Skip to content

implement SIMD float rounding functions #2028

New issue

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

Sign up for GitHub

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

Already on GitHub? Sign in to your account

Merged
merged 2 commits into from
Mar 16, 2022
Merged

implement SIMD float rounding functions #2028

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
2029641
implement fabs using soft floats
RalfJung Mar 16, 2022
1f237b3
implement SIMD float rounding functions
RalfJung Mar 16, 2022
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
66 changes: 60 additions & 6 deletions src/shims/intrinsics.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -90,9 +90,20 @@ pub trait EvalContextExt<'mir, 'tcx: 'mir>: crate::MiriEvalContextExt<'mir, 'tcx
}

// Floating-point operations
"fabsf32" => {
let &[ref f] = check_arg_count(args)?;
let f = this.read_scalar(f)?.to_f32()?;
// Can be implemented in soft-floats.
this.write_scalar(Scalar::from_f32(f.abs()), dest)?;
}
"fabsf64" => {
let &[ref f] = check_arg_count(args)?;
let f = this.read_scalar(f)?.to_f64()?;
// Can be implemented in soft-floats.
this.write_scalar(Scalar::from_f64(f.abs()), dest)?;
}
#[rustfmt::skip]
| "sinf32"
| "fabsf32"
| "cosf32"
| "sqrtf32"
| "expf32"
Expand All @@ -110,7 +121,6 @@ pub trait EvalContextExt<'mir, 'tcx: 'mir>: crate::MiriEvalContextExt<'mir, 'tcx
let f = f32::from_bits(this.read_scalar(f)?.to_u32()?);
let f = match intrinsic_name {
"sinf32" => f.sin(),
"fabsf32" => f.abs(),
"cosf32" => f.cos(),
"sqrtf32" => f.sqrt(),
"expf32" => f.exp(),
Expand All @@ -129,7 +139,6 @@ pub trait EvalContextExt<'mir, 'tcx: 'mir>: crate::MiriEvalContextExt<'mir, 'tcx

#[rustfmt::skip]
| "sinf64"
| "fabsf64"
| "cosf64"
| "sqrtf64"
| "expf64"
Expand All @@ -147,7 +156,6 @@ pub trait EvalContextExt<'mir, 'tcx: 'mir>: crate::MiriEvalContextExt<'mir, 'tcx
let f = f64::from_bits(this.read_scalar(f)?.to_u64()?);
let f = match intrinsic_name {
"sinf64" => f.sin(),
"fabsf64" => f.abs(),
"cosf64" => f.cos(),
"sqrtf64" => f.sqrt(),
"expf64" => f.exp(),
Expand Down Expand Up @@ -317,20 +325,37 @@ pub trait EvalContextExt<'mir, 'tcx: 'mir>: crate::MiriEvalContextExt<'mir, 'tcx
// SIMD operations
#[rustfmt::skip]
| "simd_neg"
| "simd_fabs" => {
| "simd_fabs"
| "simd_ceil"
| "simd_floor"
| "simd_round"
| "simd_trunc" => {
let &[ref op] = check_arg_count(args)?;
let (op, op_len) = this.operand_to_simd(op)?;
let (dest, dest_len) = this.place_to_simd(dest)?;

assert_eq!(dest_len, op_len);

#[derive(Copy, Clone)]
enum HostFloatOp {
Ceil,
Floor,
Round,
Trunc,
}
#[derive(Copy, Clone)]
enum Op {
MirOp(mir::UnOp),
Abs,
HostOp(HostFloatOp),
}
let which = match intrinsic_name {
"simd_neg" => Op::MirOp(mir::UnOp::Neg),
"simd_fabs" => Op::Abs,
"simd_ceil" => Op::HostOp(HostFloatOp::Ceil),
"simd_floor" => Op::HostOp(HostFloatOp::Floor),
"simd_round" => Op::HostOp(HostFloatOp::Round),
"simd_trunc" => Op::HostOp(HostFloatOp::Trunc),
_ => unreachable!(),
};

Expand All @@ -342,14 +367,43 @@ pub trait EvalContextExt<'mir, 'tcx: 'mir>: crate::MiriEvalContextExt<'mir, 'tcx
Op::Abs => {
// Works for f32 and f64.
let ty::Float(float_ty) = op.layout.ty.kind() else {
bug!("simd_fabs operand is not a float")
bug!("{} operand is not a float", intrinsic_name)
};
let op = op.to_scalar()?;
match float_ty {
FloatTy::F32 => Scalar::from_f32(op.to_f32()?.abs()),
FloatTy::F64 => Scalar::from_f64(op.to_f64()?.abs()),
}
}
Op::HostOp(host_op) => {
let ty::Float(float_ty) = op.layout.ty.kind() else {
bug!("{} operand is not a float", intrinsic_name)
};
// FIXME using host floats
match float_ty {
FloatTy::F32 => {
let f = f32::from_bits(op.to_scalar()?.to_u32()?);
let res = match host_op {
HostFloatOp::Ceil => f.ceil(),
HostFloatOp::Floor => f.floor(),
HostFloatOp::Round => f.round(),
HostFloatOp::Trunc => f.trunc(),
};
Scalar::from_u32(res.to_bits())
}
FloatTy::F64 => {
let f = f64::from_bits(op.to_scalar()?.to_u64()?);
let res = match host_op {
HostFloatOp::Ceil => f.ceil(),
HostFloatOp::Floor => f.floor(),
HostFloatOp::Round => f.round(),
HostFloatOp::Trunc => f.trunc(),
};
Scalar::from_u64(res.to_bits())
}
}

