Ban non-array SIMD
This commit is contained in:
parent
712463de61
commit
d2309c2a9d
@ -11,7 +11,7 @@ This will cause an error:
|
||||
#![feature(repr_simd)]
|
||||
|
||||
#[repr(simd)]
|
||||
struct Bad<T>(T, T, T, T);
|
||||
struct Bad<T>([T; 4]);
|
||||
```
|
||||
|
||||
This will not:
|
||||
@ -20,5 +20,5 @@ This will not:
|
||||
#![feature(repr_simd)]
|
||||
|
||||
#[repr(simd)]
|
||||
struct Good(u32, u32, u32, u32);
|
||||
struct Good([u32; 4]);
|
||||
```
|
||||
|
@ -1,6 +1,6 @@
|
||||
A `#[simd]` attribute was applied to an empty tuple struct.
|
||||
A `#[simd]` attribute was applied to an empty or multi-field struct.
|
||||
|
||||
Erroneous code example:
|
||||
Erroneous code examples:
|
||||
|
||||
```compile_fail,E0075
|
||||
#![feature(repr_simd)]
|
||||
@ -9,9 +9,15 @@ Erroneous code example:
|
||||
struct Bad; // error!
|
||||
```
|
||||
|
||||
The `#[simd]` attribute can only be applied to non empty tuple structs, because
|
||||
it doesn't make sense to try to use SIMD operations when there are no values to
|
||||
operate on.
|
||||
```compile_fail,E0075
|
||||
#![feature(repr_simd)]
|
||||
|
||||
#[repr(simd)]
|
||||
struct Bad([u32; 1], [u32; 1]); // error!
|
||||
```
|
||||
|
||||
The `#[simd]` attribute can only be applied to a single-field struct, because
|
||||
the one field must be the array of values in the vector.
|
||||
|
||||
Fixed example:
|
||||
|
||||
@ -19,5 +25,5 @@ Fixed example:
|
||||
#![feature(repr_simd)]
|
||||
|
||||
#[repr(simd)]
|
||||
struct Good(u32); // ok!
|
||||
struct Good([u32; 2]); // ok!
|
||||
```
|
||||
|
@ -1,4 +1,4 @@
|
||||
All types in a tuple struct aren't the same when using the `#[simd]`
|
||||
The type of the field in a tuple struct isn't an array when using the `#[simd]`
|
||||
attribute.
|
||||
|
||||
Erroneous code example:
|
||||
@ -7,12 +7,12 @@ Erroneous code example:
|
||||
#![feature(repr_simd)]
|
||||
|
||||
#[repr(simd)]
|
||||
struct Bad(u16, u32, u32 u32); // error!
|
||||
struct Bad(u16); // error!
|
||||
```
|
||||
|
||||
When using the `#[simd]` attribute to automatically use SIMD operations in tuple
|
||||
struct, the types in the struct must all be of the same type, or the compiler
|
||||
will trigger this error.
|
||||
structs, if you want a single-lane vector then the field must be a 1-element
|
||||
array, or the compiler will trigger this error.
|
||||
|
||||
Fixed example:
|
||||
|
||||
@ -20,5 +20,5 @@ Fixed example:
|
||||
#![feature(repr_simd)]
|
||||
|
||||
#[repr(simd)]
|
||||
struct Good(u32, u32, u32, u32); // ok!
|
||||
struct Good([u16; 1]); // ok!
|
||||
```
|
||||
|
@ -7,7 +7,7 @@ Erroneous code example:
|
||||
#![feature(repr_simd)]
|
||||
|
||||
#[repr(simd)]
|
||||
struct Bad(String); // error!
|
||||
struct Bad([String; 2]); // error!
|
||||
```
|
||||
|
||||
When using the `#[simd]` attribute on a tuple struct, the elements in the tuple
|
||||
@ -19,5 +19,5 @@ Fixed example:
|
||||
#![feature(repr_simd)]
|
||||
|
||||
#[repr(simd)]
|
||||
struct Good(u32, u32, u32, u32); // ok!
|
||||
struct Good([u32; 4]); // ok!
|
||||
```
|
||||
|
@ -23,11 +23,11 @@ The generic type has to be a SIMD type. Example:
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone)]
|
||||
struct i32x2(i32, i32);
|
||||
struct i32x2([i32; 2]);
|
||||
|
||||
extern "rust-intrinsic" {
|
||||
fn simd_add<T>(a: T, b: T) -> T;
|
||||
}
|
||||
|
||||
unsafe { simd_add(i32x2(0, 0), i32x2(1, 2)); } // ok!
|
||||
unsafe { simd_add(i32x2([0, 0]), i32x2([1, 2])); } // ok!
|
||||
```
|
||||
|
@ -1064,20 +1064,29 @@ fn check_simd(tcx: TyCtxt<'_>, sp: Span, def_id: LocalDefId) {
|
||||
struct_span_code_err!(tcx.dcx(), sp, E0075, "SIMD vector cannot be empty").emit();
|
||||
return;
|
||||
}
|
||||
let e = fields[FieldIdx::ZERO].ty(tcx, args);
|
||||
if !fields.iter().all(|f| f.ty(tcx, args) == e) {
|
||||
struct_span_code_err!(tcx.dcx(), sp, E0076, "SIMD vector should be homogeneous")
|
||||
.with_span_label(sp, "SIMD elements must have the same type")
|
||||
|
||||
let array_field = &fields[FieldIdx::ZERO];
|
||||
let array_ty = array_field.ty(tcx, args);
|
||||
let ty::Array(element_ty, len_const) = array_ty.kind() else {
|
||||
struct_span_code_err!(
|
||||
tcx.dcx(),
|
||||
sp,
|
||||
E0076,
|
||||
"SIMD vector's only field must be an array"
|
||||
)
|
||||
.with_span_label(tcx.def_span(array_field.did), "not an array")
|
||||
.emit();
|
||||
return;
|
||||
};
|
||||
|
||||
if let Some(second_field) = fields.get(FieldIdx::from_u32(1)) {
|
||||
struct_span_code_err!(tcx.dcx(), sp, E0075, "SIMD vector cannot have multiple fields")
|
||||
.with_span_label(tcx.def_span(second_field.did), "excess field")
|
||||
.emit();
|
||||
return;
|
||||
}
|
||||
|
||||
let len = if let ty::Array(_ty, c) = e.kind() {
|
||||
c.try_eval_target_usize(tcx, tcx.param_env(def.did()))
|
||||
} else {
|
||||
Some(fields.len() as u64)
|
||||
};
|
||||
if let Some(len) = len {
|
||||
if let Some(len) = len_const.try_eval_target_usize(tcx, tcx.param_env(def.did())) {
|
||||
if len == 0 {
|
||||
struct_span_code_err!(tcx.dcx(), sp, E0075, "SIMD vector cannot be empty").emit();
|
||||
return;
|
||||
@ -1097,16 +1106,9 @@ fn check_simd(tcx: TyCtxt<'_>, sp: Span, def_id: LocalDefId) {
|
||||
// These are scalar types which directly match a "machine" type
|
||||
// Yes: Integers, floats, "thin" pointers
|
||||
// No: char, "fat" pointers, compound types
|
||||
match e.kind() {
|
||||
ty::Param(_) => (), // pass struct<T>(T, T, T, T) through, let monomorphization catch errors
|
||||
ty::Int(_) | ty::Uint(_) | ty::Float(_) | ty::RawPtr(_, _) => (), // struct(u8, u8, u8, u8) is ok
|
||||
ty::Array(t, _) if matches!(t.kind(), ty::Param(_)) => (), // pass struct<T>([T; N]) through, let monomorphization catch errors
|
||||
ty::Array(t, _clen)
|
||||
if matches!(
|
||||
t.kind(),
|
||||
ty::Int(_) | ty::Uint(_) | ty::Float(_) | ty::RawPtr(_, _)
|
||||
) =>
|
||||
{ /* struct([f32; 4]) is ok */ }
|
||||
match element_ty.kind() {
|
||||
ty::Param(_) => (), // pass struct<T>([T; 4]) through, let monomorphization catch errors
|
||||
ty::Int(_) | ty::Uint(_) | ty::Float(_) | ty::RawPtr(_, _) => (), // struct([u8; 4]) is ok
|
||||
_ => {
|
||||
struct_span_code_err!(
|
||||
tcx.dcx(),
|
||||
|
@ -1091,29 +1091,21 @@ pub fn sequence_element_type(self, tcx: TyCtxt<'tcx>) -> Ty<'tcx> {
|
||||
}
|
||||
|
||||
pub fn simd_size_and_type(self, tcx: TyCtxt<'tcx>) -> (u64, Ty<'tcx>) {
|
||||
match self.kind() {
|
||||
Adt(def, args) => {
|
||||
assert!(def.repr().simd(), "`simd_size_and_type` called on non-SIMD type");
|
||||
let variant = def.non_enum_variant();
|
||||
let f0_ty = variant.fields[FieldIdx::ZERO].ty(tcx, args);
|
||||
|
||||
match f0_ty.kind() {
|
||||
// If the first field is an array, we assume it is the only field and its
|
||||
// elements are the SIMD components.
|
||||
Array(f0_elem_ty, f0_len) => {
|
||||
// FIXME(repr_simd): https://github.com/rust-lang/rust/pull/78863#discussion_r522784112
|
||||
// The way we evaluate the `N` in `[T; N]` here only works since we use
|
||||
// `simd_size_and_type` post-monomorphization. It will probably start to ICE
|
||||
// if we use it in generic code. See the `simd-array-trait` ui test.
|
||||
(f0_len.eval_target_usize(tcx, ParamEnv::empty()), *f0_elem_ty)
|
||||
}
|
||||
// Otherwise, the fields of this Adt are the SIMD components (and we assume they
|
||||
// all have the same type).
|
||||
_ => (variant.fields.len() as u64, f0_ty),
|
||||
}
|
||||
}
|
||||
_ => bug!("`simd_size_and_type` called on invalid type"),
|
||||
}
|
||||
let Adt(def, args) = self.kind() else {
|
||||
bug!("`simd_size_and_type` called on invalid type")
|
||||
};
|
||||
assert!(def.repr().simd(), "`simd_size_and_type` called on non-SIMD type");
|
||||
let variant = def.non_enum_variant();
|
||||
assert_eq!(variant.fields.len(), 1);
|
||||
let field_ty = variant.fields[FieldIdx::ZERO].ty(tcx, args);
|
||||
let Array(f0_elem_ty, f0_len) = field_ty.kind() else {
|
||||
bug!("Simd type has non-array field type {field_ty:?}")
|
||||
};
|
||||
// FIXME(repr_simd): https://github.com/rust-lang/rust/pull/78863#discussion_r522784112
|
||||
// The way we evaluate the `N` in `[T; N]` here only works since we use
|
||||
// `simd_size_and_type` post-monomorphization. It will probably start to ICE
|
||||
// if we use it in generic code. See the `simd-array-trait` ui test.
|
||||
(f0_len.eval_target_usize(tcx, ParamEnv::empty()), *f0_elem_ty)
|
||||
}
|
||||
|
||||
#[inline]
|
||||
|
@ -336,10 +336,10 @@ fn $name(b: &mut Bencher) {
|
||||
reverse!(reverse_u64, u64, |x| x as u64);
|
||||
reverse!(reverse_u128, u128, |x| x as u128);
|
||||
#[repr(simd)]
|
||||
struct F64x4(f64, f64, f64, f64);
|
||||
struct F64x4([f64; 4]);
|
||||
reverse!(reverse_simd_f64x4, F64x4, |x| {
|
||||
let x = x as f64;
|
||||
F64x4(x, x, x, x)
|
||||
F64x4([x, x, x, x])
|
||||
});
|
||||
|
||||
macro_rules! rotate {
|
||||
|
@ -31,36 +31,39 @@ trait Sized {}
|
||||
#[lang = "copy"]
|
||||
trait Copy {}
|
||||
|
||||
// Do we really need to use no_core for this?!?
|
||||
impl<T: Copy, const N: usize> Copy for [T; N] {}
|
||||
|
||||
type ptr = *mut u8;
|
||||
|
||||
#[repr(simd)]
|
||||
pub struct i8x8(i8, i8, i8, i8, i8, i8, i8, i8);
|
||||
pub struct i8x8([i8; 8]);
|
||||
#[repr(simd)]
|
||||
pub struct i16x4(i16, i16, i16, i16);
|
||||
pub struct i16x4([i16; 4]);
|
||||
#[repr(simd)]
|
||||
pub struct i32x2(i32, i32);
|
||||
pub struct i32x2([i32; 2]);
|
||||
#[repr(simd)]
|
||||
pub struct i64x1(i64);
|
||||
pub struct i64x1([i64; 1]);
|
||||
#[repr(simd)]
|
||||
pub struct f16x4(f16, f16, f16, f16);
|
||||
pub struct f16x4([f16; 4]);
|
||||
#[repr(simd)]
|
||||
pub struct f32x2(f32, f32);
|
||||
pub struct f32x2([f32; 2]);
|
||||
#[repr(simd)]
|
||||
pub struct f64x1(f64);
|
||||
pub struct f64x1([f64; 1]);
|
||||
#[repr(simd)]
|
||||
pub struct i8x16(i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8);
|
||||
pub struct i8x16([i8; 16]);
|
||||
#[repr(simd)]
|
||||
pub struct i16x8(i16, i16, i16, i16, i16, i16, i16, i16);
|
||||
pub struct i16x8([i16; 8]);
|
||||
#[repr(simd)]
|
||||
pub struct i32x4(i32, i32, i32, i32);
|
||||
pub struct i32x4([i32; 4]);
|
||||
#[repr(simd)]
|
||||
pub struct i64x2(i64, i64);
|
||||
pub struct i64x2([i64; 2]);
|
||||
#[repr(simd)]
|
||||
pub struct f16x8(f16, f16, f16, f16, f16, f16, f16, f16);
|
||||
pub struct f16x8([f16; 8]);
|
||||
#[repr(simd)]
|
||||
pub struct f32x4(f32, f32, f32, f32);
|
||||
pub struct f32x4([f32; 4]);
|
||||
#[repr(simd)]
|
||||
pub struct f64x2(f64, f64);
|
||||
pub struct f64x2([f64; 2]);
|
||||
|
||||
impl Copy for i8 {}
|
||||
impl Copy for i16 {}
|
||||
|
@ -28,8 +28,11 @@ trait Sized {}
|
||||
#[lang = "copy"]
|
||||
trait Copy {}
|
||||
|
||||
// Do we really need to use no_core for this?!?
|
||||
impl<T: Copy, const N: usize> Copy for [T; N] {}
|
||||
|
||||
#[repr(simd)]
|
||||
pub struct f32x4(f32, f32, f32, f32);
|
||||
pub struct f32x4([f32; 4]);
|
||||
|
||||
impl Copy for i32 {}
|
||||
impl Copy for f32 {}
|
||||
|
@ -31,32 +31,35 @@ trait Sized {}
|
||||
#[lang = "copy"]
|
||||
trait Copy {}
|
||||
|
||||
// Do we really need to use no_core for this?!?
|
||||
impl<T: Copy, const N: usize> Copy for [T; N] {}
|
||||
|
||||
type ptr = *mut u8;
|
||||
|
||||
#[repr(simd)]
|
||||
pub struct i8x8(i8, i8, i8, i8, i8, i8, i8, i8);
|
||||
pub struct i8x8([i8; 8]);
|
||||
#[repr(simd)]
|
||||
pub struct i16x4(i16, i16, i16, i16);
|
||||
pub struct i16x4([i16; 4]);
|
||||
#[repr(simd)]
|
||||
pub struct i32x2(i32, i32);
|
||||
pub struct i32x2([i32; 2]);
|
||||
#[repr(simd)]
|
||||
pub struct i64x1(i64);
|
||||
pub struct i64x1([i64; 1]);
|
||||
#[repr(simd)]
|
||||
pub struct f16x4(f16, f16, f16, f16);
|
||||
pub struct f16x4([f16; 4]);
|
||||
#[repr(simd)]
|
||||
pub struct f32x2(f32, f32);
|
||||
pub struct f32x2([f32; 2]);
|
||||
#[repr(simd)]
|
||||
pub struct i8x16(i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8);
|
||||
pub struct i8x16([i8; 16]);
|
||||
#[repr(simd)]
|
||||
pub struct i16x8(i16, i16, i16, i16, i16, i16, i16, i16);
|
||||
pub struct i16x8([i16; 8]);
|
||||
#[repr(simd)]
|
||||
pub struct i32x4(i32, i32, i32, i32);
|
||||
pub struct i32x4([i32; 4]);
|
||||
#[repr(simd)]
|
||||
pub struct i64x2(i64, i64);
|
||||
pub struct i64x2([i64; 2]);
|
||||
#[repr(simd)]
|
||||
pub struct f16x8(f16, f16, f16, f16, f16, f16, f16, f16);
|
||||
pub struct f16x8([f16; 8]);
|
||||
#[repr(simd)]
|
||||
pub struct f32x4(f32, f32, f32, f32);
|
||||
pub struct f32x4([f32; 4]);
|
||||
|
||||
impl Copy for i8 {}
|
||||
impl Copy for i16 {}
|
||||
|
@ -31,216 +31,55 @@ trait Sized {}
|
||||
#[lang = "copy"]
|
||||
trait Copy {}
|
||||
|
||||
// Do we really need to use no_core for this?!?
