r7

• Change to use Newton-Raphson refined approximate reciprocal for division calculation in SSE2 and AVX2 code paths.
• Add AVX512 optimization.

r6

• Rename nstring parameter and change type to int array. Add parameter alpha.
• Rename sstring parameter and change type to float array. Add parameter ssystem.

r5

• Fix decimal point issue in certain locales.

r4.1

• Add Meson build system.

r4

• Code refactoring and minor performance improvement.
• Add opt parameter and AVX2 code path.
• Change default sigma/sigma2 to 8.0.
• Change default sbsize to 16 and sosize to 12.

The 7z file also includes FFTW 3.3.7 dll compiled by ICC.

r3

• Support 32-bit float input.

r2

Add SIMD optimization. Speed up about 10~30%, depending on the settings(like tbsize) or 32/64 bits.
Change the default tbsize to 3 instead of 5.
Other cosmetic changes.

r1

Initial commit