Merge branch 'main' into u8widen

.github/workflows/rust.yml

@@ -11,33 +11,36 @@ jobs:
       matrix:
         rust:
         # x86 without sse/sse2 on by default
-        - { target: i586-pc-windows-msvc, toolchain: 1.56.0, os: windows-latest }
+        - { target: i586-pc-windows-msvc, toolchain: "1.61", os: windows-latest }
         - { target: i586-pc-windows-msvc, toolchain: stable, os: windows-latest }
         - { target: i586-pc-windows-msvc, toolchain: beta, os: windows-latest }
         - { target: i586-pc-windows-msvc, toolchain: nightly, os: windows-latest }
         # x86
-        - { target: i686-pc-windows-msvc, toolchain: 1.56.0, os: windows-latest }
+        - { target: i686-pc-windows-msvc, toolchain: "1.61", os: windows-latest }
         - { target: i686-pc-windows-msvc, toolchain: stable, os: windows-latest }
         - { target: i686-pc-windows-msvc, toolchain: beta, os: windows-latest }
         - { target: i686-pc-windows-msvc, toolchain: nightly, os: windows-latest }
         # x86_64
-        - { target: x86_64-pc-windows-msvc, toolchain: 1.56.0, os: windows-latest }
+        - { target: x86_64-pc-windows-msvc, toolchain: "1.61", os: windows-latest }
         - { target: x86_64-pc-windows-msvc, toolchain: stable, os: windows-latest }
         - { target: x86_64-pc-windows-msvc, toolchain: beta, os: windows-latest }
         - { target: x86_64-pc-windows-msvc, toolchain: nightly, os: windows-latest }
+        # aarch64
+        - { target: aarch64-apple-darwin, toolchain: "1.61", os: macos-latest }
+        - { target: aarch64-apple-darwin, toolchain: stable, os: macos-latest }
+        - { target: aarch64-apple-darwin, toolchain: beta, os: macos-latest }
+        - { target: aarch64-apple-darwin, toolchain: nightly, os: macos-latest }
         # wasm32
-        - { target: wasm32-wasi, toolchain: 1.56.0, os: ubuntu-latest, wasmtime: v5.0.0 }
+        - { target: wasm32-wasi, toolchain: "1.61", os: ubuntu-latest, wasmtime: v5.0.0 }
         - { target: wasm32-wasi, toolchain: stable, os: ubuntu-latest, wasmtime: v5.0.0 }
         - { target: wasm32-wasi, toolchain: beta, os: ubuntu-latest, wasmtime: v5.0.0 }
         - { target: wasm32-wasi, toolchain: nightly, os: ubuntu-latest, wasmtime: v5.0.0 }
     steps:
-    - uses: actions/checkout@v1
-    - uses: actions-rs/toolchain@v1
+    - uses: actions/checkout@v4
+    - uses: dtolnay/rust-toolchain@master
       with:
         toolchain: ${{ matrix.rust.toolchain }}
         target:  ${{ matrix.rust.target }}
-        profile: minimal
-        default: true
 
     - name: Install wasmtime
       if: matrix.rust.target == 'wasm32-wasi'
@@ -49,12 +52,10 @@ jobs:
       run: cargo build --target ${{ matrix.rust.target }}
 
     - name: Test with default CPU features + No Default Cargo Features
-      if: matrix.rust.target == 'i586-pc-windows-msvc' || matrix.rust.target == 'i686-pc-windows-msvc' || matrix.rust.target == 'x86_64-pc-windows-msvc' || matrix.rust.target == 'wasm32-wasi'
       env:
         CARGO_TARGET_WASM32_WASI_RUNNER: wasmtime run --wasm-features all --dir .
       run: cargo test --target ${{ matrix.rust.target }} --no-default-features
     - name: Test with default CPU features + All Cargo Features
-      if: matrix.rust.target == 'i586-pc-windows-msvc' || matrix.rust.target == 'i686-pc-windows-msvc' || matrix.rust.target == 'x86_64-pc-windows-msvc' || matrix.rust.target == 'wasm32-wasi'
       env:
         CARGO_TARGET_WASM32_WASI_RUNNER: wasmtime run --wasm-features all --dir .
       run: cargo test --target ${{ matrix.rust.target }} --all-features
@@ -63,19 +64,17 @@ jobs:
       run: mv .cargo-ci .cargo
 
     - name: Test with 'native' CPU features + No Default Cargo Features
-      if: matrix.rust.target == 'i586-pc-windows-msvc' || matrix.rust.target == 'i686-pc-windows-msvc' || matrix.rust.target == 'x86_64-pc-windows-msvc' || matrix.rust.target == 'wasm32-wasi'
       run: cargo test --target ${{ matrix.rust.target }} --no-default-features
 
