2 - Fully support the TRIV2 device and kernel driver
3 - Add missing device functions (e.g., unsetModel)
4 - Resolve minor bugs and svace issues
7 - Revise the whole source codes (C++ porting/refactoring)
8 - Increase the code coverage (>= 85%)
9 - Support a segment table for TRIV2 inferences
10 - Sync with the revised NPU kernel drivers
13 - Sync with the NPU kernel driver refactoring
14 - Add the version control with the NPU kernel driver
17 - Add APIs to sepecify inference constraints (e.g., timeout/priority)
18 - Add an option to control multiple buffering
19 - Support a thread pool to support concurrent ioctls
20 - Support a HW-recurring input service for runNPU_internalInput()
23 - Fix minor bugs on NPU unittests
24 - Add install_rpath for ubuntu users
25 - Disable apptests using tvn models (instead, CI will do this)
28 - Fix minor bugs on shamt handling in NPU emulation codes
29 - Add an API to get the model metadata
32 - Fix minor bugs on data mainpulation
35 - Remove NPU emulation impl. (moved to AIP/NPU_SystemService_Emulator)
36 - Remove testdata generation impl. (moved to AIP/NPU_SystemService_Testdata)
37 - Simplify debian/tizen packaging scripts
40 - Fix minor bugs in testdata generation
43 - Support simultaneous request handling from different user apps
44 - Support timeout and priority handling for multiple requests
45 - Add and verify prebuilt tvn models (TRIV v1) in FastModel
46 - Sync with the NPU driver version in NPU_Linux-kernel
49 - Verify test cases on FastModel
50 - Fix a few minor bugs
53 - Support new NPU binary format (i.e., version 2)
54 - Support non-contiguous multiple input/output tensors
55 - Support data manipulation and quantization (only for npubinfmt v2)
58 - Add a prototype for NPU-ASR
62 - Change names of sub-packages from libnpuengine to npu-engine
65 - Tarballs containing test data are removed
66 - Test data are generated at the build time
69 - NPU Vision functionality verified in the FastModel envionment
70 - Support Ubuntu 16.04 packaging & emulation of NPU executions
71 - More test cases with single/multi-layer NPU models