*******************************************************************************
* RECENT CHANGES
*******************************************************************************

=== 1.0.31 ===
* Implementation of 3D mathematics using AVX instruction set.
* Fixed improper AVX-512 optimization for lanczos kernel genration function.
* Fixed Mid/Side transformation functions bug for 32-bit and 64-bit ARM processors. 
* Added MacOS CI builds.
* Updated module versions in dependencies.

=== 1.0.30 ===
* Implemented SIMD-optimized trigonometric functions sinf and cosf.
* Implemented Lanczos filter kernel computation function.
* Updated build scripts.
* Updated module versions in dependencies.

=== 1.0.29 ===
* Implemented AVX-512 optimized direct_fft and reverse_fft functions.
* Added definition of MacOS dependencies in build scripts.
* Updated module versions in dependencies.

=== 1.0.28 ===
* The DSP library now builds for Apple M1 chips and above on MacOS.
* Implemented abs_max2, abs_min2, abs_max3 and abs_min3 functions.
* Implemented sign_min, sign_max, sign_minmax functions.
* Updated build scripts.
* Updated module versions in dependencies.

=== 1.0.27 ===
* Updated build scripts.
* Updated module versions in dependencies.

=== 1.0.26 ===
* Updated build scripts.
* Updated module versions in dependencies.

=== 1.0.25 ===
* Implemented AVX-512 optimized minimum and maximum search functions.
* Implemented AVX-512 optimized direct convolution function.
* Updated build scripts.
* Updated module versions in dependencies.

=== 1.0.24 ===
* Fixed build for x86 Clang.

=== 1.0.23 ===
* Updated module versions in dependencies.

=== 1.0.22 ===
* Implemented functions for computing correlation between two signals.
* Fixed error in AVX and AVX-512 implementation of lr_to_ms functions.
* Updated build scripts.
* Updated module versions in dependencies.

=== 1.0.21 ===
* Updated build scripts.
* Updated module versions in dependencies.

=== 1.0.20 ===
* Optimization of compressor and gate functions using AVX-512 instruction set.
* Introduced SIMD-optimized expander curve and gain functions.
* Improved performance of logarithm values calculations.
* Fixed Mid/Side conversion functions for AArch64 architecture (contributed by
  marcan at GitHub).
* Updated build scripts.
* Updated module versions in dependencies.

=== 1.0.19 ===
* AVX2 optimization of search functions for maximum and minimum.
* Implemented SIMD-optimized gate functions.
* Fixed bun in AVX-512 implementation of mid/side conversion functions.
* AVX512 optimization of packed complex functions.
* Additional optimizations of parallel mathematics functions with AVX-512 instruction set.

=== 1.0.18 ===
* Fixed compilation regression for 32-bit Clang compiler.

=== 1.0.17 ===
* Implemented SIMD-optimized 2-knee compressor curve and gain calculation functions.
* Implemented several AVX-512 optimized packed complex functions.
* Removed pcomplex_add_r function. Use pcomplex_r2c_add2 function instead.
* Fixed improper formula for complex_div and pcomplex_div functions.
* Implemented sqr1, sqr2, ssqr1 and ssqr2 functions.
* Implemented axis_apply_lin1 function.
* Updated module versions in dependencies.

=== 1.0.16 ===
* Fixed improper AVX-512 detection which causes crash for several functions.

=== 1.0.15 ===
* Several functions optimized for AVX-512 support.
* Fixed several issues reported by PVS Studio static analyzer.
* Fixed syntax error in C interface, covered with tests.
* Bugfix in horizontal summing functions (invalid register clobber list).
* Some AMD-related optimizations.
* Updated module versions in dependencies.

=== 1.0.14 ===
* Implemented pcomplex_r2c instruction set.
* Updated build scripts.
* Updated module versions in dependencies.

=== 1.0.13 ===
* Implemented AVX2-optimized hsla_to_rgba and rgba_to_hsla functions.
* Implemented high-precision lanczos 2x, 3x, 4x, 6x and 8x oversampling functions.
* Updated build scripts.
* Updated module versions in dependencies.

=== 1.0.12 ===
* Added series of functions for processing 1-bit, 2-bit, 4-bit and 8-bit grayscale
  bitmaps which can be useful for quick text rendering using glyph images.
* Removed Makefile.d. Dependencies are now automatically generated at the build stage.
* Updated build scripts.
* Updated module versions in dependencies.

=== 1.0.11 ===
* Small math optimizations for dsp::smooth_cubic_linear and dsp::smooth_cubic_log
  functions.
* Updated module versions in dependencies.

=== 1.0.10 ===
* ARM NEON-optimized code compiles now only for at least ARMv6 architecture.
* Implemented linear ramping lramp_* functions optimized for
  i686, x86_64, ARM-32 and AArch64 architectures.
* Fixed avx::dyn_biquad_process_x8_fma3 function implementation.
* Slight optimizations of avx::biquad_process_x4_fma3 and avx::dyn_biquad_process_x4_fma3. 
* Added test build for Windows using MSYS2.
* Updated build scripts.
* Updated module versions in dependencies.