|
||||
impl<T: Copy, const N: usize> Copy for [T; N] {}
|
||||
|
||||
type ptr = *mut u8;
|
||||
|
||||
#[repr(simd)]
|
||||
pub struct i8x16(i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8);
|
||||
pub struct i8x16([i8; 16]);
|
||||
#[repr(simd)]
|
||||
pub struct i16x8(i16, i16, i16, i16, i16, i16, i16, i16);
|
||||
pub struct i16x8([i16; 8]);
|
||||
#[repr(simd)]
|
||||
pub struct i32x4(i32, i32, i32, i32);
|
||||
pub struct i32x4([i32; 4]);
|
||||
#[repr(simd)]
|
||||
pub struct i64x2(i64, i64);
|
||||
pub struct i64x2([i64; 2]);
|
||||
#[repr(simd)]
|
||||
pub struct f16x8(f16, f16, f16, f16, f16, f16, f16, f16);
|
||||
pub struct f16x8([f16; 8]);
|
||||
#[repr(simd)]
|
||||
pub struct f32x4(f32, f32, f32, f32);
|
||||
pub struct f32x4([f32; 4]);
|
||||
#[repr(simd)]
|
||||
pub struct f64x2(f64, f64);
|
||||
pub struct f64x2([f64; 2]);
|
||||
|
||||
#[repr(simd)]
|
||||
pub struct i8x32(
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
);
|
||||
pub struct i8x32([i8; 32]);
|
||||
#[repr(simd)]
|
||||
pub struct i16x16(i16, i16, i16, i16, i16, i16, i16, i16, i16, i16, i16, i16, i16, i16, i16, i16);
|
||||
pub struct i16x16([i16; 16]);
|
||||
#[repr(simd)]
|
||||
pub struct i32x8(i32, i32, i32, i32, i32, i32, i32, i32);
|
||||
pub struct i32x8([i32; 8]);
|
||||
#[repr(simd)]
|
||||
pub struct i64x4(i64, i64, i64, i64);
|
||||
pub struct i64x4([i64; 4]);
|
||||
#[repr(simd)]
|
||||
pub struct f16x16(f16, f16, f16, f16, f16, f16, f16, f16, f16, f16, f16, f16, f16, f16, f16, f16);
|
||||
pub struct f16x16([f16; 16]);
|
||||
#[repr(simd)]
|
||||
pub struct f32x8(f32, f32, f32, f32, f32, f32, f32, f32);
|
||||
pub struct f32x8([f32; 8]);
|
||||
#[repr(simd)]
|
||||
pub struct f64x4(f64, f64, f64, f64);
|
||||
pub struct f64x4([f64; 4]);
|
||||
|
||||
#[repr(simd)]
|
||||
pub struct i8x64(
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
i8,
|
||||
);
|
||||
pub struct i8x64([i8; 64]);
|
||||
#[repr(simd)]
|
||||
pub struct i16x32(
|
||||
i16,
|
||||
i16,
|
||||
i16,
|
||||
i16,
|
||||
i16,
|
||||
i16,
|
||||
i16,
|
||||
i16,
|
||||
i16,
|
||||
i16,
|
||||
i16,
|
||||
i16,
|
||||
i16,
|
||||
i16,
|
||||
i16,
|
||||
i16,
|
||||
i16,
|
||||
i16,
|
||||
i16,
|
||||
i16,
|
||||
i16,
|
||||
i16,
|
||||
i16,
|
||||
i16,
|
||||
i16,
|
||||
i16,
|
||||
i16,
|
||||
i16,
|
||||
i16,
|
||||
i16,
|
||||
i16,
|
||||
i16,
|
||||
);
|
||||
pub struct i16x32([i16; 32]);
|
||||
#[repr(simd)]
|
||||
pub struct i32x16(i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32);
|
||||
pub struct i32x16([i32; 16]);
|
||||
#[repr(simd)]
|
||||
pub struct i64x8(i64, i64, i64, i64, i64, i64, i64, i64);
|
||||
pub struct i64x8([i64; 8]);
|
||||
#[repr(simd)]
|
||||
pub struct f16x32(
|
||||
f16,
|
||||
f16,
|
||||
f16,
|
||||
f16,
|
||||
f16,
|
||||
f16,
|
||||
f16,
|
||||
f16,
|
||||
f16,
|
||||
f16,
|
||||
f16,
|
||||
f16,
|
||||
f16,
|
||||
f16,
|
||||
f16,
|
||||
f16,
|
||||
f16,
|
||||
f16,
|
||||
f16,
|
||||
f16,
|
||||
f16,
|
||||
f16,
|
||||
f16,
|
||||
f16,
|
||||
f16,
|
||||
f16,
|
||||
f16,
|
||||
f16,
|
||||
f16,
|
||||
f16,
|
||||
f16,
|
||||
f16,
|
||||
);
|
||||
pub struct f16x32([f16; 32]);
|
||||
#[repr(simd)]
|
||||
pub struct f32x16(f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32, f32);
|
||||
pub struct f32x16([f32; 16]);
|
||||
#[repr(simd)]
|
||||
pub struct f64x8(f64, f64, f64, f64, f64, f64, f64, f64);
|
||||
pub struct f64x8([f64; 8]);
|
||||
|
||||
macro_rules! impl_copy {
|
||||
($($ty:ident)*) => {
|
||||
|
@ -21,7 +21,7 @@ trait Freeze {}
|
||||
trait Copy {}
|
||||
|
||||
#[repr(simd)]
|
||||
pub struct i32x4(i32, i32, i32, i32);
|
||||
pub struct i32x4([i32; 4]);
|
||||
|
||||
#[repr(C)]
|
||||
pub struct Foo {
|
||||
@ -47,12 +47,12 @@ pub struct DoubleFoo {
|
||||
}
|
||||
|
||||
pub fn main() {
|
||||
unsafe { f(Foo { a: i32x4(1, 2, 3, 4), b: 0 }) }
|
||||
unsafe { f(Foo { a: i32x4([1, 2, 3, 4]), b: 0 }) }
|
||||
|
||||
unsafe {
|
||||
g(DoubleFoo {
|
||||
one: Foo { a: i32x4(1, 2, 3, 4), b: 0 },
|
||||
two: Foo { a: i32x4(1, 2, 3, 4), b: 0 },
|
||||
one: Foo { a: i32x4([1, 2, 3, 4]), b: 0 },
|
||||
two: Foo { a: i32x4([1, 2, 3, 4]), b: 0 },
|
||||
})
|
||||
}
|
||||
}
|
||||
|
@ -13,19 +13,11 @@
|
||||
// Setting up structs that can be used as const vectors
|
||||
#[repr(simd)]
|
||||
#[derive(Clone)]
|
||||
pub struct i8x2(i8, i8);
|
||||
pub struct i8x2([i8; 2]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Clone)]
|
||||
pub struct i8x2_arr([i8; 2]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Clone)]
|
||||
pub struct f32x2(f32, f32);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Clone)]
|
||||
pub struct f32x2_arr([f32; 2]);
|
||||
pub struct f32x2([f32; 2]);
|
||||
|
||||
#[repr(simd, packed)]
|
||||
#[derive(Copy, Clone)]
|
||||
@ -35,42 +27,34 @@
|
||||
// that they are called with a const vector
|
||||
|
||||
extern "unadjusted" {
|
||||
#[no_mangle]
|
||||
fn test_i8x2(a: i8x2);
|
||||
}
|
||||
|
||||
extern "unadjusted" {
|
||||
#[no_mangle]
|
||||
fn test_i8x2_two_args(a: i8x2, b: i8x2);
|
||||
}
|
||||
|
||||
extern "unadjusted" {
|
||||
#[no_mangle]
|
||||
fn test_i8x2_mixed_args(a: i8x2, c: i32, b: i8x2);
|
||||
}
|
||||
|
||||
extern "unadjusted" {
|
||||
#[no_mangle]
|
||||
fn test_i8x2_arr(a: i8x2_arr);
|
||||
fn test_i8x2_arr(a: i8x2);
|
||||
}
|
||||
|
||||
extern "unadjusted" {
|
||||
#[no_mangle]
|
||||
fn test_f32x2(a: f32x2);
|
||||
}
|
||||
|
||||
extern "unadjusted" {
|
||||
#[no_mangle]
|
||||
fn test_f32x2_arr(a: f32x2_arr);
|
||||
fn test_f32x2_arr(a: f32x2);
|
||||
}
|
||||
|
||||
extern "unadjusted" {
|
||||
#[no_mangle]
|
||||
fn test_simd(a: Simd<i32, 4>);
|
||||
}
|
||||
|
||||
extern "unadjusted" {
|
||||
#[no_mangle]
|
||||
fn test_simd_unaligned(a: Simd<i32, 3>);
|
||||
}
|
||||
|
||||
@ -81,22 +65,22 @@
|
||||
pub fn do_call() {
|
||||
unsafe {
|
||||
// CHECK: call void @test_i8x2(<2 x i8> <i8 32, i8 64>
|
||||
test_i8x2(const { i8x2(32, 64) });
|
||||
test_i8x2(const { i8x2([32, 64]) });
|
||||
|
||||
// CHECK: call void @test_i8x2_two_args(<2 x i8> <i8 32, i8 64>, <2 x i8> <i8 8, i8 16>
|
||||
test_i8x2_two_args(const { i8x2(32, 64) }, const { i8x2(8, 16) });
|
||||
test_i8x2_two_args(const { i8x2([32, 64]) }, const { i8x2([8, 16]) });
|
||||
|
||||
// CHECK: call void @test_i8x2_mixed_args(<2 x i8> <i8 32, i8 64>, i32 43, <2 x i8> <i8 8, i8 16>
|
||||
test_i8x2_mixed_args(const { i8x2(32, 64) }, 43, const { i8x2(8, 16) });
|
||||
test_i8x2_mixed_args(const { i8x2([32, 64]) }, 43, const { i8x2([8, 16]) });
|
||||
|
||||
// CHECK: call void @test_i8x2_arr(<2 x i8> <i8 32, i8 64>
|
||||
test_i8x2_arr(const { i8x2_arr([32, 64]) });
|
||||
test_i8x2_arr(const { i8x2([32, 64]) });
|
||||
|
||||
// CHECK: call void @test_f32x2(<2 x float> <float 0x3FD47AE140000000, float 0x3FE47AE140000000>
|
||||
test_f32x2(const { f32x2(0.32, 0.64) });
|
||||
test_f32x2(const { f32x2([0.32, 0.64]) });
|
||||
|
||||
// CHECK: void @test_f32x2_arr(<2 x float> <float 0x3FD47AE140000000, float 0x3FE47AE140000000>
|
||||
test_f32x2_arr(const { f32x2_arr([0.32, 0.64]) });
|
||||
test_f32x2_arr(const { f32x2([0.32, 0.64]) });
|
||||
|
||||
// CHECK: call void @test_simd(<4 x i32> <i32 2, i32 4, i32 6, i32 8>
|
||||
test_simd(const { Simd::<i32, 4>([2, 4, 6, 8]) });
|
||||
|
@ -132,7 +132,7 @@ pub extern "C" fn test_Nested2(_: Nested2) -> Nested2 {
|
||||
}
|
||||
|
||||
#[repr(simd)]
|
||||
struct f32x4(f32, f32, f32, f32);
|
||||
struct f32x4([f32; 4]);
|
||||
|
||||
#[repr(transparent)]
|
||||
pub struct Vector(f32x4);
|
||||
|
@ -7,23 +7,19 @@
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f32x2(pub f32, pub f32);
|
||||
pub struct f32x2(pub [f32; 2]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f32x4(pub f32, pub f32, pub f32, pub f32);
|
||||
pub struct f32x4(pub [f32; 4]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f32x8(pub f32, pub f32, pub f32, pub f32,
|
||||
pub f32, pub f32, pub f32, pub f32);
|
||||
pub struct f32x8(pub [f32; 8]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f32x16(pub f32, pub f32, pub f32, pub f32,
|
||||
pub f32, pub f32, pub f32, pub f32,
|
||||
pub f32, pub f32, pub f32, pub f32,
|
||||
pub f32, pub f32, pub f32, pub f32);
|
||||
pub struct f32x16(pub [f32; 16]);
|
||||
|
||||
extern "rust-intrinsic" {
|
||||
fn simd_fabs<T>(x: T) -> T;
|
||||
@ -59,16 +55,15 @@ pub unsafe fn fabs_32x16(a: f32x16) -> f32x16 {
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f64x2(pub f64, pub f64);
|
||||
pub struct f64x2(pub [f64; 2]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f64x4(pub f64, pub f64, pub f64, pub f64);
|
||||
pub struct f64x4(pub [f64; 4]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f64x8(pub f64, pub f64, pub f64, pub f64,
|
||||
pub f64, pub f64, pub f64, pub f64);
|
||||
pub struct f64x8(pub [f64; 8]);
|
||||
|
||||
// CHECK-LABEL: @fabs_64x4
|
||||
#[no_mangle]
|
||||
|
@ -7,23 +7,19 @@
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f32x2(pub f32, pub f32);
|
||||
pub struct f32x2(pub [f32; 2]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f32x4(pub f32, pub f32, pub f32, pub f32);
|
||||
pub struct f32x4(pub [f32; 4]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f32x8(pub f32, pub f32, pub f32, pub f32,
|
||||
pub f32, pub f32, pub f32, pub f32);
|
||||
pub struct f32x8(pub [f32; 8]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f32x16(pub f32, pub f32, pub f32, pub f32,
|
||||
pub f32, pub f32, pub f32, pub f32,
|
||||
pub f32, pub f32, pub f32, pub f32,
|
||||
pub f32, pub f32, pub f32, pub f32);
|
||||
pub struct f32x16(pub [f32; 16]);
|
||||
|
||||
extern "rust-intrinsic" {
|
||||
fn simd_ceil<T>(x: T) -> T;
|
||||
@ -59,16 +55,15 @@ pub unsafe fn ceil_32x16(a: f32x16) -> f32x16 {
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f64x2(pub f64, pub f64);
|
||||
pub struct f64x2(pub [f64; 2]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f64x4(pub f64, pub f64, pub f64, pub f64);
|
||||
pub struct f64x4(pub [f64; 4]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f64x8(pub f64, pub f64, pub f64, pub f64,
|
||||
pub f64, pub f64, pub f64, pub f64);
|
||||
pub struct f64x8(pub [f64; 8]);
|
||||
|
||||
// CHECK-LABEL: @ceil_64x4
|
||||
#[no_mangle]
|
||||
|
@ -7,23 +7,19 @@
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f32x2(pub f32, pub f32);
|
||||
pub struct f32x2(pub [f32; 2]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f32x4(pub f32, pub f32, pub f32, pub f32);
|
||||
pub struct f32x4(pub [f32; 4]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f32x8(pub f32, pub f32, pub f32, pub f32,
|
||||
pub f32, pub f32, pub f32, pub f32);
|
||||
pub struct f32x8(pub [f32; 8]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f32x16(pub f32, pub f32, pub f32, pub f32,
|
||||
pub f32, pub f32, pub f32, pub f32,
|
||||
pub f32, pub f32, pub f32, pub f32,
|
||||
pub f32, pub f32, pub f32, pub f32);
|
||||
pub struct f32x16(pub [f32; 16]);
|
||||
|
||||
extern "rust-intrinsic" {
|
||||
fn simd_fcos<T>(x: T) -> T;
|
||||
@ -59,16 +55,15 @@ pub unsafe fn fcos_32x16(a: f32x16) -> f32x16 {
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f64x2(pub f64, pub f64);
|
||||
pub struct f64x2(pub [f64; 2]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f64x4(pub f64, pub f64, pub f64, pub f64);
|
||||
pub struct f64x4(pub [f64; 4]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f64x8(pub f64, pub f64, pub f64, pub f64,
|
||||
pub f64, pub f64, pub f64, pub f64);
|
||||
pub struct f64x8(pub [f64; 8]);
|
||||
|
||||
// CHECK-LABEL: @fcos_64x4
|
||||
#[no_mangle]
|
||||
|
@ -7,23 +7,19 @@
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f32x2(pub f32, pub f32);
|
||||
pub struct f32x2(pub [f32; 2]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f32x4(pub f32, pub f32, pub f32, pub f32);
|
||||
pub struct f32x4(pub [f32; 4]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f32x8(pub f32, pub f32, pub f32, pub f32,
|
||||
pub f32, pub f32, pub f32, pub f32);
|
||||
pub struct f32x8(pub [f32; 8]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f32x16(pub f32, pub f32, pub f32, pub f32,
|
||||
pub f32, pub f32, pub f32, pub f32,
|
||||
pub f32, pub f32, pub f32, pub f32,
|
||||
pub f32, pub f32, pub f32, pub f32);
|
||||
pub struct f32x16(pub [f32; 16]);
|
||||
|
||||
extern "rust-intrinsic" {
|
||||
fn simd_fexp<T>(x: T) -> T;
|
||||
@ -59,16 +55,15 @@ pub unsafe fn exp_32x16(a: f32x16) -> f32x16 {
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f64x2(pub f64, pub f64);
|
||||
pub struct f64x2(pub [f64; 2]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f64x4(pub f64, pub f64, pub f64, pub f64);
|
||||
pub struct f64x4(pub [f64; 4]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f64x8(pub f64, pub f64, pub f64, pub f64,
|
||||
pub f64, pub f64, pub f64, pub f64);
|
||||
pub struct f64x8(pub [f64; 8]);
|
||||
|
||||
// CHECK-LABEL: @exp_64x4
|
||||
#[no_mangle]
|
||||
|
@ -7,23 +7,19 @@
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f32x2(pub f32, pub f32);
|
||||
pub struct f32x2(pub [f32; 2]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f32x4(pub f32, pub f32, pub f32, pub f32);
|
||||
pub struct f32x4(pub [f32; 4]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f32x8(pub f32, pub f32, pub f32, pub f32,
|
||||
pub f32, pub f32, pub f32, pub f32);
|
||||
pub struct f32x8(pub [f32; 8]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f32x16(pub f32, pub f32, pub f32, pub f32,
|
||||
pub f32, pub f32, pub f32, pub f32,
|
||||
pub f32, pub f32, pub f32, pub f32,
|
||||
pub f32, pub f32, pub f32, pub f32);
|
||||
pub struct f32x16(pub [f32; 16]);
|
||||
|
||||
extern "rust-intrinsic" {
|
||||
fn simd_fexp2<T>(x: T) -> T;
|
||||
@ -59,16 +55,15 @@ pub unsafe fn exp2_32x16(a: f32x16) -> f32x16 {
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f64x2(pub f64, pub f64);
|
||||
pub struct f64x2(pub [f64; 2]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f64x4(pub f64, pub f64, pub f64, pub f64);
|
||||
pub struct f64x4(pub [f64; 4]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f64x8(pub f64, pub f64, pub f64, pub f64,
|
||||
pub f64, pub f64, pub f64, pub f64);
|
||||
pub struct f64x8(pub [f64; 8]);
|
||||
|
||||
// CHECK-LABEL: @exp2_64x4
|
||||
#[no_mangle]
|
||||
|
@ -7,23 +7,19 @@
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f32x2(pub f32, pub f32);
|
||||
pub struct f32x2(pub [f32; 2]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f32x4(pub f32, pub f32, pub f32, pub f32);
|
||||
pub struct f32x4(pub [f32; 4]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f32x8(pub f32, pub f32, pub f32, pub f32,
|
||||
pub f32, pub f32, pub f32, pub f32);
|
||||
pub struct f32x8(pub [f32; 8]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f32x16(pub f32, pub f32, pub f32, pub f32,
|
||||
pub f32, pub f32, pub f32, pub f32,
|
||||
pub f32, pub f32, pub f32, pub f32,
|
||||
pub f32, pub f32, pub f32, pub f32);
|
||||
pub struct f32x16(pub [f32; 16]);
|
||||
|
||||
extern "rust-intrinsic" {
|
||||
fn simd_floor<T>(x: T) -> T;
|
||||
@ -59,16 +55,15 @@ pub unsafe fn floor_32x16(a: f32x16) -> f32x16 {
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f64x2(pub f64, pub f64);
|
||||
pub struct f64x2(pub [f64; 2]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f64x4(pub f64, pub f64, pub f64, pub f64);
|
||||
pub struct f64x4(pub [f64; 4]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f64x8(pub f64, pub f64, pub f64, pub f64,
|
||||
pub f64, pub f64, pub f64, pub f64);
|
||||
pub struct f64x8(pub [f64; 8]);
|
||||
|
||||
// CHECK-LABEL: @floor_64x4
|
||||
#[no_mangle]
|
||||
|
@ -7,23 +7,19 @@
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f32x2(pub f32, pub f32);
|
||||
pub struct f32x2(pub [f32; 2]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f32x4(pub f32, pub f32, pub f32, pub f32);
|
||||
pub struct f32x4(pub [f32; 4]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f32x8(pub f32, pub f32, pub f32, pub f32,
|
||||
pub f32, pub f32, pub f32, pub f32);
|
||||
pub struct f32x8(pub [f32; 8]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f32x16(pub f32, pub f32, pub f32, pub f32,
|
||||
pub f32, pub f32, pub f32, pub f32,
|
||||
pub f32, pub f32, pub f32, pub f32,
|
||||
pub f32, pub f32, pub f32, pub f32);
|
||||
pub struct f32x16(pub [f32; 16]);
|
||||
|
||||
extern "rust-intrinsic" {
|
||||
fn simd_fma<T>(x: T, b: T, c: T) -> T;
|
||||
@ -59,16 +55,15 @@ pub unsafe fn fma_32x16(a: f32x16, b: f32x16, c: f32x16) -> f32x16 {
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f64x2(pub f64, pub f64);
|
||||
pub struct f64x2(pub [f64; 2]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f64x4(pub f64, pub f64, pub f64, pub f64);
|
||||
pub struct f64x4(pub [f64; 4]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f64x8(pub f64, pub f64, pub f64, pub f64,
|
||||
pub f64, pub f64, pub f64, pub f64);
|
||||
pub struct f64x8(pub [f64; 8]);
|
||||
|
||||
// CHECK-LABEL: @fma_64x4
|
||||
#[no_mangle]
|
||||
|
@ -7,23 +7,19 @@
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f32x2(pub f32, pub f32);
|
||||
pub struct f32x2(pub [f32; 2]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f32x4(pub f32, pub f32, pub f32, pub f32);
|
||||
pub struct f32x4(pub [f32; 4]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f32x8(pub f32, pub f32, pub f32, pub f32,
|
||||
pub f32, pub f32, pub f32, pub f32);
|
||||
pub struct f32x8(pub [f32; 8]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f32x16(pub f32, pub f32, pub f32, pub f32,
|
||||
pub f32, pub f32, pub f32, pub f32,
|
||||
pub f32, pub f32, pub f32, pub f32,
|
||||
pub f32, pub f32, pub f32, pub f32);
|
||||
pub struct f32x16(pub [f32; 16]);
|
||||
|
||||
extern "rust-intrinsic" {
|
||||
fn simd_fsqrt<T>(x: T) -> T;
|
||||
@ -59,16 +55,15 @@ pub unsafe fn fsqrt_32x16(a: f32x16) -> f32x16 {
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f64x2(pub f64, pub f64);
|
||||
pub struct f64x2(pub [f64; 2]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f64x4(pub f64, pub f64, pub f64, pub f64);
|
||||
pub struct f64x4(pub [f64; 4]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f64x8(pub f64, pub f64, pub f64, pub f64,
|
||||
pub f64, pub f64, pub f64, pub f64);
|
||||
pub struct f64x8(pub [f64; 8]);
|
||||
|
||||
// CHECK-LABEL: @fsqrt_64x4
|
||||
#[no_mangle]
|
||||
|
@ -7,23 +7,19 @@
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f32x2(pub f32, pub f32);
|
||||
pub struct f32x2(pub [f32; 2]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f32x4(pub f32, pub f32, pub f32, pub f32);
|
||||
pub struct f32x4(pub [f32; 4]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f32x8(pub f32, pub f32, pub f32, pub f32,
|
||||
pub f32, pub f32, pub f32, pub f32);
|
||||
pub struct f32x8(pub [f32; 8]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f32x16(pub f32, pub f32, pub f32, pub f32,
|
||||
pub f32, pub f32, pub f32, pub f32,
|
||||
pub f32, pub f32, pub f32, pub f32,
|
||||
pub f32, pub f32, pub f32, pub f32);
|
||||
pub struct f32x16(pub [f32; 16]);
|
||||
|
||||
extern "rust-intrinsic" {
|
||||
fn simd_flog<T>(x: T) -> T;
|
||||
@ -59,16 +55,15 @@ pub unsafe fn log_32x16(a: f32x16) -> f32x16 {
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f64x2(pub f64, pub f64);
|
||||
pub struct f64x2(pub [f64; 2]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f64x4(pub f64, pub f64, pub f64, pub f64);
|
||||
pub struct f64x4(pub [f64; 4]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f64x8(pub f64, pub f64, pub f64, pub f64,
|
||||
pub f64, pub f64, pub f64, pub f64);
|
||||
pub struct f64x8(pub [f64; 8]);
|
||||
|
||||
// CHECK-LABEL: @log_64x4
|
||||
#[no_mangle]
|
||||
|
@ -7,23 +7,19 @@
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f32x2(pub f32, pub f32);
|
||||
pub struct f32x2(pub [f32; 2]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f32x4(pub f32, pub f32, pub f32, pub f32);
|
||||
pub struct f32x4(pub [f32; 4]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f32x8(pub f32, pub f32, pub f32, pub f32,
|
||||
pub f32, pub f32, pub f32, pub f32);
|
||||
pub struct f32x8(pub [f32; 8]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f32x16(pub f32, pub f32, pub f32, pub f32,
|
||||
pub f32, pub f32, pub f32, pub f32,
|
||||
pub f32, pub f32, pub f32, pub f32,
|
||||
pub f32, pub f32, pub f32, pub f32);
|
||||
pub struct f32x16(pub [f32; 16]);
|
||||
|
||||
extern "rust-intrinsic" {
|
||||
fn simd_flog10<T>(x: T) -> T;
|
||||
@ -59,16 +55,15 @@ pub unsafe fn log10_32x16(a: f32x16) -> f32x16 {
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f64x2(pub f64, pub f64);
|
||||
pub struct f64x2(pub [f64; 2]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f64x4(pub f64, pub f64, pub f64, pub f64);
|
||||
pub struct f64x4(pub [f64; 4]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f64x8(pub f64, pub f64, pub f64, pub f64,
|
||||
pub f64, pub f64, pub f64, pub f64);
|
||||
pub struct f64x8(pub [f64; 8]);
|
||||
|
||||
// CHECK-LABEL: @log10_64x4
|
||||
#[no_mangle]
|
||||
|
@ -7,23 +7,19 @@
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f32x2(pub f32, pub f32);
|
||||
pub struct f32x2(pub [f32; 2]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f32x4(pub f32, pub f32, pub f32, pub f32);
|
||||
pub struct f32x4(pub [f32; 4]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f32x8(pub f32, pub f32, pub f32, pub f32,
|
||||
pub f32, pub f32, pub f32, pub f32);
|
||||
pub struct f32x8(pub [f32; 8]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f32x16(pub f32, pub f32, pub f32, pub f32,
|
||||
pub f32, pub f32, pub f32, pub f32,
|
||||
pub f32, pub f32, pub f32, pub f32,
|
||||
pub f32, pub f32, pub f32, pub f32);
|
||||
pub struct f32x16(pub [f32; 16]);
|
||||
|
||||
extern "rust-intrinsic" {
|
||||
fn simd_flog2<T>(x: T) -> T;
|
||||
@ -59,16 +55,15 @@ pub unsafe fn log2_32x16(a: f32x16) -> f32x16 {
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f64x2(pub f64, pub f64);
|
||||
pub struct f64x2(pub [f64; 2]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f64x4(pub f64, pub f64, pub f64, pub f64);
|
||||
pub struct f64x4(pub [f64; 4]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f64x8(pub f64, pub f64, pub f64, pub f64,
|
||||
pub f64, pub f64, pub f64, pub f64);
|
||||
pub struct f64x8(pub [f64; 8]);
|
||||
|
||||
// CHECK-LABEL: @log2_64x4
|
||||
#[no_mangle]
|
||||
|
@ -7,7 +7,7 @@
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f32x4(pub f32, pub f32, pub f32, pub f32);
|
||||
pub struct f32x4(pub [f32; 4]);
|
||||
|
||||
extern "rust-intrinsic" {
|
||||
fn simd_fmin<T>(x: T, y: T) -> T;
|
||||
|
@ -7,23 +7,19 @@
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f32x2(pub f32, pub f32);
|
||||
pub struct f32x2(pub [f32; 2]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f32x4(pub f32, pub f32, pub f32, pub f32);
