Resynced to main userland

Update core userland

Synced to raspberrypi userland 20170123. No change to raspimjpeg

dtoverlay: Add makefile updates

cmake: define PROJECT_APIVER
this helps in compiling components which use packageconfig to poke for
library versions and requiring minimum supported version

Signed-off-by: Khem Raj <

cmake: define PROJECT_APIVER
this helps in compiling components which use packageconfig to poke for
library versions and requiring minimum supported version

Signed-off-by: Khem Raj <raj.khem@gmail.com>

show more ...

cmake: Add -fPIC to avoid linking issues with Pi2 compile flags

mmal: Add MMAL container parsing code

RaspiStill example code for YUV fast paths

Update RaspiStill framework to provide access to the new EGL image
targets that provide a fastpath for mapping the individual YUV planes to

RaspiStill example code for YUV fast paths

Update RaspiStill framework to provide access to the new EGL image
targets that provide a fastpath for mapping the individual YUV planes to
one byte per pixel GL_LUMINANCE buffers. These are still accessed as OES
textures so this is effectively a greyscale image created from a single
plane of a MMAL buffer.

Also added some example code (yuv, sobel) that demonstrates how to use
the new EGL image targets.

Also, add -gc to the command line to capture the GL frame-buffer as a TGA file.

Thanks to Tim Gover

show more ...

mmal: Avoid building mmal components and add rules for mmal_vc_diag

add -D_FILE_OFFSET_BITS=64 to fix 2gb file size limit

Fix 2gb file size limit by adding missing -D_FILE_OFFSET_BITS=64.
Previous pull request applied on 'next' branch instead of 'master'

Accelerated the khrn_hashword() function.

This function has been reimplemented in assembly, so that the data to be
processed can be preloaded ahead of when it is needed. To suit the quir

Accelerated the khrn_hashword() function.

This function has been reimplemented in assembly, so that the data to be
processed can be preloaded ahead of when it is needed. To suit the quirks
of the BCM2708 L2 cache, each 32-byte cache line is preloaded only once,
and only at its lowest address. Speed improvements of up to 2.2x have been
measured over the un-optimised code.

show more ...

Update master branch to next

Initial import of userland code