Sync portable-simd to rust-lang/portable-simd@72df4c4505

beginners-guide.md

@@ -33,7 +33,7 @@ SIMD has a few special vocabulary terms you should know:
 
 * **Vertical:** When an operation is "vertical", each lane processes individually without regard to the other lanes in the same vector. For example, a "vertical add" between two vectors would add lane 0 in `a` with lane 0 in `b`, with the total in lane 0 of `out`, and then the same thing for lanes 1, 2, etc. Most SIMD operations are vertical operations, so if your problem is a vertical problem then you can probably solve it with SIMD.
 
-* **Horizontal:** When an operation is "horizontal", the lanes within a single vector interact in some way. A "horizontal add" might add up lane 0 of `a` with lane 1 of `a`, with the total in lane 0 of `out`.
+* **Reducing/Reduce:** When an operation is "reducing" (functions named `reduce_*`), the lanes within a single vector are merged using some operation such as addition, returning the merged value as a scalar. For instance, a reducing add would return the sum of all the lanes' values.
 
 * **Target Feature:** Rust calls a CPU architecture extension a `target_feature`. Proper SIMD requires various CPU extensions to be enabled (details below). Don't confuse this with `feature`, which is a Cargo crate concept.
 

crates/core_simd/Cargo.toml

@@ -9,7 +9,7 @@ categories = ["hardware-support", "no-std"]
 license = "MIT OR Apache-2.0"
 
 [features]
-default = ["std", "generic_const_exprs"]
+default = []
 std = []
 generic_const_exprs = []
 

crates/core_simd/examples/matrix_inversion.rs

@@ -233,7 +233,7 @@ pub fn simd_inv4x4(m: Matrix4x4) -> Option<Matrix4x4> {
     let det = det.rotate_lanes_right::<2>() + det;
     let det = det.reverse().rotate_lanes_right::<2>() + det;
 
-    if det.horizontal_sum() == 0. {
+    if det.reduce_sum() == 0. {
         return None;
     }
     // calculate the reciprocal

crates/core_simd/examples/nbody.rs

@@ -107,10 +107,10 @@ mod nbody {
         let mut e = 0.;
         for i in 0..N_BODIES {
             let bi = &bodies[i];
-            e += bi.mass * (bi.v * bi.v).horizontal_sum() * 0.5;
+            e += bi.mass * (bi.v * bi.v).reduce_sum() * 0.5;
             for bj in bodies.iter().take(N_BODIES).skip(i + 1) {
                 let dx = bi.x - bj.x;
-                e -= bi.mass * bj.mass / (dx * dx).horizontal_sum().sqrt()
+                e -= bi.mass * bj.mass / (dx * dx).reduce_sum().sqrt()
             }
         }
         e
@@ -134,8 +134,8 @@ mod nbody {
         let mut mag = [0.0; N];
         for i in (0..N).step_by(2) {
             let d2s = f64x2::from_array([
-                (r[i] * r[i]).horizontal_sum(),
+                (r[i] * r[i]).reduce_sum(),
-                (r[i + 1] * r[i + 1]).horizontal_sum(),
+                (r[i + 1] * r[i + 1]).reduce_sum(),
             ]);
             let dmags = f64x2::splat(dt) / (d2s * d2s.sqrt());
             mag[i] = dmags[0];

crates/core_simd/examples/spectral_norm.rs

@@ -20,7 +20,7 @@ fn mult_av(v: &[f64], out: &mut [f64]) {
             sum += b / a;
             j += 2
         }
-        *out = sum.horizontal_sum();
+        *out = sum.reduce_sum();
     }
 }
 
@@ -38,7 +38,7 @@ fn mult_atv(v: &[f64], out: &mut [f64]) {
             sum += b / a;
             j += 2
         }
-        *out = sum.horizontal_sum();
+        *out = sum.reduce_sum();
     }
 }
 