|
||||
pub struct f32x4(pub [f32; 4]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f32x8(pub f32, pub f32, pub f32, pub f32,
|
||||
pub f32, pub f32, pub f32, pub f32);
|
||||
pub struct f32x8(pub [f32; 8]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f32x16(pub f32, pub f32, pub f32, pub f32,
|
||||
pub f32, pub f32, pub f32, pub f32,
|
||||
pub f32, pub f32, pub f32, pub f32,
|
||||
pub f32, pub f32, pub f32, pub f32);
|
||||
pub struct f32x16(pub [f32; 16]);
|
||||
|
||||
extern "rust-intrinsic" {
|
||||
fn simd_fpow<T>(x: T, b: T) -> T;
|
||||
@ -59,16 +55,15 @@ pub unsafe fn fpow_32x16(a: f32x16, b: f32x16) -> f32x16 {
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f64x2(pub f64, pub f64);
|
||||
pub struct f64x2(pub [f64; 2]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f64x4(pub f64, pub f64, pub f64, pub f64);
|
||||
pub struct f64x4(pub [f64; 4]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f64x8(pub f64, pub f64, pub f64, pub f64,
|
||||
pub f64, pub f64, pub f64, pub f64);
|
||||
pub struct f64x8(pub [f64; 8]);
|
||||
|
||||
// CHECK-LABEL: @fpow_64x4
|
||||
#[no_mangle]
|
||||
|
@ -7,23 +7,19 @@
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f32x2(pub f32, pub f32);
|
||||
pub struct f32x2(pub [f32; 2]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f32x4(pub f32, pub f32, pub f32, pub f32);
|
||||
pub struct f32x4(pub [f32; 4]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f32x8(pub f32, pub f32, pub f32, pub f32,
|
||||
pub f32, pub f32, pub f32, pub f32);
|
||||
pub struct f32x8(pub [f32; 8]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f32x16(pub f32, pub f32, pub f32, pub f32,
|
||||
pub f32, pub f32, pub f32, pub f32,
|
||||
pub f32, pub f32, pub f32, pub f32,
|
||||
pub f32, pub f32, pub f32, pub f32);
|
||||
pub struct f32x16(pub [f32; 16]);
|
||||
|
||||
extern "rust-intrinsic" {
|
||||
fn simd_fpowi<T>(x: T, b: i32) -> T;
|
||||
@ -59,16 +55,15 @@ pub unsafe fn fpowi_32x16(a: f32x16, b: i32) -> f32x16 {
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f64x2(pub f64, pub f64);
|
||||
pub struct f64x2(pub [f64; 2]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f64x4(pub f64, pub f64, pub f64, pub f64);
|
||||
pub struct f64x4(pub [f64; 4]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f64x8(pub f64, pub f64, pub f64, pub f64,
|
||||
pub f64, pub f64, pub f64, pub f64);
|
||||
pub struct f64x8(pub [f64; 8]);
|
||||
|
||||
// CHECK-LABEL: @fpowi_64x4
|
||||
#[no_mangle]
|
||||
|
@ -7,23 +7,19 @@
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f32x2(pub f32, pub f32);
|
||||
pub struct f32x2(pub [f32; 2]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f32x4(pub f32, pub f32, pub f32, pub f32);
|
||||
pub struct f32x4(pub [f32; 4]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f32x8(pub f32, pub f32, pub f32, pub f32,
|
||||
pub f32, pub f32, pub f32, pub f32);
|
||||
pub struct f32x8(pub [f32; 8]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f32x16(pub f32, pub f32, pub f32, pub f32,
|
||||
pub f32, pub f32, pub f32, pub f32,
|
||||
pub f32, pub f32, pub f32, pub f32,
|
||||
pub f32, pub f32, pub f32, pub f32);
|
||||
pub struct f32x16(pub [f32; 16]);
|
||||
|
||||
extern "rust-intrinsic" {
|
||||
fn simd_fsin<T>(x: T) -> T;
|
||||
@ -59,16 +55,15 @@ pub unsafe fn fsin_32x16(a: f32x16) -> f32x16 {
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f64x2(pub f64, pub f64);
|
||||
pub struct f64x2(pub [f64; 2]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f64x4(pub f64, pub f64, pub f64, pub f64);
|
||||
pub struct f64x4(pub [f64; 4]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f64x8(pub f64, pub f64, pub f64, pub f64,
|
||||
pub f64, pub f64, pub f64, pub f64);
|
||||
pub struct f64x8(pub [f64; 8]);
|
||||
|
||||
// CHECK-LABEL: @fsin_64x4
|
||||
#[no_mangle]
|
||||
|
@ -9,107 +9,57 @@
|
||||
|
||||
// signed integer types
|
||||
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct i8x2(i8, i8);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct i8x4(i8, i8, i8, i8);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct i8x8(
|
||||
i8, i8, i8, i8, i8, i8, i8, i8,
|
||||
);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct i8x16(
|
||||
i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8,
|
||||
);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct i8x32(
|
||||
i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8,
|
||||
i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8,
|
||||
);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct i8x64(
|
||||
i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8,
|
||||
i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8,
|
||||
i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8,
|
||||
i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8,
|
||||
);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct i8x2([i8; 2]);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct i8x4([i8; 4]);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct i8x8([i8; 8]);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct i8x16([i8; 16]);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct i8x32([i8; 32]);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct i8x64([i8; 64]);
|
||||
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct i16x2(i16, i16);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct i16x4(i16, i16, i16, i16);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct i16x8(
|
||||
i16, i16, i16, i16, i16, i16, i16, i16,
|
||||
);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct i16x16(
|
||||
i16, i16, i16, i16, i16, i16, i16, i16, i16, i16, i16, i16, i16, i16, i16, i16,
|
||||
);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct i16x32(
|
||||
i16, i16, i16, i16, i16, i16, i16, i16, i16, i16, i16, i16, i16, i16, i16, i16,
|
||||
i16, i16, i16, i16, i16, i16, i16, i16, i16, i16, i16, i16, i16, i16, i16, i16,
|
||||
);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct i16x2([i16; 2]);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct i16x4([i16; 4]);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct i16x8([i16; 8]);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct i16x16([i16; 16]);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct i16x32([i16; 32]);
|
||||
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct i32x2(i32, i32);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct i32x4(i32, i32, i32, i32);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct i32x8(
|
||||
i32, i32, i32, i32, i32, i32, i32, i32,
|
||||
);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct i32x16(
|
||||
i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32,
|
||||
);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct i32x2([i32; 2]);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct i32x4([i32; 4]);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct i32x8([i32; 8]);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct i32x16([i32; 16]);
|
||||
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct i64x2(i64, i64);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct i64x4(i64, i64, i64, i64);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct i64x8(
|
||||
i64, i64, i64, i64, i64, i64, i64, i64,
|
||||
);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct i64x2([i64; 2]);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct i64x4([i64; 4]);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct i64x8([i64; 8]);
|
||||
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct i128x2(i128, i128);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct i128x4(i128, i128, i128, i128);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct i128x2([i128; 2]);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct i128x4([i128; 4]);
|
||||
|
||||
// unsigned integer types
|
||||
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct u8x2(u8, u8);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct u8x4(u8, u8, u8, u8);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct u8x8(
|
||||
u8, u8, u8, u8, u8, u8, u8, u8,
|
||||
);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct u8x16(
|
||||
u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8,
|
||||
);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct u8x32(
|
||||
u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8,
|
||||
u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8,
|
||||
);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct u8x64(
|
||||
u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8,
|
||||
u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8,
|
||||
u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8,
|
||||
u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8,
|
||||
);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct u8x2([u8; 2]);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct u8x4([u8; 4]);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct u8x8([u8; 8]);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct u8x16([u8; 16]);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct u8x32([u8; 32]);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct u8x64([u8; 64]);
|
||||
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct u16x2(u16, u16);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct u16x4(u16, u16, u16, u16);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct u16x8(
|
||||
u16, u16, u16, u16, u16, u16, u16, u16,
|
||||
);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct u16x16(
|
||||
u16, u16, u16, u16, u16, u16, u16, u16, u16, u16, u16, u16, u16, u16, u16, u16,
|
||||
);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct u16x32(
|
||||
u16, u16, u16, u16, u16, u16, u16, u16, u16, u16, u16, u16, u16, u16, u16, u16,
|
||||
u16, u16, u16, u16, u16, u16, u16, u16, u16, u16, u16, u16, u16, u16, u16, u16,
|
||||
);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct u16x2([u16; 2]);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct u16x4([u16; 4]);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct u16x8([u16; 8]);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct u16x16([u16; 16]);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct u16x32([u16; 32]);
|
||||
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct u32x2(u32, u32);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct u32x4(u32, u32, u32, u32);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct u32x8(
|
||||
u32, u32, u32, u32, u32, u32, u32, u32,
|
||||
);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct u32x16(
|
||||
u32, u32, u32, u32, u32, u32, u32, u32, u32, u32, u32, u32, u32, u32, u32, u32,
|
||||
);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct u32x2([u32; 2]);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct u32x4([u32; 4]);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct u32x8([u32; 8]);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct u32x16([u32; 16]);
|
||||
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct u64x2(u64, u64);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct u64x4(u64, u64, u64, u64);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct u64x8(
|
||||
u64, u64, u64, u64, u64, u64, u64, u64,
|
||||
);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct u64x2([u64; 2]);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct u64x4([u64; 4]);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct u64x8([u64; 8]);
|
||||
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct u128x2(u128, u128);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct u128x4(u128, u128, u128, u128);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct u128x2([u128; 2]);
|
||||
#[repr(simd)] #[derive(Copy, Clone)] pub struct u128x4([u128; 4]);
|
||||
|
||||
extern "rust-intrinsic" {
|
||||
fn simd_saturating_add<T>(x: T, y: T) -> T;
|
||||
|
@ -8,19 +8,15 @@
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone)]
|
||||
pub struct u32x2(u32, u32);
|
||||
pub struct u32x2([u32; 2]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone)]
|
||||
pub struct i32x2(i32, i32);
|
||||
pub struct i32x2([i32; 2]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone)]
|
||||
pub struct i8x16(
|
||||
i8, i8, i8, i8, i8, i8, i8, i8,
|
||||
i8, i8, i8, i8, i8, i8, i8, i8,
|
||||
);
|
||||
|
||||
pub struct i8x16([i8; 16]);
|
||||
|
||||
extern "rust-intrinsic" {
|
||||
fn simd_bitmask<T, U>(x: T) -> U;
|
||||
|
@ -9,11 +9,11 @@
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct Vec2<T>(pub T, pub T);
|
||||
pub struct Vec2<T>(pub [T; 2]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct Vec4<T>(pub T, pub T, pub T, pub T);
|
||||
pub struct Vec4<T>(pub [T; 4]);
|
||||
|
||||
extern "rust-intrinsic" {
|
||||
fn simd_gather<T, P, M>(value: T, pointers: P, mask: M) -> T;
|
||||
|
@ -7,11 +7,11 @@
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct Vec2<T>(pub T, pub T);
|
||||
pub struct Vec2<T>(pub [T; 2]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct Vec4<T>(pub T, pub T, pub T, pub T);
|
||||
pub struct Vec4<T>(pub [T; 4]);
|
||||
|
||||
extern "rust-intrinsic" {
|
||||
fn simd_masked_load<M, P, T>(mask: M, pointer: P, values: T) -> T;
|
||||
|
@ -7,11 +7,11 @@
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct Vec2<T>(pub T, pub T);
|
||||
pub struct Vec2<T>(pub [T; 2]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct Vec4<T>(pub T, pub T, pub T, pub T);
|
||||
pub struct Vec4<T>(pub [T; 4]);
|
||||
|
||||
extern "rust-intrinsic" {
|
||||
fn simd_masked_store<M, P, T>(mask: M, pointer: P, values: T) -> ();
|
||||
|
@ -9,11 +9,11 @@
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct Vec2<T>(pub T, pub T);
|
||||
pub struct Vec2<T>(pub [T; 2]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct Vec4<T>(pub T, pub T, pub T, pub T);
|
||||
pub struct Vec4<T>(pub [T; 4]);
|
||||
|
||||
extern "rust-intrinsic" {
|
||||
fn simd_scatter<T, P, M>(value: T, pointers: P, mask: M);
|
||||
|
@ -7,15 +7,15 @@
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f32x4(pub f32, pub f32, pub f32, pub f32);
|
||||
pub struct f32x4(pub [f32; 4]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct f32x8(f32, f32, f32, f32, f32, f32, f32, f32);
|
||||
pub struct f32x8([f32; 8]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
pub struct b8x4(pub i8, pub i8, pub i8, pub i8);
|
||||
pub struct b8x4(pub [i8; 4]);
|
||||
|
||||
extern "rust-intrinsic" {
|
||||
fn simd_select<T, U>(x: T, a: U, b: U) -> U;
|
||||
|
@ -13,10 +13,6 @@
|
||||
#[derive(Copy, Clone)]
|
||||
pub struct T([f32; 4]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone)]
|
||||
pub struct U(f32, f32, f32, f32);
|
||||
|
||||
// CHECK-LABEL: @array_align(
|
||||
#[no_mangle]
|
||||
pub fn array_align() -> usize {
|
||||
@ -28,7 +24,7 @@ pub fn array_align() -> usize {
|
||||
#[no_mangle]
|
||||
pub fn vector_align() -> usize {
|
||||
// CHECK: ret [[USIZE]] [[VECTOR_ALIGN:[0-9]+]]
|
||||
const { std::mem::align_of::<U>() }
|
||||
const { std::mem::align_of::<T>() }
|
||||
}
|
||||
|
||||
// CHECK-LABEL: @build_array_s
|
||||
@ -60,22 +56,3 @@ pub fn vector_align() -> usize {
|
||||
// CHECK: store <4 x float> %[[VAL:.+]], ptr %_0, align [[VECTOR_ALIGN]]
|
||||
unsafe { std::mem::transmute(x) }
|
||||
}
|
||||
|
||||
// CHECK-LABEL: @build_array_u
|
||||
#[no_mangle]
|
||||
pub fn build_array_u(x: [f32; 4]) -> U {
|
||||
// CHECK: store float %a, {{.+}}, align [[VECTOR_ALIGN]]
|
||||
// CHECK: store float %b, {{.+}}, align [[ARRAY_ALIGN]]
|
||||
// CHECK: store float %c, {{.+}}, align
|
||||
// CHECK: store float %d, {{.+}}, align [[ARRAY_ALIGN]]
|
||||
let [a, b, c, d] = x;
|
||||
U(a, b, c, d)
|
||||
}
|
||||
|
||||
// CHECK-LABEL: @build_array_transmute_u
|
||||
#[no_mangle]
|
||||
pub fn build_array_transmute_u(x: [f32; 4]) -> U {
|
||||
// CHECK: %[[VAL:.+]] = load <4 x float>, ptr %x, align [[ARRAY_ALIGN]]
|
||||
// CHECK: store <4 x float> %[[VAL:.+]], ptr %_0, align [[VECTOR_ALIGN]]
|
||||
unsafe { std::mem::transmute(x) }
|
||||
}
|
||||
|
@ -7,7 +7,7 @@
|
||||
|
||||
#[derive(Copy, Clone)]
|
||||
#[repr(simd)]
|
||||
pub struct int16x4_t(pub i16, pub i16, pub i16, pub i16);
|
||||
pub struct int16x4_t(pub [i16; 4]);
|
||||
|
||||
#[derive(Copy, Clone)]
|
||||
pub struct int16x4x2_t(pub int16x4_t, pub int16x4_t);
|
||||
|
@ -16,7 +16,7 @@ pub enum Unhab {}
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone)]
|
||||
pub struct i64x4(i64, i64, i64, i64);
|
||||
pub struct i64x4([i64; 4]);
|
||||
|
||||
#[derive(Copy, Clone)]
|
||||
pub union UnionI64x4 {
|
||||
|
@ -27,7 +27,7 @@ pub fn scalarpair_layout(s: &(u64, u32, ())) {
|
||||
}
|
||||
|
||||
#[repr(simd)]
|
||||
pub struct U64x4(u64, u64, u64, u64);
|
||||
pub struct U64x4([u64; 4]);
|
||||
|
||||
// Check that we correctly generate a GEP for a ZST that is not included in Vector layout
|
||||
// CHECK-LABEL: @vector_layout
|
||||
|
@ -10,27 +10,27 @@
|
||||
// gdb-command:run
|
||||
|
||||
// gdb-command:print vi8x16
|
||||
// gdb-check:$1 = simd::i8x16 (0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15)
|
||||
// gdb-check:$1 = simd::i8x16 ([0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15])
|
||||
// gdb-command:print vi16x8
|
||||
// gdb-check:$2 = simd::i16x8 (16, 17, 18, 19, 20, 21, 22, 23)
|
||||
// gdb-check:$2 = simd::i16x8 ([16, 17, 18, 19, 20, 21, 22, 23])
|
||||
// gdb-command:print vi32x4
|
||||
// gdb-check:$3 = simd::i32x4 (24, 25, 26, 27)
|
||||
// gdb-check:$3 = simd::i32x4 ([24, 25, 26, 27])
|
||||
// gdb-command:print vi64x2
|
||||
// gdb-check:$4 = simd::i64x2 (28, 29)
|
||||
// gdb-check:$4 = simd::i64x2 ([28, 29])
|
||||
|
||||
// gdb-command:print vu8x16
|
||||
// gdb-check:$5 = simd::u8x16 (30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45)
|
||||
// gdb-check:$5 = simd::u8x16 ([30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45])
|
||||
// gdb-command:print vu16x8
|
||||
// gdb-check:$6 = simd::u16x8 (46, 47, 48, 49, 50, 51, 52, 53)
|
||||
// gdb-check:$6 = simd::u16x8 ([46, 47, 48, 49, 50, 51, 52, 53])
|
||||
// gdb-command:print vu32x4
|
||||
// gdb-check:$7 = simd::u32x4 (54, 55, 56, 57)
|
||||
// gdb-check:$7 = simd::u32x4 ([54, 55, 56, 57])
|
||||
// gdb-command:print vu64x2
|
||||
// gdb-check:$8 = simd::u64x2 (58, 59)
|
||||
// gdb-check:$8 = simd::u64x2 ([58, 59])
|
||||
|
||||
// gdb-command:print vf32x4
|
||||
// gdb-check:$9 = simd::f32x4 (60.5, 61.5, 62.5, 63.5)
|
||||
// gdb-check:$9 = simd::f32x4 ([60.5, 61.5, 62.5, 63.5])
|
||||
// gdb-command:print vf64x2
|
||||
// gdb-check:$10 = simd::f64x2 (64.5, 65.5)
|
||||
// gdb-check:$10 = simd::f64x2 ([64.5, 65.5])
|
||||
|
||||
// gdb-command:continue
|
||||
|
||||
@ -40,43 +40,43 @@
|
||||
#![feature(repr_simd)]
|
||||
|
||||
#[repr(simd)]
|
||||
struct i8x16(i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8);
|
||||
struct i8x16([i8; 16]);
|
||||
#[repr(simd)]
|
||||
struct i16x8(i16, i16, i16, i16, i16, i16, i16, i16);
|
||||
struct i16x8([i16; 8]);
|
||||
#[repr(simd)]
|
||||
struct i32x4(i32, i32, i32, i32);
|
||||
struct i32x4([i32; 4]);
|
||||
#[repr(simd)]
|
||||
struct i64x2(i64, i64);
|
||||
struct i64x2([i64; 2]);
|
||||
#[repr(simd)]
|
||||
struct u8x16(u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8, u8);
|
||||
struct u8x16([u8; 16]);
|
||||
#[repr(simd)]
|
||||
struct u16x8(u16, u16, u16, u16, u16, u16, u16, u16);
|
||||
struct u16x8([u16; 8]);
|
||||
#[repr(simd)]
|
||||
struct u32x4(u32, u32, u32, u32);
|
||||
struct u32x4([u32; 4]);
|
||||
#[repr(simd)]
|
||||
struct u64x2(u64, u64);
|
||||
struct u64x2([u64; 2]);
|
||||
#[repr(simd)]
|
||||
struct f32x4(f32, f32, f32, f32);
|
||||
struct f32x4([f32; 4]);
|
||||
#[repr(simd)]
|
||||
struct f64x2(f64, f64);
|
||||
struct f64x2([f64; 2]);
|
||||
|
||||
fn main() {
|
||||
|
||||
let vi8x16 = i8x16(0, 1, 2, 3, 4, 5, 6, 7,
|
||||
8, 9, 10, 11, 12, 13, 14, 15);
|
||||
let vi8x16 = i8x16([0, 1, 2, 3, 4, 5, 6, 7,
|
||||
8, 9, 10, 11, 12, 13, 14, 15]);
|
||||
|
||||
let vi16x8 = i16x8(16, 17, 18, 19, 20, 21, 22, 23);
|
||||
let vi32x4 = i32x4(24, 25, 26, 27);
|
||||
let vi64x2 = i64x2(28, 29);
|
||||
let vi16x8 = i16x8([16, 17, 18, 19, 20, 21, 22, 23]);
|
||||
let vi32x4 = i32x4([24, 25, 26, 27]);
|
||||
let vi64x2 = i64x2([28, 29]);
|
||||
|
||||
let vu8x16 = u8x16(30, 31, 32, 33, 34, 35, 36, 37,
|
||||
38, 39, 40, 41, 42, 43, 44, 45);
|
||||
let vu16x8 = u16x8(46, 47, 48, 49, 50, 51, 52, 53);
|
||||
let vu32x4 = u32x4(54, 55, 56, 57);
|
||||
let vu64x2 = u64x2(58, 59);
|
||||
let vu8x16 = u8x16([30, 31, 32, 33, 34, 35, 36, 37,
|
||||
38, 39, 40, 41, 42, 43, 44, 45]);
|
||||
let vu16x8 = u16x8([46, 47, 48, 49, 50, 51, 52, 53]);
|
||||
let vu32x4 = u32x4([54, 55, 56, 57]);
|
||||
let vu64x2 = u64x2([58, 59]);
|
||||
|
||||
let vf32x4 = f32x4(60.5f32, 61.5f32, 62.5f32, 63.5f32);
|
||||
let vf64x2 = f64x2(64.5f64, 65.5f64);
|
||||
let vf32x4 = f32x4([60.5f32, 61.5f32, 62.5f32, 63.5f32]);
|
||||
let vf64x2 = f64x2([64.5f64, 65.5f64]);
|
||||
|
||||
zzz(); // #break
|
||||
}
|
||||
|
@ -3,7 +3,7 @@
|
||||
//@ revisions:rpass1 rpass2
|
||||
|
||||
#[repr(simd)]
|
||||
struct I32x2(i32, i32);
|
||||
struct I32x2([i32; 2]);
|
||||
|
||||
extern "rust-intrinsic" {
|
||||
fn simd_shuffle<T, I, U>(x: T, y: T, idx: I) -> U;
|
||||
@ -12,7 +12,7 @@
|
||||
fn main() {
|
||||
unsafe {
|
||||
const IDX: [u32; 2] = [0, 0];
|
||||
let _: I32x2 = simd_shuffle(I32x2(1, 2), I32x2(3, 4), IDX);
|
||||
let _: I32x2 = simd_shuffle(I32x2(1, 2), I32x2(3, 4), IDX);
|
||||
let _: I32x2 = simd_shuffle(I32x2([1, 2]), I32x2([3, 4]), IDX);
|
||||
let _: I32x2 = simd_shuffle(I32x2([1, 2]), I32x2([3, 4]), IDX);
|
||||
}
|
||||
}
|
||||
|
@ -8,7 +8,7 @@
|
||||
|
||||
#[derive(Copy)]
|
||||
#[repr(simd)]
|
||||
pub struct f32x4(f32, f32, f32, f32);
|
||||
pub struct f32x4([f32; 4]);
|
||||
|
||||
extern "C" {
|
||||
#[link_name = "llvm.sqrt.v4f32"]
|
||||
@ -21,7 +21,7 @@ pub fn foo(x: f32x4) -> f32x4 {
|
||||
|
||||
#[derive(Copy)]
|
||||
#[repr(simd)]
|
||||
pub struct i32x4(i32, i32, i32, i32);
|
||||
pub struct i32x4([i32; 4]);
|
||||
|
||||
extern "C" {
|
||||
// _mm_sll_epi32
|
||||
@ -62,6 +62,8 @@ pub trait Copy {}
|
||||
|
||||
impl Copy for f32 {}
|
||||
impl Copy for i32 {}
|
||||
impl Copy for [f32; 4] {}
|
||||
impl Copy for [i32; 4] {}
|
||||
|
||||
pub mod marker {
|
||||
pub use Copy;
|
||||
|
@ -6,7 +6,7 @@ pub struct ReprC {
|
||||
}
|
||||
#[repr(simd, packed(2))]
|
||||
pub struct ReprSimd {
|
||||
field: u8,
|
||||
field: [u8; 1],
|
||||
}
|
||||
#[repr(transparent)]
|
||||
pub struct ReprTransparent {
|
||||
|
@ -25,16 +25,13 @@ impl Copy for i8 {}
|
||||
impl<T: ?Sized> Copy for *const T {}
|
||||
impl<T: ?Sized> Copy for *mut T {}
|
||||
|
||||
// I hate no_core tests!
