Auto merge of #93146 - workingjubilee:use-std-simd, r=Mark-Simulacrum

pub use std::simd::StdFloat;

Syncs portable-simd up to commit rust-lang/portable-simd@03f6fbb,
Diff: rust-lang/portable-simd@533f0fc...03f6fbb

This sync requires a little bit more legwork because it also introduces a trait into `std::simd`, so that it is no longer simply a reexport of `core::simd`. Out of simple-minded consistency and to allow more options, I replicated the pattern for the way `core::simd` is integrated in the first place, however this is not necessary if it doesn't acquire any interdependencies inside `std`: it could be a simple crate reexport. I just don't know yet if that will happen or not.

To summarize other misc changes:
- Shifts no longer panic, now wrap on too-large shifts (like `Simd` integers usually do!)
- mask16x32 will now be many i16s, not many i32s... 🙃
- `#[must_use]` is spread around generously
- Adjusts division, float min/max, and `Mask::{from,to}_array` internally to be faster
- Adds the much-requested `Simd::cast::<U>` function (equivalent to `simd.to_array().map(|lane| lane as U)`)

Author: bors

library/portable-simd/Cargo.toml

@@ -2,5 +2,6 @@
 
 members = [
     "crates/core_simd",
+    "crates/std_float",
     "crates/test_helpers",
 ]

library/portable-simd/crates/core_simd/Cargo.toml

@@ -26,3 +26,6 @@ features = ["alloc"]
 
 [dev-dependencies.test_helpers]
 path = "../test_helpers"
+
+[dev-dependencies]
+std_float = { path = "../std_float/", features = ["as_crate"] }

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

@@ -1,11 +1,13 @@
-#![cfg_attr(feature = "std", feature(portable_simd))]
+#![feature(portable_simd)]
+extern crate std_float;
 
 /// Benchmarks game nbody code
 /// Taken from the `packed_simd` crate
 /// Run this benchmark with `cargo test --example nbody`
-#[cfg(feature = "std")]
 mod nbody {
-    use core_simd::*;
+    use core_simd::simd::*;
+    #[allow(unused)] // False positive?
+    use std_float::StdFloat;
 
     use std::f64::consts::PI;
     const SOLAR_MASS: f64 = 4.0 * PI * PI;
@@ -167,7 +169,6 @@ mod nbody {
     }
 }
 
-#[cfg(feature = "std")]
 #[cfg(test)]
 mod tests {
     // Good enough for demonstration purposes, not going for strictness here.
@@ -184,7 +185,6 @@ mod tests {
 }
 
 fn main() {
-    #[cfg(feature = "std")]
     {
         let (energy_before, energy_after) = nbody::run(1000);
         println!("Energy before: {}", energy_before);

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

@@ -39,13 +39,21 @@ extern "platform-intrinsic" {
 
     /// fptoui/fptosi/uitofp/sitofp
     pub(crate) fn simd_cast<T, U>(x: T) -> U;
+    /// follows Rust's `T as U` semantics, including saturating float casts
+    /// which amounts to the same as `simd_cast` for many cases
+    #[cfg(not(bootstrap))]
+    pub(crate) fn simd_as<T, U>(x: T) -> U;
 
     /// neg/fneg
     pub(crate) fn simd_neg<T>(x: T) -> T;
 
     /// fabs
     pub(crate) fn simd_fabs<T>(x: T) -> T;
 
