* remove local caching
* remove local caching
* reduce camerad power
* break stupid imx390 black level support
* handle edges
* now 13ms, vignetting is 'slightly' less correct
* halfs->floats and inlines, down to 12.9ms
* oops, fix float
* val from 12 is ushort
* don't decide vignetting in the debayer kernel
* 7.77 ms
* adding back black level support was free
* Revert "adding back black level support was free"
This reverts commit a841d17727886807a040dcf856ab22480fa93eff.
* minor
* rip out unused gain and black level, remove print
* save 150mW
* fix replay test
* fix top/bottom rows
* lame left right edge fix
Co-authored-by: Comma Device <device@comma.ai>
Co-authored-by: Joost Wooning <jwooning@gmail.com>
old-commit-hash: 72b52f8b7d74bf3ea6fe8bd4d59af1c44c83afdb
* 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>
old-commit-hash: 94b9972eb798408a3c412716b8d9eb8ddb9354bc
* 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>
old-commit-hash: 489fbb74b0f532c738cf8c06afe1a5f9ce1e1651
when only uploading no need to print success status or update diff
don't print succeed/fail when only uploading
old-commit-hash: 934bc1e6fb2ab9b9476438b2f9aea45a4dcb5298
* test failure()
* let's just change a tune here
* debug
revert
* debug
* use current commit, not ref_commit
fix
* need to figure out better place for this
* quick test
* test without upload
* temp
* use azure token
* fixes
* shouldn't need this
* debug
* debug
* not getting anything?
* does this mean nothing gets envvars?
* add azure token to docker environment variables
* quote
* move back
* clean up a bit
* more clean up
* like this sorting better
* replace flags with CI and clean up
* test FULL_TEST and minimalize diff a bit
* now test all
* revert tests comments
* remove flags
* revert this
revert this
* now make it fail
* now update ref_commit to last commit (make sure we can re-start this test if we commit before last one finishes uploading)
* fix
fix
fix
fix
* bad commit
* why is it not throwing an exception?
* debug
* URLFile returns empty bytes if using cache and remote file doesn't exist
* we always need to download anyway
* debug...
* duh, wrong file. but neither should have it
* add that back and just check explicitly
* check both
* clean up and make a diff
* stylize
* see if this is a better diff on files changed
* update refs
* revert changes
* only for owners or members
* if we have token access
* if we have token access
* if we have token access
* move up
* clean up
* revert
* move update refs to test_processes
* clean up
* update messages
* update msg
* update README and delete update_refs
* this isn't possible to reach anymore
* fix readme
* better help message
better help message
better help message
* only show basename when uploading, only if failed to find
* test diff
* fix printing old ref commit
* change to using
* update refs
* Revert "update refs"
This reverts commit 2e352a736a6de68e2c7064daa4e2e9409ce77686.
* revert
* ref refers to reference commit/logs, cur refers to current logs/commit (future ref)
* like for better
* Apply suggestions from code review
Co-authored-by: Adeeb Shihadeh <adeebshihadeh@gmail.com>
* Update selfdrive/test/process_replay/test_processes.py
* every time lol
Co-authored-by: Adeeb Shihadeh <adeebshihadeh@gmail.com>
old-commit-hash: d2d3b7b823bac0d24e2136f7a97c3aa4c6d9ae47
* test for comparing debayering results to ref commit
* fix run on device
* print more useful diff
* limit to diff output
* remove env var
* ref
* more useful diff
* use real local work size
* remove pyopencl install
* update ref
* include rgb_to_yuv
* fix diff
* get kernel build out of loop
Co-authored-by: Comma Device <device@comma.ai>
old-commit-hash: 9da9b773eba344e9e59aeaa8a2c0235aadbd1f44
* 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
old-commit-hash: 77a6f3d034a7b17d70b750b9ea454310f4b8edf1
* bump panda & cereal
* bump panda
* rest of references
* regen TOYOTA with safety param 578, TOYOTA2 with 329
* regen rest of routes
* update ref
* bump cereal
old-commit-hash: 4190d2608a536e1b28fb999a72bbf5b80810743d
* controlsd: better alert precedence for system malfunctions
* down to 20
* move that
* check can flags
* update refs
* update refs
old-commit-hash: be748c0d6a1a931a26b4ec741bbba71d4269b821
* add panda flag for toyota stock long with camera
* clean up
* Add 2022 RAV4 Hybrid from Philly
* fix wrong fw in interface, did this ever work?
* Must be a hybrid
* no radar parsing
* fix can error
* move to own platform
* generate docs
* fix
* Add 2022 Rav4 XSE Australia fingerprint parameters (#24303)
* Update values.py
Add 2022 Rav4 XSE Australia
* add commas
Co-authored-by: Shane Smiskol <shane@smiskol.com>
* bump panda
* wait, the camera doesn't even send 0x343, right?
* use a set instead, more obvious
* don't test without a parser
* bump panda
* flip panda flag
* bump panda
* add commas
* regen and update refs
* set to none by default
* revert parenthesis
* update comment
* bump panda
* regen and update refs
* add test models and update readme
* bump to master
Co-authored-by: BrettLynch123 <34538435+BrettLynch123@users.noreply.github.com>
old-commit-hash: c1caca104f8553d485334e33446bd7b8ec1fb8ca
* only use half of the available bytes
* move to set_safety_model
* regen and update refs
* last two
* bump to master
old-commit-hash: 66dd8f934cc84cdbd3e48e7b49836ec95f82c23a
* use deque for prev_cruise_buttons
* stash
* disallow engagement if resume or decel isn't pressed
* fix buttons
* reduce chances of causing 30-frame fault
Fix clu11
drive down cancel times
* Revert "reduce chances of causing 30-frame fault"
This reverts commit 0f54c051bf5a22b1f584718ec46e78bc3a82a0dd.
* consider pause/resume button
* stash
* Revert "stash"
This reverts commit 551ca7be6c45729f0747abc81cd81894cd621c32.
* sadly some cars need 8 op frames (identical to 4 updates from CLU11)
* add main button
* 6 should be safe
* use max
* clean up
* change to 4 samples and process all messages received, like panda
* bump panda
* test: replay segments with known re-engagements on-device
* Revert "test: replay segments with known re-engagements on-device"
This reverts commit 9730c3c14f942b82b6ed5ef2e81b8ae0126f3006.
* need prev_cruise_buttons as we don't run every CAN message through carstate
* more generic
* extend
* Update selfdrive/car/interfaces.py
Co-authored-by: Adeeb Shihadeh <adeebshihadeh@gmail.com>
* 3% faster at 1000000 loops
* update refs
Co-authored-by: Adeeb Shihadeh <adeebshihadeh@gmail.com>
old-commit-hash: efa98d2a0694b9a33a2fda3bebfae50fb4ba33cc
* initialize CUDA runtime before predictions
* dmond pauses modeld execution...
* done in other PR
old-commit-hash: 60b7114024181f79819c9a8b14863a8f23fb0252
* tici downclock
* fix that
* gp ref
* lower that a bit
* that was a scam
* perf
* update power
* increase that
Co-authored-by: Comma Device <device@comma.ai>
old-commit-hash: b88fb037aba5582a4c669fe972ddde17d0c4d9b6
Lukas Petersson