|
||||
impl<T: Copy, const N: usize> Copy for [T; N] {}
|
||||
|
||||
// Regression test for https://github.com/rust-lang/rust/issues/118124.
|
||||
|
||||
#[repr(simd)]
|
||||
pub struct int8x16_t(
|
||||
pub(crate) i8, pub(crate) i8, pub(crate) i8, pub(crate) i8,
|
||||
pub(crate) i8, pub(crate) i8, pub(crate) i8, pub(crate) i8,
|
||||
pub(crate) i8, pub(crate) i8, pub(crate) i8, pub(crate) i8,
|
||||
pub(crate) i8, pub(crate) i8, pub(crate) i8, pub(crate) i8,
|
||||
);
|
||||
pub struct int8x16_t(pub(crate) [i8; 16]);
|
||||
impl Copy for int8x16_t {}
|
||||
|
||||
#[repr(C)]
|
||||
|
@ -65,13 +65,13 @@ fn is_sigill(status: ExitStatus) -> bool {
|
||||
#[allow(nonstandard_style)]
|
||||
mod test {
|
||||
#[derive(PartialEq, Debug, Clone, Copy)]
|
||||
struct f32x2(f32, f32);
|
||||
struct f32x2([f32; 2]);
|
||||
|
||||
#[derive(PartialEq, Debug, Clone, Copy)]
|
||||
struct f32x4(f32, f32, f32, f32);
|
||||
struct f32x4([f32; 4]);
|
||||
|
||||
#[derive(PartialEq, Debug, Clone, Copy)]
|
||||
struct f32x8(f32, f32, f32, f32, f32, f32, f32, f32);
|
||||
struct f32x8([f32; 8]);
|
||||
|
||||
pub fn main(level: &str) {
|
||||
unsafe {
|
||||
@ -97,9 +97,9 @@ unsafe fn $main:ident(level: &str) {
|
||||
)*) => ($(
|
||||
$(#[$attr])*
|
||||
unsafe fn $main(level: &str) {
|
||||
let m128 = f32x2(1., 2.);
|
||||
let m256 = f32x4(3., 4., 5., 6.);
|
||||
let m512 = f32x8(7., 8., 9., 10., 11., 12., 13., 14.);
|
||||
let m128 = f32x2([1., 2.]);
|
||||
let m256 = f32x4([3., 4., 5., 6.]);
|
||||
let m512 = f32x8([7., 8., 9., 10., 11., 12., 13., 14.]);
|
||||
assert_eq!(id_sse_128(m128), m128);
|
||||
assert_eq!(id_sse_256(m256), m256);
|
||||
assert_eq!(id_sse_512(m512), m512);
|
||||
@ -133,55 +133,55 @@ unsafe fn main_avx512(level: &str) { ... }
|
||||
|
||||
#[target_feature(enable = "sse2")]
|
||||
unsafe fn id_sse_128(a: f32x2) -> f32x2 {
|
||||
assert_eq!(a, f32x2(1., 2.));
|
||||
assert_eq!(a, f32x2([1., 2.]));
|
||||
a.clone()
|
||||
}
|
||||
|
||||
#[target_feature(enable = "sse2")]
|
||||
unsafe fn id_sse_256(a: f32x4) -> f32x4 {
|
||||
assert_eq!(a, f32x4(3., 4., 5., 6.));
|
||||
assert_eq!(a, f32x4([3., 4., 5., 6.]));
|
||||
a.clone()
|
||||
}
|
||||
|
||||
#[target_feature(enable = "sse2")]
|
||||
unsafe fn id_sse_512(a: f32x8) -> f32x8 {
|
||||
assert_eq!(a, f32x8(7., 8., 9., 10., 11., 12., 13., 14.));
|
||||
assert_eq!(a, f32x8([7., 8., 9., 10., 11., 12., 13., 14.]));
|
||||
a.clone()
|
||||
}
|
||||
|
||||
#[target_feature(enable = "avx")]
|
||||
unsafe fn id_avx_128(a: f32x2) -> f32x2 {
|
||||
assert_eq!(a, f32x2(1., 2.));
|
||||
assert_eq!(a, f32x2([1., 2.]));
|
||||
a.clone()
|
||||
}
|
||||
|
||||
#[target_feature(enable = "avx")]
|
||||
unsafe fn id_avx_256(a: f32x4) -> f32x4 {
|
||||
assert_eq!(a, f32x4(3., 4., 5., 6.));
|
||||
assert_eq!(a, f32x4([3., 4., 5., 6.]));
|
||||
a.clone()
|
||||
}
|
||||
|
||||
#[target_feature(enable = "avx")]
|
||||
unsafe fn id_avx_512(a: f32x8) -> f32x8 {
|
||||
assert_eq!(a, f32x8(7., 8., 9., 10., 11., 12., 13., 14.));
|
||||
assert_eq!(a, f32x8([7., 8., 9., 10., 11., 12., 13., 14.]));
|
||||
a.clone()
|
||||
}
|
||||
|
||||
#[target_feature(enable = "avx512bw")]
|
||||
unsafe fn id_avx512_128(a: f32x2) -> f32x2 {
|
||||
assert_eq!(a, f32x2(1., 2.));
|
||||
assert_eq!(a, f32x2([1., 2.]));
|
||||
a.clone()
|
||||
}
|
||||
|
||||
#[target_feature(enable = "avx512bw")]
|
||||
unsafe fn id_avx512_256(a: f32x4) -> f32x4 {
|
||||
assert_eq!(a, f32x4(3., 4., 5., 6.));
|
||||
assert_eq!(a, f32x4([3., 4., 5., 6.]));
|
||||
a.clone()
|
||||
}
|
||||
|
||||
#[target_feature(enable = "avx512bw")]
|
||||
unsafe fn id_avx512_512(a: f32x8) -> f32x8 {
|
||||
assert_eq!(a, f32x8(7., 8., 9., 10., 11., 12., 13., 14.));
|
||||
assert_eq!(a, f32x8([7., 8., 9., 10., 11., 12., 13., 14.]));
|
||||
a.clone()
|
||||
}
|
||||
}
|
||||
|
@ -6,10 +6,10 @@
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Clone, Copy)]
|
||||
struct SimdType(f32, f32, f32, f32);
|
||||
struct SimdType([f32; 4]);
|
||||
|
||||
#[repr(simd)]
|
||||
struct SimdNonCopy(f32, f32, f32, f32);
|
||||
struct SimdNonCopy([f32; 4]);
|
||||
|
||||
fn main() {
|
||||
unsafe {
|
||||
|
@ -6,10 +6,10 @@
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Clone, Copy)]
|
||||
struct SimdType(f32, f32, f32, f32);
|
||||
struct SimdType([f32; 4]);
|
||||
|
||||
#[repr(simd)]
|
||||
struct SimdNonCopy(f32, f32, f32, f32);
|
||||
struct SimdNonCopy([f32; 4]);
|
||||
|
||||
fn main() {
|
||||
unsafe {
|
||||
@ -17,7 +17,7 @@ fn main() {
|
||||
|
||||
// Register operands must be Copy
|
||||
|
||||
asm!("{:v}", in(vreg) SimdNonCopy(0.0, 0.0, 0.0, 0.0));
|
||||
asm!("{:v}", in(vreg) SimdNonCopy([0.0, 0.0, 0.0, 0.0]));
|
||||
//~^ ERROR arguments for inline assembly must be copyable
|
||||
|
||||
// Register operands must be integers, floats, SIMD vectors, pointers or
|
||||
@ -25,7 +25,7 @@ fn main() {
|
||||
|
||||
asm!("{}", in(reg) 0i64);
|
||||
asm!("{}", in(reg) 0f64);
|
||||
asm!("{:v}", in(vreg) SimdType(0.0, 0.0, 0.0, 0.0));
|
||||
asm!("{:v}", in(vreg) SimdType([0.0, 0.0, 0.0, 0.0]));
|
||||
asm!("{}", in(reg) 0 as *const u8);
|
||||
asm!("{}", in(reg) 0 as *mut u8);
|
||||
asm!("{}", in(reg) main as fn());
|
||||
|
@ -1,8 +1,8 @@
|
||||
error: arguments for inline assembly must be copyable
|
||||
--> $DIR/type-check-2.rs:20:31
|
||||
|
|
||||
LL | asm!("{:v}", in(vreg) SimdNonCopy(0.0, 0.0, 0.0, 0.0));
|
||||
| ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
|
||||
LL | asm!("{:v}", in(vreg) SimdNonCopy([0.0, 0.0, 0.0, 0.0]));
|
||||
| ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
|
||||
|
|
||||
= note: `SimdNonCopy` does not implement the Copy trait
|
||||
|
||||
|
@ -8,11 +8,11 @@
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone)]
|
||||
struct Simd256bit(f64, f64, f64, f64);
|
||||
struct Simd256bit([f64; 4]);
|
||||
|
||||
fn main() {
|
||||
let f64x2: float64x2_t = unsafe { std::mem::transmute(0i128) };
|
||||
let f64x4 = Simd256bit(0.0, 0.0, 0.0, 0.0);
|
||||
let f64x4 = Simd256bit([0.0, 0.0, 0.0, 0.0]);
|
||||
|
||||
unsafe {
|
||||
// Types must be listed in the register class.
|
||||
|
@ -8,7 +8,7 @@
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone)]
|
||||
struct Simd256bit(f64, f64, f64, f64);
|
||||
struct Simd256bit([f64; 4]);
|
||||
|
||||
fn main() {}
|
||||
|
||||
|
@ -5,7 +5,7 @@
|
||||
use std::arch::{asm, global_asm};
|
||||
|
||||
#[repr(simd)]
|
||||
struct SimdNonCopy(f32, f32, f32, f32);
|
||||
struct SimdNonCopy([f32; 4]);
|
||||
|
||||
fn main() {
|
||||
unsafe {
|
||||
@ -29,7 +29,7 @@ fn main() {
|
||||
|
||||
// Register operands must be Copy
|
||||
|
||||
asm!("{}", in(xmm_reg) SimdNonCopy(0.0, 0.0, 0.0, 0.0));
|
||||
asm!("{}", in(xmm_reg) SimdNonCopy([0.0, 0.0, 0.0, 0.0]));
|
||||
//~^ ERROR arguments for inline assembly must be copyable
|
||||
|
||||
// Register operands must be integers, floats, SIMD vectors, pointers or
|
||||
|
@ -1,8 +1,8 @@
|
||||
error: arguments for inline assembly must be copyable
|
||||
--> $DIR/type-check-2.rs:32:32
|
||||
|
|
||||
LL | asm!("{}", in(xmm_reg) SimdNonCopy(0.0, 0.0, 0.0, 0.0));
|
||||
| ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
|
||||
LL | asm!("{}", in(xmm_reg) SimdNonCopy([0.0, 0.0, 0.0, 0.0]));
|
||||
| ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
|
||||
|
|
||||
= note: `SimdNonCopy` does not implement the Copy trait
|
||||
|
||||
|
@ -1,12 +1,9 @@
|
||||
//@ only-x86_64
|
||||
|
||||
#![feature(repr_simd, never_type)]
|
||||
#![feature(never_type)]
|
||||
|
||||
use std::arch::asm;
|
||||
|
||||
#[repr(simd)]
|
||||
struct SimdNonCopy(f32, f32, f32, f32);
|
||||
|
||||
fn main() {
|
||||
unsafe {
|
||||
// Inputs must be initialized
|
||||
|
@ -1,5 +1,5 @@
|
||||
error[E0381]: used binding `x` isn't initialized
|
||||
--> $DIR/type-check-5.rs:15:28
|
||||
--> $DIR/type-check-5.rs:12:28
|
||||
|
|
||||
LL | let x: u64;
|
||||
| - binding declared here but left uninitialized
|
||||
@ -12,7 +12,7 @@ LL | let x: u64 = 42;
|
||||
| ++++
|
||||
|
||||
error[E0381]: used binding `y` isn't initialized
|
||||
--> $DIR/type-check-5.rs:18:9
|
||||
--> $DIR/type-check-5.rs:15:9
|
||||
|
|
||||
LL | let mut y: u64;
|
||||
| ----- binding declared here but left uninitialized
|
||||
@ -25,7 +25,7 @@ LL | let mut y: u64 = 42;
|
||||
| ++++
|
||||
|
||||
error[E0596]: cannot borrow `v` as mutable, as it is not declared as mutable
|
||||
--> $DIR/type-check-5.rs:24:13
|
||||
--> $DIR/type-check-5.rs:21:13
|
||||
|
|
||||
LL | let v: Vec<u64> = vec![0, 1, 2];
|
||||
| ^ not mutable
|
||||
|
@ -5,10 +5,9 @@
|
||||
#![stable(feature = "foo", since = "1.3.37")]
|
||||
#![allow(non_camel_case_types)]
|
||||
|
||||
#[repr(simd)] struct i8x1(i8);
|
||||
#[repr(simd)] struct u16x2(u16, u16);
|
||||
// Make some of them array types to ensure those also work.
|
||||
#[repr(simd)] struct i8x1_arr([i8; 1]);
|
||||
// repr(simd) now only supports array types
|
||||
#[repr(simd)] struct i8x1([i8; 1]);
|
||||
#[repr(simd)] struct u16x2([u16; 2]);
|
||||
#[repr(simd)] struct f32x4([f32; 4]);
|
||||
|
||||
extern "rust-intrinsic" {
|
||||
@ -20,26 +19,18 @@
|
||||
|
||||
fn main() {
|
||||
{
|
||||
const U: i8x1 = i8x1(13);
|
||||
const U: i8x1 = i8x1([13]);
|
||||
const V: i8x1 = unsafe { simd_insert(U, 0_u32, 42_i8) };
|
||||
const X0: i8 = V.0;
|
||||
const Y0: i8 = unsafe { simd_extract(V, 0) };
|
||||
assert_eq!(X0, 42);
|
||||
assert_eq!(Y0, 42);
|
||||
}
|
||||
{
|
||||
const U: i8x1_arr = i8x1_arr([13]);
|
||||
const V: i8x1_arr = unsafe { simd_insert(U, 0_u32, 42_i8) };
|
||||
const X0: i8 = V.0[0];
|
||||
const Y0: i8 = unsafe { simd_extract(V, 0) };
|
||||
assert_eq!(X0, 42);
|
||||
assert_eq!(Y0, 42);
|
||||
}
|
||||
{
|
||||
const U: u16x2 = u16x2(13, 14);
|
||||
const U: u16x2 = u16x2([13, 14]);
|
||||
const V: u16x2 = unsafe { simd_insert(U, 1_u32, 42_u16) };
|
||||
const X0: u16 = V.0;
|
||||
const X1: u16 = V.1;
|
||||
const X0: u16 = V.0[0];
|
||||
const X1: u16 = V.0[1];
|
||||
const Y0: u16 = unsafe { simd_extract(V, 0) };
|
||||
const Y1: u16 = unsafe { simd_extract(V, 1) };
|
||||
assert_eq!(X0, 13);
|
||||
|
@ -3,5 +3,8 @@
|
||||
#[repr(simd)]
|
||||
struct Bad; //~ ERROR E0075
|
||||
|
||||
#[repr(simd)]
|
||||
struct AlsoBad([i32; 1], [i32; 1]); //~ ERROR E0075
|
||||
|
||||
fn main() {
|
||||
}
|
||||
|
@ -4,6 +4,12 @@ error[E0075]: SIMD vector cannot be empty
|
||||
LL | struct Bad;
|
||||
| ^^^^^^^^^^
|
||||
|
||||
error: aborting due to 1 previous error
|
||||
error[E0075]: SIMD vector cannot have multiple fields
|
||||
--> $DIR/E0075.rs:7:1
|
||||
|
|
||||
LL | struct AlsoBad([i32; 1], [i32; 1]);
|
||||
| ^^^^^^^^^^^^^^ -------- excess field
|
||||
|
||||
error: aborting due to 2 previous errors
|
||||
|
||||
For more information about this error, try `rustc --explain E0075`.
