faster benchmark

benchmark.rs

#[macro_use]
extern crate criterion;
extern crate faster;
extern crate simdeez;
use criterion::Criterion;
use criterion::Fun;
use faster::*;
use simdeez::*;
use simdeez::sse2::*;

#[inline(always)]
unsafe fn simdeezfunc<S:Simd>() -> Vec<f32> {
       let slice = &[-123.456f32; 128][..];
       let vector_width = S::WIDTH_BYTES/4; 
       let mut result : Vec<f32> = Vec::with_capacity(128);
       result.set_len(128);
       let mut i = 0;
       while i < 128 {
            let v = S::loadu_ps(&slice[i]);
            let mut r = S::mul_ps(S::set1_ps(9.0),S::sqrt_ps(S::ceil_ps(S::rsqrt_ps(S::sqrt_ps(S::abs_ps(v))))));
            r = S::sub_ps(S::sub_ps(r,S::set1_ps(4.0)),S::set1_ps(2.0));
            S::storeu_ps(&mut result[i],r);
            i += vector_width
       }
       result
}

#[target_feature(enable="sse2")]
unsafe fn simdeez_sse2() -> Vec<f32> {
    simdeezfunc::<Sse2>()
}

fn b(c: &mut Criterion) {
    let f = Fun::new("faster 3333", |b, _i| {
        b.iter(|| {
            (&[-123.456f32; 128][..])
                .simd_iter(f32s(0.0))
                .simd_map(|v| {
                    f32s(9.0) * v.abs().sqrt().rsqrt().ceil().sqrt() - f32s(4.0) - f32s(2.0)
                })
                .scalar_collect()
        })
    });
    let s = Fun::new("simdeez 3333", |b, _i| b.iter(|| 
      unsafe {  simdeez_sse2() }
    ));

    let functions = vec![f, s];
    c.bench_functions("b", functions, 0);
}
criterion_group!(benches, b);
criterion_main!(benches);