Hello when using memory mapping single planar when and why would I want more than 1 memory buffer? Is there a way to check how many buffers I need?
eligijus: you typically need 3 buffers: most HW needs 2 buffers (one that's being filled, one ready for the next frame) and the third is the one userspace is processing.
More buffers may be needed if userspace processing time fluctuates so you need more buffering to prevent dropping frames.
For HW codecs even more buffers may be needed due to codec requirements.
but for regular video capture most applications allocate 3 or 4 buffers.
If the driver needs more buffers it will automatically increase the requested number of buffers to the minimum it really needs.
hverkuil, so allocating 5 buffers would be a safe bet?
yes
hey
since we have a standard api for cropping, composing and scaling, wouldn't it make sense to also incorporate api for pixel format conversion?
i mean, the selection api already operates on source data rectangle, but what if the source data is not in the same pixel format as the target buffer?
then e.g. the rounding constraints may differ depending on what comes on the input
(as an example, consider a framegrabber that can be fed with whatever pix format and convert it to the memory one, but there are some limitations e.g. depending on the component count of the format)