|
||||
|
@ -1,7 +1,7 @@
|
||||
#![feature(repr_simd)]
|
||||
|
||||
#[repr(simd)]
|
||||
struct Bad(u16, u32, u32);
|
||||
struct Bad(u32);
|
||||
//~^ ERROR E0076
|
||||
|
||||
fn main() {
|
||||
|
@ -1,8 +1,8 @@
|
||||
error[E0076]: SIMD vector should be homogeneous
|
||||
error[E0076]: SIMD vector's only field must be an array
|
||||
--> $DIR/E0076.rs:4:1
|
||||
|
|
||||
LL | struct Bad(u16, u32, u32);
|
||||
| ^^^^^^^^^^ SIMD elements must have the same type
|
||||
LL | struct Bad(u32);
|
||||
| ^^^^^^^^^^ --- not an array
|
||||
|
||||
error: aborting due to 1 previous error
|
||||
|
||||
|
@ -1,7 +1,7 @@
|
||||
#![feature(repr_simd)]
|
||||
|
||||
#[repr(simd)]
|
||||
struct Bad(String); //~ ERROR E0077
|
||||
struct Bad([String; 2]); //~ ERROR E0077
|
||||
|
||||
fn main() {
|
||||
}
|
||||
|
@ -1,7 +1,7 @@
|
||||
error[E0077]: SIMD vector element type should be a primitive scalar (integer/float/pointer) type
|
||||
--> $DIR/E0077.rs:4:1
|
||||
|
|
||||
LL | struct Bad(String);
|
||||
LL | struct Bad([String; 2]);
|
||||
| ^^^^^^^^^^
|
||||
|
||||
error: aborting due to 1 previous error
|
||||
|
@ -1,9 +1,9 @@
|
||||
#[repr(simd)] //~ error: SIMD types are experimental
|
||||
struct Foo(u64, u64);
|
||||
struct Foo([u64; 2]);
|
||||
|
||||
#[repr(C)] //~ ERROR conflicting representation hints
|
||||
//~^ WARN this was previously accepted
|
||||
#[repr(simd)] //~ error: SIMD types are experimental
|
||||
struct Bar(u64, u64);
|
||||
struct Bar([u64; 2]);
|
||||
|
||||
fn main() {}
|
||||
|
@ -3,7 +3,7 @@
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone)]
|
||||
struct LocalSimd(u8, u8);
|
||||
struct LocalSimd([u8; 2]);
|
||||
|
||||
extern "C" {
|
||||
fn baz() -> LocalSimd; //~ ERROR use of SIMD type
|
||||
|
@ -2,10 +2,7 @@
|
||||
|
||||
#[repr(simd)] //~ ERROR SIMD types are experimental
|
||||
struct RGBA {
|
||||
r: f32,
|
||||
g: f32,
|
||||
b: f32,
|
||||
a: f32
|
||||
rgba: [f32; 4],
|
||||
}
|
||||
|
||||
pub fn main() {}
|
||||
|
@ -49,7 +49,7 @@ union P2 { x: (u32, u32) } //~ ERROR: layout_of
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone)]
|
||||
struct F32x4(f32, f32, f32, f32);
|
||||
struct F32x4([f32; 4]);
|
||||
|
||||
#[rustc_layout(debug)]
|
||||
#[repr(packed(1))]
|
||||
|
@ -10,7 +10,7 @@ fn f() {}
|
||||
struct SPacked(f64, f64);
|
||||
|
||||
#[repr(simd)]
|
||||
struct SSimd(f64, f64);
|
||||
struct SSimd([f64; 2]);
|
||||
|
||||
#[repr(i8)] //~ ERROR: attribute should be applied to an enum
|
||||
struct SInt(f64, f64);
|
||||
|
@ -18,6 +18,6 @@ enum U {
|
||||
#[repr(Rust, simd)]
|
||||
//~^ ERROR conflicting representation hints
|
||||
//~| ERROR SIMD types are experimental and possibly buggy
|
||||
struct F32x4(f32, f32, f32, f32);
|
||||
struct F32x4([f32; 4]);
|
||||
|
||||
fn main() {}
|
||||
|
@ -10,7 +10,7 @@
|
||||
|
||||
#[repr(simd)]
|
||||
#[allow(non_camel_case_types)]
|
||||
struct int8x4_t(u8,u8,u8,u8);
|
||||
struct int8x4_t([u8; 4]);
|
||||
|
||||
fn get_elem<const LANE: u32>(a: int8x4_t) -> u8 {
|
||||
const { assert!(LANE < 4); } // the error should be here...
|
||||
@ -20,5 +20,5 @@ fn get_elem<const LANE: u32>(a: int8x4_t) -> u8 {
|
||||
}
|
||||
|
||||
fn main() {
|
||||
get_elem::<4>(int8x4_t(0,0,0,0));
|
||||
get_elem::<4>(int8x4_t([0, 0, 0, 0]));
|
||||
}
|
||||
|
@ -15,8 +15,8 @@ LL | const { assert!(LANE < 4); } // the error should be here...
|
||||
note: the above error was encountered while instantiating `fn get_elem::<4>`
|
||||
--> $DIR/const-err-trumps-simd-err.rs:23:5
|
||||
|
|
||||
LL | get_elem::<4>(int8x4_t(0,0,0,0));
|
||||
| ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
|
||||
LL | get_elem::<4>(int8x4_t([0, 0, 0, 0]));
|
||||
| ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
|
||||
|
||||
error: aborting due to 1 previous error
|
||||
|
||||
|
@ -6,7 +6,7 @@
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone)]
|
||||
struct f32x4(f32, f32, f32, f32);
|
||||
struct f32x4([f32; 4]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone)]
|
||||
@ -67,8 +67,8 @@ pub fn main() {
|
||||
let y = [2.0f32, 4.0f32, 6.0f32, 8.0f32];
|
||||
|
||||
// lame-o
|
||||
let a = f32x4(1.0f32, 2.0f32, 3.0f32, 4.0f32);
|
||||
let f32x4(a0, a1, a2, a3) = add(a, a);
|
||||
let a = f32x4([1.0f32, 2.0f32, 3.0f32, 4.0f32]);
|
||||
let f32x4([a0, a1, a2, a3]) = add(a, a);
|
||||
assert_eq!(a0, 2.0f32);
|
||||
assert_eq!(a1, 4.0f32);
|
||||
assert_eq!(a2, 6.0f32);
|
||||
|
@ -13,7 +13,7 @@
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
struct f32x4(pub f32, pub f32, pub f32, pub f32);
|
||||
struct f32x4(pub [f32; 4]);
|
||||
|
||||
extern "rust-intrinsic" {
|
||||
fn simd_fsqrt<T>(x: T) -> T;
|
||||
@ -47,19 +47,19 @@ macro_rules! assert_approx_eq {
|
||||
($a:expr, $b:expr) => ({
|
||||
let a = $a;
|
||||
let b = $b;
|
||||
assert_approx_eq_f32!(a.0, b.0);
|
||||
assert_approx_eq_f32!(a.1, b.1);
|
||||
assert_approx_eq_f32!(a.2, b.2);
|
||||
assert_approx_eq_f32!(a.3, b.3);
|
||||
assert_approx_eq_f32!(a.0[0], b.0[0]);
|
||||
assert_approx_eq_f32!(a.0[1], b.0[1]);
|
||||
assert_approx_eq_f32!(a.0[2], b.0[2]);
|
||||
assert_approx_eq_f32!(a.0[3], b.0[3]);
|
||||
})
|
||||
}
|
||||
|
||||
fn main() {
|
||||
let x = f32x4(1.0, 1.0, 1.0, 1.0);
|
||||
let y = f32x4(-1.0, -1.0, -1.0, -1.0);
|
||||
let z = f32x4(0.0, 0.0, 0.0, 0.0);
|
||||
let x = f32x4([1.0, 1.0, 1.0, 1.0]);
|
||||
let y = f32x4([-1.0, -1.0, -1.0, -1.0]);
|
||||
let z = f32x4([0.0, 0.0, 0.0, 0.0]);
|
||||
|
||||
let h = f32x4(0.5, 0.5, 0.5, 0.5);
|
||||
let h = f32x4([0.5, 0.5, 0.5, 0.5]);
|
||||
|
||||
unsafe {
|
||||
let r = simd_fabs(y);
|
||||
|
@ -8,13 +8,13 @@
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
struct f32x4(pub f32, pub f32, pub f32, pub f32);
|
||||
struct f32x4(pub [f32; 4]);
|
||||
|
||||
use std::intrinsics::simd::*;
|
||||
|
||||
fn main() {
|
||||
let x = f32x4(1.0, 2.0, 3.0, 4.0);
|
||||
let y = f32x4(2.0, 1.0, 4.0, 3.0);
|
||||
let x = f32x4([1.0, 2.0, 3.0, 4.0]);
|
||||
let y = f32x4([2.0, 1.0, 4.0, 3.0]);
|
||||
|
||||
#[cfg(not(any(target_arch = "mips", target_arch = "mips64")))]
|
||||
let nan = f32::NAN;
|
||||
@ -23,13 +23,13 @@ fn main() {
|
||||
#[cfg(any(target_arch = "mips", target_arch = "mips64"))]
|
||||
let nan = f32::from_bits(f32::NAN.to_bits() - 1);
|
||||
|
||||
let n = f32x4(nan, nan, nan, nan);
|
||||
let n = f32x4([nan, nan, nan, nan]);
|
||||
|
||||
unsafe {
|
||||
let min0 = simd_fmin(x, y);
|
||||
let min1 = simd_fmin(y, x);
|
||||
assert_eq!(min0, min1);
|
||||
let e = f32x4(1.0, 1.0, 3.0, 3.0);
|
||||
let e = f32x4([1.0, 1.0, 3.0, 3.0]);
|
||||
assert_eq!(min0, e);
|
||||
let minn = simd_fmin(x, n);
|
||||
assert_eq!(minn, x);
|
||||
@ -39,7 +39,7 @@ fn main() {
|
||||
let max0 = simd_fmax(x, y);
|
||||
let max1 = simd_fmax(y, x);
|
||||
assert_eq!(max0, max1);
|
||||
let e = f32x4(2.0, 2.0, 4.0, 4.0);
|
||||
let e = f32x4([2.0, 2.0, 4.0, 4.0]);
|
||||
assert_eq!(max0, e);
|
||||
let maxn = simd_fmax(x, n);
|
||||
assert_eq!(maxn, x);
|
||||
|
@ -4,15 +4,15 @@
|
||||
#![allow(non_camel_case_types)]
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone)]
|
||||
pub struct i32x4(pub i32, pub i32, pub i32, pub i32);
|
||||
pub struct i32x4(pub [i32; 4]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone)]
|
||||
pub struct u32x4(pub u32, pub u32, pub u32, pub u32);
|
||||
pub struct u32x4(pub [u32; 4]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone)]
|
||||
pub struct f32x4(pub f32, pub f32, pub f32, pub f32);
|
||||
pub struct f32x4(pub [f32; 4]);
|
||||
|
||||
extern "rust-intrinsic" {
|
||||
fn simd_add<T>(x: T, y: T) -> T;
|
||||
@ -35,9 +35,9 @@
|
||||
}
|
||||
|
||||
fn main() {
|
||||
let x = i32x4(0, 0, 0, 0);
|
||||
let y = u32x4(0, 0, 0, 0);
|
||||
let z = f32x4(0.0, 0.0, 0.0, 0.0);
|
||||
let x = i32x4([0, 0, 0, 0]);
|
||||
let y = u32x4([0, 0, 0, 0]);
|
||||
let z = f32x4([0.0, 0.0, 0.0, 0.0]);
|
||||
|
||||
unsafe {
|
||||
simd_add(x, x);
|
||||
|
@ -5,7 +5,7 @@
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone)]
|
||||
struct i32x4(pub i32, pub i32, pub i32, pub i32);
|
||||
struct i32x4(pub [i32; 4]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone)]
|
||||
@ -13,17 +13,9 @@
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone)]
|
||||
struct f32x4(pub f32, pub f32, pub f32, pub f32);
|
||||
struct f32x4(pub [f32; 4]);
|
||||
|
||||
macro_rules! all_eq {
|
||||
($a: expr, $b: expr) => {{
|
||||
let a = $a;
|
||||
let b = $b;
|
||||
assert!(a.0 == b.0 && a.1 == b.1 && a.2 == b.2 && a.3 == b.3);
|
||||
}};
|
||||
}
|
||||
|
||||
macro_rules! all_eq_ {
|
||||
($a: expr, $b: expr) => {{
|
||||
let a = $a;
|
||||
let b = $b;
|
||||
@ -52,112 +44,112 @@ macro_rules! all_eq_ {
|
||||
}
|
||||
|
||||
fn main() {
|
||||
let x1 = i32x4(1, 2, 3, 4);
|
||||
let x1 = i32x4([1, 2, 3, 4]);
|
||||
let y1 = U32::<4>([1, 2, 3, 4]);
|
||||
let z1 = f32x4(1.0, 2.0, 3.0, 4.0);
|
||||
let x2 = i32x4(2, 3, 4, 5);
|
||||
let z1 = f32x4([1.0, 2.0, 3.0, 4.0]);
|
||||
let x2 = i32x4([2, 3, 4, 5]);
|
||||
let y2 = U32::<4>([2, 3, 4, 5]);
|
||||
let z2 = f32x4(2.0, 3.0, 4.0, 5.0);
|
||||
let x3 = i32x4(0, i32::MAX, i32::MIN, -1_i32);
|
||||
let z2 = f32x4([2.0, 3.0, 4.0, 5.0]);
|
||||
let x3 = i32x4([0, i32::MAX, i32::MIN, -1_i32]);
|
||||
let y3 = U32::<4>([0, i32::MAX as _, i32::MIN as _, -1_i32 as _]);
|
||||
|
||||
unsafe {
|
||||
all_eq!(simd_add(x1, x2), i32x4(3, 5, 7, 9));
|
||||
all_eq!(simd_add(x2, x1), i32x4(3, 5, 7, 9));
|
||||
all_eq_!(simd_add(y1, y2), U32::<4>([3, 5, 7, 9]));
|
||||
all_eq_!(simd_add(y2, y1), U32::<4>([3, 5, 7, 9]));
|
||||
all_eq!(simd_add(z1, z2), f32x4(3.0, 5.0, 7.0, 9.0));
|
||||
all_eq!(simd_add(z2, z1), f32x4(3.0, 5.0, 7.0, 9.0));
|
||||
all_eq!(simd_add(x1, x2), i32x4([3, 5, 7, 9]));
|
||||
all_eq!(simd_add(x2, x1), i32x4([3, 5, 7, 9]));
|
||||
all_eq!(simd_add(y1, y2), U32::<4>([3, 5, 7, 9]));
|
||||
all_eq!(simd_add(y2, y1), U32::<4>([3, 5, 7, 9]));
|
||||
all_eq!(simd_add(z1, z2), f32x4([3.0, 5.0, 7.0, 9.0]));
|
||||
all_eq!(simd_add(z2, z1), f32x4([3.0, 5.0, 7.0, 9.0]));
|
||||
|
||||
all_eq!(simd_mul(x1, x2), i32x4(2, 6, 12, 20));
|
||||
all_eq!(simd_mul(x2, x1), i32x4(2, 6, 12, 20));
|
||||
all_eq_!(simd_mul(y1, y2), U32::<4>([2, 6, 12, 20]));
|
||||
all_eq_!(simd_mul(y2, y1), U32::<4>([2, 6, 12, 20]));
|
||||
all_eq!(simd_mul(z1, z2), f32x4(2.0, 6.0, 12.0, 20.0));
|
||||
all_eq!(simd_mul(z2, z1), f32x4(2.0, 6.0, 12.0, 20.0));
|
||||
all_eq!(simd_mul(x1, x2), i32x4([2, 6, 12, 20]));
|
||||
all_eq!(simd_mul(x2, x1), i32x4([2, 6, 12, 20]));
|
||||
all_eq!(simd_mul(y1, y2), U32::<4>([2, 6, 12, 20]));
|
||||
all_eq!(simd_mul(y2, y1), U32::<4>([2, 6, 12, 20]));
|
||||
all_eq!(simd_mul(z1, z2), f32x4([2.0, 6.0, 12.0, 20.0]));
|
||||
all_eq!(simd_mul(z2, z1), f32x4([2.0, 6.0, 12.0, 20.0]));
|
||||
|
||||
all_eq!(simd_sub(x2, x1), i32x4(1, 1, 1, 1));
|
||||
all_eq!(simd_sub(x1, x2), i32x4(-1, -1, -1, -1));
|
||||
all_eq_!(simd_sub(y2, y1), U32::<4>([1, 1, 1, 1]));
|
||||
all_eq_!(simd_sub(y1, y2), U32::<4>([!0, !0, !0, !0]));
|
||||
all_eq!(simd_sub(z2, z1), f32x4(1.0, 1.0, 1.0, 1.0));
|
||||
all_eq!(simd_sub(z1, z2), f32x4(-1.0, -1.0, -1.0, -1.0));
|
||||
all_eq!(simd_sub(x2, x1), i32x4([1, 1, 1, 1]));
|
||||
all_eq!(simd_sub(x1, x2), i32x4([-1, -1, -1, -1]));
|
||||
all_eq!(simd_sub(y2, y1), U32::<4>([1, 1, 1, 1]));
|
||||
all_eq!(simd_sub(y1, y2), U32::<4>([!0, !0, !0, !0]));
|
||||
all_eq!(simd_sub(z2, z1), f32x4([1.0, 1.0, 1.0, 1.0]));
|
||||
all_eq!(simd_sub(z1, z2), f32x4([-1.0, -1.0, -1.0, -1.0]));
|
||||
|
||||
all_eq!(simd_div(x1, x1), i32x4(1, 1, 1, 1));
|
||||
all_eq!(simd_div(i32x4(2, 4, 6, 8), i32x4(2, 2, 2, 2)), x1);
|
||||
all_eq_!(simd_div(y1, y1), U32::<4>([1, 1, 1, 1]));
|
||||
all_eq_!(simd_div(U32::<4>([2, 4, 6, 8]), U32::<4>([2, 2, 2, 2])), y1);
|
||||
all_eq!(simd_div(z1, z1), f32x4(1.0, 1.0, 1.0, 1.0));
|
||||
all_eq!(simd_div(z1, z2), f32x4(1.0 / 2.0, 2.0 / 3.0, 3.0 / 4.0, 4.0 / 5.0));
|
||||
all_eq!(simd_div(z2, z1), f32x4(2.0 / 1.0, 3.0 / 2.0, 4.0 / 3.0, 5.0 / 4.0));
|
||||
all_eq!(simd_div(x1, x1), i32x4([1, 1, 1, 1]));
|
||||
all_eq!(simd_div(i32x4([2, 4, 6, 8]), i32x4([2, 2, 2, 2])), x1);
|
||||
all_eq!(simd_div(y1, y1), U32::<4>([1, 1, 1, 1]));
|
||||
all_eq!(simd_div(U32::<4>([2, 4, 6, 8]), U32::<4>([2, 2, 2, 2])), y1);
|
||||
all_eq!(simd_div(z1, z1), f32x4([1.0, 1.0, 1.0, 1.0]));
|
||||
all_eq!(simd_div(z1, z2), f32x4([1.0 / 2.0, 2.0 / 3.0, 3.0 / 4.0, 4.0 / 5.0]));
|
||||
all_eq!(simd_div(z2, z1), f32x4([2.0 / 1.0, 3.0 / 2.0, 4.0 / 3.0, 5.0 / 4.0]));
|
||||
|
||||
all_eq!(simd_rem(x1, x1), i32x4(0, 0, 0, 0));
|
||||
all_eq!(simd_rem(x2, x1), i32x4(0, 1, 1, 1));
|
||||
all_eq_!(simd_rem(y1, y1), U32::<4>([0, 0, 0, 0]));
|
||||
all_eq_!(simd_rem(y2, y1), U32::<4>([0, 1, 1, 1]));
|
||||
all_eq!(simd_rem(z1, z1), f32x4(0.0, 0.0, 0.0, 0.0));
|
||||
all_eq!(simd_rem(x1, x1), i32x4([0, 0, 0, 0]));
|
||||
all_eq!(simd_rem(x2, x1), i32x4([0, 1, 1, 1]));
|
||||
all_eq!(simd_rem(y1, y1), U32::<4>([0, 0, 0, 0]));
|
||||
all_eq!(simd_rem(y2, y1), U32::<4>([0, 1, 1, 1]));
|
||||
all_eq!(simd_rem(z1, z1), f32x4([0.0, 0.0, 0.0, 0.0]));
|
||||
all_eq!(simd_rem(z1, z2), z1);
|
||||
all_eq!(simd_rem(z2, z1), f32x4(0.0, 1.0, 1.0, 1.0));
|
||||
all_eq!(simd_rem(z2, z1), f32x4([0.0, 1.0, 1.0, 1.0]));
|
||||
|
||||
all_eq!(simd_shl(x1, x2), i32x4(1 << 2, 2 << 3, 3 << 4, 4 << 5));
|
||||
all_eq!(simd_shl(x2, x1), i32x4(2 << 1, 3 << 2, 4 << 3, 5 << 4));
|
||||
all_eq_!(simd_shl(y1, y2), U32::<4>([1 << 2, 2 << 3, 3 << 4, 4 << 5]));
|
||||
all_eq_!(simd_shl(y2, y1), U32::<4>([2 << 1, 3 << 2, 4 << 3, 5 << 4]));
|
||||
all_eq!(simd_shl(x1, x2), i32x4([1 << 2, 2 << 3, 3 << 4, 4 << 5]));
|
||||
all_eq!(simd_shl(x2, x1), i32x4([2 << 1, 3 << 2, 4 << 3, 5 << 4]));
|
||||
all_eq!(simd_shl(y1, y2), U32::<4>([1 << 2, 2 << 3, 3 << 4, 4 << 5]));
|
||||
all_eq!(simd_shl(y2, y1), U32::<4>([2 << 1, 3 << 2, 4 << 3, 5 << 4]));
|
||||
|
||||
// test right-shift by assuming left-shift is correct
|
||||
all_eq!(simd_shr(simd_shl(x1, x2), x2), x1);
|
||||
all_eq!(simd_shr(simd_shl(x2, x1), x1), x2);
|
||||
all_eq_!(simd_shr(simd_shl(y1, y2), y2), y1);
|
||||
all_eq_!(simd_shr(simd_shl(y2, y1), y1), y2);
|
||||
all_eq!(simd_shr(simd_shl(y1, y2), y2), y1);
|
||||
all_eq!(simd_shr(simd_shl(y2, y1), y1), y2);
|
||||
|
||||
// ensure we get logical vs. arithmetic shifts correct
|
||||
let (a, b, c, d) = (-12, -123, -1234, -12345);
|
||||
all_eq!(simd_shr(i32x4(a, b, c, d), x1), i32x4(a >> 1, b >> 2, c >> 3, d >> 4));
|
||||
all_eq_!(
|
||||
all_eq!(simd_shr(i32x4([a, b, c, d]), x1), i32x4([a >> 1, b >> 2, c >> 3, d >> 4]));
|
||||
all_eq!(
|
||||
simd_shr(U32::<4>([a as u32, b as u32, c as u32, d as u32]), y1),
|
||||
U32::<4>([(a as u32) >> 1, (b as u32) >> 2, (c as u32) >> 3, (d as u32) >> 4])
|
||||
);
|
||||
|
||||
all_eq!(simd_and(x1, x2), i32x4(0, 2, 0, 4));
|
||||
all_eq!(simd_and(x2, x1), i32x4(0, 2, 0, 4));
|
||||
all_eq_!(simd_and(y1, y2), U32::<4>([0, 2, 0, 4]));
|
||||
all_eq_!(simd_and(y2, y1), U32::<4>([0, 2, 0, 4]));
|
||||
all_eq!(simd_and(x1, x2), i32x4([0, 2, 0, 4]));
|
||||
all_eq!(simd_and(x2, x1), i32x4([0, 2, 0, 4]));
|
||||