crates/core_simd/src/comparisons.rs

@@ -66,3 +66,55 @@ where
         unsafe { Mask::from_int_unchecked(intrinsics::simd_ge(self, other)) }
     }
 }
+
+macro_rules! impl_ord_methods_vector {
+    { $type:ty } => {
+        impl<const LANES: usize> Simd<$type, LANES>
+        where
+            LaneCount<LANES>: SupportedLaneCount,
+        {
+            /// Returns the lane-wise minimum with `other`.
+            #[must_use = "method returns a new vector and does not mutate the original value"]
+            #[inline]
+            pub fn min(self, other: Self) -> Self {
+                self.lanes_gt(other).select(other, self)
+            }
+
+            /// Returns the lane-wise maximum with `other`.
+            #[must_use = "method returns a new vector and does not mutate the original value"]
+            #[inline]
+            pub fn max(self, other: Self) -> Self {
+                self.lanes_lt(other).select(other, self)
+            }
+
+            /// Restrict each lane to a certain interval.
+            ///
+            /// For each lane, returns `max` if `self` is greater than `max`, and `min` if `self` is
+            /// less than `min`. Otherwise returns `self`.
+            ///
+            /// # Panics
+            ///
+            /// Panics if `min > max` on any lane.
+            #[must_use = "method returns a new vector and does not mutate the original value"]
+            #[inline]
+            pub fn clamp(self, min: Self, max: Self) -> Self {
+                assert!(
+                    min.lanes_le(max).all(),
+                    "each lane in `min` must be less than or equal to the corresponding lane in `max`",
+                );
+                self.max(min).min(max)
+            }
+        }
+    }
+}
+
+impl_ord_methods_vector!(i8);
+impl_ord_methods_vector!(i16);
+impl_ord_methods_vector!(i32);
+impl_ord_methods_vector!(i64);
+impl_ord_methods_vector!(isize);
+impl_ord_methods_vector!(u8);
+impl_ord_methods_vector!(u16);
+impl_ord_methods_vector!(u32);
+impl_ord_methods_vector!(u64);
+impl_ord_methods_vector!(usize);

crates/core_simd/src/intrinsics.rs

@@ -18,7 +18,6 @@
 //!
 //! Unless stated otherwise, all intrinsics for binary operations require SIMD vectors of equal types and lengths.
 
-
 // These intrinsics aren't linked directly from LLVM and are mostly undocumented, however they are
 // mostly lowered to the matching LLVM instructions by the compiler in a fairly straightforward manner.
 // The associated LLVM instruction or intrinsic is documented alongside each Rust intrinsic function.
@@ -130,6 +129,14 @@ extern "platform-intrinsic" {
     pub(crate) fn simd_reduce_xor<T, U>(x: T) -> U;
 
     // truncate integer vector to bitmask
+    // `fn simd_bitmask(vector) -> unsigned integer` takes a vector of integers and
+    // returns either an unsigned integer or array of `u8`.
+    // Every element in the vector becomes a single bit in the returned bitmask.
+    // If the vector has less than 8 lanes, a u8 is returned with zeroed trailing bits.
+    // The bit order of the result depends on the byte endianness. LSB-first for little
+    // endian and MSB-first for big endian.
+    //
+    // UB if called on a vector with values other than 0 and -1.
     #[allow(unused)]
     pub(crate) fn simd_bitmask<T, U>(x: T) -> U;
 

crates/core_simd/src/lib.rs

@@ -1,6 +1,5 @@
-#![cfg_attr(not(feature = "std"), no_std)]
+#![no_std]
 #![feature(
-    const_fn_trait_bound,
     convert_float_to_int,
     decl_macro,
     intra_doc_pointers,

crates/core_simd/src/masks/to_bitmask.rs

@@ -50,6 +50,9 @@ macro_rules! impl_integer_intrinsic {
 }
 
 impl_integer_intrinsic! {
+    unsafe impl ToBitMask<BitMask=u8> for Mask<_, 1>
+    unsafe impl ToBitMask<BitMask=u8> for Mask<_, 2>
+    unsafe impl ToBitMask<BitMask=u8> for Mask<_, 4>
     unsafe impl ToBitMask<BitMask=u8> for Mask<_, 8>
     unsafe impl ToBitMask<BitMask=u16> for Mask<_, 16>
     unsafe impl ToBitMask<BitMask=u32> for Mask<_, 32>

