KFR 7 — Major Update¶

KFR 7 introduces a range of new capabilities, platform support, and performance improvements for modern audio and signal processing workflows.

KFR 7.2 Highlights¶

• The C API now covers biquad filter design, window functions, complex and real DFT, convolution and filtering, IIR and FIR filter design, and sample rate conversion.
• iOS build jobs and kfr.xcframework packaging are included in the release workflow.
• KFR_ENABLE_DFT is enabled by default for MSVC, as for all other compilers.
• GCC is supported for non-x86 builds and uses the vector extension backend.
• New CMake options: KFR_INSTALL_LICENSE and KFR_FORCE_INLINE_MSVC.
• Breaking changes: LICENSE.txt and README.md are installed to CMAKE_INSTALL_DOCDIR, the cmake/ios.cmake toolchain file is removed, and the default KFR_ARCHS doesn't include avx512 on 32-bit x86.

KFR 7.1 Highlights¶

• ngfft is a new low-overhead FFT API that gives applications manual control over memory management.
• Power-of-two FFTs are up to 40% faster, and FFTs whose sizes have many prime factors also use a more optimized algorithm.
• DFT is now supported with GCC and MSVC, in addition to Clang.
• Bit-index permutation operations are optimized for Clang and the generic backend.

Multi-Channel Audio Filtering¶

The new audio_filter processes planar audio_data with independent filters for each channel. It supports FIR, IIR, and convolution filtering, making multi-channel processing simpler without coupling channel state.

Core Audio Decoding¶

KFR can decode audio with Core Audio on macOS and iOS, with binary_reader input, floating-point output, and precise frame seeking.

Reworked Audio I/O¶

Audio input and output have been completely reworked. The new system supports WAV, AIFF, Apple CAF, FLAC, raw formats, and MP3 reading. It is more reliable, faster, and tolerant to format variations.

Highlights:

• Read and write directly from memory or custom streams
• Handle multiple bit depths and floating-point formats
• Work with planar or interleaved audio
• Built-in dithering
• Reading arbitrary chunks from RIFF-based formats
• Full UTF-8 file path support on Windows

RISC-V Architecture Support¶

KFR now runs on RISC-V, with optimized implementations that take advantage of the new vector extensions for better performance on emerging hardware.

Elliptic Filters¶

A new family of elliptic filters provides sharper frequency roll-off than Butterworth or Chebyshev designs, offering higher precision for demanding applications.

Universal macOS Binaries¶

KFR can now build universal binaries that run natively on both Intel and Apple Silicon (ARM64) Macs. These are built automatically and available for download.

Zero-Phase IIR Filtering¶

A new zero-phase IIR filter applies forward-backward processing for phase-accurate results, useful in offline and analysis workflows.

C++20 as the New Baseline¶

KFR now requires C++20, unlocking cleaner syntax and better use of modern language features.

New High-Level Audio Module¶

A new audio module simplifies working with multi-channel data. The audio_data class provides a unified interface for:

• Format conversion
• Interleaving and deinterleaving
• Zero-copy construction from user-provided pointers

Additional Improvements¶

• Scoped control over denormal flushing for stable floating-point behavior
• Managed memory allocation turned off by default for better speed (can be re-enabled via KFR_MANAGED_ALLOCATION)
• All tests migrated to Catch2
• Expanded documentation with more examples and tutorials

Summary of Progress Since Version 6¶

• Multidimensional DFT via the C API
• DFT performance improved by up to 80% on ARM and ARM64
• New Android x86/x64 and Linux ARM/AArch64 builds
• Matrix transpose up to 30% faster
• More stable Bluestein algorithm across all transform sizes