Conversation
This is needed to compile correcly without earlier header inclusion. Signed-off-by: Naushir Patuck <naush@raspberrypi.com>
- Add Buffer and BufferRef classes for managing V4L2 buffers - Add DmaHeap helper for dmabuf allocation - Add caching for v4l2_format to reduce ioctl calls - Improve buffer queue tracking - Add dmabuf allocator support to BackendDevice - Optimize stream on/off handling - Update convert example for new buffer API Signed-off-by: Naushir Patuck <naush@raspberrypi.com>
Correct the stride calucation and add YUV420SP, YUV420SP_COL128, and YUV420SP10_COL128 identifiers. Signed-off-by: Naushir Patuck <naush@raspberrypi.com>
Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>
Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>
Signed-off-by: Naushir Patuck <naush@raspberrypi.com>
Add a GStreamer element for hardware-accelerated image scaling and format conversion using the PiSP Backend. Features: - Single and dual output support - DMABuf input/output for zero-copy pipelines - Software buffer fallback - Crop support with per-output configuration - Format conversion between RGB, YUV420, YUV422, etc. Signed-off-by: Naushir Patuck <naush@raspberrypi.com>
|
Does any of this only work on 2712D0? |
|
It runs on C1, but doesn't give correct results when given limited range colorimetry or for 32bpp RGB output. You've implied the colorimetry issue could be fixed using other blocks, but I'll claim ignorance on that. AIUI the initial user wanting this will be on Compute Modules, so those are all D0. |
|
CI failing due to apt issues Fun! |
|
This week hasn't gone quite to plan, so I'll leave this for @naushir to sort out when he's back. There is one further tweak required to the component anyway to copy the input width and height to the output if not specified on the output, otherwise the PiSP setup fails as the width/height are both 0. |
In @naushir 's absence, @davidplowman / @njhollinghurst would one of you be able to review the first 7 patches as they affect the main libpisp library.
The last is the squash of all the development GStreamer patches, and I'm not expecting you to be able to review it. Hopefully I've now kept it clean and it's only touching the gst directory or meson build files, so if there are any major goofs then they have limited effect.