CMSIS-DSP 1.17.0

* Internal rework of lots of files to be able to build Neon versions when targetting Cortex-A 
* Cortex-A build is not part of the pack release and should not impact Cortex-M build
* Other changes in the github repository : Cortex-A build with Neon using cmake and Zephyr module
* Correction of arm_biquad_cascade_df1_f32 (#290) : out of buffer access with Helium
* Fix overflow risk in arm_div_int64_to_int32 (#287)
* New transform APIs to get the required size of buffers depending on the target architecture
* Fix a small bug regarding copying history samples in taps_5-8 processing of FIRF32 when targetting Helium (#245)
* Fix issue #270 : Vector type casting issues for Helium
* Corrected problem in arm_biquad_cascade_df1_32x64_q31: Workaround because of gcc bug when targetting Helium (#265)
* Adding "deprecated" comments (#263)
* Corrected out of bound access for coefficients in fir interpolate f32 (#261)
* FilteringFunctions: Avoid uninitialized value warnings from GCC 14.3 (#262)
* Correction of issue #258 : Remove use of INFINITY from the float code
* Fix the issues in Controller benchmark testing (#239)
* Fix typos discovered by codespell (#236)
* Fix compile issue caused by the incorrect name of round-up macro (#237)
* Add support for running tests for Support benchmarks (#238)
* Create arm_mat_init_q7 (#235)
* Improve float representation accuracy by increasing precision from 8 to 9 decimal digits (#226)
* Correct issue #217 : vector type issues when targeting Helium

CMSIS-DSP 1.16.2

• Improvements to documentation
• Added arm_fir_decimate_f64
• Removed some gcc workarounds that are no more needed for Helium
• Removed redundant test in arm_atan2_q15 and arm_atan2_q31
• Replaced INFINITY by maximum positive float
• Replaced hex constants with 2's complement equivalents
• Added C++ template extension (outside of pack - headers available from github repository)
• Added new compilation defines (ARM_DSP_ATTRIBUTE, ARM_DSP_TABLE_ATTRIBUTE and ARM_DSP_CUSTOM_CONFIG)
• Solved several github issues
• Corrected wrong version number in pack for v1.16.0
• Corrected issue #201 in v1.16.1 (related to new ARM_DSP_ATTRIBUTE)

CMSIS-DSP 1.15.0

• Switched to semantic versioning
• Corrections to MFCC
• Updated support for cmsis build tools 2.0
• Compute graph removed. It is now part of CMSIS-Stream
• Github issues:
  • 95,96,105,106,108,111,112
• New FFT initializations to allow linker to do code size optimizations
  • If you know the size of the transform at build time, prefer functions like
    arm_cfft_init_64_q15 to functions like arm_cfft_init_q15.
    It applies for CFFT, RFFT and MFCC and other datatypes.

CMSIS-DSP 1.14.4

What's Changed

• Missing header added to arm_math.h
• Documentation improvements
• Improvements to the generic nodes of the compute graph

CMSIS-DSP 1.14.3

What's Changed

• Converted atan2_coefs_q15 to signed int constants by @petertorelli in #75
• Add braces around MAX macro by @a1lu in #77
• fix: typo from algorthms to algorithms by @charansai612 in #84
• use 32-bit bitreversal loop index by @rsaxvc in #81

New Contributors

• @petertorelli made their first contribution in #75
• @a1lu made their first contribution in #77
• @charansai612 made their first contribution in #84
• @rsaxvc made their first contribution in #81

Full Changelog: v1.14.2...v1.14.3

CMSIS-DSP 1.14.2

Bug fixes
Improvement to compute graph

CMSIS-DSP 1.14.1

Improvements to the Compute Graph with new customization options.
Some Neon acceleration for aarch64 for f64 kernels.

CMSIS-DSP 1.14.0

Add support for CMSIS Event Recorder to the CMSIS-DSP Compute Graph.
Add more customization options to the CMSIS-DSP Compute Graph.

CMSIS-DSP 1.12.0

Renaming of SDF -> Compute Graph - Static Flow
New features to make the compute graph framework more developer friendly.
More f64 functions

CMSIS-DSP 1.11.0

Improved documentation

QR decomposition with Householder transform (float only).