all_eq!(simd_and(y1, y2), U32::<4>([0, 2, 0, 4]));
|
||||
all_eq!(simd_and(y2, y1), U32::<4>([0, 2, 0, 4]));
|
||||
|
||||
all_eq!(simd_or(x1, x2), i32x4(3, 3, 7, 5));
|
||||
all_eq!(simd_or(x2, x1), i32x4(3, 3, 7, 5));
|
||||
all_eq_!(simd_or(y1, y2), U32::<4>([3, 3, 7, 5]));
|
||||
all_eq_!(simd_or(y2, y1), U32::<4>([3, 3, 7, 5]));
|
||||
all_eq!(simd_or(x1, x2), i32x4([3, 3, 7, 5]));
|
||||
all_eq!(simd_or(x2, x1), i32x4([3, 3, 7, 5]));
|
||||
all_eq!(simd_or(y1, y2), U32::<4>([3, 3, 7, 5]));
|
||||
all_eq!(simd_or(y2, y1), U32::<4>([3, 3, 7, 5]));
|
||||
|
||||
all_eq!(simd_xor(x1, x2), i32x4(3, 1, 7, 1));
|
||||
all_eq!(simd_xor(x2, x1), i32x4(3, 1, 7, 1));
|
||||
all_eq_!(simd_xor(y1, y2), U32::<4>([3, 1, 7, 1]));
|
||||
all_eq_!(simd_xor(y2, y1), U32::<4>([3, 1, 7, 1]));
|
||||
all_eq!(simd_xor(x1, x2), i32x4([3, 1, 7, 1]));
|
||||
all_eq!(simd_xor(x2, x1), i32x4([3, 1, 7, 1]));
|
||||
all_eq!(simd_xor(y1, y2), U32::<4>([3, 1, 7, 1]));
|
||||
all_eq!(simd_xor(y2, y1), U32::<4>([3, 1, 7, 1]));
|
||||
|
||||
all_eq!(simd_neg(x1), i32x4(-1, -2, -3, -4));
|
||||
all_eq!(simd_neg(x2), i32x4(-2, -3, -4, -5));
|
||||
all_eq!(simd_neg(z1), f32x4(-1.0, -2.0, -3.0, -4.0));
|
||||
all_eq!(simd_neg(z2), f32x4(-2.0, -3.0, -4.0, -5.0));
|
||||
all_eq!(simd_neg(x1), i32x4([-1, -2, -3, -4]));
|
||||
all_eq!(simd_neg(x2), i32x4([-2, -3, -4, -5]));
|
||||
all_eq!(simd_neg(z1), f32x4([-1.0, -2.0, -3.0, -4.0]));
|
||||
all_eq!(simd_neg(z2), f32x4([-2.0, -3.0, -4.0, -5.0]));
|
||||
|
||||
all_eq!(simd_bswap(x1), i32x4(0x01000000, 0x02000000, 0x03000000, 0x04000000));
|
||||
all_eq_!(simd_bswap(y1), U32::<4>([0x01000000, 0x02000000, 0x03000000, 0x04000000]));
|
||||
all_eq!(simd_bswap(x1), i32x4([0x01000000, 0x02000000, 0x03000000, 0x04000000]));
|
||||
all_eq!(simd_bswap(y1), U32::<4>([0x01000000, 0x02000000, 0x03000000, 0x04000000]));
|
||||
|
||||
all_eq!(
|
||||
simd_bitreverse(x1),
|
||||
i32x4(0x80000000u32 as i32, 0x40000000, 0xc0000000u32 as i32, 0x20000000)
|
||||
i32x4([0x80000000u32 as i32, 0x40000000, 0xc0000000u32 as i32, 0x20000000])
|
||||
);
|
||||
all_eq_!(simd_bitreverse(y1), U32::<4>([0x80000000, 0x40000000, 0xc0000000, 0x20000000]));
|
||||
all_eq!(simd_bitreverse(y1), U32::<4>([0x80000000, 0x40000000, 0xc0000000, 0x20000000]));
|
||||
|
||||
all_eq!(simd_ctlz(x1), i32x4(31, 30, 30, 29));
|
||||
all_eq_!(simd_ctlz(y1), U32::<4>([31, 30, 30, 29]));
|
||||
all_eq!(simd_ctlz(x1), i32x4([31, 30, 30, 29]));
|
||||
all_eq!(simd_ctlz(y1), U32::<4>([31, 30, 30, 29]));
|
||||
|
||||
all_eq!(simd_ctpop(x1), i32x4(1, 1, 2, 1));
|
||||
all_eq_!(simd_ctpop(y1), U32::<4>([1, 1, 2, 1]));
|
||||
all_eq!(simd_ctpop(x2), i32x4(1, 2, 1, 2));
|
||||
all_eq_!(simd_ctpop(y2), U32::<4>([1, 2, 1, 2]));
|
||||
all_eq!(simd_ctpop(x3), i32x4(0, 31, 1, 32));
|
||||
all_eq_!(simd_ctpop(y3), U32::<4>([0, 31, 1, 32]));
|
||||
all_eq!(simd_ctpop(x1), i32x4([1, 1, 2, 1]));
|
||||
all_eq!(simd_ctpop(y1), U32::<4>([1, 1, 2, 1]));
|
||||
all_eq!(simd_ctpop(x2), i32x4([1, 2, 1, 2]));
|
||||
all_eq!(simd_ctpop(y2), U32::<4>([1, 2, 1, 2]));
|
||||
all_eq!(simd_ctpop(x3), i32x4([0, 31, 1, 32]));
|
||||
all_eq!(simd_ctpop(y3), U32::<4>([0, 31, 1, 32]));
|
||||
|
||||
all_eq!(simd_cttz(x1), i32x4(0, 1, 0, 2));
|
||||
all_eq_!(simd_cttz(y1), U32::<4>([0, 1, 0, 2]));
|
||||
all_eq!(simd_cttz(x1), i32x4([0, 1, 0, 2]));
|
||||
all_eq!(simd_cttz(y1), U32::<4>([0, 1, 0, 2]));
|
||||
}
|
||||
}
|
||||
|
@ -4,15 +4,15 @@
|
||||
#![allow(non_camel_case_types)]
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone)]
|
||||
pub struct i32x4(pub i32, pub i32, pub i32, pub i32);
|
||||
pub struct i32x4(pub [i32; 4]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone)]
|
||||
pub struct x4<T>(pub T, pub T, pub T, pub T);
|
||||
pub struct x4<T>(pub [T; 4]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone)]
|
||||
pub struct f32x4(pub f32, pub f32, pub f32, pub f32);
|
||||
pub struct f32x4(pub [f32; 4]);
|
||||
|
||||
extern "rust-intrinsic" {
|
||||
fn simd_saturating_add<T>(x: T, y: T) -> T;
|
||||
@ -20,9 +20,9 @@
|
||||
}
|
||||
|
||||
fn main() {
|
||||
let x = i32x4(0, 0, 0, 0);
|
||||
let y = x4(0_usize, 0, 0, 0);
|
||||
let z = f32x4(0.0, 0.0, 0.0, 0.0);
|
||||
let x = i32x4([0, 0, 0, 0]);
|
||||
let y = x4([0_usize, 0, 0, 0]);
|
||||
let z = f32x4([0.0, 0.0, 0.0, 0.0]);
|
||||
|
||||
unsafe {
|
||||
simd_saturating_add(x, x);
|
||||
|
@ -6,7 +6,7 @@
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
struct u32x4(pub u32, pub u32, pub u32, pub u32);
|
||||
struct u32x4(pub [u32; 4]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone)]
|
||||
@ -22,11 +22,11 @@ fn main() {
|
||||
{
|
||||
const M: u32 = u32::MAX;
|
||||
|
||||
let a = u32x4(1, 2, 3, 4);
|
||||
let b = u32x4(2, 4, 6, 8);
|
||||
let m = u32x4(M, M, M, M);
|
||||
let m1 = u32x4(M - 1, M - 1, M - 1, M - 1);
|
||||
let z = u32x4(0, 0, 0, 0);
|
||||
let a = u32x4([1, 2, 3, 4]);
|
||||
let b = u32x4([2, 4, 6, 8]);
|
||||
let m = u32x4([M, M, M, M]);
|
||||
let m1 = u32x4([M - 1, M - 1, M - 1, M - 1]);
|
||||
let z = u32x4([0, 0, 0, 0]);
|
||||
|
||||
unsafe {
|
||||
assert_eq!(simd_saturating_add(z, z), z);
|
||||
|
@ -11,36 +11,36 @@
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
struct u32x4(pub u32, pub u32, pub u32, pub u32);
|
||||
struct u32x4(pub [u32; 4]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
struct u8x4(pub u8, pub u8, pub u8, pub u8);
|
||||
struct u8x4(pub [u8; 4]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
struct Tx4<T>(pub T, pub T, pub T, pub T);
|
||||
struct Tx4<T>(pub [T; 4]);
|
||||
|
||||
extern "rust-intrinsic" {
|
||||
fn simd_bitmask<T, U>(x: T) -> U;
|
||||
}
|
||||
|
||||
fn main() {
|
||||
let z = u32x4(0, 0, 0, 0);
|
||||
let z = u32x4([0, 0, 0, 0]);
|
||||
let ez = 0_u8;
|
||||
|
||||
let o = u32x4(!0, !0, !0, !0);
|
||||
let o = u32x4([!0, !0, !0, !0]);
|
||||
let eo = 0b_1111_u8;
|
||||
|
||||
let m0 = u32x4(!0, 0, !0, 0);
|
||||
let m0 = u32x4([!0, 0, !0, 0]);
|
||||
let e0 = 0b_0000_0101_u8;
|
||||
|
||||
// Check that the MSB is extracted:
|
||||
let m = u8x4(0b_1000_0000, 0b_0100_0001, 0b_1100_0001, 0b_1111_1111);
|
||||
let m = u8x4([0b_1000_0000, 0b_0100_0001, 0b_1100_0001, 0b_1111_1111]);
|
||||
let e = 0b_1101;
|
||||
|
||||
// Check usize / isize
|
||||
let msize: Tx4<usize> = Tx4(usize::MAX, 0, usize::MAX, usize::MAX);
|
||||
let msize: Tx4<usize> = Tx4([usize::MAX, 0, usize::MAX, usize::MAX]);
|
||||
|
||||
unsafe {
|
||||
let r: u8 = simd_bitmask(z);
|
||||
|
@ -5,22 +5,20 @@
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone)]
|
||||
#[allow(non_camel_case_types)]
|
||||
struct i32x4(i32, i32, i32, i32);
|
||||
struct i32x4([i32; 4]);
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone)]
|
||||
#[allow(non_camel_case_types)]
|
||||
struct i32x8(i32, i32, i32, i32,
|
||||
i32, i32, i32, i32);
|
||||
struct i32x8([i32; 8]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone)]
|
||||
#[allow(non_camel_case_types)]
|
||||
struct f32x4(f32, f32, f32, f32);
|
||||
struct f32x4([f32; 4]);
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone)]
|
||||
#[allow(non_camel_case_types)]
|
||||
struct f32x8(f32, f32, f32, f32,
|
||||
f32, f32, f32, f32);
|
||||
struct f32x8([f32; 8]);
|
||||
|
||||
|
||||
extern "rust-intrinsic" {
|
||||
@ -28,7 +26,7 @@ struct f32x8(f32, f32, f32, f32,
|
||||
}
|
||||
|
||||
fn main() {
|
||||
let x = i32x4(0, 0, 0, 0);
|
||||
let x = i32x4([0, 0, 0, 0]);
|
||||
|
||||
unsafe {
|
||||
simd_cast::<i32, i32>(0);
|
||||
|
@ -1,23 +1,23 @@
|
||||
error[E0511]: invalid monomorphization of `simd_cast` intrinsic: expected SIMD input type, found non-SIMD `i32`
|
||||
--> $DIR/generic-cast.rs:34:9
|
||||
--> $DIR/generic-cast.rs:32:9
|
||||
|
|
||||
LL | simd_cast::<i32, i32>(0);
|
||||
| ^^^^^^^^^^^^^^^^^^^^^^^^
|
||||
|
||||
error[E0511]: invalid monomorphization of `simd_cast` intrinsic: expected SIMD input type, found non-SIMD `i32`
|
||||
--> $DIR/generic-cast.rs:36:9
|
||||
--> $DIR/generic-cast.rs:34:9
|
||||
|
|
||||
LL | simd_cast::<i32, i32x4>(0);
|
||||
| ^^^^^^^^^^^^^^^^^^^^^^^^^^
|
||||
|
||||
error[E0511]: invalid monomorphization of `simd_cast` intrinsic: expected SIMD return type, found non-SIMD `i32`
|
||||
--> $DIR/generic-cast.rs:38:9
|
||||
--> $DIR/generic-cast.rs:36:9
|
||||
|
|
||||
LL | simd_cast::<i32x4, i32>(x);
|
||||
| ^^^^^^^^^^^^^^^^^^^^^^^^^^
|
||||
|
||||
error[E0511]: invalid monomorphization of `simd_cast` intrinsic: expected return type with length 4 (same as input type `i32x4`), found `i32x8` with length 8
|
||||
--> $DIR/generic-cast.rs:40:9
|
||||
--> $DIR/generic-cast.rs:38:9
|
||||
|
|
||||
LL | simd_cast::<_, i32x8>(x);
|
||||
| ^^^^^^^^^^^^^^^^^^^^^^^^
|
||||
|
@ -6,13 +6,13 @@
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone)]
|
||||
struct i32x4(i32, i32, i32, i32);
|
||||
struct i32x4([i32; 4]);
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone)]
|
||||
struct u32x4(pub u32, pub u32, pub u32, pub u32);
|
||||
struct u32x4(pub [u32; 4]);
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone)]
|
||||
struct f32x4(pub f32, pub f32, pub f32, pub f32);
|
||||
struct f32x4(pub [f32; 4]);
|
||||
|
||||
extern "rust-intrinsic" {
|
||||
fn simd_eq<T, U>(x: T, y: T) -> U;
|
||||
@ -29,10 +29,10 @@ macro_rules! cmp {
|
||||
let rhs = $rhs;
|
||||
let e: u32x4 = concat_idents!(simd_, $method)($lhs, $rhs);
|
||||
// assume the scalar version is correct/the behaviour we want.
|
||||
assert!((e.0 != 0) == lhs.0 .$method(&rhs.0));
|
||||
assert!((e.1 != 0) == lhs.1 .$method(&rhs.1));
|
||||
assert!((e.2 != 0) == lhs.2 .$method(&rhs.2));
|
||||
assert!((e.3 != 0) == lhs.3 .$method(&rhs.3));
|
||||
assert!((e.0[0] != 0) == lhs.0[0] .$method(&rhs.0[0]));
|
||||
assert!((e.0[1] != 0) == lhs.0[1] .$method(&rhs.0[1]));
|
||||
assert!((e.0[2] != 0) == lhs.0[2] .$method(&rhs.0[2]));
|
||||
assert!((e.0[3] != 0) == lhs.0[3] .$method(&rhs.0[3]));
|
||||
}}
|
||||
}
|
||||
macro_rules! tests {
|
||||
@ -61,17 +61,17 @@ macro_rules! tests {
|
||||
fn main() {
|
||||
// 13 vs. -100 tests that we get signed vs. unsigned comparisons
|
||||
// correct (i32: 13 > -100, u32: 13 < -100). let i1 = i32x4(10, -11, 12, 13);
|
||||
let i1 = i32x4(10, -11, 12, 13);
|
||||
let i2 = i32x4(5, -5, 20, -100);
|
||||
let i3 = i32x4(10, -11, 20, -100);
|
||||
let i1 = i32x4([10, -11, 12, 13]);
|
||||
let i2 = i32x4([5, -5, 20, -100]);
|
||||
let i3 = i32x4([10, -11, 20, -100]);
|
||||
|
||||
let u1 = u32x4(10, !11+1, 12, 13);
|
||||
let u2 = u32x4(5, !5+1, 20, !100+1);
|
||||
let u3 = u32x4(10, !11+1, 20, !100+1);
|
||||
let u1 = u32x4([10, !11+1, 12, 13]);
|
||||
let u2 = u32x4([5, !5+1, 20, !100+1]);
|
||||
let u3 = u32x4([10, !11+1, 20, !100+1]);
|
||||
|
||||
let f1 = f32x4(10.0, -11.0, 12.0, 13.0);
|
||||
let f2 = f32x4(5.0, -5.0, 20.0, -100.0);
|
||||
let f3 = f32x4(10.0, -11.0, 20.0, -100.0);
|
||||
let f1 = f32x4([10.0, -11.0, 12.0, 13.0]);
|
||||
let f2 = f32x4([5.0, -5.0, 20.0, -100.0]);
|
||||
let f3 = f32x4([10.0, -11.0, 20.0, -100.0]);
|
||||
|
||||
unsafe {
|
||||
tests! {
|
||||
@ -92,7 +92,7 @@ fn main() {
|
||||
// NAN comparisons are special:
|
||||
// -11 (*) 13
|
||||
// -5 -100 (*)
|
||||
let f4 = f32x4(f32::NAN, f1.1, f32::NAN, f2.3);
|
||||
let f4 = f32x4([f32::NAN, f1.0[1], f32::NAN, f2.0[3]]);
|
||||
|
||||
unsafe {
|
||||
tests! {
|
||||
|
@ -5,12 +5,11 @@
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone)]
|
||||
#[allow(non_camel_case_types)]
|
||||
struct i32x4(i32, i32, i32, i32);
|
||||
struct i32x4([i32; 4]);
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone)]
|
||||
#[allow(non_camel_case_types)]
|
||||
struct i16x8(i16, i16, i16, i16,
|
||||
i16, i16, i16, i16);
|
||||
struct i16x8([i16; 8]);
|
||||
|
||||
extern "rust-intrinsic" {
|
||||
fn simd_eq<T, U>(x: T, y: T) -> U;
|
||||
@ -22,7 +21,7 @@ struct i16x8(i16, i16, i16, i16,
|
||||
}
|
||||
|
||||
fn main() {
|
||||
let x = i32x4(0, 0, 0, 0);
|
||||
let x = i32x4([0, 0, 0, 0]);
|
||||
|
||||
unsafe {
|
||||
simd_eq::<i32, i32>(0, 0);
|
||||
|
@ -1,107 +1,107 @@
|
||||
error[E0511]: invalid monomorphization of `simd_eq` intrinsic: expected SIMD input type, found non-SIMD `i32`
|
||||
--> $DIR/generic-comparison.rs:28:9
|
||||
--> $DIR/generic-comparison.rs:27:9
|
||||
|
|
||||
LL | simd_eq::<i32, i32>(0, 0);
|
||||
| ^^^^^^^^^^^^^^^^^^^^^^^^^
|
||||
|
||||
error[E0511]: invalid monomorphization of `simd_ne` intrinsic: expected SIMD input type, found non-SIMD `i32`
|
||||
--> $DIR/generic-comparison.rs:30:9
|
||||
--> $DIR/generic-comparison.rs:29:9
|
||||
|
|
||||
LL | simd_ne::<i32, i32>(0, 0);
|
||||
| ^^^^^^^^^^^^^^^^^^^^^^^^^
|
||||
|
||||
error[E0511]: invalid monomorphization of `simd_lt` intrinsic: expected SIMD input type, found non-SIMD `i32`
|
||||
--> $DIR/generic-comparison.rs:32:9
|
||||
--> $DIR/generic-comparison.rs:31:9
|
||||
|
|
||||
LL | simd_lt::<i32, i32>(0, 0);
|
||||
| ^^^^^^^^^^^^^^^^^^^^^^^^^
|
||||
|
||||
error[E0511]: invalid monomorphization of `simd_le` intrinsic: expected SIMD input type, found non-SIMD `i32`
|
||||
--> $DIR/generic-comparison.rs:34:9
|
||||
--> $DIR/generic-comparison.rs:33:9
|
||||
|
|
||||
LL | simd_le::<i32, i32>(0, 0);
|
||||
| ^^^^^^^^^^^^^^^^^^^^^^^^^
|
||||
|
||||
error[E0511]: invalid monomorphization of `simd_gt` intrinsic: expected SIMD input type, found non-SIMD `i32`
|
||||
--> $DIR/generic-comparison.rs:36:9
|
||||
--> $DIR/generic-comparison.rs:35:9
|
||||
|
|
||||
LL | simd_gt::<i32, i32>(0, 0);
|
||||
| ^^^^^^^^^^^^^^^^^^^^^^^^^
|
||||
|
||||
error[E0511]: invalid monomorphization of `simd_ge` intrinsic: expected SIMD input type, found non-SIMD `i32`
|
||||
--> $DIR/generic-comparison.rs:38:9
|
||||
--> $DIR/generic-comparison.rs:37:9
|
||||
|
|
||||
LL | simd_ge::<i32, i32>(0, 0);
|
||||
| ^^^^^^^^^^^^^^^^^^^^^^^^^
|
||||
|
||||
error[E0511]: invalid monomorphization of `simd_eq` intrinsic: expected SIMD return type, found non-SIMD `i32`
|
||||
--> $DIR/generic-comparison.rs:41:9
|
||||
--> $DIR/generic-comparison.rs:40:9
|
||||
|
|
||||
LL | simd_eq::<_, i32>(x, x);
|
||||
| ^^^^^^^^^^^^^^^^^^^^^^^
|
||||
|
||||
error[E0511]: invalid monomorphization of `simd_ne` intrinsic: expected SIMD return type, found non-SIMD `i32`
|
||||
--> $DIR/generic-comparison.rs:43:9
|
||||
--> $DIR/generic-comparison.rs:42:9
|
||||
|
|
||||
LL | simd_ne::<_, i32>(x, x);
|
||||
| ^^^^^^^^^^^^^^^^^^^^^^^
|
||||
|
||||
error[E0511]: invalid monomorphization of `simd_lt` intrinsic: expected SIMD return type, found non-SIMD `i32`
|
||||
--> $DIR/generic-comparison.rs:45:9
|
||||
--> $DIR/generic-comparison.rs:44:9
|
||||
|
|
||||
LL | simd_lt::<_, i32>(x, x);
|
||||
| ^^^^^^^^^^^^^^^^^^^^^^^
|
||||
|
||||
error[E0511]: invalid monomorphization of `simd_le` intrinsic: expected SIMD return type, found non-SIMD `i32`
|
||||
--> $DIR/generic-comparison.rs:47:9
|
||||
--> $DIR/generic-comparison.rs:46:9
|
||||
|
|
||||
LL | simd_le::<_, i32>(x, x);
|
||||
| ^^^^^^^^^^^^^^^^^^^^^^^
|
||||
|
||||
error[E0511]: invalid monomorphization of `simd_gt` intrinsic: expected SIMD return type, found non-SIMD `i32`
|
