2 - Add ioctl() to get the number of requests for trinity devices
3 - Add ioctl() to stop all requests for trinity devices
4 - Support log verbose mode for debugging
5 - Version sync with npu-engine-emul
8 - Fix minor bugs (e.g., weight offset setting and data size calc)
9 - Add user-level APIs to get status (e.g., memory/device status)
10 - Support memory status ioctl for trinity devices
11 - Support the stop feature to terminate submitted requests
14 - Fully support the TRIV2 device and kernel driver
15 - Add missing device functions (e.g., unsetModel)
16 - Resolve minor bugs and svace issues
19 - Revise the whole source codes (C++ porting/refactoring)
20 - Increase the code coverage (>= 85%)
21 - Support a segment table for TRIV2 inferences
22 - Sync with the revised NPU kernel drivers
25 - Sync with the NPU kernel driver refactoring
26 - Add the version control with the NPU kernel driver
29 - Add APIs to sepecify inference constraints (e.g., timeout/priority)
30 - Add an option to control multiple buffering
31 - Support a thread pool to support concurrent ioctls
32 - Support a HW-recurring input service for runNPU_internalInput()
35 - Fix minor bugs on NPU unittests
36 - Add install_rpath for ubuntu users
37 - Disable apptests using tvn models (instead, CI will do this)
40 - Fix minor bugs on shamt handling in NPU emulation codes
41 - Add an API to get the model metadata
44 - Fix minor bugs on data mainpulation
47 - Remove NPU emulation impl. (moved to AIP/NPU_SystemService_Emulator)
48 - Remove testdata generation impl. (moved to AIP/NPU_SystemService_Testdata)
49 - Simplify debian/tizen packaging scripts
52 - Fix minor bugs in testdata generation
55 - Support simultaneous request handling from different user apps
56 - Support timeout and priority handling for multiple requests
57 - Add and verify prebuilt tvn models (TRIV v1) in FastModel
58 - Sync with the NPU driver version in NPU_Linux-kernel
61 - Verify test cases on FastModel
62 - Fix a few minor bugs
65 - Support new NPU binary format (i.e., version 2)
66 - Support non-contiguous multiple input/output tensors
67 - Support data manipulation and quantization (only for npubinfmt v2)
70 - Add a prototype for NPU-ASR
74 - Change names of sub-packages from libnpuengine to npu-engine
77 - Tarballs containing test data are removed
78 - Test data are generated at the build time
81 - NPU Vision functionality verified in the FastModel envionment
82 - Support Ubuntu 16.04 packaging & emulation of NPU executions
83 - More test cases with single/multi-layer NPU models