Release Notes¶
This page contains the release notes for Qualcomm® AI Engine Direct Delegate:
Version
Release Date
Key Updates
v2.39.0
09/30/2025
Support new SoC
Minor fixes
v2.38.0
08/31/2025
Support new SoC
Add new java interface option: HTP_OPTIMIZE_FOR_INFERENCE_O3
Support Broadcast_to Op
Minor fixes
v2.37.0
07/31/2025
Support new SoC
Minor fixes
v2.36.0
06/30/2025
Support new SoC
Minor fixes
v2.35.0
05/31/2025
Support new SoC
Minor fixes
v2.34.0
04/30/2025
Support new SoC
Minor fixes
v2.33.0
03/31/2025
Support new SoC
Minor fixes
v2.32.0
02/28/2025
Support new SoC
Minor fixes
v2.31.0
01/31/2025
Support IrBackend
Support EmbeddingLookup op
v2.30.0
12/31/2024
Libraries for Android are compiled to be 16kb page-aligned
Support profiler in java interface
Support QNN context binary version 3
NOTE: The default HTP precision is changed to HtpFp16
NOTE: some chips don’t support HTP FP16 runtime
v2.29.0
11/30/2024
Minor fixes
v2.28.0
10/31/2024
TFLite renamed to LiteRT (TBD)
Revamped QNN Delegate logging
Added optimization strategy O3 to java interface
Mentioned Maven package and Proguard settings
v2.27.0
09/30/2024
Added $ORIGIN to rpath
Added ‘packagin=aar’ in Android aar pom file
Fixed bug with HTP device_id affecting cache SHA
v2.26.0
08/31/2024
Added O3 optimization option
Added linting profiling option
Added device ID setting
v2.25.0
07/31/2024
Finished op-version check
Optimized power mode settings
Added license files into AAR
v2.24.0
06/30/2024
OP version checks
Support TransposeConv2d with fusion
JNI statically linked
Support StridedSlice offset
Removed dynamic symbols
Plan to implement TFLite Stable Delegate APIs
v2.23.0
05/31/2024
TF 2.15.0 and NDK26c used
Fixed supported operators docs
Optimized Square operator
v2.22.0
04/30/2024
Added –htp_disable_conv_hmx option
Fixed QnnSaver bug
Support 16bit QuantizeOp
Support Square op
Improved capability detection
Plan to upgrade TF and NDK
v2.21.0
03/31/2024
useConvHmx default True
TfLiteQnnDelegateDelete() not thread-safe
v2.20.0
02/29/2024
Bug fixes around HTP Prepare
useConvHmx to be default True
TF upgrade planned
v2.19.0
01/31/2024
Fixed RPC polling
Fixed skip nodes warnings
v2.18.0
12/31/2023
Added java API checkCapability()
Compiled with NDK r25c
Added HTP vtcm size and hvx thread options
v2.17.0
11/30/2023
Added graph priority
Snapdragon 8 Gen 3 support
Added java interface options
Improved fp16 fallback
v2.16.0
10/06/2023
Added Atan2, AddN, Ceil
Toolchain change to NDK25c planned
v2.15.0
09/30/2023
Added Conv3d, GatherNd, Elu, L2Pool2D, ScatternNd, Relu_0_to_1, TransposeConv3d, PadV2, OneHot, Cos, Round
Added Skip OP API example
v2.14.0
08/31/2023
Added qtld-release.aar
Added platform support for aarch64-oe-linux-gcc11.2 and aarch64-ubuntu-gcc9.4
v2.13.0
07/31/2023
Fixed preparation errors
Supported implicit group Conv2D
v2.12.0
06/30/2023
Initial release
Built with ndk19c
Plan to switch to ndk25c