     - name: Test with 'native' CPU features + All Cargo Features
-      if: matrix.rust.target == 'i586-pc-windows-msvc' || matrix.rust.target == 'i686-pc-windows-msvc' || matrix.rust.target == 'x86_64-pc-windows-msvc' || matrix.rust.target == 'wasm32-wasi'
       run: cargo test --target ${{ matrix.rust.target }} --all-features
 
   cross_compile_aarch64:
     runs-on: ubuntu-latest
 
     strategy:
       matrix:
-        rust: [1.59,stable,nightly]
+        rust: ["1.60", stable, nightly]
         features: ["", "std"]
 
     steps:
@@ -85,12 +84,11 @@ jobs:
           sudo apt-get install qemu binfmt-support qemu-user-static gcc-aarch64-linux-gnu binutils-aarch64-linux-gnu
 
       - name: Installing Rust toolchain
-        uses: actions-rs/toolchain@v1
+        uses: dtolnay/rust-toolchain@master
         with:
-          override: true
           toolchain: ${{ matrix.rust }}
           target: aarch64-unknown-linux-musl
-      - uses: actions/checkout@v2
+      - uses: actions/checkout@v4
       - name: build
         run: >
           cargo build --verbose --no-default-features --target aarch64-unknown-linux-musl --features "$FEATURES"

Cargo.toml

@@ -1,7 +1,7 @@
 [package]
 name = "wide"
 description = "A crate to help you go wide."
-version = "0.7.25"
+version = "0.7.26"
 authors = ["Lokathor <zefria@gmail.com>"]
 repository = "https://github.com/Lokathor/wide"
 readme = "README.md"
@@ -10,8 +10,9 @@ categories = ["data-structures", "hardware-support"]
 edition = "2018"
 license = "Zlib OR Apache-2.0 OR MIT"
 
-# minimum version supported with WASM and all intrinsics for safe_arch (also first version to support this config option)
-rust-version = "1.56"
+# Aarch64 needs 1.59 while others need 1.56
+# When updating, also update CI workflows and the badge in the README.
+rust-version = "1.61"
 
 [features]
 default = ["std"]

README.md

@@ -1,5 +1,5 @@
 [![License:Zlib](https://img.shields.io/badge/License-Zlib-brightgreen.svg)](https://opensource.org/licenses/Zlib)
-![Minimum Rust Version](https://img.shields.io/badge/Min%20Rust-1.52-green.svg)
+![Minimum Rust Version](https://img.shields.io/badge/Min%20Rust-1.61-green.svg)
 [![crates.io](https://img.shields.io/crates/v/wide.svg)](https://crates.io/crates/wide)
 [![docs.rs](https://docs.rs/wide/badge.svg)](https://docs.rs/wide/)
 

src/f32x4_.rs

@@ -4,13 +4,13 @@ pick! {
   if #[cfg(target_feature="sse")] {
     #[derive(Default, Clone, Copy, PartialEq)]
     #[repr(C, align(16))]
-    pub struct f32x4 { sse: m128 }
+    pub struct f32x4 { pub(crate) sse: m128 }
   } else if #[cfg(target_feature="simd128")] {
     use core::arch::wasm32::*;
 
     #[derive(Clone, Copy)]
     #[repr(transparent)]
-    pub struct f32x4 { simd: v128 }
+    pub struct f32x4 { pub(crate) simd: v128 }
 
     impl Default for f32x4 {
       fn default() -> Self {
@@ -27,7 +27,7 @@ pick! {
     use core::arch::aarch64::*;
     #[repr(C)]
     #[derive(Copy, Clone)]
-    pub struct f32x4 { neon : float32x4_t }
+    pub struct f32x4 { pub(crate) neon : float32x4_t }
 
     impl Default for f32x4 {
       #[inline]
@@ -48,7 +48,7 @@ pick! {
     } else {
     #[derive(Default, Clone, Copy, PartialEq)]
     #[repr(C, align(16))]
-    pub struct f32x4 { arr: [f32;4] }
+    pub struct f32x4 { pub(crate) arr: [f32;4] }
   }
 }
 
