Skip to content

Make mask types opaque #218

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
Shnatsel merged 9 commits into from
May 23, 2026
Merged

Make mask types opaque #218

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
9 commits
Select commit Hold shift + click to select a range
8dead9c
Make mask types opaque
Shnatsel May 18, 2026
505a4f3
Merge branch 'main' into opaque-mask-representation-minimal
Shnatsel May 19, 2026
55c4db5
Convert mask splat() to accept booleans instead of integers, mirrorin…
Shnatsel May 19, 2026
ce79eec
Doc comments: instead of just 'all zeroes' or 'all ones', specify the…
Shnatsel May 19, 2026
e2a1d52
Expose the high-level splat() level in masks once again, now that we'…
Shnatsel May 19, 2026
ebc8f17
drop now-unnecessary explicit conversion; fixed build and test
Shnatsel May 19, 2026
5981ef3
Drop SimdFrom generic bound now that we expose a public splat() metho…
Shnatsel May 19, 2026
fcbe60c
Merge branch 'main' into opaque-mask-representation-minimal
Shnatsel May 23, 2026
dcca45b
Reduce code duplication in Mask.splat()
Shnatsel May 23, 2026
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
1,028 changes: 58 additions & 970 deletions fearless_simd/src/generated/avx2.rs
View file
Open in desktop

Large diffs are not rendered by default.

1,770 changes: 679 additions & 1,091 deletions fearless_simd/src/generated/fallback.rs
View file
Open in desktop

Large diffs are not rendered by default.

1,188 changes: 32 additions & 1,156 deletions fearless_simd/src/generated/neon.rs
View file
Open in desktop

Large diffs are not rendered by default.

720 changes: 0 additions & 720 deletions fearless_simd/src/generated/ops.rs
View file
Open in desktop

Large diffs are not rendered by default.

904 changes: 316 additions & 588 deletions fearless_simd/src/generated/simd_trait.rs
View file
Open in desktop

Large diffs are not rendered by default.

1,346 changes: 183 additions & 1,163 deletions fearless_simd/src/generated/simd_types.rs
View file
Open in desktop

Large diffs are not rendered by default.

970 changes: 39 additions & 931 deletions fearless_simd/src/generated/sse4_2.rs
View file
Open in desktop

Large diffs are not rendered by default.

946 changes: 23 additions & 923 deletions fearless_simd/src/generated/wasm.rs
View file
Open in desktop

Large diffs are not rendered by default.

9 changes: 5 additions & 4 deletions fearless_simd/src/traits.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -9,11 +9,12 @@ use crate::{Level, Simd, SimdBase, seal::Seal};

/// Element-wise selection between two SIMD vectors using `self`.
pub trait Select<T: Seal>: Seal {
/// For each element of this mask, select the first operand if the element is all ones, and select the second
/// operand if the element is all zeroes.
/// For each logical lane of this mask, select the first operand if the lane is true, and select the second
/// operand if the lane is false.
///
/// If a mask element is *not* all ones or all zeroes, the result is unspecified. It may vary depending on
/// architecture, feature level, the mask elements' width, the mask vector's width, or library version.
/// Masks may be converted to and from signed integer lane arrays for compatibility with older APIs. For those
/// conversions, false is encoded as all zeroes (integer value 0) and true is encoded as all ones (integer value -1).
/// If a mask is constructed from any other integer bit pattern, the result of this operation is unspecified.
fn select(self, if_true: T, if_false: T) -> T;
}

Expand Down
13 changes: 13 additions & 0 deletions fearless_simd_gen/src/generic.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -13,6 +13,19 @@ pub(crate) fn generic_op_name(op: &str, ty: &VecType) -> Ident {
Ident::new(&format!("{op}_{}", ty.rust_name()), Span::call_site())
}

/// For backends that store masks as all-zero/all-one integer lanes, convert the public
/// `bool` mask splat argument into the backend's lane representation.
pub(crate) fn integer_lane_mask_splat_arg(vec_ty: &VecType) -> TokenStream {
if vec_ty.scalar != ScalarType::Mask {
return TokenStream::new();
}

let scalar = vec_ty.scalar.rust(vec_ty.scalar_bits);
quote! {
let val: #scalar = if val { !0 } else { 0 };
}
}

