Psychtoolbox>Screen.{mex*} subfunction

info = Screen(‘GetWindowInfo’, windowPtr [, infoType=0] [, auxArg1]);

Returns a struct with miscellaneous info for the specified onscreen window.

“windowPtr” is the handle of the onscreen window for which info should be
returned.

“infoType” If left out or set to zero, all available information for the
‘windowPtr’ is returned.

If set to -1, only the OpenGL context of the onscreen window is activated
(Expert use only!).

If set to 1, only the rasterbeam position of the associated display device is
returned (or -1 if unsupported).

If set to 2, information about the window server is returned (or -1 if
unsupported).

If set to 3, low-level window server settings are changed according to
‘auxArg1’. Do *not* use, unless you really know what you’re doing and have read
the relevant PTB source code!

If set to 4, returns a single value with the current activity status of
asynchronous flips. 1 if a Screen(‘AsyncFlipBegin’) was called and the flip is
still active, ie., hasn’t been finished with a matching Screen(‘AsyncFlipEnd’)
or Screen(‘AsyncFlipCheckEnd’);, zero otherwise.You can call this function with
an infoType of zero only if no async flips are active, or if the imaging
pipeline is fully enabled. This is why you need to use the special infoType 4 to
find out if async flips are active.

If set to 5, will start measurement of GPU time for render operations. The clock
will start on the next drawing command after this call. The clock will by
default stop at the next call to Screen(‘Flip’), Screen(‘AsyncFlipBegin’), or
Screen(‘DrawingFinished’). Measured time will include all the time spent by the
GPU for preparing the final visual stimulus image for the next flip, including
all post-processing operations performed by the imaging pipeline.

If you want to exclude time spent in image post-processing or just measure the
time spent for a defined set of drawing commands, you can stop the clock earlier
by calling this function with infoType set to 6. In that case, only GPU time
spent between the infoType=5 call and the infoType=6 call will be reported,
excluding any later drawing commands or imaging pipeline post-processing.

After the measured GPU operations complete, the elapsed rendertime will be
returned in the ‘GPULastFrameRenderTime’ field of the struct that you get when
calling with infoType=0.
Due to the asynchronous nature of GPU rendering, the measured time may not be
immediately available after the clock is stopped. In this case,
‘GPULastFrameRenderTime’ will be zero and you will need to repeat the infoType=0
query later.
Please note that not all GPU’s and operating systems support this function. If
the function is unsupported, a value of zero will be returned in the info struct
and by any call with ‘infoType’ of 5 or 6.

An ‘infoType’ of 7 does return the same information as the default ‘infoType’ 0,
but does not set the window ‘windowPtr’ as drawing target, does not activate its
OpenGL context and only returns information that is safe to return without
setting the window as drawing target.

An ‘infoType’ of 8 returns 1 if the X-Screens primary gpu uses the
modesetting-ddx under Linux.

An ‘infoType’ of 9 returns a struct with interop info needed for interop with
certain clients, currently tailored to the needs of OpenGL interop with the
OpenXR api on Linux and Windows.

The default info struct for ‘infoType’ 7 and the default ‘infoType’ 0 contains
all kinds of information. Just check its output to see what is returned. Most of
this info is not interesting for normal users, mostly provided for internal use
by M-Files belonging to Psychtoolbox itself, e.g., display tests.

