* Added modeld.py (WIP)
* No more VisionIpcBufExtra
* Started work on cython bindings for runmodel
* Got ONNXModel cython bindings mostly working, added ModelFrame bindings
* Got modeld main loop running without model eval
* Move everything into ModelState
* Doesn't crash!
* Moved ModelState into modeld.py
* Added driving_pyx
* Added cython bindings for message generation
* Moved CLContext definition to visionipc.pxd
* *facepalm*
* Move cl_pyx into commonmodel_pyx
* Split out ONNXModel into a subclass of RunModel
* Added snpemodel/thneedmodel bindings
* Removed modeld.cc
* Fixed scons for macOS
* Fixed sconscript
* Added flag for thneedmodel
* paths are now relative to openpilot root dir
* Set cl kernel paths in SConscript
* Set LD_PRELOAD=libthneed.so to fix ioctl interception
* Run from root dir
* A few more fixes
* A few more minor fixes
* Use C update_calibration for now to exactly match refs
* Add nav_instructions input
* Link driving_pyx.pyx with transformations
* Checked python FirstOrderFilter against C++ FirstOrderFilter
* Set process name to fix test_onroad
* Revert changes to onnxmodel.cc
* Fixed bad onnx_runner.py path in onnxmodel.cc
* Import all constants from driving.h
* logging -> cloudlog
* pylint import-error suppressions no longer needed?
* Loop in SConscript
* Added parens
* Bump modeld cpu usage in test_onroad
* Get rid of use_nav
* use config_realtime_process
* error message from ioctl sniffer was messing up pyenv
* cast distance_idx to int
* Removed cloudlog.infos in model.run
* Fixed rebase conflicts
* Clean up driving.pxd/pyx
* Fixed linter error
* Enable nav features in modeld
* Enable mapsd/navmodeld
* Updated model_replay_ref_commit and added mapsd/navmodeld to test_onroad
* fixed process name
* always publish from mapsd
* time to onroad test
* run few times
* test on stripped builds
* manager
* cleanup
* split those out
* run first
---------
Co-authored-by: Comma Device <device@comma.ai>
* 10s lat
* Full length MPC
* redfine N
* Leave controls the same for now
* Updates
* use long plan in lat plan
* interp plan
* add new interp
* simplergit add selfdrive/controls/plannerd.py selfdrive/controls/
* expand to 10s
* revert this
* fix linter
* Update sconscripts
* fix test
* fix test
* fix test
* Revert "Update sconscripts"
This reverts commit 6e23c69dcebd5ed003e37e01921f6af7c31de0db.
* Dont import drive helpers
* better compile deps
* fix compile
* comment
* update replay
* Update plannerd time
* wip torqued
* add basic logic
* setup in manager
* check sanity and publish msg
* add first order filter to outputs
* wire up controlsd, and update gains
* rename intercept to offset
* add cloudlog, live values are not updated
* fix bugs, do not reset points for now
* fix crashes
* rename to main
* fix bugs, works offline
* fix float in cereal bug
* add latacc filter
* randomly choose points, approx for iid
* add variable decay
* local param to capnp instead of dict
* verify works in replay
* use torqued output in controlsd
* use in controlsd; use points from past routes
* controlsd bugfix
* filter before updating gains, needs to be replaced
* save all points to ensure smooth transition across routes, revert friction factor to 1.5
* add filters to prevent noisy low-speed data points; improve fit sanity
* add engaged buffer
* revert lat_acc thresh
* use paramsd realtime process config
* make latacc-to-torque generic, and overrideable
* move freq to 4Hz, avoid storing in np.array, don't publish points in the message
* float instead of np
* remove constant while storing pts
* rename slope, offset to lat_accet_factor, offset
* resolve issues
* use camelcase in all capnp params
* use camelcase everywhere
* reduce latacc threshold or sanity, add car_sane todo, save points properly
* add and check tag
* write param to disk at end of route
* remove args
* rebase op, cereal
* save on exit
* restore default handler
* cpu usage check
* add to process replay
* handle reset better, reduce unnecessary computation