=== 1.0.9 ===
* Added Clang build for the CI.

=== 1.0.8 ===
* Better optimizations of dsp::copy and dsp::move for SSE, SSE3 and AVX.
* The function dsp::init() is now thread safe.
* Additional optimization of dsp::normalize() function, added dsp::normalize1()
  and dsp::normalize2() functions.
* Updated build scripts.

=== 1.0.7 ===
* Implemented axis_apply_log1 and axis_apply_log2 optimized for AArch64 ASIMD.
* Implemented fill_rgba and fill_hsla for AArch64 ASIMD.
* Implemented rgba_to_hsla, hsla_to_rgba, rgba_to_bgra32, rgba32_to_bgra32 for AArch64 ASIMD.
* Implemented eff_hsla_hue, eff_hsla_sat, eff_hsla_light, eff_hsla_alpha for AArch64 ASIMD.

=== 1.0.6 ===
* Updated build scripts.

=== 1.0.5 ===
* Implemented dsp::prgba32_set_alpha and dsp::pbgra32_set_alpha functions.
* Better assembly code form 32-bit ARM, now compatible with Clang.
* Better assembly code form 64-bit ARM, now compatible with Clang.
* Updated build scripts.

=== 1.0.4 ===
* Updated build scripts.

=== 1.0.3 ===
* Migrating to C++ 11 support.
* Updated library dependencies.
* Updated build scripts.

=== 1.0.2 ===
* Updated build scripts.

=== 1.0.1 ===
* Updated build scripts.

=== 1.0.0 ===
* Updated build scripts.
* Added Eclipse project files.

=== 0.5.14 ===
* Updated build system.

=== 0.5.13 ===
* Added x86/x86_64 4-lobe oversampling functions.
* Added ARMv7 NEON-optimized 4-lobe oversampling functions.
* Added AArch64 ASIMD-optimized 4-lobe oversampling functions.
* Fixed target compiler architecture detection.

=== 0.5.12 ===
* Submitted x86/x86_64 SSE2-optimized linear interpolation functions.
* Submitted x86/x86_64 AVX2-optimized linear interpolation functions.
* Submitted ARMv7 NEON-optimized linear interpolation functions.
* Submitted AArch64 ASIMD-optimized linear interpolation functions.
* Updated build system.

=== 0.5.11 ===
* Improved build: fixed many GCC warnings.
* DSP library can now be compiled with Clang C++ compiler.

=== 0.5.10 ===
* Updated build scripts.
* Reviewed generic abgr32_to_bgrff function implementation for Big endian platforms.
* Submitted x86/x86_64 SSE2-optimized function abgr32_to_bgrff.
* Submitted x86/x86_64 AVX2-optimized function abgr32_to_bgrff.
* Submitted ARMv7 NEON-optimized function abgr32_to_bgrff.
* Submitted AArch64 ASIMD-optimized function abgr32_to_bgrff.
* Fixed lacking export for SSE2-optimized functions support: fmmod3, fmrmod3, fmmod4, fmrmod4.
* Added support of AMD Zen 2 architecture for CPU tuning.
* More careful AVX2 support detection for AMD processors.
* Fixed improper RGBA to HSLA conversion in rgba_to_hsla routine.

=== 0.5.9 ===
* Updated build scripts that now use tags without prefixes first.

=== 0.5.8 ===
* Fixed excessive header installation with proper one.
* Fixed installation and deinstallation of shared object and library files.
* Added `make distsrc` target to build tarball with all required source code.

=== 0.5.7 ===
* Updated makefiles: added `make tree` target.

=== 0.5.6 ===
* Added build of static library.
* Source code now more properly references LGPL3+ license.

=== 0.5.5 ===
* Provided C-compatible function interface and symbol exports. All C functions and
  types are prefixed with lsp_dsp_ prefix to not to clash with other types. C++
  interface has not changed, all functions and data types are still in lsp::dsp namespace.

=== 0.5.4 ===
* Added pkg-config file generation.

=== 0.5.3 ===
* Added pmin, pmax, psmin, psmax, pamin, pamax functions.
* Added SSE optimizations for pmin, pmax, psmin, psmax, pamin, pamax functions.
* Added AVX optimizations for pmin, pmax, psmin, psmax, pamin, pamax functions.
* Added ARMv7 optimizations for pmin, pmax, psmin, psmax, pamin, pamax functions.
* Added AArch64 optimizations for pmin, pmax, psmin, psmax, pamin, pamax functions.
* Fixed bugs in msmatrix conversion functions for AVX and SSE.
* Fixed bugs in sse::bilinear_transform_x1 routine.

=== 0.5.2 ===
* Updated build system.
* Code refactoring.
* Added openBSD support in build.

=== 0.5.1 ===
* Some minor code refactoring.

=== 0.5.0 ===
* Initial release.
* Supports SIMD optimizations for 32-bit i586 architecture.
* Supports SIMD optimizations for 64-bit x86_64 architecture.
* Supports SIMD optimizations for 32-bit ARMv7 architecture.
* Supports SIMD optimizations for 64-bit AArch64 architecture.

