Hi all,
I've observed some strange timing issues when drawing off-screen
buffers to the screen in Windows. My expectation is that if I wait
for vertical refresh, and then call CopyWindow, it should take about
the same number of MS to blit the offscreen window. But this does not
seem to be the case; it takes either ~1 ms, or ~13 ms, regularly
alternating between these two as a function of when the offscreen
buffers were created. Specifically, the first 10 or so offscreen
buffers can be rendered in ~1ms, and the rest take ~13ms.
I've attached code that reproduces this effect. It creates 100 off
screen windows, and then draws them a total of 4 time each, and then
plots how long GetSecs thinks it took to blit them.
The alternating behavior seems to show up on all (2) machines that
I've been able to test it on.
Of course, 13ms jitter isn't much to worry about, except that I'm
doing priming experiments and I'd like to have all the noise on the
side of the human response, and as little on the part of the
computer...
Any ideas what's going on here?
----------------------------------------------
clear('buffer');
clear('timePerImage');
%!reschange -width=800 -height=600 -depth=32 -refresh=60
screen('closeall'); % close any orphaned windows
[window, screenRect] = screen(0, 'OpenWindow', [0 0 0],[],32); %
create a new window, 32 bit depth
HideCursor; % turn off mouse cursor
for i=1:100 % load images before displaying them on screen
buffer(i)=screen(window,'OpenOffscreenWindow',0);
screen(buffer(i), 'DrawRect', i*25, [10,10,780,580]);
end
screen(window, 'WaitBlanking'); % make sure that waitblanking has
been called at least once before entering loop
priority(2); % set high priority for max performance
for index=1:length(buffer)*4
screen(window, 'WaitBlanking'); % wait until refresh blanking
startAt = GetSecs;
screen('CopyWindow', buffer( 1+mod(index,length(buffer)) ),
window); % draw on screen (hopefully finishing before blanking
period),
timePerImage(index) = GetSecs - startAt;
end
priority(0);
screen('closeall');
%!reschange -width=1280 -height=1024 -depth=32 -refresh=85
plot(timePerImage*1000);
title('time to draw each frame in ms');
average_display_of_single_image_inMS = 1000*mean(timePerImage)
I've observed some strange timing issues when drawing off-screen
buffers to the screen in Windows. My expectation is that if I wait
for vertical refresh, and then call CopyWindow, it should take about
the same number of MS to blit the offscreen window. But this does not
seem to be the case; it takes either ~1 ms, or ~13 ms, regularly
alternating between these two as a function of when the offscreen
buffers were created. Specifically, the first 10 or so offscreen
buffers can be rendered in ~1ms, and the rest take ~13ms.
I've attached code that reproduces this effect. It creates 100 off
screen windows, and then draws them a total of 4 time each, and then
plots how long GetSecs thinks it took to blit them.
The alternating behavior seems to show up on all (2) machines that
I've been able to test it on.
Of course, 13ms jitter isn't much to worry about, except that I'm
doing priming experiments and I'd like to have all the noise on the
side of the human response, and as little on the part of the
computer...
Any ideas what's going on here?
----------------------------------------------
clear('buffer');
clear('timePerImage');
%!reschange -width=800 -height=600 -depth=32 -refresh=60
screen('closeall'); % close any orphaned windows
[window, screenRect] = screen(0, 'OpenWindow', [0 0 0],[],32); %
create a new window, 32 bit depth
HideCursor; % turn off mouse cursor
for i=1:100 % load images before displaying them on screen
buffer(i)=screen(window,'OpenOffscreenWindow',0);
screen(buffer(i), 'DrawRect', i*25, [10,10,780,580]);
end
screen(window, 'WaitBlanking'); % make sure that waitblanking has
been called at least once before entering loop
priority(2); % set high priority for max performance
for index=1:length(buffer)*4
screen(window, 'WaitBlanking'); % wait until refresh blanking
startAt = GetSecs;
screen('CopyWindow', buffer( 1+mod(index,length(buffer)) ),
window); % draw on screen (hopefully finishing before blanking
period),
timePerImage(index) = GetSecs - startAt;
end
priority(0);
screen('closeall');
%!reschange -width=1280 -height=1024 -depth=32 -refresh=85
plot(timePerImage*1000);
title('time to draw each frame in ms');
average_display_of_single_image_inMS = 1000*mean(timePerImage)