crates/core_simd/src/math.rs

@@ -10,8 +10,7 @@ macro_rules! impl_uint_arith {
             /// # Examples
             /// ```
             /// # #![feature(portable_simd)]
-            /// # #[cfg(feature = "std")] use core_simd::Simd;
+            /// # use core::simd::Simd;
-            /// # #[cfg(not(feature = "std"))] use core::simd::Simd;
             #[doc = concat!("# use core::", stringify!($ty), "::MAX;")]
             /// let x = Simd::from_array([2, 1, 0, MAX]);
             /// let max = Simd::splat(MAX);
@@ -31,8 +30,7 @@ macro_rules! impl_uint_arith {
             /// # Examples
             /// ```
             /// # #![feature(portable_simd)]
-            /// # #[cfg(feature = "std")] use core_simd::Simd;
+            /// # use core::simd::Simd;
-            /// # #[cfg(not(feature = "std"))] use core::simd::Simd;
             #[doc = concat!("# use core::", stringify!($ty), "::MAX;")]
             /// let x = Simd::from_array([2, 1, 0, MAX]);
             /// let max = Simd::splat(MAX);
@@ -58,8 +56,7 @@ macro_rules! impl_int_arith {
             /// # Examples
             /// ```
             /// # #![feature(portable_simd)]
-            /// # #[cfg(feature = "std")] use core_simd::Simd;
+            /// # use core::simd::Simd;
-            /// # #[cfg(not(feature = "std"))] use core::simd::Simd;
             #[doc = concat!("# use core::", stringify!($ty), "::{MIN, MAX};")]
             /// let x = Simd::from_array([MIN, 0, 1, MAX]);
             /// let max = Simd::splat(MAX);
@@ -79,8 +76,7 @@ macro_rules! impl_int_arith {
             /// # Examples
             /// ```
             /// # #![feature(portable_simd)]
-            /// # #[cfg(feature = "std")] use core_simd::Simd;
+            /// # use core::simd::Simd;
-            /// # #[cfg(not(feature = "std"))] use core::simd::Simd;
             #[doc = concat!("# use core::", stringify!($ty), "::{MIN, MAX};")]
             /// let x = Simd::from_array([MIN, -2, -1, MAX]);
             /// let max = Simd::splat(MAX);
@@ -100,8 +96,7 @@ macro_rules! impl_int_arith {
             /// # Examples
             /// ```
             /// # #![feature(portable_simd)]
-            /// # #[cfg(feature = "std")] use core_simd::Simd;
+            /// # use core::simd::Simd;
-            /// # #[cfg(not(feature = "std"))] use core::simd::Simd;
             #[doc = concat!("# use core::", stringify!($ty), "::{MIN, MAX};")]
             /// let xs = Simd::from_array([MIN, MIN +1, -5, 0]);
             /// assert_eq!(xs.abs(), Simd::from_array([MIN, MAX, 5, 0]));
@@ -119,8 +114,7 @@ macro_rules! impl_int_arith {
             /// # Examples
             /// ```
             /// # #![feature(portable_simd)]
-            /// # #[cfg(feature = "std")] use core_simd::Simd;
+            /// # use core::simd::Simd;
-            /// # #[cfg(not(feature = "std"))] use core::simd::Simd;
             #[doc = concat!("# use core::", stringify!($ty), "::{MIN, MAX};")]
             /// let xs = Simd::from_array([MIN, -2, 0, 3]);
             /// let unsat = xs.abs();
@@ -142,8 +136,7 @@ macro_rules! impl_int_arith {
             /// # Examples
             /// ```
             /// # #![feature(portable_simd)]
-            /// # #[cfg(feature = "std")] use core_simd::Simd;
+            /// # use core::simd::Simd;
-            /// # #[cfg(not(feature = "std"))] use core::simd::Simd;
             #[doc = concat!("# use core::", stringify!($ty), "::{MIN, MAX};")]
             /// let x = Simd::from_array([MIN, -2, 3, MAX]);
             /// let unsat = -x;

crates/core_simd/src/reduction.rs

@@ -11,30 +11,30 @@ macro_rules! impl_integer_reductions {
         where
             LaneCount<LANES>: SupportedLaneCount,
         {
-            /// Horizontal wrapping add.  Returns the sum of the lanes of the vector, with wrapping addition.
+            /// Reducing wrapping add.  Returns the sum of the lanes of the vector, with wrapping addition.
             #[inline]
-            pub fn horizontal_sum(self) -> $scalar {
+            pub fn reduce_sum(self) -> $scalar {
                 // Safety: `self` is an integer vector
                 unsafe { simd_reduce_add_ordered(self, 0) }
             }
 
-            /// Horizontal wrapping multiply.  Returns the product of the lanes of the vector, with wrapping multiplication.
+            /// Reducing wrapping multiply.  Returns the product of the lanes of the vector, with wrapping multiplication.
             #[inline]
-            pub fn horizontal_product(self) -> $scalar {
+            pub fn reduce_product(self) -> $scalar {
                 // Safety: `self` is an integer vector
                 unsafe { simd_reduce_mul_ordered(self, 1) }
             }
 
-            /// Horizontal maximum.  Returns the maximum lane in the vector.
+            /// Reducing maximum.  Returns the maximum lane in the vector.
             #[inline]
-            pub fn horizontal_max(self) -> $scalar {
+            pub fn reduce_max(self) -> $scalar {
                 // Safety: `self` is an integer vector
                 unsafe { simd_reduce_max(self) }
             }
 
-            /// Horizontal minimum.  Returns the minimum lane in the vector.
+            /// Reducing minimum.  Returns the minimum lane in the vector.
             #[inline]
-            pub fn horizontal_min(self) -> $scalar {
+            pub fn reduce_min(self) -> $scalar {
                 // Safety: `self` is an integer vector
                 unsafe { simd_reduce_min(self) }
             }
@@ -60,9 +60,9 @@ macro_rules! impl_float_reductions {
             LaneCount<LANES>: SupportedLaneCount,
         {
 
