* zero ll patched big model
* probe in a subprocess so usbgpu lock gets released
* compiles
* runs
* num_jobs gets overwritten, use side effect
* poll tg devices
* make sure build crashes on missing gpu
* fine not to rely on Device.default
* seperate tg env for each model runner
* comment
* Revert "seperate tg env for each model runner"
This reverts commit f6470cc4258eaeb3e8e37907ef370871c9af5aa4.
* env is shared, gate on flag
* no fallback warp dev must be set
* build for current device only, unless pc/release
* comment
* list
* listen for plug in
* add icon to status bar, read params on every frame (?)
* log available devices
* try copy out when loading?
* Revert "log available devices"
This reverts commit e8c52a5d59456d4820ecb13b99a6c46ea1386a20.
* Revert "try copy out when loading?"
This reverts commit 518f403aa03faeda1950fe3dbce0d9e4c1584455.
* don't trigger device probe/caching on modeld prepare
* re-export with ll and road edges
* dont cache devices in manager process
* get USBGPU from params
* no usbgpu env
* missed one
* sconscript don't poll
* unconditional env
* always explicitely set devices on input tensors
* set DEV so amd uses right compiler and iface??
* fix flag
* bump tg
* rm xdg_cache_home
* tg don't bump all the way
* missing gmmu=0 at compile time
* dm set dev
* tg backend
* update gitignore
* missing import
* unused imports
* rely on Device.DEFAULT at compile time (already the case bc onnxrunner)
* comments
* dm warp needs DEV set too
* build both smol and big
* misc typos
* set dev at compile time
* don't need
* DEV=CPU when getting metadata, ensure we don't grab gpu lock
* this would also grab lock
* put bool
* warp compile always prepare only
* missed one
* poll ui
* missing here
* don't force usbgpu at build time
* tmp patch fetch_fw
* catch all, follow hardwared patterns
* simpler
* compile make input queues
* revert this
* group this more readable
* rm empty line
* make dummy frame using numpy
* revert compile make input queues
* no compiler at runtime
* cleanup
* fine to rebuild all on change to device node for now
* fix usbgpu_present
* fix sconscript
* no size in header stream decompress
* DEBUG=2
* minimal viable feedback
* egpu gray
* oops
* gotta do this actually
* modeld build only depends on modeld devices
* don't ship onnx to release? or chunk
* don't need
* can only set compiler on dev=
* none device works, will use default
* make linter happy
* chunk agnostic onnx input to compile_modeld
* chunk big onnx
* +x chunker
* fix #!
* and don't ship chunked onnx to release
* firmware now in correct location
* better err on missing onnx/chunk
* SConscript also need to accept chunked onnx
* metadata also need to load maybe chunked
* dedupe cmd
* this needs to be on cpu
* devices are set in the tgflags, we already depend on them
* rebuilding on changed order is fine
* read file chunked can already load either chunked or not
* chunk all big onnx
* less confusing
* unused import
* python device to load onnx bytes
* default device for runners, python for metadata
* why not
* chunked to shm
ui: nonblocking writes for ExperimentalMode + DriverView toggles + cycle-restart
All four put calls fire from the main render thread on user interaction
and block on disk fsync, causing visible UI frame spikes.
Each consumer is safe under nonblocking:
- onboarding inactivity_callback: write-and-forget (~25 ms saved)
- home long-press exp toggle: ui_state.experimental_mode owns visual state (~10 ms)
- onroad exp_button: $held_mode + selfdriveState owns visual state
- restart_needed_callback (OnroadCycleRequested): cross-process signal,
consumer is selfdrived which polls the param
BigParamControl-driven toggles in settings (developer.py, toggles.py)
are intentionally left blocking — those widgets refresh visual state
from disk every frame to mirror external changes, which would race a
nonblocking write.
* log raylib fps
* log fps from frame time
* whitespace
* or just log frame time?