||||
--> $DIR/generic-comparison.rs:49:9
|
||||
--> $DIR/generic-comparison.rs:48:9
|
||||
|
|
||||
LL | simd_gt::<_, i32>(x, x);
|
||||
| ^^^^^^^^^^^^^^^^^^^^^^^
|
||||
|
||||
error[E0511]: invalid monomorphization of `simd_ge` intrinsic: expected SIMD return type, found non-SIMD `i32`
|
||||
--> $DIR/generic-comparison.rs:51:9
|
||||
--> $DIR/generic-comparison.rs:50:9
|
||||
|
|
||||
LL | simd_ge::<_, i32>(x, x);
|
||||
| ^^^^^^^^^^^^^^^^^^^^^^^
|
||||
|
||||
error[E0511]: invalid monomorphization of `simd_eq` intrinsic: expected return type with length 4 (same as input type `i32x4`), found `i16x8` with length 8
|
||||
--> $DIR/generic-comparison.rs:54:9
|
||||
--> $DIR/generic-comparison.rs:53:9
|
||||
|
|
||||
LL | simd_eq::<_, i16x8>(x, x);
|
||||
| ^^^^^^^^^^^^^^^^^^^^^^^^^
|
||||
|
||||
error[E0511]: invalid monomorphization of `simd_ne` intrinsic: expected return type with length 4 (same as input type `i32x4`), found `i16x8` with length 8
|
||||
--> $DIR/generic-comparison.rs:56:9
|
||||
--> $DIR/generic-comparison.rs:55:9
|
||||
|
|
||||
LL | simd_ne::<_, i16x8>(x, x);
|
||||
| ^^^^^^^^^^^^^^^^^^^^^^^^^
|
||||
|
||||
error[E0511]: invalid monomorphization of `simd_lt` intrinsic: expected return type with length 4 (same as input type `i32x4`), found `i16x8` with length 8
|
||||
--> $DIR/generic-comparison.rs:58:9
|
||||
--> $DIR/generic-comparison.rs:57:9
|
||||
|
|
||||
LL | simd_lt::<_, i16x8>(x, x);
|
||||
| ^^^^^^^^^^^^^^^^^^^^^^^^^
|
||||
|
||||
error[E0511]: invalid monomorphization of `simd_le` intrinsic: expected return type with length 4 (same as input type `i32x4`), found `i16x8` with length 8
|
||||
--> $DIR/generic-comparison.rs:60:9
|
||||
--> $DIR/generic-comparison.rs:59:9
|
||||
|
|
||||
LL | simd_le::<_, i16x8>(x, x);
|
||||
| ^^^^^^^^^^^^^^^^^^^^^^^^^
|
||||
|
||||
error[E0511]: invalid monomorphization of `simd_gt` intrinsic: expected return type with length 4 (same as input type `i32x4`), found `i16x8` with length 8
|
||||
--> $DIR/generic-comparison.rs:62:9
|
||||
--> $DIR/generic-comparison.rs:61:9
|
||||
|
|
||||
LL | simd_gt::<_, i16x8>(x, x);
|
||||
| ^^^^^^^^^^^^^^^^^^^^^^^^^
|
||||
|
||||
error[E0511]: invalid monomorphization of `simd_ge` intrinsic: expected return type with length 4 (same as input type `i32x4`), found `i16x8` with length 8
|
||||
--> $DIR/generic-comparison.rs:64:9
|
||||
--> $DIR/generic-comparison.rs:63:9
|
||||
|
|
||||
LL | simd_ge::<_, i16x8>(x, x);
|
||||
| ^^^^^^^^^^^^^^^^^^^^^^^^^
|
||||
|
@ -6,16 +6,15 @@
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, Debug, PartialEq)]
|
||||
#[allow(non_camel_case_types)]
|
||||
struct i32x2(i32, i32);
|
||||
struct i32x2([i32; 2]);
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, Debug, PartialEq)]
|
||||
#[allow(non_camel_case_types)]
|
||||
struct i32x4(i32, i32, i32, i32);
|
||||
struct i32x4([i32; 4]);
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, Debug, PartialEq)]
|
||||
#[allow(non_camel_case_types)]
|
||||
struct i32x8(i32, i32, i32, i32,
|
||||
i32, i32, i32, i32);
|
||||
struct i32x8([i32; 8]);
|
||||
|
||||
extern "rust-intrinsic" {
|
||||
fn simd_insert<T, E>(x: T, idx: u32, y: E) -> T;
|
||||
@ -37,26 +36,26 @@ macro_rules! all_eq {
|
||||
}
|
||||
|
||||
fn main() {
|
||||
let x2 = i32x2(20, 21);
|
||||
let x4 = i32x4(40, 41, 42, 43);
|
||||
let x8 = i32x8(80, 81, 82, 83, 84, 85, 86, 87);
|
||||
let x2 = i32x2([20, 21]);
|
||||
let x4 = i32x4([40, 41, 42, 43]);
|
||||
let x8 = i32x8([80, 81, 82, 83, 84, 85, 86, 87]);
|
||||
unsafe {
|
||||
all_eq!(simd_insert(x2, 0, 100), i32x2(100, 21));
|
||||
all_eq!(simd_insert(x2, 1, 100), i32x2(20, 100));
|
||||
all_eq!(simd_insert(x2, 0, 100), i32x2([100, 21]));
|
||||
all_eq!(simd_insert(x2, 1, 100), i32x2([20, 100]));
|
||||
|
||||
all_eq!(simd_insert(x4, 0, 100), i32x4(100, 41, 42, 43));
|
||||
all_eq!(simd_insert(x4, 1, 100), i32x4(40, 100, 42, 43));
|
||||
all_eq!(simd_insert(x4, 2, 100), i32x4(40, 41, 100, 43));
|
||||
all_eq!(simd_insert(x4, 3, 100), i32x4(40, 41, 42, 100));
|
||||
all_eq!(simd_insert(x4, 0, 100), i32x4([100, 41, 42, 43]));
|
||||
all_eq!(simd_insert(x4, 1, 100), i32x4([40, 100, 42, 43]));
|
||||
all_eq!(simd_insert(x4, 2, 100), i32x4([40, 41, 100, 43]));
|
||||
all_eq!(simd_insert(x4, 3, 100), i32x4([40, 41, 42, 100]));
|
||||
|
||||
all_eq!(simd_insert(x8, 0, 100), i32x8(100, 81, 82, 83, 84, 85, 86, 87));
|
||||
all_eq!(simd_insert(x8, 1, 100), i32x8(80, 100, 82, 83, 84, 85, 86, 87));
|
||||
all_eq!(simd_insert(x8, 2, 100), i32x8(80, 81, 100, 83, 84, 85, 86, 87));
|
||||
all_eq!(simd_insert(x8, 3, 100), i32x8(80, 81, 82, 100, 84, 85, 86, 87));
|
||||
all_eq!(simd_insert(x8, 4, 100), i32x8(80, 81, 82, 83, 100, 85, 86, 87));
|
||||
all_eq!(simd_insert(x8, 5, 100), i32x8(80, 81, 82, 83, 84, 100, 86, 87));
|
||||
all_eq!(simd_insert(x8, 6, 100), i32x8(80, 81, 82, 83, 84, 85, 100, 87));
|
||||
all_eq!(simd_insert(x8, 7, 100), i32x8(80, 81, 82, 83, 84, 85, 86, 100));
|
||||
all_eq!(simd_insert(x8, 0, 100), i32x8([100, 81, 82, 83, 84, 85, 86, 87]));
|
||||
all_eq!(simd_insert(x8, 1, 100), i32x8([80, 100, 82, 83, 84, 85, 86, 87]));
|
||||
all_eq!(simd_insert(x8, 2, 100), i32x8([80, 81, 100, 83, 84, 85, 86, 87]));
|
||||
all_eq!(simd_insert(x8, 3, 100), i32x8([80, 81, 82, 100, 84, 85, 86, 87]));
|
||||
all_eq!(simd_insert(x8, 4, 100), i32x8([80, 81, 82, 83, 100, 85, 86, 87]));
|
||||
all_eq!(simd_insert(x8, 5, 100), i32x8([80, 81, 82, 83, 84, 100, 86, 87]));
|
||||
all_eq!(simd_insert(x8, 6, 100), i32x8([80, 81, 82, 83, 84, 85, 100, 87]));
|
||||
all_eq!(simd_insert(x8, 7, 100), i32x8([80, 81, 82, 83, 84, 85, 86, 100]));
|
||||
|
||||
all_eq!(simd_extract(x2, 0), 20);
|
||||
all_eq!(simd_extract(x2, 1), 21);
|
||||
@ -76,24 +75,24 @@ fn main() {
|
||||
all_eq!(simd_extract(x8, 7), 87);
|
||||
}
|
||||
|
||||
let y2 = i32x2(120, 121);
|
||||
let y4 = i32x4(140, 141, 142, 143);
|
||||
let y8 = i32x8(180, 181, 182, 183, 184, 185, 186, 187);
|
||||
let y2 = i32x2([120, 121]);
|
||||
let y4 = i32x4([140, 141, 142, 143]);
|
||||
let y8 = i32x8([180, 181, 182, 183, 184, 185, 186, 187]);
|
||||
unsafe {
|
||||
all_eq!(simd_shuffle(x2, y2, const { [3u32, 0] }), i32x2(121, 20));
|
||||
all_eq!(simd_shuffle(x2, y2, const { [3u32, 0, 1, 2] }), i32x4(121, 20, 21, 120));
|
||||
all_eq!(simd_shuffle(x2, y2, const { [3u32, 0] }), i32x2([121, 20]));
|
||||
all_eq!(simd_shuffle(x2, y2, const { [3u32, 0, 1, 2] }), i32x4([121, 20, 21, 120]));
|
||||
all_eq!(simd_shuffle(x2, y2, const { [3u32, 0, 1, 2, 1, 2, 3, 0] }),
|
||||
i32x8(121, 20, 21, 120, 21, 120, 121, 20));
|
||||
i32x8([121, 20, 21, 120, 21, 120, 121, 20]));
|
||||
|
||||
all_eq!(simd_shuffle(x4, y4, const { [7u32, 2] }), i32x2(143, 42));
|
||||
all_eq!(simd_shuffle(x4, y4, const { [7u32, 2, 5, 0] }), i32x4(143, 42, 141, 40));
|
||||
all_eq!(simd_shuffle(x4, y4, const { [7u32, 2] }), i32x2([143, 42]));
|
||||
all_eq!(simd_shuffle(x4, y4, const { [7u32, 2, 5, 0] }), i32x4([143, 42, 141, 40]));
|
||||
all_eq!(simd_shuffle(x4, y4, const { [7u32, 2, 5, 0, 3, 6, 4, 1] }),
|
||||
i32x8(143, 42, 141, 40, 43, 142, 140, 41));
|
||||
i32x8([143, 42, 141, 40, 43, 142, 140, 41]));
|
||||
|
||||
all_eq!(simd_shuffle(x8, y8, const { [11u32, 5] }), i32x2(183, 85));
|
||||
all_eq!(simd_shuffle(x8, y8, const { [11u32, 5, 15, 0] }), i32x4(183, 85, 187, 80));
|
||||
all_eq!(simd_shuffle(x8, y8, const { [11u32, 5] }), i32x2([183, 85]));
|
||||
all_eq!(simd_shuffle(x8, y8, const { [11u32, 5, 15, 0] }), i32x4([183, 85, 187, 80]));
|
||||
all_eq!(simd_shuffle(x8, y8, const { [11u32, 5, 15, 0, 3, 8, 12, 1] }),
|
||||
i32x8(183, 85, 187, 80, 83, 180, 184, 81));
|
||||
i32x8([183, 85, 187, 80, 83, 180, 184, 81]));
|
||||
}
|
||||
|
||||
}
|
||||
|
@ -6,28 +6,28 @@
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone)]
|
||||
#[allow(non_camel_case_types)]
|
||||
struct i32x2(i32, i32);
|
||||
struct i32x2([i32; 2]);
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone)]
|
||||
#[allow(non_camel_case_types)]
|
||||
struct i32x4(i32, i32, i32, i32);
|
||||
struct i32x4([i32; 4]);
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone)]
|
||||
#[allow(non_camel_case_types)]
|
||||
struct i32x8(i32, i32, i32, i32, i32, i32, i32, i32);
|
||||
struct i32x8([i32; 8]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone)]
|
||||
#[allow(non_camel_case_types)]
|
||||
struct f32x2(f32, f32);
|
||||
struct f32x2([f32; 2]);
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone)]
|
||||
#[allow(non_camel_case_types)]
|
||||
struct f32x4(f32, f32, f32, f32);
|
||||
struct f32x4([f32; 4]);
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone)]
|
||||
#[allow(non_camel_case_types)]
|
||||
struct f32x8(f32, f32, f32, f32, f32, f32, f32, f32);
|
||||
struct f32x8([f32; 8]);
|
||||
|
||||
extern "rust-intrinsic" {
|
||||
fn simd_insert<T, E>(x: T, idx: u32, y: E) -> T;
|
||||
@ -38,7 +38,7 @@
|
||||
}
|
||||
|
||||
fn main() {
|
||||
let x = i32x4(0, 0, 0, 0);
|
||||
let x = i32x4([0, 0, 0, 0]);
|
||||
|
||||
unsafe {
|
||||
simd_insert(0, 0, 0);
|
||||
|
@ -8,7 +8,7 @@
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
struct x4<T>(pub T, pub T, pub T, pub T);
|
||||
struct x4<T>(pub [T; 4]);
|
||||
|
||||
extern "rust-intrinsic" {
|
||||
fn simd_gather<T, U, V>(x: T, y: U, z: V) -> T;
|
||||
@ -18,19 +18,19 @@
|
||||
fn main() {
|
||||
let mut x = [0_f32, 1., 2., 3., 4., 5., 6., 7.];
|
||||
|
||||
let default = x4(-3_f32, -3., -3., -3.);
|
||||
let s_strided = x4(0_f32, 2., -3., 6.);
|
||||
let mask = x4(-1_i32, -1, 0, -1);
|
||||
let default = x4([-3_f32, -3., -3., -3.]);
|
||||
let s_strided = x4([0_f32, 2., -3., 6.]);
|
||||
let mask = x4([-1_i32, -1, 0, -1]);
|
||||
|
||||
// reading from *const
|
||||
unsafe {
|
||||
let pointer = x.as_ptr();
|
||||
let pointers = x4(
|
||||
let pointers = x4([
|
||||
pointer.offset(0),
|
||||
pointer.offset(2),
|
||||
pointer.offset(4),
|
||||
pointer.offset(6)
|
||||
);
|
||||
]);
|
||||
|
||||
let r_strided = simd_gather(default, pointers, mask);
|
||||
|
||||
@ -40,12 +40,12 @@ fn main() {
|
||||
// reading from *mut
|
||||
unsafe {
|
||||
let pointer = x.as_mut_ptr();
|
||||
let pointers = x4(
|
||||
let pointers = x4([
|
||||
pointer.offset(0),
|
||||
pointer.offset(2),
|
||||
pointer.offset(4),
|
||||
pointer.offset(6)
|
||||
);
|
||||
]);
|
||||
|
||||
let r_strided = simd_gather(default, pointers, mask);
|
||||
|
||||
@ -55,14 +55,14 @@ fn main() {
|
||||
// writing to *mut
|
||||
unsafe {
|
||||
let pointer = x.as_mut_ptr();
|
||||
let pointers = x4(
|
||||
let pointers = x4([
|
||||
pointer.offset(0),
|
||||
pointer.offset(2),
|
||||
pointer.offset(4),
|
||||
pointer.offset(6)
|
||||
);
|
||||
]);
|
||||
|
||||
let values = x4(42_f32, 43_f32, 44_f32, 45_f32);
|
||||
let values = x4([42_f32, 43_f32, 44_f32, 45_f32]);
|
||||
simd_scatter(values, pointers, mask);
|
||||
|
||||
assert_eq!(x, [42., 1., 43., 3., 4., 5., 45., 7.]);
|
||||
@ -80,18 +80,18 @@ fn main() {
|
||||
&x[7] as *const f32
|
||||
];
|
||||
|
||||
let default = x4(y[0], y[0], y[0], y[0]);
|
||||
let s_strided = x4(y[0], y[2], y[0], y[6]);
|
||||
let default = x4([y[0], y[0], y[0], y[0]]);
|
||||
let s_strided = x4([y[0], y[2], y[0], y[6]]);
|
||||
|
||||
// reading from *const
|
||||
unsafe {
|
||||
let pointer = y.as_ptr();
|
||||
let pointers = x4(
|
||||
let pointers = x4([
|
||||
pointer.offset(0),
|
||||
pointer.offset(2),
|
||||
pointer.offset(4),
|
||||
pointer.offset(6)
|
||||
);
|
||||
]);
|
||||
|
||||
let r_strided = simd_gather(default, pointers, mask);
|
||||
|
||||
@ -101,12 +101,12 @@ fn main() {
|
||||
// reading from *mut
|
||||
unsafe {
|
||||
let pointer = y.as_mut_ptr();
|
||||
let pointers = x4(
|
||||
let pointers = x4([
|
||||
pointer.offset(0),
|
||||
pointer.offset(2),
|
||||
pointer.offset(4),
|
||||
pointer.offset(6)
|
||||
);
|
||||
]);
|
||||
|
||||
let r_strided = simd_gather(default, pointers, mask);
|
||||
|
||||
@ -116,14 +116,14 @@ fn main() {
|
||||
// writing to *mut
|
||||
unsafe {
|
||||
let pointer = y.as_mut_ptr();
|
||||
let pointers = x4(
|
||||
let pointers = x4([
|
||||
pointer.offset(0),
|
||||
pointer.offset(2),
|
||||
pointer.offset(4),
|
||||
pointer.offset(6)
|
||||
);
|
||||
]);
|
||||
|
||||
let values = x4(y[7], y[6], y[5], y[1]);
|
||||
let values = x4([y[7], y[6], y[5], y[1]]);
|
||||
simd_scatter(values, pointers, mask);
|
||||
|
||||
let s = [
|
||||
|
@ -10,19 +10,19 @@
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone)]
|
||||
struct i32x4(pub i32, pub i32, pub i32, pub i32);
|
||||
struct i32x4(pub [i32; 4]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone)]
|
||||
struct u32x4(pub u32, pub u32, pub u32, pub u32);
|
||||
struct u32x4(pub [u32; 4]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone)]
|
||||
struct f32x4(pub f32, pub f32, pub f32, pub f32);
|
||||
struct f32x4(pub [f32; 4]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone)]
|
||||
struct b8x4(pub i8, pub i8, pub i8, pub i8);
|
||||
struct b8x4(pub [i8; 4]);
|
||||
|
||||
extern "rust-intrinsic" {
|
||||
fn simd_reduce_add_unordered<T, U>(x: T) -> U;
|
||||
@ -40,7 +40,7 @@
|
||||
|
||||
fn main() {
|
||||
unsafe {
|
||||
let x = i32x4(1, -2, 3, 4);
|
||||
let x = i32x4([1, -2, 3, 4]);
|
||||
let r: i32 = simd_reduce_add_unordered(x);
|
||||
assert_eq!(r, 6_i32);
|
||||
let r: i32 = simd_reduce_mul_unordered(x);
|
||||
@ -55,7 +55,7 @@ fn main() {
|
||||
let r: i32 = simd_reduce_max(x);
|
||||
assert_eq!(r, 4_i32);
|
||||
|
||||
let x = i32x4(-1, -1, -1, -1);
|
||||
let x = i32x4([-1, -1, -1, -1]);
|
||||
let r: i32 = simd_reduce_and(x);
|
||||
assert_eq!(r, -1_i32);
|
||||
let r: i32 = simd_reduce_or(x);
|
||||
@ -63,7 +63,7 @@ fn main() {
|
||||
let r: i32 = simd_reduce_xor(x);
|
||||
assert_eq!(r, 0_i32);
|
||||
|
||||
let x = i32x4(-1, -1, 0, -1);
|
||||
let x = i32x4([-1, -1, 0, -1]);
|
||||
let r: i32 = simd_reduce_and(x);
|
||||
assert_eq!(r, 0_i32);
|
||||
let r: i32 = simd_reduce_or(x);
|
||||
@ -73,7 +73,7 @@ fn main() {
|
||||
}
|
||||
|
||||
unsafe {
|
||||
let x = u32x4(1, 2, 3, 4);
|
||||
let x = u32x4([1, 2, 3, 4]);
|
||||
let r: u32 = simd_reduce_add_unordered(x);
|
||||
assert_eq!(r, 10_u32);
|
||||
let r: u32 = simd_reduce_mul_unordered(x);
|
||||
@ -89,7 +89,7 @@ fn main() {
|
||||
assert_eq!(r, 4_u32);
|
||||
|
||||
let t = u32::MAX;
|
||||
let x = u32x4(t, t, t, t);
|
||||
let x = u32x4([t, t, t, t]);
|
||||
let r: u32 = simd_reduce_and(x);
|
||||
assert_eq!(r, t);
|
||||
let r: u32 = simd_reduce_or(x);
|
||||
@ -97,7 +97,7 @@ fn main() {
|
||||
let r: u32 = simd_reduce_xor(x);
|
||||
assert_eq!(r, 0_u32);
|
||||
|
||||
let x = u32x4(t, t, 0, t);
|
||||
let x = u32x4([t, t, 0, t]);
|
||||
let r: u32 = simd_reduce_and(x);
|
||||
assert_eq!(r, 0_u32);
|
||||
let r: u32 = simd_reduce_or(x);
|
||||
@ -107,7 +107,7 @@ fn main() {
|
||||
}
|
||||
|
||||
unsafe {
|
||||
let x = f32x4(1., -2., 3., 4.);
|
||||
let x = f32x4([1., -2., 3., 4.]);
|
||||
let r: f32 = simd_reduce_add_unordered(x);
|
||||
assert_eq!(r, 6_f32);
|
||||
let r: f32 = simd_reduce_mul_unordered(x);
|
||||
@ -128,19 +128,19 @@ fn main() {