+    // minnum/maxnum
+    pub(crate) fn simd_fmin<T>(x: T, y: T) -> T;
+    pub(crate) fn simd_fmax<T>(x: T, y: T) -> T;
+
     pub(crate) fn simd_eq<T, U>(x: T, y: T) -> U;
     pub(crate) fn simd_ne<T, U>(x: T, y: T) -> U;
     pub(crate) fn simd_lt<T, U>(x: T, y: T) -> U;
@@ -87,29 +95,3 @@ extern "platform-intrinsic" {
     #[allow(unused)]
     pub(crate) fn simd_select_bitmask<M, T>(m: M, a: T, b: T) -> T;
 }
-
-#[cfg(feature = "std")]
-mod std {
-    extern "platform-intrinsic" {
-        // ceil
-        pub(crate) fn simd_ceil<T>(x: T) -> T;
-
-        // floor
-        pub(crate) fn simd_floor<T>(x: T) -> T;
-
-        // round
-        pub(crate) fn simd_round<T>(x: T) -> T;
-
-        // trunc
-        pub(crate) fn simd_trunc<T>(x: T) -> T;
-
-        // fsqrt
-        pub(crate) fn simd_fsqrt<T>(x: T) -> T;
-
-        // fma
-        pub(crate) fn simd_fma<T>(x: T, y: T, z: T) -> T;
-    }
-}
-
-#[cfg(feature = "std")]
-pub(crate) use crate::simd::intrinsics::std::*;

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

@@ -12,9 +12,10 @@
 )]
 mod mask_impl;
 
+use crate::simd::intrinsics;
 use crate::simd::{LaneCount, Simd, SimdElement, SupportedLaneCount};
 use core::cmp::Ordering;
-use core::fmt;
+use core::{fmt, mem};
 
 mod sealed {
     use super::*;
@@ -105,22 +106,39 @@ where
         Self(mask_impl::Mask::splat(value))
     }
 
-    /// Converts an array to a SIMD vector.
+    /// Converts an array of bools to a SIMD mask.
     pub fn from_array(array: [bool; LANES]) -> Self {
-        let mut vector = Self::splat(false);
-        for (i, v) in array.iter().enumerate() {
-            vector.set(i, *v);
+        // SAFETY: Rust's bool has a layout of 1 byte (u8) with a value of
+        //     true:    0b_0000_0001
+        //     false:   0b_0000_0000
+        // Thus, an array of bools is also a valid array of bytes: [u8; N]
+        // This would be hypothetically valid as an "in-place" transmute,
+        // but these are "dependently-sized" types, so copy elision it is!
+        unsafe {
+            let bytes: [u8; LANES] = mem::transmute_copy(&array);
+            let bools: Simd<i8, LANES> =
+                intrinsics::simd_ne(Simd::from_array(bytes), Simd::splat(0u8));
+            Mask::from_int_unchecked(intrinsics::simd_cast(bools))
         }
-        vector
     }
 
-    /// Converts a SIMD vector to an array.
+    /// Converts a SIMD mask to an array of bools.
     pub fn to_array(self) -> [bool; LANES] {
-        let mut array = [false; LANES];
-        for (i, v) in array.iter_mut().enumerate() {
-            *v = self.test(i);
+        // This follows mostly the same logic as from_array.
+        // SAFETY: Rust's bool has a layout of 1 byte (u8) with a value of
+        //     true:    0b_0000_0001
+        //     false:   0b_0000_0000
+        // Thus, an array of bools is also a valid array of bytes: [u8; N]
+        // Since our masks are equal to integers where all bits are set,
+        // we can simply convert them to i8s, and then bitand them by the
+        // bitpattern for Rust's "true" bool.
+        // This would be hypothetically valid as an "in-place" transmute,
+        // but these are "dependently-sized" types, so copy elision it is!
+        unsafe {
+            let mut bytes: Simd<i8, LANES> = intrinsics::simd_cast(self.to_int());
+            bytes &= Simd::splat(1i8);
+            mem::transmute_copy(&bytes)
         }
-        array
     }
 
     /// Converts a vector of integers to a mask, where 0 represents `false` and -1
@@ -516,7 +534,7 @@ pub type mask16x8 = Mask<i16, 8>;
 pub type mask16x16 = Mask<i16, 16>;
 
 /// Vector of 32 16-bit masks
-pub type mask16x32 = Mask<i32, 32>;
+pub type mask16x32 = Mask<i16, 32>;
 
 /// Vector of two 32-bit masks
 pub type mask32x2 = Mask<i32, 2>;