote this for fun. Streching the limits of perl (or Goo::Canvas specifically) was the point.
In fact, the slowness is caused by the fact that I create 360300 = 10800 rectangle objects, and every time the mouse moves, all of them gets updated. I wrote a modified version that uses C for the Julia iteration:

#!/usr/bin/perl use strict; use warnings; use Gtk2 -init; use Gtk2::Ex::Dialogs; use Gtk2::Gdk::Keysyms; use Glib ':constants'; use Goo::Canvas; use Inline C => Config => CCFLAGS => '-O3 -msse3 -mfpmath=sse -march=c +ore2 -ffast-math'; use Inline C => 'DATA'; # setting up default sizes my $xbox = 360; my $ybox = 300; my $boxsize = 2; my $centerx = 0; my $centery = 0; my $wsize = 4; my $pmax = $centerx + $wsize / 2; my $pmin = $centerx - $wsize / 2; my $qmax = $centery + $wsize / 2 * $ybox/$xbox; my $qmin = $centery - $wsize / 2 * $ybox/$xbox; my $A = ($pmax-$pmin)/$xbox; my $C = ($qmax-$qmin)/$ybox; #print "$pmax, $pmin, $qmax, $qmin, $A, $C\n"; my $kmax=100; # setting up palette my @palette = map { '#'.sprintf("%02x", int(sqrt($_/$kmax)*256)). sprintf("%02x", int(($_/$kmax)**3*256)). sprintf("%02x", int(sin(3.14159*$_/$kmax)*256)) } +(0..$kmax-1); $palette[-1] = '#000000'; # Create the main window my $win = new Gtk2::Window ( "toplevel" ); $win->signal_connect ("delete_event", sub { Gtk2->main_quit; }); #$win->signal_connect ("configure_event", \&win_expose); $win->set_title( "Julia set demo" ); $win->set_border_width (6); #$win->maximize; $win->set_resizable (0); $win->resize(700, 500); my $vbox = Gtk2::VBox->new (0, 6); $win->add ($vbox); # Creating the canvas my $canvas = Goo::Canvas->new(); $canvas->set_size_request($xbox*$boxsize, $ybox*$boxsize); $canvas->set_bounds(0, 0, $xbox*$boxsize, $ybox*$boxsize); my $white = Gtk2::Gdk::Color->new (0xFFFF, 0xFFFF, 0xFFFF); $canvas->modify_base('normal', $white ); $vbox->pack_start($canvas, 1, 1, 0); my $root = $canvas->get_root_item(); my $g = Goo::Canvas::Group->new($root); my $rect = Goo::Canvas::Rect->new( $g, 0, 0, $xbox*$boxsize, $ybox*$boxsize, 'line-width' => 1, 'stroke-color' => 'white', #invisible on white bg 'fill-color' => 'white', # must be filled for mouse event sensit +ivity 'title' => 'bg_white_rect', ); # Creating the colored rectangles that make up the Julia set my @pixels; foreach my $x (0..$xbox-1) { foreach my $y (0..$ybox-1) { $pixels[$x][$y] = Goo::Canvas::Rect->new( $g, $x*$boxsize, $y*$boxsize, $boxsize, $boxsize, 'line-width' => 0, #'stroke-color' => 'white', 'fill-color' => 'white', 'pointer-events' => 'GOO_CANVAS_EVENTS_NONE', #'title' => 'bg_white_rect', ); } } $g->signal_connect('button-release-event', \&on_background_motion_noti +fy); #'motion-notify-event', \&on_background_motion_notify); $canvas->can_focus(TRUE); $canvas->grab_focus($root); $win->show_all; main Gtk2; # replotting on mouse movement sub on_background_motion_notify { my ($group, $item, $ev) = @_; my $cy = $C*int(($ev->y)/$boxsize) + $qmin; #/// my $cx = $A*int(($ev->x)/$boxsize) + $pmin; #print $ev->x," ,",$ev->y," ,$cx ,$cy\n"; my ($k, $x, $y, $x2, $y2); foreach my $p (0..$xbox-1) { foreach my $q (0..$ybox-1) { $k = julia_iter($p, $q, $A, $C, $pmin, $qmin, $kmax, $cx, +$cy); $pixels[$p][$q]->set( 'fill-color' => $palette[$k], ); } } return TRUE; } __DATA__ __C__ int julia_iter (int p, int q, double A, double C, double pmin, double +qmin, int kmax, double cx, double cy) { int k = 0; double x = A*p + pmin; double y = C*q + qmin; double x2 = x*x; double y2 = y*y; while (x2+y2<4 && k<kmax) { y=2*x*y+cy; x=x2-y2+cx; x2=x*x; y2=y*y; k++; } return k; }
Profiling this shows that the julia_iter function (that calculates the new color) needs 3 us while the set method for the rectangles takes 12 us.


Comment on Re^2: The horribly slow Julia set viewer with the extremely inefficient plotting method
Download Code
Re^3: The horribly slow Julia set viewer with the extremely inefficient plotting method
by kikuchiyo (Monk) on Dec 17, 2009 at 17:01 UTC
    To illustrate the point, here is a version that does not use Goo::Canvas and rectangles at all, it builds up a string of rgb pixels instead and uses that to create a pixbuf, which in turn gets displayed on a DrawingArea.

    This is actually fast enough to be usable, as the animation is quite smooth on my machine.

      To illustrate the point

      .....now you're taiking..... nice example


      I'm not really a human, but I play one on earth.
      Old Perl Programmer Haiku

Log In?
Username:
Password:

What's my password?
Create A New User
Node Status?
node history
Node Type: note [id://813194]
help
Chatterbox?
and the web crawler heard nothing...

How do I use this? | Other CB clients
Other Users?
Others musing on the Monastery: (13)
As of 2014-04-18 20:02 GMT
Sections?
Information?
Find Nodes?
Leftovers?
    Voting Booth?

    April first is:







    Results (471 votes), past polls