/// Implementation based on split/combine
///
/// Only suitable for lane-wise and block-wise operations
Expand Down
60 changes: 40 additions & 20 deletions fearless_simd_gen/src/mk_fallback.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2,7 +2,9 @@
// SPDX-License-Identifier: Apache-2.0 OR MIT

use crate::arch::fallback;
use crate::generic::{generic_from_bytes, generic_op_name, generic_to_bytes};
use crate::generic::{
generic_from_bytes, generic_op_name, generic_to_bytes, integer_lane_mask_splat_arg,
};
use crate::level::Level;
use crate::ops::{Op, OpSig, RefKind, valid_reinterpret};
use crate::types::{ScalarType, VecType};
Expand Down Expand Up @@ -136,8 +138,10 @@ impl Level for Fallback {
match sig {
OpSig::Splat => {
let num_elements = vec_ty.len;
let normalize_mask = integer_lane_mask_splat_arg(vec_ty);
quote! {
#method_sig {
#normalize_mask
[val; #num_elements].simd_into(self)
}
}
Expand All @@ -146,7 +150,7 @@ impl Level for Fallback {
let items = make_list(
(0..vec_ty.len)
.map(|idx| {
let args = [quote! { a[#idx] }];
let args = [lane(quote! { a }, vec_ty, idx)];
let expr = fallback::expr(method, vec_ty, &args);
quote! { #expr }
})
Expand All @@ -164,7 +168,8 @@ impl Level for Fallback {
(0..vec_ty.len)
.map(|idx| {
let scalar_ty = target_ty.scalar.rust(target_ty.scalar_bits);
quote! { a[#idx] as #scalar_ty }
let a = lane(quote! { a }, vec_ty, idx);
quote! { #a as #scalar_ty }
})
.collect::<Vec<_>>(),
);
Expand All @@ -179,19 +184,20 @@ impl Level for Fallback {
let items = make_list(
(0..vec_ty.len)
.map(|idx| {
let b_lane = lane(quote! { b }, vec_ty, idx);
let b = if fallback::translate_op(
method,
vec_ty.scalar == ScalarType::Float,
)
.map(rhs_reference)
.unwrap_or(true)
{
quote! { &b[#idx] }
quote! { &#b_lane }
} else {
quote! { b[#idx] }
b_lane
};

let args = [quote! { a[#idx] }, quote! { #b }];
let args = [lane(quote! { a }, vec_ty, idx), quote! { #b }];
let expr = fallback::expr(method, vec_ty, &args);
quote! { #expr }
})
Expand All @@ -208,7 +214,7 @@ impl Level for Fallback {
let items = make_list(
(0..vec_ty.len)
.map(|idx| {
let args = [quote! { a[#idx] }, quote! { shift }];
let args = [lane(quote! { a }, vec_ty, idx), quote! { shift }];
let expr = fallback::expr(method, vec_ty, &args);
quote! { #expr }
})
Expand Down Expand Up @@ -254,7 +260,9 @@ impl Level for Fallback {
let items = make_list(
(0..vec_ty.len)
.map(|idx: usize| {
let args = [quote! { &a[#idx] }, quote! { &b[#idx] }];
let a = lane(quote! { a }, vec_ty, idx);
let b = lane(quote! { b }, vec_ty, idx);
let args = [quote! { &#a }, quote! { &#b }];
let expr = fallback::expr(method, vec_ty, &args);
let mask_ty = mask_type.scalar.rust(vec_ty.scalar_bits);
quote! { -(#expr as #mask_ty) }
Expand All @@ -269,10 +277,14 @@ impl Level for Fallback {
}
}
OpSig::Select => {
let mask_type = vec_ty.mask_ty();
let items = make_list(
(0..vec_ty.len)
.map(|idx| {
quote! { if a[#idx] != 0 { b[#idx] } else { c[#idx] } }
let a = lane(quote! { a }, &mask_type, idx);
let b = lane(quote! { b }, vec_ty, idx);
let c = lane(quote! { c }, vec_ty, idx);
quote! { if #a != 0 { #b } else { #c } }
})
.collect::<Vec<_>>(),
);
Expand Down Expand Up @@ -326,7 +338,9 @@ impl Level for Fallback {
let zip = make_list(
indices
.map(|idx| {
quote! {a[#idx], b[#idx] }
let a = lane(quote! { a }, vec_ty, idx);
let b = lane(quote! { b }, vec_ty, idx);
quote! { #a, #b }
})
.collect::<Vec<_>>(),
);
Expand All @@ -347,12 +361,8 @@ impl Level for Fallback {
let unzip = make_list(
indices
.clone()
.map(|idx| {
quote! {a[#idx]}
})
.chain(indices.map(|idx| {
quote! {b[#idx]}
}))
.map(|idx| lane(quote! { a }, vec_ty, idx))
.chain(indices.map(|idx| lane(quote! { b }, vec_ty, idx)))
.collect::<Vec<_>>(),
);

