>Psychtoolbox>PsychTests

GraphicsDisplaySyncAcrossDualHeadsTestLinux([screenids][, nrtrials=6000][, syncmethod=0])

Test synchronizity between the scanout/refresh cycles of the first two
display heads on screen ‘screenids’. If ‘screenids’ is omitted, we test the
two heads on the screen with maximal id. ‘nrtrial’ sample passes
with a sampling interval of roughly 1 msecs are conducted. ‘nrtrials’
defaults to 6000 samples.

Each sample consists of querying the current rasterbeam position on each
display head. At the end, the samples of both heads are plotted against each
other for comparison. On graphics cards without beamposition query support, e.g.,
Intel graphics, RaspberryPi, or modern AMD gpu’s with DCN display engine, instead
timestamps of the end of vblank are collected and plotted against each other for
comparison. In this case, the default duration if ‘nrtrials’ is omitted will be
roughly 10 seconds of runtime.

In case of older AMD gpu’s with DCE display engines (AMD Radeon Vega and earlier),
the following applies:

The optional parameter ‘syncmethod’ if set to 1 will try to synchronize all
display heads on AMD graphics cards, provided the ‘ScreenToHead’ mapping
(see Screen(‘Preference’,’ScreenToHead’,…); is set up properly for your setup.
On a single x-screen setup this will always be the case, whereas on a multi
x-screen setup tweaking will often be required.

A ‘syncmethod’ of 2 is also implemented for non-AMD graphics cards, but this
implementation is so experimental that it is guaranteed to fail miserably in
practice, so don’t bother trying.

The help text printed to the command window tells you how to interpret the plots.