The info struct contains the following fields

Beamposition: Current rasterbeam position of the video scanout cycle.
LastVBLTimeOfFlip: VBL timestamp of last finished Screen(‘Flip’) operation.
TimeAtSwapRequest: Timestamp taken prior to submission of the low-level swap
command. Useful for micro-benchmarking.
TimePostSwapRequest: Timestamp taken after submission of the low-level swap
command. Useful for micro-benchmarking.
VBLTimePostFlip: Optional flip completion timestamp from VBLANK timestamping.
Useful for micro-benchmarking.
OSSwapTimestamp: Optional flip completion timestamp from OS-Builtin
timestamping. Useful for micro-benchmarking.
GPULastFrameRenderTime: Duration of all rendering operations, as measured by
GPU, if infoType=5 was used.
RawSwapTimeOfFlip: Raw (uncorrected by high-precision timestamping) timestamp of
last finished Screen(‘Flip’) operation.
LastVBLTime: System time when last vertical blank happened, or the same as
LastVBLTimeOfFlip if the system doesn’t support queries of this property
(currently only OS/X does.)
VBLCount: Running count of vertical blank intervals since (graphics)system
startup. Or zero if notsupported by system. Currently only OS/X and Linux do
support this with some GPU’s.
VideoRefreshFromBeamposition: Estimate of video refresh cycle from beamposition
measurement method.
GLVendor, GLRenderer, GLVersion: Vendor name, renderer name and version of the
OpenGL implementation.
GLDeviceUUID: The unique device id if supported by the OpenGL implementation, an
empty field otherwise.
StereoMode: Currently selected stereomode, as requested in call to
Screen(‘OpenWindow’, …);
StereoDrawBuffer: Current drawbuffer for stereo display (0 = left eye, 1 = right
eye, 2 = None in mono mode).
ImagingMode: Currently selected imging pipeline mode, as requested in call to
Screen(‘OpenWindow’, …);
MultiSampling: Currently selected multisample anti-aliasing mode, as requested
in call to Screen(‘OpenWindow’, …);
MissedDeadlines: Number of missed Screen(‘Flip’) stimulus onset deadlines,
according to internal skip detector.
FlipCount: Total number of flip command executions, ie., of stimulus updates.
GuesstimatedMemoryUsageMB: Estimated memory usage of window or texture in
Megabytes. Can be very inaccurate or unavailable!
VBLStartLine, VBLEndline: Start/Endline of vertical blanking interval. The
VBLEndline value is not available/valid on all GPU’s.
SwapGroup: Swap group id of the swap group to which this window is assigned.
Zero for none.
SwapBarrier: Swap barrier id of the swap barrier to which this windows swap
group is assigned. Zero for none.
SysWindowHandle: Low-level windowing system specific window handle of the
onscreen window.
SysWindowInteropHandle: Low-level windowing system specific window-related
auxiliary handle of the onscreen window.
ExternalMouseMultFactor: Scaling factor to apply for remapping input coordinates
on some systems, e.g., by RemapMouse.m.
VRRMode: Actual selected mode for VRR stimulus onset scheduling (1 = auto maps
to actual choice): 0 = Off, 2 = Simple, 3 = OwnScheduled.
VRRStyleHint: Style hint code for the current active VRR stimulation timing
style, ie. what is assumed about timing behaviour of the paradigm.
VRRLatencyCompensation: Current estimate of average VRR swapbuffers latency,
used for compensating during VRR scheduling in ‘OwnScheduled’ mode.

The following settings are derived from a builtin detection heuristic, which
works on most common GPU’s:

GPUCoreId: Symbolic name string that roughly describes the name of the GPU core
of the graphics card. This string is arbitrarily
chosen to roughly group the cores by common capabilities (and quirks). Currently
defined are:
R100 = Very old ATI GPUs, R300 = GPU’s roughly starting at Radeon 9000, R500 =
Radeon X1000 or later, R600 = Radeon HD2000 or later.
NV10 = Very old NVidia GPUs, NV30 = NV30 or later, NV40 = Geforce6000/7000 or
later, G80 = Geforce8000 or later.
An empty GPUCoreId string means a different, unspecified core.

DisplayCoreId: Vendor of the display engine / display gpu, NVidia, AMD, Intel,
or same as ‘GPUCoreId’. May differ from ‘GPUCoreId’ in hybrid graphics laptops.
BitsPerColorComponent: Effective color depths of window/framebuffer in bits per
color channel component (bpc).
GLSupportsFBOUpToBpc: 0 = No support for framebuffer objects. Otherwise maximum
supported bpc (8, 16, 32).
GLSupportsBlendingUpToBpc: Maximum supported bpc for hardware accelerated
framebuffer blending (alpha blending).
GLSupportsTexturesUpToBpc: Maximum supported bpc for textures (8, 16, 32).
GLSupportsFilteringUpToBpc: Maximum supported bpc for hardware accelerated
linear filtering of textures (8, 16, 32).
GLSupportsPrecisionColors: 1 = Hardware can be fully trusted to rasterize
perfect 32 bpc colors in floating point color mode without special support of
PTB. 0 = Needs special (slower) support from PTB to work.
GLSupportsFP32Shading: 1 = All internal calculations of the GPU are done with
IEEE single precision 32 bit floating point, i.e., very accurate.
GPUMinorType: Numeric vendor specific gpu id. Chip family on NVidia, display
engine revision on AMD atm. -1 if unknown. Subject to change without notice!

###See also: OpenWindow, Flip, NominalFrameRate