10年動かしても劣化しないプログラムの仕組み：
Rust 2024とWebAssembly（WASM）が中小企業のIT固定費を劇的に下げる理由

#![no_std]

#[cfg(target_arch = "wasm32")]
use core::arch::wasm32::*;

/// 16-byte aligned 256-sample frame for WebAssembly SIMD128 (`v128`) processing.
#[repr(C, align(16))]
pub struct AlignedSignalFrame256 {
    pub samples: [f32; 256],
    pub window: [f32; 256],
}

impl AlignedSignalFrame256 {
    pub const fn new() -> Self {
        Self {
            samples: [0.0; 256],
            window: [1.0; 256],
        }
    }

    /// Exports linear memory pointer for zero-copy JS `Float32Array(memory.buffer, ptr, 256)`
    #[inline]
    pub const fn samples_ptr(&self) -> *const f32 {
        self.samples.as_ptr()
    }

    /// Computes windowed energy sum(samples[i] * window[i])^2 using 4-lane f32x4 SIMD128
    #[cfg(target_arch = "wasm32")]
    #[target_feature(enable = "simd128")]
    pub unsafe fn compute_windowed_energy_simd(&self) -> f32 {
        let mut acc = f32x4_splat(0.0);
        let s_ptr = self.samples.as_ptr() as *const v128;
        let w_ptr = self.window.as_ptr() as *const v128;
        let mut i = 0usize;
        // 256 floats = 64 vectors of 4 lanes (f32x4)
        while i < 64 {
            let sv = v128_load(s_ptr.add(i));
            let wv = v128_load(w_ptr.add(i));
            let windowed = f32x4_mul(sv, wv);
            let sq = f32x4_mul(windowed, windowed);
            acc = f32x4_add(acc, sq);
            i += 1;
        }
        f32x4_extract_lane::<0>(acc)
            + f32x4_extract_lane::<1>(acc)
            + f32x4_extract_lane::<2>(acc)
            + f32x4_extract_lane::<3>(acc)
    }

    /// Portable scalar fallback for host unit testing (identical numerical semantics)
    pub fn compute_windowed_energy_scalar(&self) -> f32 {
        let mut sum = 0.0f32;
        let mut i = 0;
        while i < 256 {
            let w = self.samples[i] * self.window[i];
            sum += w * w;
            i += 1;
        }
        sum
    }
}