|
||||
}
|
||||
|
||||
unsafe {
|
||||
let x = b8x4(!0, !0, !0, !0);
|
||||
let x = b8x4([!0, !0, !0, !0]);
|
||||
let r: bool = simd_reduce_all(x);
|
||||
assert_eq!(r, true);
|
||||
let r: bool = simd_reduce_any(x);
|
||||
assert_eq!(r, true);
|
||||
|
||||
let x = b8x4(!0, !0, 0, !0);
|
||||
let x = b8x4([!0, !0, 0, !0]);
|
||||
let r: bool = simd_reduce_all(x);
|
||||
assert_eq!(r, false);
|
||||
let r: bool = simd_reduce_any(x);
|
||||
assert_eq!(r, true);
|
||||
|
||||
let x = b8x4(0, 0, 0, 0);
|
||||
let x = b8x4([0, 0, 0, 0]);
|
||||
let r: bool = simd_reduce_all(x);
|
||||
assert_eq!(r, false);
|
||||
let r: bool = simd_reduce_any(x);
|
||||
|
@ -9,11 +9,11 @@
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone)]
|
||||
pub struct f32x4(pub f32, pub f32, pub f32, pub f32);
|
||||
pub struct f32x4(pub [f32; 4]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone)]
|
||||
pub struct u32x4(pub u32, pub u32, pub u32, pub u32);
|
||||
pub struct u32x4(pub [u32; 4]);
|
||||
|
||||
|
||||
extern "rust-intrinsic" {
|
||||
@ -27,8 +27,8 @@
|
||||
}
|
||||
|
||||
fn main() {
|
||||
let x = u32x4(0, 0, 0, 0);
|
||||
let z = f32x4(0.0, 0.0, 0.0, 0.0);
|
||||
let x = u32x4([0, 0, 0, 0]);
|
||||
let z = f32x4([0.0, 0.0, 0.0, 0.0]);
|
||||
|
||||
unsafe {
|
||||
simd_reduce_add_ordered(z, 0);
|
||||
|
@ -11,23 +11,23 @@
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
struct i32x4(pub i32, pub i32, pub i32, pub i32);
|
||||
struct i32x4(pub [i32; 4]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
struct u32x4(pub u32, pub u32, pub u32, pub u32);
|
||||
struct u32x4(pub [u32; 4]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
struct u32x8(u32, u32, u32, u32, u32, u32, u32, u32);
|
||||
struct u32x8([u32; 8]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
struct f32x4(pub f32, pub f32, pub f32, pub f32);
|
||||
struct f32x4(pub [f32; 4]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
struct b8x4(pub i8, pub i8, pub i8, pub i8);
|
||||
struct b8x4(pub [i8; 4]);
|
||||
|
||||
extern "rust-intrinsic" {
|
||||
fn simd_select<T, U>(x: T, a: U, b: U) -> U;
|
||||
@ -35,15 +35,15 @@
|
||||
}
|
||||
|
||||
fn main() {
|
||||
let m0 = b8x4(!0, !0, !0, !0);
|
||||
let m1 = b8x4(0, 0, 0, 0);
|
||||
let m2 = b8x4(!0, !0, 0, 0);
|
||||
let m3 = b8x4(0, 0, !0, !0);
|
||||
let m4 = b8x4(!0, 0, !0, 0);
|
||||
let m0 = b8x4([!0, !0, !0, !0]);
|
||||
let m1 = b8x4([0, 0, 0, 0]);
|
||||
let m2 = b8x4([!0, !0, 0, 0]);
|
||||
let m3 = b8x4([0, 0, !0, !0]);
|
||||
let m4 = b8x4([!0, 0, !0, 0]);
|
||||
|
||||
unsafe {
|
||||
let a = i32x4(1, -2, 3, 4);
|
||||
let b = i32x4(5, 6, -7, 8);
|
||||
let a = i32x4([1, -2, 3, 4]);
|
||||
let b = i32x4([5, 6, -7, 8]);
|
||||
|
||||
let r: i32x4 = simd_select(m0, a, b);
|
||||
let e = a;
|
||||
@ -54,21 +54,21 @@ fn main() {
|
||||
assert_eq!(r, e);
|
||||
|
||||
let r: i32x4 = simd_select(m2, a, b);
|
||||
let e = i32x4(1, -2, -7, 8);
|
||||
let e = i32x4([1, -2, -7, 8]);
|
||||
assert_eq!(r, e);
|
||||
|
||||
let r: i32x4 = simd_select(m3, a, b);
|
||||
let e = i32x4(5, 6, 3, 4);
|
||||
let e = i32x4([5, 6, 3, 4]);
|
||||
assert_eq!(r, e);
|
||||
|
||||
let r: i32x4 = simd_select(m4, a, b);
|
||||
let e = i32x4(1, 6, 3, 8);
|
||||
let e = i32x4([1, 6, 3, 8]);
|
||||
assert_eq!(r, e);
|
||||
}
|
||||
|
||||
unsafe {
|
||||
let a = u32x4(1, 2, 3, 4);
|
||||
let b = u32x4(5, 6, 7, 8);
|
||||
let a = u32x4([1, 2, 3, 4]);
|
||||
let b = u32x4([5, 6, 7, 8]);
|
||||
|
||||
let r: u32x4 = simd_select(m0, a, b);
|
||||
let e = a;
|
||||
@ -79,21 +79,21 @@ fn main() {
|
||||
assert_eq!(r, e);
|
||||
|
||||
let r: u32x4 = simd_select(m2, a, b);
|
||||
let e = u32x4(1, 2, 7, 8);
|
||||
let e = u32x4([1, 2, 7, 8]);
|
||||
assert_eq!(r, e);
|
||||
|
||||
let r: u32x4 = simd_select(m3, a, b);
|
||||
let e = u32x4(5, 6, 3, 4);
|
||||
let e = u32x4([5, 6, 3, 4]);
|
||||
assert_eq!(r, e);
|
||||
|
||||
let r: u32x4 = simd_select(m4, a, b);
|
||||
let e = u32x4(1, 6, 3, 8);
|
||||
let e = u32x4([1, 6, 3, 8]);
|
||||
assert_eq!(r, e);
|
||||
}
|
||||
|
||||
unsafe {
|
||||
let a = f32x4(1., 2., 3., 4.);
|
||||
let b = f32x4(5., 6., 7., 8.);
|
||||
let a = f32x4([1., 2., 3., 4.]);
|
||||
let b = f32x4([5., 6., 7., 8.]);
|
||||
|
||||
let r: f32x4 = simd_select(m0, a, b);
|
||||
let e = a;
|
||||
@ -104,23 +104,23 @@ fn main() {
|
||||
assert_eq!(r, e);
|
||||
|
||||
let r: f32x4 = simd_select(m2, a, b);
|
||||
let e = f32x4(1., 2., 7., 8.);
|
||||
let e = f32x4([1., 2., 7., 8.]);
|
||||
assert_eq!(r, e);
|
||||
|
||||
let r: f32x4 = simd_select(m3, a, b);
|
||||
let e = f32x4(5., 6., 3., 4.);
|
||||
let e = f32x4([5., 6., 3., 4.]);
|
||||
assert_eq!(r, e);
|
||||
|
||||
let r: f32x4 = simd_select(m4, a, b);
|
||||
let e = f32x4(1., 6., 3., 8.);
|
||||
let e = f32x4([1., 6., 3., 8.]);
|
||||
assert_eq!(r, e);
|
||||
}
|
||||
|
||||
unsafe {
|
||||
let t = !0 as i8;
|
||||
let f = 0 as i8;
|
||||
let a = b8x4(t, f, t, f);
|
||||
let b = b8x4(f, f, f, t);
|
||||
let a = b8x4([t, f, t, f]);
|
||||
let b = b8x4([f, f, f, t]);
|
||||
|
||||
let r: b8x4 = simd_select(m0, a, b);
|
||||
let e = a;
|
||||
@ -131,21 +131,21 @@ fn main() {
|
||||
assert_eq!(r, e);
|
||||
|
||||
let r: b8x4 = simd_select(m2, a, b);
|
||||
let e = b8x4(t, f, f, t);
|
||||
let e = b8x4([t, f, f, t]);
|
||||
assert_eq!(r, e);
|
||||
|
||||
let r: b8x4 = simd_select(m3, a, b);
|
||||
let e = b8x4(f, f, t, f);
|
||||
let e = b8x4([f, f, t, f]);
|
||||
assert_eq!(r, e);
|
||||
|
||||
let r: b8x4 = simd_select(m4, a, b);
|
||||
let e = b8x4(t, f, t, t);
|
||||
let e = b8x4([t, f, t, t]);
|
||||
assert_eq!(r, e);
|
||||
}
|
||||
|
||||
unsafe {
|
||||
let a = u32x8(0, 1, 2, 3, 4, 5, 6, 7);
|
||||
let b = u32x8(8, 9, 10, 11, 12, 13, 14, 15);
|
||||
let a = u32x8([0, 1, 2, 3, 4, 5, 6, 7]);
|
||||
let b = u32x8([8, 9, 10, 11, 12, 13, 14, 15]);
|
||||
|
||||
let r: u32x8 = simd_select_bitmask(0u8, a, b);
|
||||
let e = b;
|
||||
@ -156,21 +156,21 @@ fn main() {
|
||||
assert_eq!(r, e);
|
||||
|
||||
let r: u32x8 = simd_select_bitmask(0b01010101u8, a, b);
|
||||
let e = u32x8(0, 9, 2, 11, 4, 13, 6, 15);
|
||||
let e = u32x8([0, 9, 2, 11, 4, 13, 6, 15]);
|
||||
assert_eq!(r, e);
|
||||
|
||||
let r: u32x8 = simd_select_bitmask(0b10101010u8, a, b);
|
||||
let e = u32x8(8, 1, 10, 3, 12, 5, 14, 7);
|
||||
let e = u32x8([8, 1, 10, 3, 12, 5, 14, 7]);
|
||||
assert_eq!(r, e);
|
||||
|
||||
let r: u32x8 = simd_select_bitmask(0b11110000u8, a, b);
|
||||
let e = u32x8(8, 9, 10, 11, 4, 5, 6, 7);
|
||||
let e = u32x8([8, 9, 10, 11, 4, 5, 6, 7]);
|
||||
assert_eq!(r, e);
|
||||
}
|
||||
|
||||
unsafe {
|
||||
let a = u32x4(0, 1, 2, 3);
|
||||
let b = u32x4(4, 5, 6, 7);
|
||||
let a = u32x4([0, 1, 2, 3]);
|
||||
let b = u32x4([4, 5, 6, 7]);
|
||||
|
||||
let r: u32x4 = simd_select_bitmask(0u8, a, b);
|
||||
let e = b;
|
||||
@ -181,15 +181,15 @@ fn main() {
|
||||
assert_eq!(r, e);
|
||||
|
||||
let r: u32x4 = simd_select_bitmask(0b0101u8, a, b);
|
||||
let e = u32x4(0, 5, 2, 7);
|
||||
let e = u32x4([0, 5, 2, 7]);
|
||||
assert_eq!(r, e);
|
||||
|
||||
let r: u32x4 = simd_select_bitmask(0b1010u8, a, b);
|
||||
let e = u32x4(4, 1, 6, 3);
|
||||
let e = u32x4([4, 1, 6, 3]);
|
||||
assert_eq!(r, e);
|
||||
|
||||
let r: u32x4 = simd_select_bitmask(0b1100u8, a, b);
|
||||
let e = u32x4(4, 5, 2, 3);
|
||||
let e = u32x4([4, 5, 2, 3]);
|
||||
assert_eq!(r, e);
|
||||
}
|
||||
}
|
||||
|
@ -8,19 +8,19 @@
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone)]
|
||||
pub struct f32x4(pub f32, pub f32, pub f32, pub f32);
|
||||
pub struct f32x4(pub [f32; 4]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone)]
|
||||
pub struct u32x4(pub u32, pub u32, pub u32, pub u32);
|
||||
pub struct u32x4(pub [u32; 4]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq)]
|
||||
struct b8x4(pub i8, pub i8, pub i8, pub i8);
|
||||
struct b8x4(pub [i8; 4]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq)]
|
||||
struct b8x8(pub i8, pub i8, pub i8, pub i8, pub i8, pub i8, pub i8, pub i8);
|
||||
struct b8x8(pub [i8; 8]);
|
||||
|
||||
extern "rust-intrinsic" {
|
||||
fn simd_select<T, U>(x: T, a: U, b: U) -> U;
|
||||
@ -28,10 +28,10 @@
|
||||
}
|
||||
|
||||
fn main() {
|
||||
let m4 = b8x4(0, 0, 0, 0);
|
||||
let m8 = b8x8(0, 0, 0, 0, 0, 0, 0, 0);
|
||||
let x = u32x4(0, 0, 0, 0);
|
||||
let z = f32x4(0.0, 0.0, 0.0, 0.0);
|
||||
let m4 = b8x4([0, 0, 0, 0]);
|
||||
let m8 = b8x8([0, 0, 0, 0, 0, 0, 0, 0]);
|
||||
let x = u32x4([0, 0, 0, 0]);
|
||||
let z = f32x4([0.0, 0.0, 0.0, 0.0]);
|
||||
|
||||
unsafe {
|
||||
simd_select(m4, x, x);
|
||||
|
@ -11,7 +11,7 @@
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Debug, PartialEq)]
|
||||
struct Simd2(u8, u8);
|
||||
struct Simd2([u8; 2]);
|
||||
|
||||
fn main() {
|
||||
unsafe {
|
||||
@ -22,5 +22,5 @@ fn main() {
|
||||
#[inline(always)]
|
||||
unsafe fn inline_me() -> Simd2 {
|
||||
const IDX: [u32; 2] = [0, 3];
|
||||
simd_shuffle(Simd2(10, 11), Simd2(12, 13), IDX)
|
||||
simd_shuffle(Simd2([10, 11]), Simd2([12, 13]), IDX)
|
||||
}
|
||||
|
@ -11,12 +11,12 @@
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Debug, PartialEq)]
|
||||
struct Simd2(u8, u8);
|
||||
struct Simd2([u8; 2]);
|
||||
|
||||
fn main() {
|
||||
unsafe {
|
||||
const IDX: [u32; 2] = [0, 1];
|
||||
let p_res: Simd2 = simd_shuffle(Simd2(10, 11), Simd2(12, 13), IDX);
|
||||
let p_res: Simd2 = simd_shuffle(Simd2([10, 11]), Simd2([12, 13]), IDX);
|
||||
let a_res: Simd2 = inline_me();
|
||||
|
||||
assert_10_11(p_res);
|
||||
@ -26,17 +26,17 @@ fn main() {
|
||||
|
||||
#[inline(never)]
|
||||
fn assert_10_11(x: Simd2) {
|
||||
assert_eq!(x, Simd2(10, 11));
|
||||
assert_eq!(x, Simd2([10, 11]));
|
||||
}
|
||||
|
||||
#[inline(never)]
|
||||
fn assert_10_13(x: Simd2) {
|
||||
assert_eq!(x, Simd2(10, 13));
|
||||
assert_eq!(x, Simd2([10, 13]));
|
||||
}
|
||||
|
||||
|
||||
#[inline(always)]
|
||||
unsafe fn inline_me() -> Simd2 {
|
||||
const IDX: [u32; 2] = [0, 3];
|
||||
simd_shuffle(Simd2(10, 11), Simd2(12, 13), IDX)
|
||||
simd_shuffle(Simd2([10, 11]), Simd2([12, 13]), IDX)
|
||||
}
|
||||
|
@ -2,9 +2,9 @@
|
||||
#![feature(repr_simd)]
|
||||
|
||||
#[repr(simd)]
|
||||
struct T(f64, f64, f64);
|
||||
struct T([f64; 3]);
|
||||
|
||||
static X: T = T(0.0, 0.0, 0.0);
|
||||
static X: T = T([0.0, 0.0, 0.0]);
|
||||
|
||||
fn main() {
|
||||
let _ = X;
|
||||
|
@ -6,7 +6,7 @@
|
||||
extern crate test;
|
||||
|
||||
#[repr(simd)]
|
||||
pub struct Mu64(pub u64, pub u64, pub u64, pub u64);
|
||||
pub struct Mu64(pub [u64; 4]);
|
||||
|
||||
fn main() {
|
||||
// This ensures an unaligned pointer even in optimized builds, though LLVM
|
||||
@ -18,7 +18,7 @@ fn main() {
|
||||
let misaligned_ptr: &mut [u8; 32] = {
|
||||
std::mem::transmute(memory.offset(1))
|
||||
};
|
||||
*misaligned_ptr = std::mem::transmute(Mu64(1, 1, 1, 1));
|
||||
*misaligned_ptr = std::mem::transmute(Mu64([1, 1, 1, 1]));
|
||||
test::black_box(memory);
|
||||
}
|
||||
}
|
||||
|
@ -5,18 +5,18 @@
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, Debug)]
|
||||
pub struct Char3(pub i8, pub i8, pub i8);
|
||||
pub struct Char3(pub [i8; 3]);
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, Debug)]
|
||||
pub struct Short3(pub i16, pub i16, pub i16);
|
||||
pub struct Short3(pub [i16; 3]);
|
||||
|
||||
extern "rust-intrinsic" {
|
||||
fn simd_cast<T, U>(x: T) -> U;
|
||||
}
|
||||
|
||||
fn main() {
|
||||
let cast: Short3 = unsafe { simd_cast(Char3(10, -3, -9)) };
|
||||
let cast: Short3 = unsafe { simd_cast(Char3([10, -3, -9])) };
|
||||
|
||||
println!("{:?}", cast);
|
||||
}
|
||||
|
@ -8,7 +8,7 @@
|
||||
|
||||
#[repr(simd)]
|
||||
#[derive(Copy, Clone, PartialEq, Debug)]
|
||||
struct x4<T>(pub T, pub T, pub T, pub T);
|
||||
struct x4<T>(pub [T; 4]);
|
||||
|
||||
extern "rust-intrinsic" {
|
||||
fn simd_gather<T, U, V>(x: T, y: U, z: V) -> T;
|
||||
@ -16,36 +16,36 @@
|
||||
|
||||
fn main() {
|
||||
let x: [usize; 4] = [10, 11, 12, 13];
|
||||
let default = x4(0_usize, 1, 2, 3);
|
||||
let default = x4([0_usize, 1, 2, 3]);
|
||||
let all_set = u8::MAX as i8; // aka -1
|
||||
let mask = x4(all_set, all_set, all_set, all_set);
|
||||
let expected = x4(10_usize, 11, 12, 13);
|
||||
let mask = x4([all_set, all_set, all_set, all_set]);
|
||||
let expected = x4([10_usize, 11, 12, 13]);
|
||||
|
||||
unsafe {
|
||||
let pointer = x.as_ptr();
|
||||
let pointers = x4(
|
||||
let pointers = x4([
|
||||
pointer.offset(0),
|
||||
pointer.offset(1),
|
||||
pointer.offset(2),
|
||||
pointer.offset(3)
|
||||
);
|
||||
]);
|
||||
let result = simd_gather(default, pointers, mask);
|
||||
assert_eq!(result, expected);
|
||||
}
|
||||
|
||||
// and again for isize
|
||||
let x: [isize; 4] = [10, 11, 12, 13];
|
||||
let default = x4(0_isize, 1, 2, 3);
|
||||
let expected = x4(10_isize, 11, 12, 13);
|
||||
let default = x4([0_isize, 1, 2, 3]);
|
||||
let expected = x4([10_isize, 11, 12, 13]);
|
||||
|
||||
unsafe {
|
||||
let pointer = x.as_ptr();
|
||||
let pointers = x4(
|
||||
let pointers = x4([
|
||||
pointer.offset(0),
|
||||
pointer.offset(1),
|
||||
pointer.offset(2),
|
||||
pointer.offset(3)
|
||||
);
|
||||
]);
|
||||
let result = simd_gather(default, pointers, mask);
|
||||
assert_eq!(result, expected);
|
||||
}
|
||||
|
@ -2,7 +2,11 @@
|
||||
|
||||
#[repr(simd)]
|
||||
struct I64F64(i64, f64);
|
||||
//~^ ERROR SIMD vector should be homogeneous
|
||||
//~^ ERROR SIMD vector's only field must be an array
|
||||
|
||||
#[repr(simd)]
|
||||
struct I64x4F64x0([i64; 4], [f64; 0]);
|
||||
//~^ ERROR SIMD vector cannot have multiple fields
|
||||
|
||||
static X: I64F64 = I64F64(1, 2.0);
|
||||
|
||||
|
@ -1,9 +1,16 @@
|
||||
error[E0076]: SIMD vector should be homogeneous
|
||||
error[E0076]: SIMD vector's only field must be an array
|
||||
--> $DIR/monomorphize-heterogeneous.rs:4:1
|
||||
|
|
||||
LL | struct I64F64(i64, f64);
|
||||
| ^^^^^^^^^^^^^ SIMD elements must have the same type
|
||||
| ^^^^^^^^^^^^^ --- not an array
|
||||
|
||||
error: aborting due to 1 previous error
|
||||
error[E0075]: SIMD vector cannot have multiple fields
|
||||
--> $DIR/monomorphize-heterogeneous.rs:8:1
|
||||
|
|
||||
LL | struct I64x4F64x0([i64; 4], [f64; 0]);
|
||||
| ^^^^^^^^^^^^^^^^^ -------- excess field
|
||||
|
||||
For more information about this error, try `rustc --explain E0076`.
|
||||
error: aborting due to 2 previous errors
|
||||
|
||||
Some errors have detailed explanations: E0075, E0076.
|
||||
For more information about an error, try `rustc --explain E0075`.
|
||||
|
11
tests/ui/simd/monomorphize-too-long.rs
Normal file
11
tests/ui/simd/monomorphize-too-long.rs
Normal file
@ -0,0 +1,11 @@
|
||||
//@ build-fail
|
||||
//@ error-pattern: monomorphising SIMD type `Simd<u16, 54321>` of length greater than 32768
|
||||
|
||||
#![feature(repr_simd)]
|
||||
|
||||
#[repr(simd)]
|
||||
struct Simd<T, const N: usize>([T; N]);
|
||||
|
||||
fn main() {
|
||||
let _too_big = Simd([1_u16; 54321]);
|
||||
}
|
Some files were not shown because too many files have changed in this diff Show More
Loading…
Reference in New Issue
Block a user