}
};
this.write_scalar(val, &dest.into())?;
}
Expand Down
65 changes: 61 additions & 4 deletions tests/run-pass/portable-simd.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -106,19 +106,39 @@ fn simd_ops_i32() {
assert_eq!(a.min(b * i32x4::splat(4)), i32x4::from_array([4, 8, 10, -16]));

assert_eq!(
i8x4::from_array([i8::MAX, -23, 23, i8::MIN]).saturating_add(i8x4::from_array([1, i8::MIN, i8::MAX, 28])),
i8x4::from_array([i8::MAX, -23, 23, i8::MIN]).saturating_add(i8x4::from_array([
1,
i8::MIN,
i8::MAX,
28
])),
i8x4::from_array([i8::MAX, i8::MIN, i8::MAX, -100])
);
assert_eq!(
i8x4::from_array([i8::MAX, -28, 27, 42]).saturating_sub(i8x4::from_array([1, i8::MAX, i8::MAX, -80])),
i8x4::from_array([i8::MAX, -28, 27, 42]).saturating_sub(i8x4::from_array([
1,
i8::MAX,
i8::MAX,
-80
])),
i8x4::from_array([126, i8::MIN, -100, 122])
);
assert_eq!(
u8x4::from_array([u8::MAX, 0, 23, 42]).saturating_add(u8x4::from_array([1, 1, u8::MAX, 200])),
u8x4::from_array([u8::MAX, 0, 23, 42]).saturating_add(u8x4::from_array([
1,
1,
u8::MAX,
200
])),
u8x4::from_array([u8::MAX, 1, u8::MAX, 242])
);
assert_eq!(
u8x4::from_array([u8::MAX, 0, 23, 42]).saturating_sub(u8x4::from_array([1, 1, u8::MAX, 200])),
u8x4::from_array([u8::MAX, 0, 23, 42]).saturating_sub(u8x4::from_array([
1,
1,
u8::MAX,
200
])),
u8x4::from_array([254, 0, 0, 0])
);

Expand Down Expand Up @@ -259,6 +279,42 @@ fn simd_gather_scatter() {
assert_eq!(vec, vec![124, 11, 12, 82, 14, 15, 16, 17, 18]);
}

fn simd_round() {
assert_eq!(
f32x4::from_array([0.9, 1.001, 2.0, -4.5]).ceil(),
f32x4::from_array([1.0, 2.0, 2.0, -4.0])
);
assert_eq!(
f32x4::from_array([0.9, 1.001, 2.0, -4.5]).floor(),
f32x4::from_array([0.0, 1.0, 2.0, -5.0])
);
assert_eq!(
f32x4::from_array([0.9, 1.001, 2.0, -4.5]).round(),
f32x4::from_array([1.0, 1.0, 2.0, -5.0])
);
assert_eq!(
f32x4::from_array([0.9, 1.001, 2.0, -4.5]).trunc(),
f32x4::from_array([0.0, 1.0, 2.0, -4.0])
);

assert_eq!(
f64x4::from_array([0.9, 1.001, 2.0, -4.5]).ceil(),
f64x4::from_array([1.0, 2.0, 2.0, -4.0])
);
assert_eq!(
f64x4::from_array([0.9, 1.001, 2.0, -4.5]).floor(),
f64x4::from_array([0.0, 1.0, 2.0, -5.0])
);
assert_eq!(
f64x4::from_array([0.9, 1.001, 2.0, -4.5]).round(),
f64x4::from_array([1.0, 1.0, 2.0, -5.0])
);
assert_eq!(
f64x4::from_array([0.9, 1.001, 2.0, -4.5]).trunc(),
f64x4::from_array([0.0, 1.0, 2.0, -4.0])
);
}

fn simd_intrinsics() {
extern "platform-intrinsic" {
fn simd_eq<T, U>(x: T, y: T) -> U;
Expand Down Expand Up @@ -299,5 +355,6 @@ fn main() {
simd_cast();
simd_swizzle();
simd_gather_scatter();
simd_round();
simd_intrinsics();
}