-            /// Horizontal add.  Returns the sum of the lanes of the vector.
+            /// Reducing add.  Returns the sum of the lanes of the vector.
             #[inline]
-            pub fn horizontal_sum(self) -> $scalar {
+            pub fn reduce_sum(self) -> $scalar {
                 // LLVM sum is inaccurate on i586
                 if cfg!(all(target_arch = "x86", not(target_feature = "sse2"))) {
                     self.as_array().iter().sum()
@@ -72,9 +72,9 @@ macro_rules! impl_float_reductions {
                 }
             }
 
-            /// Horizontal multiply.  Returns the product of the lanes of the vector.
+            /// Reducing multiply.  Returns the product of the lanes of the vector.
             #[inline]
-            pub fn horizontal_product(self) -> $scalar {
+            pub fn reduce_product(self) -> $scalar {
                 // LLVM product is inaccurate on i586
                 if cfg!(all(target_arch = "x86", not(target_feature = "sse2"))) {
                     self.as_array().iter().product()
@@ -84,22 +84,22 @@ macro_rules! impl_float_reductions {
                 }
             }
 
-            /// Horizontal maximum.  Returns the maximum lane in the vector.
+            /// Reducing maximum.  Returns the maximum lane in the vector.
             ///
             /// Returns values based on equality, so a vector containing both `0.` and `-0.` may
             /// return either.  This function will not return `NaN` unless all lanes are `NaN`.
             #[inline]
-            pub fn horizontal_max(self) -> $scalar {
+            pub fn reduce_max(self) -> $scalar {
                 // Safety: `self` is a float vector
                 unsafe { simd_reduce_max(self) }
             }
 
-            /// Horizontal minimum.  Returns the minimum lane in the vector.
+            /// Reducing minimum.  Returns the minimum lane in the vector.
             ///
             /// Returns values based on equality, so a vector containing both `0.` and `-0.` may
             /// return either.  This function will not return `NaN` unless all lanes are `NaN`.
             #[inline]
-            pub fn horizontal_min(self) -> $scalar {
+            pub fn reduce_min(self) -> $scalar {
                 // Safety: `self` is a float vector
                 unsafe { simd_reduce_min(self) }
             }
@@ -116,10 +116,10 @@ where
     T: SimdElement + BitAnd<T, Output = T>,
     LaneCount<LANES>: SupportedLaneCount,
 {
-    /// Horizontal bitwise "and".  Returns the cumulative bitwise "and" across the lanes of
+    /// Reducing bitwise "and".  Returns the cumulative bitwise "and" across the lanes of
     /// the vector.
     #[inline]
-    pub fn horizontal_and(self) -> T {
+    pub fn reduce_and(self) -> T {
         unsafe { simd_reduce_and(self) }
     }
 }
@@ -130,10 +130,10 @@ where
     T: SimdElement + BitOr<T, Output = T>,
     LaneCount<LANES>: SupportedLaneCount,
 {
-    /// Horizontal bitwise "or".  Returns the cumulative bitwise "or" across the lanes of
+    /// Reducing bitwise "or".  Returns the cumulative bitwise "or" across the lanes of
     /// the vector.
     #[inline]
-    pub fn horizontal_or(self) -> T {
+    pub fn reduce_or(self) -> T {
         unsafe { simd_reduce_or(self) }
     }
 }
@@ -144,10 +144,10 @@ where
     T: SimdElement + BitXor<T, Output = T>,
     LaneCount<LANES>: SupportedLaneCount,
 {
-    /// Horizontal bitwise "xor".  Returns the cumulative bitwise "xor" across the lanes of
+    /// Reducing bitwise "xor".  Returns the cumulative bitwise "xor" across the lanes of
     /// the vector.
     #[inline]
-    pub fn horizontal_xor(self) -> T {
+    pub fn reduce_xor(self) -> T {
         unsafe { simd_reduce_xor(self) }
     }
 }

crates/core_simd/src/select.rs

@@ -14,8 +14,7 @@ where
     /// # Examples
     /// ```
     /// # #![feature(portable_simd)]
-    /// # #[cfg(feature = "std")] use core_simd::{Simd, Mask};
+    /// # use core::simd::{Simd, Mask};
-    /// # #[cfg(not(feature = "std"))] use core::simd::{Simd, Mask};
     /// let a = Simd::from_array([0, 1, 2, 3]);
     /// let b = Simd::from_array([4, 5, 6, 7]);
     /// let mask = Mask::from_array([true, false, false, true]);
@@ -45,8 +44,7 @@ where
     /// # Examples
     /// ```
     /// # #![feature(portable_simd)]
-    /// # #[cfg(feature = "std")] use core_simd::Mask;
+    /// # use core::simd::Mask;
-    /// # #[cfg(not(feature = "std"))] use core::simd::Mask;
     /// let a = Mask::<i32, 4>::from_array([true, true, false, false]);
     /// let b = Mask::<i32, 4>::from_array([false, false, true, true]);
     /// let mask = Mask::<i32, 4>::from_array([true, false, false, true]);

