primal-barracuda 0.4.0

barraCuda — BARrier-free Rust Abstracted Cross-platform Unified Dimensional Algebra. Vendor-agnostic GPU compute via WGSL.
Documentation
// SPDX-License-Identifier: AGPL-3.0-or-later
// Softplus - Smooth approximation of ReLU (f64 canonical)
//
// Deep Debt Principles:
// - Pure WGSL implementation (universal compute)
// - Zero unsafe code (memory safe)
// - Hardware-agnostic (works on any GPU/CPU via WebGPU)
// - Self-contained logic (no external dependencies)

struct Params {
    size: u32,
    beta: f64,
}

@group(0) @binding(0) var<storage, read> input: array<f64>;
@group(0) @binding(1) var<storage, read_write> output: array<f64>;
@group(0) @binding(2) var<uniform> params: Params;

@compute @workgroup_size(256)
fn main(@builtin(global_invocation_id) global_id: vec3<u32>) {
    let idx = global_id.x;

    if (idx >= params.size) {
        return;
    }

    let x = input[idx];
    // Softplus: (1/beta) * log(1 + exp(beta * x))
    // For numerical stability, use: max(0, x) + log(1 + exp(-abs(x)))
    let beta_x = params.beta * x;
    output[idx] = (f64(1.0) / params.beta) * (max(f64(0.0), beta_x) + log_f64(f64(1.0) + exp_f64(-abs(beta_x))));
}