2 - Support profiling APIs in FastModel/FPGA env.
3 - Fix some package dependency issues
6 - Implement profiling APIs with the libmrpsim support
7 - Fix minor bugs mostly related to NPU emulation
10 - Revise NPU driver APIs to use Trinity API level 2
11 - Extend trinity-smi to monitor the status of apps and their tasks
12 - Seperate libnpuhost codes from core handler codes
15 - Support 2-TOPS NPU devices for emulation with different command files
16 - Add the command line utility (trinity-smi) for monitoring trinity devices
17 - Support the Docker envionment (xenial, bionic) for npu-engine users
20 - Revise the utility for trinity-testing applications
21 - Replace emulation library from libnpu-engine-emul to libmrpsim
22 - Resolve minor issues from static analyzers (i.e., svace, coverity)
25 - Fix minor bugs related to layout conversion
28 - Add a prototype of profiling interface as libnpuhost APIs
29 - Add some apptests to check a multi-tasking capability
30 - Skip to install dummy (or uncompatible) apptests in non-emulated env.
31 - Fix minor bugs related to model unregistration
34 - Fix external segment handling bugs
35 - Add npu-engine-utils package
36 - Support TRIV2's data manipulation (layout conversion)
39 - Resolve issues reported by static analyzer (svace/coverity)
40 - Add driver-assisted memory copy to avoid FPGA memcpy bug
41 - Prepare the code release
44 - Revise the user APIs for HW-recurring tasks
45 - Sync with the kernel driver's changed ioctl definition
46 - Add testcases for external hwmem and HW-recurring tasks
49 - Fix tensor size calculations on TRIV2 layout
50 - Add some testcases to run infernces using an async thread
53 - Add ioctl() to get the number of requests for trinity devices
54 - Add ioctl() to stop all requests for trinity devices
55 - Support log verbose mode for debugging
56 - Version sync with npu-engine-emul
59 - Fix minor bugs (e.g., weight offset setting and data size calc)
60 - Add user-level APIs to get status (e.g., memory/device status)
61 - Support memory status ioctl for trinity devices
62 - Support the stop feature to terminate submitted requests
65 - Fully support the TRIV2 device and kernel driver
66 - Add missing device functions (e.g., unsetModel)
67 - Resolve minor bugs and svace issues
70 - Revise the whole source codes (C++ porting/refactoring)
71 - Increase the code coverage (>= 85%)
72 - Support a segment table for TRIV2 inferences
73 - Sync with the revised NPU kernel drivers
76 - Sync with the NPU kernel driver refactoring
77 - Add the version control with the NPU kernel driver
80 - Add APIs to sepecify inference constraints (e.g., timeout/priority)
81 - Add an option to control multiple buffering
82 - Support a thread pool to support concurrent ioctls
83 - Support a HW-recurring input service for runNPU_internalInput()
86 - Fix minor bugs on NPU unittests
87 - Add install_rpath for ubuntu users
88 - Disable apptests using tvn models (instead, CI will do this)
91 - Fix minor bugs on shamt handling in NPU emulation codes
92 - Add an API to get the model metadata
95 - Fix minor bugs on data mainpulation
98 - Remove NPU emulation impl. (moved to AIP/NPU_SystemService_Emulator)
99 - Remove testdata generation impl. (moved to AIP/NPU_SystemService_Testdata)
100 - Simplify debian/tizen packaging scripts
103 - Fix minor bugs in testdata generation
106 - Support simultaneous request handling from different user apps
107 - Support timeout and priority handling for multiple requests
108 - Add and verify prebuilt tvn models (TRIV v1) in FastModel
109 - Sync with the NPU driver version in NPU_Linux-kernel
112 - Verify test cases on FastModel
113 - Fix a few minor bugs
116 - Support new NPU binary format (i.e., version 2)
117 - Support non-contiguous multiple input/output tensors
118 - Support data manipulation and quantization (only for npubinfmt v2)
121 - Add a prototype for NPU-ASR
125 - Change names of sub-packages from libnpuengine to npu-engine
128 - Tarballs containing test data are removed
129 - Test data are generated at the build time
132 - NPU Vision functionality verified in the FastModel envionment
133 - Support Ubuntu 16.04 packaging & emulation of NPU executions
134 - More test cases with single/multi-layer NPU models