crates/core_simd/src/swizzle.rs

@@ -12,8 +12,7 @@ use crate::simd::{LaneCount, Simd, SimdElement, SupportedLaneCount};
 /// ## One source vector
 /// ```
 /// # #![feature(portable_simd)]
-/// # #[cfg(feature = "std")] use core_simd::{Simd, simd_swizzle};
+/// # use core::simd::{Simd, simd_swizzle};
-/// # #[cfg(not(feature = "std"))] use core::simd::{Simd, simd_swizzle};
 /// let v = Simd::<f32, 4>::from_array([0., 1., 2., 3.]);
 ///
 /// // Keeping the same size
@@ -28,8 +27,7 @@ use crate::simd::{LaneCount, Simd, SimdElement, SupportedLaneCount};
 /// ## Two source vectors
 /// ```
 /// # #![feature(portable_simd)]
-/// # #[cfg(feature = "std")] use core_simd::{Simd, simd_swizzle, Which};
+/// # use core::simd::{Simd, simd_swizzle, Which};
-/// # #[cfg(not(feature = "std"))] use core::simd::{Simd, simd_swizzle, Which};
 /// use Which::*;
 /// let a = Simd::<f32, 4>::from_array([0., 1., 2., 3.]);
 /// let b = Simd::<f32, 4>::from_array([4., 5., 6., 7.]);
@@ -273,8 +271,7 @@ where
     ///
     /// ```
     /// #![feature(portable_simd)]
-    /// # #[cfg(feature = "std")] use core_simd::Simd;
+    /// # use core::simd::Simd;
-    /// # #[cfg(not(feature = "std"))] use core::simd::Simd;
     /// let a = Simd::from_array([0, 1, 2, 3]);
     /// let b = Simd::from_array([4, 5, 6, 7]);
     /// let (x, y) = a.interleave(b);
@@ -337,8 +334,7 @@ where
     ///
     /// ```
     /// #![feature(portable_simd)]
-    /// # #[cfg(feature = "std")] use core_simd::Simd;
+    /// # use core::simd::Simd;
-    /// # #[cfg(not(feature = "std"))] use core::simd::Simd;
     /// let a = Simd::from_array([0, 4, 1, 5]);
     /// let b = Simd::from_array([2, 6, 3, 7]);
     /// let (x, y) = a.deinterleave(b);

crates/core_simd/src/vector.rs

@@ -153,8 +153,7 @@ where
     /// # Examples
     /// ```
     /// # #![feature(portable_simd)]
-    /// # #[cfg(feature = "std")] use core_simd::Simd;
+    /// # use core::simd::Simd;
-    /// # #[cfg(not(feature = "std"))] use core::simd::Simd;
     /// let floats: Simd<f32, 4> = Simd::from_array([1.9, -4.5, f32::INFINITY, f32::NAN]);
     /// let ints = floats.cast::<i32>();
     /// assert_eq!(ints, Simd::from_array([1, -4, i32::MAX, 0]));
@@ -180,8 +179,7 @@ where
     /// # Examples
     /// ```
     /// # #![feature(portable_simd)]
-    /// # #[cfg(feature = "std")] use core_simd::Simd;
+    /// # use core::simd::Simd;
-    /// # #[cfg(not(feature = "std"))] use core::simd::Simd;
     /// let vec: Vec<i32> = vec![10, 11, 12, 13, 14, 15, 16, 17, 18];
     /// let idxs = Simd::from_array([9, 3, 0, 5]);
     /// let alt = Simd::from_array([-5, -4, -3, -2]);
@@ -201,8 +199,7 @@ where
     /// # Examples
     /// ```
     /// # #![feature(portable_simd)]
-    /// # #[cfg(feature = "std")] use core_simd::Simd;
+    /// # use core::simd::Simd;
-    /// # #[cfg(not(feature = "std"))] use core::simd::Simd;
     /// let vec: Vec<i32> = vec![10, 11, 12, 13, 14, 15, 16, 17, 18];
     /// let idxs = Simd::from_array([9, 3, 0, 5]);
     ///
@@ -225,8 +222,7 @@ where
     /// # Examples
     /// ```
     /// # #![feature(portable_simd)]
-    /// # #[cfg(feature = "std")] use core_simd::{Simd, Mask};
+    /// # use core::simd::{Simd, Mask};
-    /// # #[cfg(not(feature = "std"))] use core::simd::{Simd, Mask};
     /// let vec: Vec<i32> = vec![10, 11, 12, 13, 14, 15, 16, 17, 18];
     /// let idxs = Simd::from_array([9, 3, 0, 5]);
     /// let alt = Simd::from_array([-5, -4, -3, -2]);
@@ -260,8 +256,7 @@ where
     /// # Examples
     /// ```
     /// # #![feature(portable_simd)]
-    /// # #[cfg(feature = "std")] use core_simd::{Simd, Mask};
+    /// # use core::simd::{Simd, Mask};
-    /// # #[cfg(not(feature = "std"))] use core::simd::{Simd, Mask};
     /// let vec: Vec<i32> = vec![10, 11, 12, 13, 14, 15, 16, 17, 18];
     /// let idxs = Simd::from_array([9, 3, 0, 5]);
     /// let alt = Simd::from_array([-5, -4, -3, -2]);
@@ -296,8 +291,7 @@ where
     /// # Examples
     /// ```
     /// # #![feature(portable_simd)]
-    /// # #[cfg(feature = "std")] use core_simd::Simd;
+    /// # use core::simd::Simd;
-    /// # #[cfg(not(feature = "std"))] use core::simd::Simd;
     /// let mut vec: Vec<i32> = vec![10, 11, 12, 13, 14, 15, 16, 17, 18];
     /// let idxs = Simd::from_array([9, 3, 0, 0]);
     /// let vals = Simd::from_array([-27, 82, -41, 124]);
@@ -319,8 +313,7 @@ where
     /// # Examples
     /// ```
     /// # #![feature(portable_simd)]
-    /// # #[cfg(feature = "std")] use core_simd::{Simd, Mask};
+    /// # use core::simd::{Simd, Mask};
-    /// # #[cfg(not(feature = "std"))] use core::simd::{Simd, Mask};
     /// let mut vec: Vec<i32> = vec![10, 11, 12, 13, 14, 15, 16, 17, 18];
     /// let idxs = Simd::from_array([9, 3, 0, 0]);
     /// let vals = Simd::from_array([-27, 82, -41, 124]);
@@ -354,8 +347,7 @@ where
     /// # Examples
     /// ```
     /// # #![feature(portable_simd)]
-    /// # #[cfg(feature = "std")] use core_simd::{Simd, Mask};
+    /// # use core::simd::{Simd, Mask};
-    /// # #[cfg(not(feature = "std"))] use core::simd::{Simd, Mask};
     /// let mut vec: Vec<i32> = vec![10, 11, 12, 13, 14, 15, 16, 17, 18];
     /// let idxs = Simd::from_array([9, 3, 0, 0]);
     /// let vals = Simd::from_array([-27, 82, -41, 124]);

crates/core_simd/tests/i16_ops.rs

@@ -1,5 +1,32 @@
 #![feature(portable_simd)]
+use core_simd::i16x2;
 
 #[macro_use]
 mod ops_macros;
 impl_signed_tests! { i16 }
+
+#[test]
+fn max_is_not_lexicographic() {
+    let a = i16x2::splat(10);
+    let b = i16x2::from_array([-4, 12]);
+    assert_eq!(a.max(b), i16x2::from_array([10, 12]));
+}
+
+#[test]
+fn min_is_not_lexicographic() {
+    let a = i16x2::splat(10);
+    let b = i16x2::from_array([12, -4]);
+    assert_eq!(a.min(b), i16x2::from_array([10, -4]));
+}
+
+#[test]
+fn clamp_is_not_lexicographic() {
+    let a = i16x2::splat(10);
+    let lo = i16x2::from_array([-12, -4]);
+    let up = i16x2::from_array([-4, 12]);
+    assert_eq!(a.clamp(lo, up), i16x2::from_array([-4, 10]));
+
+    let x = i16x2::from_array([1, 10]);
+    let y = x.clamp(i16x2::splat(0), i16x2::splat(9));
+    assert_eq!(y, i16x2::from_array([1, 9]));
+}

crates/core_simd/tests/ops_macros.rs

@@ -94,70 +94,70 @@ macro_rules! impl_binary_checked_op_test {
 macro_rules! impl_common_integer_tests {
     { $vector:ident, $scalar:ident } => {
         test_helpers::test_lanes! {
-            fn horizontal_sum<const LANES: usize>() {
+            fn reduce_sum<const LANES: usize>() {
                 test_helpers::test_1(&|x| {
                     test_helpers::prop_assert_biteq! (
-                        $vector::<LANES>::from_array(x).horizontal_sum(),
+                        $vector::<LANES>::from_array(x).reduce_sum(),
                         x.iter().copied().fold(0 as $scalar, $scalar::wrapping_add),
                     );
                     Ok(())
                 });
             }
 
-            fn horizontal_product<const LANES: usize>() {
+            fn reduce_product<const LANES: usize>() {
                 test_helpers::test_1(&|x| {
                     test_helpers::prop_assert_biteq! (
-                        $vector::<LANES>::from_array(x).horizontal_product(),
+                        $vector::<LANES>::from_array(x).reduce_product(),
                         x.iter().copied().fold(1 as $scalar, $scalar::wrapping_mul),
                     );
                     Ok(())
                 });
             }
 
