Function dot_product_f64 

pub fn dot_product_f64(a: &[f32], b: &[f32]) -> f64

Compute the dot product of two f32 slices using f64 accumulation.

Uses a 4-wide accumulator to exploit instruction-level parallelism and reduce the dependency chain length, preventing precision loss in the inner products that drive the CG recurrence.

Panics

Debug-asserts that a.len() == b.len().