Skip to content
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

Jump to bottom

Increasing the SIMD size improves the vectorization possibilities #116018

Merged
merged 3 commits into from Oct 6, 2023
Merged

Increasing the SIMD size improves the vectorization possibilities #116018

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
d30f210
Increasing the SIMD size improves the vectorization possibilities
DianQK Sep 21, 2023
910674f
Only check for successful vectorization on wider_reduce_into_iter
DianQK Sep 24, 2023
e300847
Add a wishlist FIXME
scottmcm Oct 6, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Jump to
Jump to file
Failed to load files.
Diff view
Diff view
26 changes: 13 additions & 13 deletions tests/codegen/simd/simd-wide-sum.rs
View file
Open in desktop
Expand Up @@ -11,24 +11,24 @@
#![feature(portable_simd)]

use std::simd::{Simd, SimdUint};
const N: usize = 8;
const N: usize = 16;

#[no_mangle]
// CHECK-LABEL: @wider_reduce_simd
pub fn wider_reduce_simd(x: Simd<u8, N>) -> u16 {
// CHECK: zext <8 x i8>
// CHECK-SAME: to <8 x i16>
// CHECK: call i16 @llvm.vector.reduce.add.v8i16(<8 x i16>
// CHECK: zext <16 x i8>
// CHECK-SAME: to <16 x i16>
// CHECK: call i16 @llvm.vector.reduce.add.v16i16(<16 x i16>
let x: Simd<u16, N> = x.cast();
x.reduce_sum()
}

#[no_mangle]
// CHECK-LABEL: @wider_reduce_loop
pub fn wider_reduce_loop(x: Simd<u8, N>) -> u16 {
// CHECK: zext <8 x i8>
// CHECK-SAME: to <8 x i16>
// CHECK: call i16 @llvm.vector.reduce.add.v8i16(<8 x i16>
// CHECK: zext <16 x i8>
// CHECK-SAME: to <16 x i16>
// CHECK: call i16 @llvm.vector.reduce.add.v16i16(<16 x i16>
let mut sum = 0_u16;
for i in 0..N {
sum += u16::from(x[i]);
Expand All @@ -39,9 +39,9 @@ pub fn wider_reduce_loop(x: Simd<u8, N>) -> u16 {
#[no_mangle]
// CHECK-LABEL: @wider_reduce_iter
pub fn wider_reduce_iter(x: Simd<u8, N>) -> u16 {
// CHECK: zext <8 x i8>
// CHECK-SAME: to <8 x i16>
// CHECK: call i16 @llvm.vector.reduce.add.v8i16(<8 x i16>
// CHECK: zext <16 x i8>
// CHECK-SAME: to <16 x i16>
// CHECK: call i16 @llvm.vector.reduce.add.v16i16(<16 x i16>
x.as_array().iter().copied().map(u16::from).sum()
}

Expand All @@ -52,8 +52,8 @@ pub fn wider_reduce_iter(x: Simd<u8, N>) -> u16 {
#[no_mangle]
// CHECK-LABEL: @wider_reduce_into_iter
pub fn wider_reduce_into_iter(x: Simd<u8, N>) -> u16 {
// CHECK: zext <8 x i8>
// CHECK-SAME: to <8 x i16>
// CHECK: call i16 @llvm.vector.reduce.add.v8i16(<8 x i16>
// FIXME: It would be nice if this was exactly the same as the above tests,
// but at the time of writing this comment, that didn't happen on LLVM main.
// CHECK: call i16 @llvm.vector.reduce.add
scottmcm marked this conversation as resolved.
Show resolved Hide resolved
x.to_array().into_iter().map(u16::from).sum()
}