* always publish raw values - useful for debug
* regen routes
* update refs
* checks on cache restore
* check tuning vals too
* clean that up
* reduce cpu usage
* reduce cpu usage by 75%
* cleanup
* optimize further
* handle reset condition better, don't put points in init, use only in corolla
* bump cereal after rebasing
* update refs
* Update common/params.cc
Co-authored-by: Adeeb Shihadeh <adeebshihadeh@gmail.com>
* remove unnecessary checks
* Update RELEASES.md
Co-authored-by: Adeeb Shihadeh <adeebshihadeh@gmail.com>
* ui_blit working
* simpler and working
* more believable that it's real
* working on device
* build on pc
* use hardware pc
* reduce cpu usage
* yuv conversion to EGL
* move everything to cameraview
* some cleanup
* more cleanup
* init array
* init images with std::map
* dont destroy images
* do destroy images
Co-authored-by: Comma Device <device@comma.ai>
* get log
* simplify two nonsense
* not needed
* libyuv is a joke
* clean up
* try small
* fast but not bad
* working
* clean up driverview
* simplified
* thats mirrored
* smol
* tweak
* ref is screen
* w/ ee
* update camera model
* no if TICI
* start
* update pose thresh
* less cpu more dsp
* new libyuv
* new snpe
* add files
* test
* should be fast
* update out len
* trigger test
* use master snpe
* add cereal
* update cereal
* refactor parsing
* missing ;
* get
* wrong type
* test model
* use driver data
* 10829278-72fe-4283-a118-2cef959ce174/1550
* no pf
* adapt driverview
* ;
* rhd learner
* update libyuv buildi x64
* ad4337ea
* remove blink slack
* test
* no
* use toggle
* b16
* fix for nv12
* 5b02cff5 both
* update test
* update cereal
* update cereal
* update cereal
* v2 packets
* revert libyuv
* no /
* update snpemodel
* ;
* memcpy
* fix test
* use toggle in driverview
* update power
* update replay
* Revert "update replay"
This reverts commit 1d0979ca59dbc89bc5890656e9501e83f0556d50.
* update model ref
* halve cpu
* fake 8bit onnx runner
* same thresh as report
* cereal master
Co-authored-by: Comma Device <device@comma.ai>
* rgb to nv12
* nv12 works (w memcpy)
* correct now
* no copy
* fix nv12 with fast debayer
* reverts of unused stuff
* ui use nv12
* comment out thumbnails for now
* rebase fix
* dm read nv12
* model read nv12
* fix ffmpeg encoder
* thumbnails from nv12
* replay to nv12
* python framereader support nv12
* remove hardcoded frame/buffer sizes
* fix build
* ffmpeg encoder fix buffers
* small cleanup
* reduce power usage test
* fix cpu usage test
* fix snapshot
* fix loggerd test
* bump cereal
Co-authored-by: Comma Device <device@comma.ai>
Co-authored-by: Joost Wooning <jwooning@gmail.com>
* enable encoderd
* correct enable line
* fix loggerd tests
* fix power draw and cpu tests
* correct cpu for encoderd
* fix a bug, video_writer is shared
* fix issue with not recording dcam
* add recording state
* wooo tests pass. encode id keeps counting
* core 3
* loggerd then encoderd
* stop loggerd first
* core 3 always online
* see the camera when we see encoder packet
* encoderd on small core uses 37%
* remove encoder logic from loggerd
* delete unit test that doesn't really make sense anymore
Co-authored-by: Comma Device <device@comma.ai>
* camerad: combine debayering and rgb_to_yuv opencl kernels
* fix border
* fix snapshot
* rename function
* update camerad cpu usage
* update camerad power draw
* vignetting equal to previously
* test other local worksize
* use less floats
* reduce amount of code
* move barrier back
* make faster
* fix corners
* cleanup
* cleanup
* allow more jitter on driverState timing
Co-authored-by: Comma Device <device@comma.ai>
* remove log_name
* log without compression
* fix tests
* remove extension for bootlog
* another test fix
* uploader compresses
* also compress in athena
* only compress qlog
* more generic check in do_upload
* fix bootlog compression
* lower loggerd cpu usage
* dont link against bz2
* set core affinity to little cluster
* handle old files