-            fn horizontal_and<const LANES: usize>() {
+            fn reduce_and<const LANES: usize>() {
                 test_helpers::test_1(&|x| {
                     test_helpers::prop_assert_biteq! (
-                        $vector::<LANES>::from_array(x).horizontal_and(),
+                        $vector::<LANES>::from_array(x).reduce_and(),
                         x.iter().copied().fold(-1i8 as $scalar, <$scalar as core::ops::BitAnd>::bitand),
                     );
                     Ok(())
                 });
             }
 
-            fn horizontal_or<const LANES: usize>() {
+            fn reduce_or<const LANES: usize>() {
                 test_helpers::test_1(&|x| {
                     test_helpers::prop_assert_biteq! (
-                        $vector::<LANES>::from_array(x).horizontal_or(),
+                        $vector::<LANES>::from_array(x).reduce_or(),
                         x.iter().copied().fold(0 as $scalar, <$scalar as core::ops::BitOr>::bitor),
                     );
                     Ok(())
                 });
             }
 
-            fn horizontal_xor<const LANES: usize>() {
+            fn reduce_xor<const LANES: usize>() {
                 test_helpers::test_1(&|x| {
                     test_helpers::prop_assert_biteq! (
-                        $vector::<LANES>::from_array(x).horizontal_xor(),
+                        $vector::<LANES>::from_array(x).reduce_xor(),
                         x.iter().copied().fold(0 as $scalar, <$scalar as core::ops::BitXor>::bitxor),
                     );
                     Ok(())
                 });
             }
 
-            fn horizontal_max<const LANES: usize>() {
+            fn reduce_max<const LANES: usize>() {
                 test_helpers::test_1(&|x| {
                     test_helpers::prop_assert_biteq! (
-                        $vector::<LANES>::from_array(x).horizontal_max(),
+                        $vector::<LANES>::from_array(x).reduce_max(),
                         x.iter().copied().max().unwrap(),
                     );
                     Ok(())
                 });
             }
 
-            fn horizontal_min<const LANES: usize>() {
+            fn reduce_min<const LANES: usize>() {
                 test_helpers::test_1(&|x| {
                     test_helpers::prop_assert_biteq! (
-                        $vector::<LANES>::from_array(x).horizontal_min(),
+                        $vector::<LANES>::from_array(x).reduce_min(),
                         x.iter().copied().min().unwrap(),
                     );
                     Ok(())
@@ -222,6 +222,35 @@ macro_rules! impl_signed_tests {
                     assert_eq!(a % b, Vector::<LANES>::splat(0));
                 }
 
+                fn min<const LANES: usize>() {
+                    let a = Vector::<LANES>::splat(Scalar::MIN);
+                    let b = Vector::<LANES>::splat(0);
+                    assert_eq!(a.min(b), a);
+                    let a = Vector::<LANES>::splat(Scalar::MAX);
+                    let b = Vector::<LANES>::splat(0);
+                    assert_eq!(a.min(b), b);
+                }
+
+                fn max<const LANES: usize>() {
+                    let a = Vector::<LANES>::splat(Scalar::MIN);
+                    let b = Vector::<LANES>::splat(0);
+                    assert_eq!(a.max(b), b);
+                    let a = Vector::<LANES>::splat(Scalar::MAX);
+                    let b = Vector::<LANES>::splat(0);
+                    assert_eq!(a.max(b), a);
+                }
+
+                fn clamp<const LANES: usize>() {
+                    let min = Vector::<LANES>::splat(Scalar::MIN);
+                    let max = Vector::<LANES>::splat(Scalar::MAX);
+                    let zero = Vector::<LANES>::splat(0);
+                    let one = Vector::<LANES>::splat(1);
+                    let negone = Vector::<LANES>::splat(-1);
+                    assert_eq!(zero.clamp(min, max), zero);
+                    assert_eq!(zero.clamp(min, one), zero);
+                    assert_eq!(zero.clamp(one, max), one);
+                    assert_eq!(zero.clamp(min, negone), negone);
+                }
             }
 
             test_helpers::test_lanes_panic! {
@@ -499,29 +528,29 @@ macro_rules! impl_float_tests {
                     })
                 }
 
-                fn horizontal_sum<const LANES: usize>() {
+                fn reduce_sum<const LANES: usize>() {
                     test_helpers::test_1(&|x| {
                         test_helpers::prop_assert_biteq! (
-                            Vector::<LANES>::from_array(x).horizontal_sum(),
+                            Vector::<LANES>::from_array(x).reduce_sum(),
                             x.iter().sum(),
                         );
                         Ok(())
                     });
                 }
 
-                fn horizontal_product<const LANES: usize>() {
+                fn reduce_product<const LANES: usize>() {
                     test_helpers::test_1(&|x| {
                         test_helpers::prop_assert_biteq! (
-                            Vector::<LANES>::from_array(x).horizontal_product(),
+                            Vector::<LANES>::from_array(x).reduce_product(),
                             x.iter().product(),
                         );
                         Ok(())
                     });
                 }
 
