2 - Add APIs to sepecify inference constraints (e.g., timeout/priority)
3 - Add an option to control multiple buffering
4 - Support a thread pool to support concurrent ioctls
5 - Support a HW-recurring input service for runNPU_internalInput()
8 - Fix minor bugs on NPU unittests
9 - Add install_rpath for ubuntu users
10 - Disable apptests using tvn models (instead, CI will do this)
13 - Fix minor bugs on shamt handling in NPU emulation codes
14 - Add an API to get the model metadata
17 - Fix minor bugs on data mainpulation
20 - Remove NPU emulation impl. (moved to AIP/NPU_SystemService_Emulator)
21 - Remove testdata generation impl. (moved to AIP/NPU_SystemService_Testdata)
22 - Simplify debian/tizen packaging scripts
25 - Fix minor bugs in testdata generation
28 - Support simultaneous request handling from different user apps
29 - Support timeout and priority handling for multiple requests
30 - Add and verify prebuilt tvn models (TRIV v1) in FastModel
31 - Sync with the NPU driver version in NPU_Linux-kernel
34 - Verify test cases on FastModel
35 - Fix a few minor bugs
38 - Support new NPU binary format (i.e., version 2)
39 - Support non-contiguous multiple input/output tensors
40 - Support data manipulation and quantization (only for npubinfmt v2)
43 - Add a prototype for NPU-ASR
47 - Change names of sub-packages from libnpuengine to npu-engine
50 - Tarballs containing test data are removed
51 - Test data are generated at the build time
54 - NPU Vision functionality verified in the FastModel envionment
55 - Support Ubuntu 16.04 packaging & emulation of NPU executions
56 - More test cases with single/multi-layer NPU models