Created
April 6, 2023 15:08
-
-
Save shadeglare/cd8b3b7256dcbd89b59cc63c8e96339f to your computer and use it in GitHub Desktop.
Benchmark for plain shift op vs arm optimised one
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| fn shl_plain(v: [u32; 4], o: i32) -> [u32; 4] { | |
| [v[0] << o, v[1] << o, v[2] << o, v[3] << o] | |
| } | |
| #[cfg(target_arch = "aarch64")] | |
| fn shl_neon(v: [u32; 4], o: i32) -> [u32; 4] { | |
| use std::arch::aarch64::{vld1q_s32, vld1q_u32, vqshlq_u32, vst1q_u32}; | |
| unsafe { | |
| let v = vld1q_u32(v.as_ptr()); | |
| let o = vld1q_s32([o, o, o, o].as_ptr()); | |
| let r = vqshlq_u32(v, o); | |
| let mut p = std::mem::MaybeUninit::<[u32; 4]>::uninit(); | |
| vst1q_u32(p.as_mut_ptr() as *mut u32, r); | |
| p.assume_init() | |
| } | |
| } | |
| use criterion::{criterion_group, criterion_main, BenchmarkId, Criterion}; | |
| pub fn criterion_benchmark(c: &mut Criterion) { | |
| let mut group = c.benchmark_group("shl"); | |
| for i in 0..10 { | |
| group.bench_with_input(BenchmarkId::new("plain", i), &i, |b, i| { | |
| b.iter(|| shl_plain([1, 2, 3, 4], 1)) | |
| }); | |
| group.bench_with_input(BenchmarkId::new("neon", i), &i, |b, i| { | |
| b.iter(|| shl_neon([1, 2, 3, 4], 1)) | |
| }); | |
| } | |
| group.finish(); | |
| } | |
| criterion_group!(benches, criterion_benchmark); | |
| criterion_main!(benches); |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment