Auto merge of #91484 - workingjubilee:simd-remove-autosplats, r=Mark-…

…Simulacrum

Sync portable-simd to remove autosplats

This PR syncs portable-simd in up to rust-lang/portable-simd@a838552 in order to address the type inference breakages documented on nightly in #90904 by removing the vector + scalar binary operations (called "autosplats", "broadcasting", or "rank promotion", depending on who you ask) that allow `{scalar} + &'_ {scalar}` to fail in some cases, because it becomes possible the programmer may have meant `{scalar} + &'_ {vector}`.

A few quality-of-life improvements make their way in as well:
- Lane counts can now go to 64, as LLVM seems to have fixed their miscompilation for those.
- `{i,u}8x64` to `__m512i` is now available.
- a bunch of `#[must_use]` notes appear throughout the module.
- Some implementations, mostly instances of `impl core::ops::{Op}<Simd> for Simd` that aren't `{vector} + {vector}` (e.g. `{vector} + &'_ {vector}`), leverage some generics and `where` bounds now to make them easier to understand by reducing a dozen implementations into one (and make it possible for people to open the docs on less burly devices).
- And some internal-only improvements.

None of these changes should affect a beta backport, only actual users of `core::simd` (and most aren't even visible in the programmatic sense), though I can extract an even more minimal changeset for beta if necessary. It seemed simpler to just keep moving forward.

library/core/tests/simd.rs

@@ -7,7 +7,7 @@ fn testing() {
     let x = f32x4::from_array([1.0, 1.0, 1.0, 1.0]);
     let y = -x;
 
-    let h = x * 0.5;
+    let h = x * f32x4::splat(0.5);
 
     let r = y.abs();
     assert_eq!(x, r);

library/portable-simd/CONTRIBUTING.md

@@ -15,7 +15,7 @@ SIMD can be quite complex, and even a "simple" issue can be huge. If an issue is
 
 ## CI
 
-We currently have 2 CI matrices through Travis CI and GitHub Actions that will automatically build and test your change in order to verify that `std::simd`'s portable API is, in fact, portable. If your change builds locally, but does not build on either, this is likely due to a platform-specific concern that your code has not addressed. Please consult the build logs and address the error, or ask for help if you need it.
+We currently use GitHub Actions which will automatically build and test your change in order to verify that `std::simd`'s portable API is, in fact, portable. If your change builds locally, but does not build in CI, this is likely due to a platform-specific concern that your code has not addressed. Please consult the build logs and address the error, or ask for help if you need it.
 
 ## Beyond stdsimd
 

library/portable-simd/README.md

@@ -1,5 +1,5 @@
 # The Rust standard library's portable SIMD API
-[![Build Status](https://travis-ci.com/rust-lang/portable-simd.svg?branch=master)](https://travis-ci.com/rust-lang/portable-simd)
+![Build Status](https://github.com/rust-lang/portable-simd/actions/workflows/ci.yml/badge.svg?branch=master)
 
 Code repository for the [Portable SIMD Project Group](https://github.com/rust-lang/project-portable-simd).
 Please refer to [CONTRIBUTING.md](./CONTRIBUTING.md) for our contributing guidelines.

library/portable-simd/crates/core_simd/examples/nbody.rs

@@ -97,7 +97,7 @@ mod nbody {
         let sun = &mut sun[0];
         for body in rest {
             let m_ratio = body.mass / SOLAR_MASS;
-            sun.v -= body.v * m_ratio;
+            sun.v -= body.v * Simd::splat(m_ratio);
         }
     }
 
@@ -143,14 +143,14 @@ mod nbody {
         let mut i = 0;
         for j in 0..N_BODIES {
             for k in j + 1..N_BODIES {
-                let f = r[i] * mag[i];
-                bodies[j].v -= f * bodies[k].mass;
-                bodies[k].v += f * bodies[j].mass;
+                let f = r[i] * Simd::splat(mag[i]);
+                bodies[j].v -= f * Simd::splat(bodies[k].mass);
+                bodies[k].v += f * Simd::splat(bodies[j].mass);
                 i += 1
             }
         }
         for body in bodies {
-            body.x += dt * body.v
+            body.x += Simd::splat(dt) * body.v
         }
     }
 

library/portable-simd/crates/core_simd/src/comparisons.rs

@@ -8,12 +8,14 @@ where
 {
     /// Test if each lane is equal to the corresponding lane in `other`.
     #[inline]
+    #[must_use = "method returns a new mask and does not mutate the original value"]
     pub fn lanes_eq(self, other: Self) -> Mask<T::Mask, LANES> {
         unsafe { Mask::from_int_unchecked(intrinsics::simd_eq(self, other)) }
     }
 
     /// Test if each lane is not equal to the corresponding lane in `other`.
     #[inline]
+    #[must_use = "method returns a new mask and does not mutate the original value"]
     pub fn lanes_ne(self, other: Self) -> Mask<T::Mask, LANES> {
         unsafe { Mask::from_int_unchecked(intrinsics::simd_ne(self, other)) }
     }
@@ -26,24 +28,28 @@ where
 {
     /// Test if each lane is less than the corresponding lane in `other`.
     #[inline]
+    #[must_use = "method returns a new mask and does not mutate the original value"]
     pub fn lanes_lt(self, other: Self) -> Mask<T::Mask, LANES> {
         unsafe { Mask::from_int_unchecked(intrinsics::simd_lt(self, other)) }
     }
 
     /// Test if each lane is greater than the corresponding lane in `other`.
     #[inline]
+    #[must_use = "method returns a new mask and does not mutate the original value"]
     pub fn lanes_gt(self, other: Self) -> Mask<T::Mask, LANES> {
         unsafe { Mask::from_int_unchecked(intrinsics::simd_gt(self, other)) }
     }
 
     /// Test if each lane is less than or equal to the corresponding lane in `other`.
     #[inline]
+    #[must_use = "method returns a new mask and does not mutate the original value"]
     pub fn lanes_le(self, other: Self) -> Mask<T::Mask, LANES> {
         unsafe { Mask::from_int_unchecked(intrinsics::simd_le(self, other)) }
     }
 
     /// Test if each lane is greater than or equal to the corresponding lane in `other`.
     #[inline]
+    #[must_use = "method returns a new mask and does not mutate the original value"]
     pub fn lanes_ge(self, other: Self) -> Mask<T::Mask, LANES> {
         unsafe { Mask::from_int_unchecked(intrinsics::simd_ge(self, other)) }
     }

library/portable-simd/crates/core_simd/src/lane_count.rs

@@ -15,34 +15,28 @@ impl<const LANES: usize> LaneCount<LANES> {
 pub trait SupportedLaneCount: Sealed {
     #[doc(hidden)]
     type BitMask: Copy + Default + AsRef<[u8]> + AsMut<[u8]>;
-
-    #[doc(hidden)]
-    type IntBitMask;
 }
 
 impl<const LANES: usize> Sealed for LaneCount<LANES> {}
 
 impl SupportedLaneCount for LaneCount<1> {
     type BitMask = [u8; 1];
-    type IntBitMask = u8;
 }
 impl SupportedLaneCount for LaneCount<2> {
     type BitMask = [u8; 1];
-    type IntBitMask = u8;
 }
 impl SupportedLaneCount for LaneCount<4> {
     type BitMask = [u8; 1];
-    type IntBitMask = u8;
 }
 impl SupportedLaneCount for LaneCount<8> {
     type BitMask = [u8; 1];
-    type IntBitMask = u8;
 }
 impl SupportedLaneCount for LaneCount<16> {
     type BitMask = [u8; 2];
-    type IntBitMask = u16;
 }
 impl SupportedLaneCount for LaneCount<32> {
     type BitMask = [u8; 4];
-    type IntBitMask = u32;
+}
+impl SupportedLaneCount for LaneCount<64> {
+    type BitMask = [u8; 8];
 }

library/portable-simd/crates/core_simd/src/masks.rs

@@ -129,6 +129,7 @@ where
     /// # Safety
     /// All lanes must be either 0 or -1.
     #[inline]
+    #[must_use = "method returns a new mask and does not mutate the original value"]
     pub unsafe fn from_int_unchecked(value: Simd<T, LANES>) -> Self {
         unsafe { Self(mask_impl::Mask::from_int_unchecked(value)) }
     }
@@ -139,6 +140,7 @@ where
     /// # Panics
     /// Panics if any lane is not 0 or -1.
     #[inline]
+    #[must_use = "method returns a new mask and does not mutate the original value"]
     pub fn from_int(value: Simd<T, LANES>) -> Self {
         assert!(T::valid(value), "all values must be either 0 or -1",);
         unsafe { Self::from_int_unchecked(value) }
@@ -147,6 +149,7 @@ where
     /// Converts the mask to a vector of integers, where 0 represents `false` and -1
     /// represents `true`.
     #[inline]
+    #[must_use = "method returns a new vector and does not mutate the original value"]
     pub fn to_int(self) -> Simd<T, LANES> {
         self.0.to_int()
     }
@@ -156,6 +159,7 @@ where
     /// # Safety
     /// `lane` must be less than `LANES`.
     #[inline]
+    #[must_use = "method returns a new bool and does not mutate the original value"]
     pub unsafe fn test_unchecked(&self, lane: usize) -> bool {
         unsafe { self.0.test_unchecked(lane) }
     }
@@ -165,6 +169,7 @@ where
     /// # Panics
     /// Panics if `lane` is greater than or equal to the number of lanes in the vector.
     #[inline]
+    #[must_use = "method returns a new bool and does not mutate the original value"]
     pub fn test(&self, lane: usize) -> bool {
         assert!(lane < LANES, "lane index out of range");
         unsafe { self.test_unchecked(lane) }
@@ -195,24 +200,30 @@ where
 
