* remove the clCreateProgramWithSource interceptor
* that's old code, thneed is better
* label them thneed_, we shouldn't need to touch CL for anything not SNPE related
old-commit-hash: b8571710e09e58b9e67170a9924eef79f32c480b
* add PubMaster & SubMaster
remove 'delete msg'
remove headers
* use constructor to initial all submster
* modify drain sockets
* fix typo in ssconscript.remove lines
no checkValid in loggerd
last modify
handle_message:event->&event
fix type
remove heads
event to auto
* new interface
* api changed
* Revert "use constructor to initial all submster"
This reverts commit 73be7ea46250a325ce41d3a0445e34395a2ae692.
* change to new api
* revert loggerd
* dd
* use new PubSub api
* update to new interface
don't modify loggerd
reset panda
reset opendbc
remove empty lines
* switch to new pubMaster
* update to the new inteface
change
remove error code
. to ->
merge paramsd.cc
update panda
fix typo
simplify
fix typo
* Fix build
* always conflate
Co-authored-by: deanlee <deanlee3@gmail.com>
old-commit-hash: ab5af232b243ba8137695f68c125f922fb412d44
* add thneed self test
* don't do the memset in thneed, shouldn't matter though
Co-authored-by: Comma Device <device@comma.ai>
old-commit-hash: 6c0ad1e67567e149cdf9e00ab7ae43f4cc5bc579
* bigmodel
* more debug print
* debugging bigmodel
* remove the tanh, debugging
* print images/buffers
* disassemble the command queues
* decompiler
* dump the shaders
* full disasm
* support patching kernel and fixing convolution_horizontal_reduced_reads_1x1
* microbenchmark
* 42 GFLOPS, 1 GB/s
* gemm benchmark
* 75 GFLOPS vs 42 GFLOPS
* 115 GFLOPS
* oops, never mind
* gemm image is slow
* this is pretty hopeless
* gemm image gets 62 GFLOPS
* this is addictive and still a waste of time
* cleanup cleanup
* that hook was dumb
* tabbing
* more tabbing
Co-authored-by: Comma Device <device@comma.ai>
old-commit-hash: 78a352a8ca8a948e86e7c752732e470f89d92280
* thneed runs the model
* thneed is doing the hooking
* set kernel args
* thneeding the bufferS
* print the images well
* thneeds with better buffers
* includes
* disasm adreno
* parse packets
* disasm works
* disasm better
* more thneeding
* much thneeding
* much more thneeding
* thneed works i think
* thneed is patient
* thneed works
* 7.7%
* gpuobj sync
* yay, it mallocs now
* cleaning it up, Thneed
* sync objs and set power
* thneed needs inputs and outputs
* thneed in modeld
* special modeld runs
* can't thneed the DSP
* test is weird
* thneed modeld uses 6.4% CPU
* add thneed to release
* move to debug
* delete some junk from the pr
* always track the timestamp
* timestamp hacks in thneed
* create a new command queue
* fix timestamp
* pretty much back to what we had, you can't use SNPE with thneed
* improve thneed test
* disable save log
Co-authored-by: Comma Device <device@comma.ai>
old-commit-hash: 302d06ee709cb65dba39e836917f5693f0819efb
- better polyfitting
- no mpc cost change during lane change
- model trained with better ll gt, sim noise and guaranteed memories
old-commit-hash: 65fcc7c45ff1d4d747dc52ff5791afb84a18744a
* convert c-capnp to c++
change paramater type
* remove c-capnp header from commonmodel.c
* unsigned short -> uint16_t
* MSM_CAMERA_I2C_ADDR_TYPE_MAX->4
old-commit-hash: b6d6f520327226004c328a4b7cbbb3b855b2909d
* lgtm in readme, and mac nui fixes
* call super inits in radars
* unused imports, dup lines
* more radars, more unused imports
* pass CP into RadarInterfaceBase
* more fixups
* unused imports
* delete unused lines
* ugh, new unused import
Co-authored-by: George Hotz <geohot@gmail.com>
old-commit-hash: 1295cfe06cd80a2c2f40e0af97697ee818037f08
* release resources when thread is done
* add white space
* add white space
Co-authored-by: deanlee <deanlee3@gmail.com>
old-commit-hash: 02e474580ee90713bc3426ea512278c5eb1c6b15
* add traffic convention
* hope this work
* no comment
* latest and gratest
* big gru model
* 1af55c7d-ee15-414a-9e98-a0cb08c3441f/75
* much later in training
* wrong temporal size
* converged
* fix lane changes
old-commit-hash: d3edc594cef776847645a6cd14e9e3ce6893bd29
* capnparm
* building in progress
* scons build works
* that script fixes opencl
* start new camera code
* includes and more camera scripts
* control c works now
* no device control yet
* phy too
* just one camera for now
* fix capnparm
* hmm, the inits are needed
* more cameras
* link stop start
* doesn't work yet
* fix ion on qcom2
* start poll ish
* 4 pictures and done
* no jpeg
* it works to picture
* destroy sync obj
* both work for now
* defined QCOM2
* fix fd leak
* run modeld
* 10 bit mode
* real frame stride
* needs digital gain
* dnew
* no color correcting on new
* that snpe doesn't work
* qcom2 gate
* cleanups
* oops, fix aarch64 detector
* update cereal
* modeld works with SNPE
* fix driver monitoring model
Co-authored-by: Tici <robbe@comma.ai>
old-commit-hash: 8db0993663d7f16cddde00a2dfac8347a152b34e
* snpe version 1.35.0.698
* fix files common for new snpe
* efficient net with normal names
* compiling with 1.32 works
* actual snpe files
* fix monitoring, delete unneeded
* point the symlinks to the right places
* err, modeld needs the CPU
* benchmark for 50 its
* put back old model for now
Co-authored-by: Comma Device <device@comma.ai>
Co-authored-by: Harald Schafer <harald.the.engineer@gmail.com>
Co-authored-by: George Hotz <geohot@gmail.com>
old-commit-hash: 6e5a2d404bfbc33e5a63986f7d618d07d9c9c07f
* cleanup simulator files
* minor updates
* update readme
* keras runner builds
* hmm, still doesn't work
* keras runner works
* should work with python3 keras mod
* touchups
old-commit-hash: c50c718293
Willem Melching