This is a misleading headline for a page makes the headline claim in the midst of one ranty paragraph mainly about how their Matlab code isn't performing as expected. If they want precise (in terms of timing and levels) control of visual and audio outputs on the computer, why are they using Matlab?
That's a good question and a reasonable response to my dismissive critique of their rant. Unfortunately I'm not sure what the answer is.
Relative to Psychtoolbox's longeviity, there are new ways of showing things on a computer, particularly APIs like WebGL and High Resolution Time for showing and timing things from a browser. On the desktop side, the variety of media libraries with python bindings has definitely grown during Psychtoolbox's adherence to Matlab.
Whatever the right approach is, I think the hard problem for Psychtoolbox is that a rather different kind of developer skillset will be required, both for the maintainers and the users. But that's not the fault of MacOS.
I think the maintainers and users are choosing to dedicate their skillset to the technology that serve their purpose the best. It is not MacOS's fault, and given Apple is a trillion-dollar company, it probably doesn't need participation from a niche scientific community anyway.
I happen to be conducting a brain imaging workshop today, and so far most of the hiccups have been caused by the quarantine rules in Catalina and up. Thankfully someone found a sudo command to fix it for a some of the participants, but I regretfully had to ask the rest to find a windows/linux machine to work on.
They're using Matlab because it's well known within the target community. Also, Psychtoolbox has been around for a long time and has been subject to studies that look specifically at precision.
Sure, but when you use a high-level language for low-level functionality you're always going to be vulnerable to changes in how the OS interacts with the run-time library of the language. I'm amazed that Psychtoolbox's use of Matlab has worked for this long, but, like Flash's obsolescence, maybe it's time to use a different language. In any case the problem is not OSX but the implementation choices made by Mathworks in Matlab for OSX.
> The stimulus presentation on MacOS used to be very good, up until version 10.12. In MacOS 10.13 something changed and it appears that a form of triple buffering has been added and, to date, none of the major experiment generators have managed to turn this off. As a result, since MacOS 10.13 stimuli appear always to be presented a screen refresh period later than expected, resulting in a delay of 16.66 ms in the apparent response times to visual stimuli.
8 comments
[ 3838 ms ] story [ 1808 ms ] threadRelative to Psychtoolbox's longeviity, there are new ways of showing things on a computer, particularly APIs like WebGL and High Resolution Time for showing and timing things from a browser. On the desktop side, the variety of media libraries with python bindings has definitely grown during Psychtoolbox's adherence to Matlab.
Whatever the right approach is, I think the hard problem for Psychtoolbox is that a rather different kind of developer skillset will be required, both for the maintainers and the users. But that's not the fault of MacOS.
I happen to be conducting a brain imaging workshop today, and so far most of the hiccups have been caused by the quarantine rules in Catalina and up. Thankfully someone found a sudo command to fix it for a some of the participants, but I regretfully had to ask the rest to find a windows/linux machine to work on.
> The stimulus presentation on MacOS used to be very good, up until version 10.12. In MacOS 10.13 something changed and it appears that a form of triple buffering has been added and, to date, none of the major experiment generators have managed to turn this off. As a result, since MacOS 10.13 stimuli appear always to be presented a screen refresh period later than expected, resulting in a delay of 16.66 ms in the apparent response times to visual stimuli.