     /// Convert this mask to a bitmask, with one bit set per lane.
     #[cfg(feature = "generic_const_exprs")]
+    #[inline]
+    #[must_use = "method returns a new array and does not mutate the original value"]
     pub fn to_bitmask(self) -> [u8; LaneCount::<LANES>::BITMASK_LEN] {
         self.0.to_bitmask()
     }
 
     /// Convert a bitmask to a mask.
     #[cfg(feature = "generic_const_exprs")]
+    #[inline]
+    #[must_use = "method returns a new mask and does not mutate the original value"]
     pub fn from_bitmask(bitmask: [u8; LaneCount::<LANES>::BITMASK_LEN]) -> Self {
         Self(mask_impl::Mask::from_bitmask(bitmask))
     }
 
     /// Returns true if any lane is set, or false otherwise.
     #[inline]
+    #[must_use = "method returns a new bool and does not mutate the original value"]
     pub fn any(self) -> bool {
         self.0.any()
     }
 
     /// Returns true if all lanes are set, or false otherwise.
     #[inline]
+    #[must_use = "method returns a new bool and does not mutate the original value"]
     pub fn all(self) -> bool {
         self.0.all()
     }
@@ -245,6 +256,7 @@ where
     LaneCount<LANES>: SupportedLaneCount,
 {
     #[inline]
+    #[must_use = "method returns a defaulted mask with all lanes set to false (0)"]
     fn default() -> Self {
         Self::splat(false)
     }
@@ -256,6 +268,7 @@ where
     LaneCount<LANES>: SupportedLaneCount,
 {
     #[inline]
+    #[must_use = "method returns a new bool and does not mutate the original value"]
     fn eq(&self, other: &Self) -> bool {
         self.0 == other.0
     }
@@ -267,6 +280,7 @@ where
     LaneCount<LANES>: SupportedLaneCount,
 {
     #[inline]
+    #[must_use = "method returns a new Ordering and does not mutate the original value"]
     fn partial_cmp(&self, other: &Self) -> Option<Ordering> {
         self.0.partial_cmp(&other.0)
     }
@@ -291,6 +305,7 @@ where
 {
     type Output = Self;
     #[inline]
+    #[must_use = "method returns a new mask and does not mutate the original value"]
     fn bitand(self, rhs: Self) -> Self {
         Self(self.0 & rhs.0)
     }
@@ -303,6 +318,7 @@ where
 {
     type Output = Self;
     #[inline]
+    #[must_use = "method returns a new mask and does not mutate the original value"]
     fn bitand(self, rhs: bool) -> Self {
         self & Self::splat(rhs)
     }
@@ -315,6 +331,7 @@ where
 {
     type Output = Mask<T, LANES>;
     #[inline]
+    #[must_use = "method returns a new mask and does not mutate the original value"]
     fn bitand(self, rhs: Mask<T, LANES>) -> Mask<T, LANES> {
         Mask::splat(self) & rhs
     }
@@ -327,6 +344,7 @@ where
 {
     type Output = Self;
     #[inline]
+    #[must_use = "method returns a new mask and does not mutate the original value"]
     fn bitor(self, rhs: Self) -> Self {
         Self(self.0 | rhs.0)
     }
@@ -339,6 +357,7 @@ where
 {
     type Output = Self;
     #[inline]
+    #[must_use = "method returns a new mask and does not mutate the original value"]
     fn bitor(self, rhs: bool) -> Self {
         self | Self::splat(rhs)
     }
@@ -351,6 +370,7 @@ where
 {
     type Output = Mask<T, LANES>;
     #[inline]
+    #[must_use = "method returns a new mask and does not mutate the original value"]
     fn bitor(self, rhs: Mask<T, LANES>) -> Mask<T, LANES> {
         Mask::splat(self) | rhs
     }
@@ -363,6 +383,7 @@ where
 {
     type Output = Self;
     #[inline]
+    #[must_use = "method returns a new mask and does not mutate the original value"]
     fn bitxor(self, rhs: Self) -> Self::Output {
         Self(self.0 ^ rhs.0)
     }
@@ -375,6 +396,7 @@ where
 {
     type Output = Self;
     #[inline]
+    #[must_use = "method returns a new mask and does not mutate the original value"]
     fn bitxor(self, rhs: bool) -> Self::Output {
         self ^ Self::splat(rhs)
     }
@@ -387,6 +409,7 @@ where
 {
     type Output = Mask<T, LANES>;
     #[inline]
+    #[must_use = "method returns a new mask and does not mutate the original value"]
     fn bitxor(self, rhs: Mask<T, LANES>) -> Self::Output {
         Mask::splat(self) ^ rhs
     }
@@ -399,6 +422,7 @@ where
 {
     type Output = Mask<T, LANES>;
     #[inline]
+    #[must_use = "method returns a new mask and does not mutate the original value"]
     fn not(self) -> Self::Output {
         Self(!self.0)
     }