* init pubmaster in init window
* yield timings
* bump ordinal
* dont log on screen off
* UInt
* lint
* /0
* oops
* oops2
* more precise raylib frame time, can get fps with 1/ft
* don't crash on screen off
* NL
* no _
---------
Co-authored-by: Shane Smiskol <shane@smiskol.com>
* ui: speed up `mici/AugmentedRoadView` by optimizing _calc_frame_matrix caching (#36669)
speed up AugmentedRoadView by optimizing _calc_frame_matrix caching
* ui: apply rect.x/y as a 2D screen offset post-projection
Removes the parent rect's screen position from the cached
video_transform passed to ModelRenderer. Instead, ModelRenderer
applies (rect.x, rect.y) as a 2D offset to projected_points at draw
time.
Why this works: the rect.x/y term in video_transform gets multiplied
by P_calib[2] before the perspective divide, then divided by the same
value, which cancels out to a simple additive shift on the final
screen coordinate. So adding x to video_transform[0,2] is equivalent
to adding x to screen_x post-projection.
Net effect: the cache key in _calc_frame_matrix no longer needs to
include rect.x/y. Cache stays hot under translation (e.g. swiping
between layouts), and the model overlay tracks the camera at 60Hz
because the offset is updated cheaply each frame.
This addresses the original revert reason for #36669 (model overlay
visually desyncs from camera during a home<->onroad swipe).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* ui: update model screen offset every frame, not just on cache miss
set_screen_offset() was called inside the cache-miss path of
_calc_frame_matrix, so the offset only updated at ~20Hz (calib publish
rate). The model overlay visibly lagged the camera during a swipe.
Move it out of the cached path so it updates each frame.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fully clean up
* shorter cmts
* make non
* clean up
* fix ty
---------
Co-authored-by: Dean Lee <deanlee3@gmail.com>
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* ui: keep arc_bar_pts cache hot when bar is translated
The LRU cache key included (cx, cy), which change every frame when
the parent widget's rect translates (e.g. during a scroll animation
in MainLayout). That made every cache lookup miss, recomputing the
arc geometry twice per frame and dropping fps.
Compute the shape at origin (so the cache key only depends on the
geometry: radius, thickness, angles), then translate to (cx, cy)
after. Cache stays hot under translation.
Measured on comma four (engaged, scrolling): ~5fps recovered while
moving between layouts.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fully clean up
* fully clean up
---------
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix frame under text
* claude always taking the hard path
* cmt
* it still shows on start
* revert
* fix
* wait to show
* already a bug, don't fix here
* cmt
* cmt
* spellings
* unused
* no roll
* lil more
* lil more
* one more
* policy enum
* better trans
* set_timer -> set_policy
* set_timer -> set_policy
* no yaonet
* del redundant code
---------
Co-authored-by: ZwX1616 <zwx1616@gmail.com>
* c4 body ui
* clean up diff
* clean up
* default bodyview debug with True
* remove battery indicator and fix close settings bug
* organize debug file
* clean
* clean up frame index
* remove unneccessary is body check
* update bodyview
* remove joystick_debug_mode based events
* remove debug script
* apply suggestions
* clean diff
* clean diff
* move ignition message
* save a line
* remove visibility set and fix tici body face when sidebar open
* remove explicit textColor offroad message
* remove unused imports
* revert pairing dialog
* apply suggestions
* add body specific icon
* add body icon for homescreen
* set mode for state on tici after on body changed
* tiny
* tweak
* v
* tweaks
* icon ratio was wrong!
* same order
* rm
* apply suggestions
* remove commented lines in animation
* onroad click callback was on home bug and fix setup widget settings call back hack
* fix body changed
* one liner
* clean up
* formatting
* if false
* revert to master + reimplement
* close sidebar on body home tici
* make ignition message bigger on c3
* flip eye direction when turning
---------
Co-authored-by: Nick <nickorie@gmail.com>
Co-authored-by: Shane Smiskol <shane@smiskol.com>
Adeeb Shihadeh