Skip to content

llvm 16 finally reconizes some additional vec in-place conversions as noops #109891

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
merged 1 commit into from
Apr 4, 2023
Merged

llvm 16 finally reconizes some additional vec in-place conversions as noops #109891

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
42 changes: 31 additions & 11 deletions tests/codegen/vec-in-place.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,11 +1,13 @@
// ignore-debug: the debug assertions get in the way
// compile-flags: -O -Z merge-functions=disabled
// min-llvm-version: 16
#![crate_type = "lib"]

// Ensure that trivial casts of vec elements are O(1)

pub struct Wrapper<T>(T);

// previously repr(C) caused the optimization to fail
#[repr(C)]
pub struct Foo {
a: u64,
Expand All @@ -14,9 +16,8 @@ pub struct Foo {
d: u64,
}

// Going from an aggregate struct to another type currently requires Copy to
// enable the TrustedRandomAccess specialization. Without it optimizations do not yet
// reliably recognize the loops as noop for repr(C) or non-Copy structs.
// implementing Copy exercises the TrustedRandomAccess specialization inside the in-place
// specialization
#[derive(Copy, Clone)]
pub struct Bar {
a: u64,
Expand All @@ -25,6 +26,14 @@ pub struct Bar {
d: u64,
}

// this exercises the try-fold codepath
pub struct Baz {
a: u64,
b: u64,
c: u64,
d: u64,
}

// CHECK-LABEL: @vec_iterator_cast_primitive
#[no_mangle]
pub fn vec_iterator_cast_primitive(vec: Vec<i8>) -> Vec<u8> {
Expand Down Expand Up @@ -52,18 +61,29 @@ pub fn vec_iterator_cast_unwrap(vec: Vec<Wrapper<u8>>) -> Vec<u8> {
// CHECK-LABEL: @vec_iterator_cast_aggregate
#[no_mangle]
pub fn vec_iterator_cast_aggregate(vec: Vec<[u64; 4]>) -> Vec<Foo> {
// FIXME These checks should be the same as other functions.
// CHECK-NOT: @__rust_alloc
// CHECK-NOT: @__rust_alloc
// CHECK-NOT: loop
// CHECK-NOT: call
vec.into_iter().map(|e| unsafe { std::mem::transmute(e) }).collect()
}

// CHECK-LABEL: @vec_iterator_cast_deaggregate
// CHECK-LABEL: @vec_iterator_cast_deaggregate_tra
#[no_mangle]
pub fn vec_iterator_cast_deaggregate(vec: Vec<Bar>) -> Vec<[u64; 4]> {
// FIXME These checks should be the same as other functions.
// CHECK-NOT: @__rust_alloc
// CHECK-NOT: @__rust_alloc
pub fn vec_iterator_cast_deaggregate_tra(vec: Vec<Bar>) -> Vec<[u64; 4]> {
// CHECK-NOT: loop
// CHECK-NOT: call

// Safety: For the purpose of this test we assume that Bar layout matches [u64; 4].
// This currently is not guaranteed for repr(Rust) types, but it happens to work here and
// the UCG may add additional guarantees for homogenous types in the future that would make this
// correct.
vec.into_iter().map(|e| unsafe { std::mem::transmute(e) }).collect()
}

// CHECK-LABEL: @vec_iterator_cast_deaggregate_fold
#[no_mangle]
pub fn vec_iterator_cast_deaggregate_fold(vec: Vec<Baz>) -> Vec<[u64; 4]> {
// CHECK-NOT: loop
// CHECK-NOT: call

// Safety: For the purpose of this test we assume that Bar layout matches [u64; 4].
// This currently is not guaranteed for repr(Rust) types, but it happens to work here and
Expand Down