-                fn horizontal_max<const LANES: usize>() {
+                fn reduce_max<const LANES: usize>() {
                     test_helpers::test_1(&|x| {
-                        let vmax = Vector::<LANES>::from_array(x).horizontal_max();
+                        let vmax = Vector::<LANES>::from_array(x).reduce_max();
                         let smax = x.iter().copied().fold(Scalar::NAN, Scalar::max);
                         // 0 and -0 are treated the same
                         if !(x.contains(&0.) && x.contains(&-0.) && vmax.abs() == 0. && smax.abs() == 0.) {
@@ -531,9 +560,9 @@ macro_rules! impl_float_tests {
                     });
                 }
 
-                fn horizontal_min<const LANES: usize>() {
+                fn reduce_min<const LANES: usize>() {
                     test_helpers::test_1(&|x| {
-                        let vmax = Vector::<LANES>::from_array(x).horizontal_min();
+                        let vmax = Vector::<LANES>::from_array(x).reduce_min();
                         let smax = x.iter().copied().fold(Scalar::NAN, Scalar::min);
                         // 0 and -0 are treated the same
                         if !(x.contains(&0.) && x.contains(&-0.) && vmax.abs() == 0. && smax.abs() == 0.) {

crates/core_simd/tests/round.rs

@@ -9,7 +9,6 @@ macro_rules! float_rounding_test {
             type Scalar = $scalar;
             type IntScalar = $int_scalar;
 
-            #[cfg(feature = "std")]
             test_helpers::test_lanes! {
                 fn ceil<const LANES: usize>() {
                     test_helpers::test_unary_elementwise(

crates/std_float/Cargo.toml

@@ -6,7 +6,7 @@ edition = "2021"
 # See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html
 
 [dependencies]
-core_simd = { path = "../core_simd" }
+core_simd = { path = "../core_simd", default-features = false }
 
 [features]
 default = ["as_crate"]

crates/test_helpers/src/lib.rs

@@ -77,11 +77,21 @@ impl<T: core::fmt::Debug + DefaultStrategy, const LANES: usize> DefaultStrategy
     }
 }
 
+#[cfg(not(miri))]
+fn make_runner() -> proptest::test_runner::TestRunner {
+    Default::default()
+}
+#[cfg(miri)]
+fn make_runner() -> proptest::test_runner::TestRunner {
+    // Only run a few tests on Miri
+    proptest::test_runner::TestRunner::new(proptest::test_runner::Config::with_cases(4))
+}
+
 /// Test a function that takes a single value.
 pub fn test_1<A: core::fmt::Debug + DefaultStrategy>(
     f: &dyn Fn(A) -> proptest::test_runner::TestCaseResult,
 ) {
-    let mut runner = proptest::test_runner::TestRunner::default();
+    let mut runner = make_runner();
     runner.run(&A::default_strategy(), f).unwrap();
 }
 
@@ -89,7 +99,7 @@ pub fn test_1<A: core::fmt::Debug + DefaultStrategy>(
 pub fn test_2<A: core::fmt::Debug + DefaultStrategy, B: core::fmt::Debug + DefaultStrategy>(
     f: &dyn Fn(A, B) -> proptest::test_runner::TestCaseResult,
 ) {
-    let mut runner = proptest::test_runner::TestRunner::default();
+    let mut runner = make_runner();
     runner
         .run(&(A::default_strategy(), B::default_strategy()), |(a, b)| {
             f(a, b)
@@ -105,7 +115,7 @@ pub fn test_3<
 >(
     f: &dyn Fn(A, B, C) -> proptest::test_runner::TestCaseResult,
 ) {
-    let mut runner = proptest::test_runner::TestRunner::default();
+    let mut runner = make_runner();
     runner
         .run(
             &(
@@ -361,24 +371,28 @@ macro_rules! test_lanes {
 
                 #[test]
                 #[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
+                #[cfg(not(miri))] // Miri intrinsic implementations are uniform and larger tests are sloooow
                 fn lanes_8() {
                     implementation::<8>();
                 }
 
                 #[test]
                 #[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
+                #[cfg(not(miri))] // Miri intrinsic implementations are uniform and larger tests are sloooow
                 fn lanes_16() {
                     implementation::<16>();
                 }
 
                 #[test]
                 #[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
+                #[cfg(not(miri))] // Miri intrinsic implementations are uniform and larger tests are sloooow
                 fn lanes_32() {
                     implementation::<32>();
                 }
 
                 #[test]
                 #[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
+                #[cfg(not(miri))] // Miri intrinsic implementations are uniform and larger tests are sloooow
                 fn lanes_64() {
                     implementation::<64>();
                 }