Sign up Calendar Latest Topics
 
 
 


Reply
  Author   Comment   Page 2 of 2      Prev   1   2
Edward

Administrator
Registered:
Posts: 245
Reply with quote  #26 
the nanmean issue can be resolved in any one of three simple ways:

a) install the matlab statstoolbox, which contains a nanmean function

b) create  your own nanmean and place it into the MonkeyLogic root folder, here is a link to a nanmean code sample - http://www.mathworks.com/matlabcentral/fileexchange/6837-
nan-suite/content/nansuite/nanmean.m

c) Update to the Git repository revision for January 13, 2016. This contains a nanmean.m file based on the link above.
0
Edward

Administrator
Registered:
Posts: 245
Reply with quote  #27 
Here are my MonkeyLogic latency results. In summary, using the new limitrate command appears to be a good solution for Matlab R2014 and later.

Controls: Matlab R2011_A_32BIT_WIN7 and R2013_B_32BIT_WIN7, latency < 1ms (running an nVidia Quadro FX570)

Before the limitrate change:
R2015_A_32BIT_WIN7 latency was ~20ms (running an nVidia Quadro FX570)
R2015_A_32BIT_WIN10 latency was ~ 9 ms (running geforce gtx960 graphics card)

After the limitrate change:
R2015_A_32BIT_WIN7 latency was < 1 ms, maybe even < ~0.1 ms (running an nVidia Quadro FX570)
R2015_A_32BIT_WIN10 latency was < 2 ms (running geforce gtx960 graphics card)

The only surprise is that the older Windows7 hardware and software ran faster. One likely explanation might be that in Win7 my video stimulus display (second monitor) was set to a resolution of 1360x768 @ 60Hz, but in Win10 I was testing on a ViewPixx monitor running at 1920x1200 @120hz. Nonetheless, these latency values are now within acceptable range on either system.

Final note, I was not able to test on a 64 Bit version of MonkeyLogic because there appears to be a codec problem.

I have updated the Git repository with a new release:
https://github.com/ryklin/MonkeyLogic_stable/releases/tag/v1.1.9

Here are the screenshots, please refer to the labels.

R2011_A_32BIT_WIN7 :

R2011_A_32BIT_WIN7.png 

R2013_B_32BIT_WIN7:
R2013_B_32BIT_WIN7.png 

R2015_A_32BIT_WIN7: (notice extremely long latency)
R2015_A_32BIT_WIN7.png 

R2015_A_32BIT_WIN7_LIMITRATE : (notice the latency is now very small after the limitrate correction)
R2015_A_32BIT_WIN7_LIMITRATE.png 

R2015_A_32BIT_WIN10: again notice the very long latency
R2015_A_32BIT_WIN10.png 

R2015_A_32BIT_WIN10_LIMITRATE: after the correction the latency is back to normal
R2015_A_32BIT_WIN10_LIMITRATE.png 

  



0
Edward

Administrator
Registered:
Posts: 245
Reply with quote  #28 
I repeated the Windows 10 latency test, this time matching resolution and refresh rate with the Win7 test. However, I did not see a performance increase. So at this point, I can't explain why my older hardware runs faster.

R2015_A_32BIT_WIN10_LIMITRATE_60Hz_low_res:

R2015_A_32BIT_WIN10_LIMITRATE_60Hz_low_res.png 

0
Edward

Administrator
Registered:
Posts: 245
Reply with quote  #29 
Here is another result from Win7 running Matlab 2013a 64 Bit using regular drawnow (no limitrate)

R2013_A_64BIT_WIN7: (not to run this test I renamed benchmarkmov.avi in monkeylogic.m -> initializing.avi in order for the indeo_codec to work).

R2013_A_64BIT_WIN7.png

0
Edward

Administrator
Registered:
Posts: 245
Reply with quote  #30 
I have made an additional update that allows this code to run properly on Matlab versions prior to 2014. It passes all latency tests, and should work on all versions of Matlab since 2011.

Here is the link to download:

https://github.com/ryklin/MonkeyLogic_stable/releases/tag/1.2.0

0
Previous Topic | Next Topic
Print
Reply

Quick Navigation:

Easily create a Forum Website with Website Toolbox.