Expand Down Expand Up @@ -392,7 +402,8 @@ impl Level for Fallback {
let items = make_list(
(0..vec_ty.len)
.map(|idx| {
quote! { a[#idx] as #scalar }
let a = lane(quote! { a }, vec_ty, idx);
quote! { #a as #scalar }
})
.collect::<Vec<_>>(),
);
Expand Down Expand Up @@ -421,7 +432,6 @@ impl Level for Fallback {
quantifier,
condition,
} => {
let indices = (0..vec_ty.len).map(|idx| quote! { #idx });
let check = if condition {
quote! { != }
} else {
Expand All @@ -430,10 +440,12 @@ impl Level for Fallback {

let expr = match quantifier {
crate::ops::Quantifier::Any => {
quote! { #(a[#indices] #check 0)||* }
let lanes = (0..vec_ty.len).map(|idx| lane(quote! { a }, vec_ty, idx));
quote! { #(#lanes #check 0)||* }
}
crate::ops::Quantifier::All => {
quote! { #(a[#indices] #check 0)&&* }
let lanes = (0..vec_ty.len).map(|idx| lane(quote! { a }, vec_ty, idx));
quote! { #(#lanes #check 0)&&* }
}
};

Expand Down Expand Up @@ -540,6 +552,14 @@ fn interleave_indices(
make_list(indices.into_iter().map(func).collect::<Vec<_>>())
}

fn lane(value: TokenStream, vec_ty: &VecType, idx: usize) -> TokenStream {
if vec_ty.scalar == ScalarType::Mask {
quote! { #value.val.0[#idx] }
} else {
quote! { #value[#idx] }
}
}

/// Whether the second argument of the function needs to be passed by reference.
fn rhs_reference(method: &str) -> bool {
!matches!(
Expand Down
4 changes: 3 additions & 1 deletion fearless_simd_gen/src/mk_neon.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -6,7 +6,7 @@ use quote::{ToTokens as _, format_ident, quote};

use crate::generic::{
generic_as_array, generic_from_array, generic_from_bytes, generic_op_name, generic_store_array,
generic_to_bytes,
generic_to_bytes, integer_lane_mask_splat_arg,
};
use crate::level::Level;
use crate::ops::{Op, SlideGranularity, valid_reinterpret};
Expand Down Expand Up @@ -82,9 +82,11 @@ impl Level for Neon {
match sig {
OpSig::Splat => {
let expr = neon::expr(method, vec_ty, &[quote! { val }]);
let normalize_mask = integer_lane_mask_splat_arg(vec_ty);
quote! {
#method_sig {
unsafe {
#normalize_mask
#expr.simd_into(self)
}
}
Expand Down
51 changes: 28 additions & 23 deletions fearless_simd_gen/src/mk_ops.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -7,7 +7,7 @@ use quote::{format_ident, quote};
use crate::{
generic::generic_op_name,
ops::{CoreOpTrait, OpKind, OpSig, TyFlavor, overloaded_ops_for},
types::{SIMD_TYPES, type_imports},
types::{SIMD_TYPES, ScalarType, type_imports},
};

pub(crate) fn mk_ops() -> TokenStream {
Expand Down Expand Up @@ -85,6 +85,32 @@ pub(crate) fn mk_ops() -> TokenStream {
}
_ => {
let scalar = ty.scalar.rust(ty.scalar_bits);
let scalar_overloads = (ty.scalar != ScalarType::Mask).then(|| {
quote! {
impl<S: Simd> core::ops::#trait_id<#scalar> for #simd<S> {
type Output = Self;
#[inline(always)]
fn #opfn(self, rhs: #scalar) -> Self::Output {
self.simd.#simd_fn(self, rhs.simd_into(self.simd))
}
}

impl<S: Simd> core::ops::#trait_assign_id<#scalar> for #simd<S> {
#[inline(always)]
fn #op_assign_fn(&mut self, rhs: #scalar) {
*self = self.simd.#simd_fn(*self, rhs.simd_into(self.simd));
}
}

impl<S: Simd> core::ops::#trait_id<#simd<S>> for #scalar {
type Output = #simd<S>;
#[inline(always)]
fn #opfn(self, rhs: #simd<S>) -> Self::Output {
rhs.simd.#simd_fn(self.simd_into(rhs.simd), rhs)
}
}
}
});
impls.push(quote! {
impl<S: Simd> core::ops::#trait_id for #simd<S> {
type Output = Self;
Expand All @@ -103,28 +129,7 @@ pub(crate) fn mk_ops() -> TokenStream {
}
}

impl<S: Simd> core::ops::#trait_id<#scalar> for #simd<S> {
type Output = Self;
#[inline(always)]
fn #opfn(self, rhs: #scalar) -> Self::Output {
self.simd.#simd_fn(self, rhs.simd_into(self.simd))
}
}

impl<S: Simd> core::ops::#trait_assign_id<#scalar> for #simd<S> {
#[inline(always)]
fn #op_assign_fn(&mut self, rhs: #scalar) {
*self = self.simd.#simd_fn(*self, rhs.simd_into(self.simd));
}
}

impl<S: Simd> core::ops::#trait_id<#simd<S>> for #scalar {
type Output = #simd<S>;
#[inline(always)]
fn #opfn(self, rhs: #simd<S>) -> Self::Output {
rhs.simd.#simd_fn(self.simd_into(rhs.simd), rhs)
}
}
#scalar_overloads
});
}
}
Expand Down
Loading
Loading