user0/rust
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Auto merge of #91838 - scottmcm:array-slice-eq-via-arrays-not-slices, r=dtolnay

Browse source 
Do array-slice equality via array equality, rather than always via slices

~~Draft because it needs a rebase after #91766 eventually gets through bors.~~

This enables the optimizations from #85828 to be used for array-to-slice comparisons too, not just array-to-array.

For example, <https://play.rust-lang.org/?version=nightly&mode=release&edition=2021&gist=5f9ba69b3d5825a782f897c830d3a6aa>
```rust
pub fn demo(x: &[u8], y: [u8; 4]) -> bool {
    *x == y
}
```
Currently writes the array to stack for no reason:
```nasm
	sub	rsp, 4
	mov	dword ptr [rsp], edx
	cmp	rsi, 4
	jne	.LBB0_1
	mov	eax, dword ptr [rdi]
	cmp	eax, dword ptr [rsp]
	sete	al
	add	rsp, 4
	ret

.LBB0_1:
	xor	eax, eax
	add	rsp, 4
	ret
```
Whereas with the change in this PR it just compares it directly:
```nasm
	cmp	rsi, 4
	jne	.LBB1_1
	cmp	dword ptr [rdi], edx
	sete	al
	ret

.LBB1_1:
	xor	eax, eax
	ret
```
This commit is contained in:
bors 2021-12-17 19:17:29 +00:00
commit 7abab1efb2
3 changed files with 89 additions and 15 deletions
Download patch file Download diff file Expand all files Collapse all files

19
src/test/codegen/slice-ref-equality.rs
View file

@ -4,18 +4,31 @@
// #71602 reported a simple array comparison just generating a loop.
// This was originally fixed by ensuring it generates a single bcmp,
// but we now generate it as a load instead. `is_zero_slice` was
// but we now generate it as a load+icmp instead. `is_zero_slice` was
// tweaked to still test the case of comparison against a slice,
// and `is_zero_array` tests the new array-specific behaviour.
// The optimization was then extended to short slice-to-array comparisons,
// so the first test here now has a long slice to still get the bcmp.
// CHECK-LABEL: @is_zero_slice
// CHECK-LABEL: @is_zero_slice_long
#[no_mangle]
pub fn is_zero_slice(data: &[u8; 4]) -> bool {
pub fn is_zero_slice_long(data: &[u8; 456]) -> bool {
// CHECK: :
// CHECK-NEXT: %{{.+}} = getelementptr {{.+}}
// CHECK-NEXT: %[[BCMP:.+]] = tail call i32 @{{bcmp|memcmp}}({{.+}})
// CHECK-NEXT: %[[EQ:.+]] = icmp eq i32 %[[BCMP]], 0
// CHECK-NEXT: ret i1 %[[EQ]]
&data[..] == [0; 456]
}
// CHECK-LABEL: @is_zero_slice_short
#[no_mangle]
pub fn is_zero_slice_short(data: &[u8; 4]) -> bool {
// CHECK: :
// CHECK-NEXT: %[[PTR:.+]] = bitcast [4 x i8]* {{.+}} to i32*
// CHECK-NEXT: %[[LOAD:.+]] = load i32, i32* %[[PTR]], align 1
// CHECK-NEXT: %[[EQ:.+]] = icmp eq i32 %[[LOAD]], 0
// CHECK-NEXT: ret i1 %[[EQ]]
&data[..] == [0; 4]
}