Skip to content

Double the chunk size for f64 #86

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
Shnatsel merged 1 commit into from
Mar 23, 2026
Merged

Double the chunk size for f64 #86

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
10 changes: 6 additions & 4 deletions src/algorithms/bravo.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -10,7 +10,7 @@
/// The initial implementation was translated from mathematical notation in the paper
/// to Rust by Claude 4.5 Opus.
use fearless_simd::prelude::*;
use fearless_simd::{f32x4, f32x8, f64x2, f64x4, Simd};
use fearless_simd::{f32x4, f32x8, f64x4, f64x8, Simd};

/// Macro to generate bit_rev_bravo implementations for concrete types.
/// Used instead of generics because `fearless_simd` doesn't let us be generic over the exact float type.
Expand Down Expand Up @@ -138,8 +138,8 @@ macro_rules! impl_bit_rev_bravo {
// which is necessary for using the native vector width
impl_bit_rev_bravo!(bit_rev_bravo_chunk_4_f32, f32, f32x4<S>, 4);
impl_bit_rev_bravo!(bit_rev_bravo_chunk_8_f32, f32, f32x8<S>, 8);
impl_bit_rev_bravo!(bit_rev_bravo_chunk_2_f64, f64, f64x2<S>, 2);
impl_bit_rev_bravo!(bit_rev_bravo_chunk_4_f64, f64, f64x4<S>, 4);
impl_bit_rev_bravo!(bit_rev_bravo_chunk_8_f64, f64, f64x8<S>, 8);

/// Performs in-place bit-reversal permutation using the CO-BRAVO algorithm.
///
Expand All @@ -163,8 +163,10 @@ pub fn bit_rev_bravo_f32<S: Simd>(simd: S, data: &mut [f32], n: usize) {
#[inline(always)] // required by fearless_simd
pub fn bit_rev_bravo_f64<S: Simd>(simd: S, data: &mut [f64], n: usize) {
match <S::f64s>::N {
2 => bit_rev_bravo_chunk_2_f64(simd, data, n), // SSE, NEON and fallback
_ => bit_rev_bravo_chunk_4_f64(simd, data, n),
// despite exceeding the native vector width, it is profitable to use larger chunks
// according to benchmarks on both Zen4 and Apple M4
2 => bit_rev_bravo_chunk_4_f64(simd, data, n), // SSE, NEON and fallback
_ => bit_rev_bravo_chunk_8_f64(simd, data, n),
// fearless_simd has no native support for AVX-512 yet
}
}
Expand Down
Loading