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