@@ -1305,7 +1305,7 @@ impl f32x4 {
     cast::<_, f32x4>(c)
   }
 
-  /// Calculate the exponent of a packed f32x4
+  /// Calculate the exponent of a packed `f32x4`
   #[inline]
   #[must_use]
   #[allow(non_upper_case_globals)]
@@ -1578,4 +1578,24 @@ impl f32x4 {
   pub fn as_array_mut(&mut self) -> &mut [f32; 4] {
     cast_mut(self)
   }
+
+  #[inline]
+  pub fn from_i32x4(v: i32x4) -> Self {
+    pick! {
+      if #[cfg(target_feature="sse2")] {
+        Self { sse: convert_to_m128_from_i32_m128i(v.sse) }
+      } else if #[cfg(target_feature="simd128")] {
+        Self { simd: f32x4_convert_i32x4(v.simd) }
+      } else if #[cfg(all(target_feature="neon",target_arch="aarch64"))] {
+        Self { neon: unsafe { vcvtq_f32_s32(v.neon) }}
+      } else {
+        Self { arr: [
+            v.as_array_ref()[0] as f32,
+            v.as_array_ref()[1] as f32,
+            v.as_array_ref()[2] as f32,
+            v.as_array_ref()[3] as f32,
+          ] }
+      }
+    }
+  }
 }

src/f32x8_.rs

@@ -1062,7 +1062,7 @@ impl f32x8 {
     cast::<_, f32x8>(c)
   }
 
-  /// Calculate the exponent of a packed f32x8
+  /// Calculate the exponent of a packed `f32x8`
   #[inline]
   #[must_use]
   #[allow(non_upper_case_globals)]
@@ -1330,7 +1330,7 @@ impl f32x8 {
     Self::pow_f32x8(self, f32x8::splat(y))
   }
 
-  /// Transpose matrix of 8x8 f32 matrix. Currently only accelerated on AVX.
+  /// Transpose matrix of 8x8 `f32` matrix. Currently only accelerated on AVX.
   #[must_use]
   #[inline]
   pub fn transpose(data: [f32x8; 8]) -> [f32x8; 8] {
@@ -1418,6 +1418,26 @@ impl f32x8 {
   pub fn as_array_mut(&mut self) -> &mut [f32; 8] {
     cast_mut(self)
   }
+
+  #[inline]
+  pub fn from_i32x8(v: i32x8) -> Self {
+    pick! {
+      if #[cfg(target_feature="avx2")] {
+        Self { avx: convert_to_m256_from_i32_m256i(v.avx2) }
+      } else {
+        Self::new([
+            v.as_array_ref()[0] as f32,
+            v.as_array_ref()[1] as f32,
+            v.as_array_ref()[2] as f32,
+            v.as_array_ref()[3] as f32,
+            v.as_array_ref()[4] as f32,
+            v.as_array_ref()[5] as f32,
+            v.as_array_ref()[6] as f32,
+            v.as_array_ref()[7] as f32,
+          ])
+      }
+    }
+  }
 }
 
 impl Not for f32x8 {

src/f64x2_.rs

@@ -4,13 +4,13 @@ pick! {
   if #[cfg(target_feature="sse2")] {
     #[derive(Default, Clone, Copy, PartialEq)]
     #[repr(C, align(16))]
-    pub struct f64x2 { sse: m128d }
+    pub struct f64x2 { pub(crate) sse: m128d }
   } else if #[cfg(target_feature="simd128")] {
     use core::arch::wasm32::*;
 
     #[derive(Clone, Copy)]
     #[repr(transparent)]
-    pub struct f64x2 { simd: v128 }
+    pub struct f64x2 { pub(crate) simd: v128 }
 
     impl Default for f64x2 {
       fn default() -> Self {
@@ -27,7 +27,7 @@ pick! {
     use core::arch::aarch64::*;
     #[repr(C)]
     #[derive(Copy, Clone)]
-    pub struct f64x2 { neon : float64x2_t }
+    pub struct f64x2 { pub(crate) neon: float64x2_t }
 
     impl Default for f64x2 {
       #[inline]
@@ -51,7 +51,7 @@ pick! {
   } else {
     #[derive(Default, Clone, Copy, PartialEq)]
     #[repr(C, align(16))]
-    pub struct f64x2 { arr: [f64;2] }
+    pub struct f64x2 { pub(crate) arr: [f64;2] }
   }
 }
 
@@ -1304,7 +1304,7 @@ impl f64x2 {
     cast::<_, f64x2>(c)
   }
 
-  /// Calculate the exponent of a packed f64x2
+  /// Calculate the exponent of a packed `f64x2`
   #[inline]
   #[must_use]
   #[allow(non_upper_case_globals)]
@@ -1608,6 +1608,33 @@ impl f64x2 {
   pub fn as_array_mut(&mut self) -> &mut [f64; 2] {
     cast_mut(self)
   }
+
+  /// Converts the lower two `i32` lanes to two `f64` lanes (and dropping the higher two `i32` lanes)
+  #[inline]
+  pub fn from_i32x4_lower2(v: i32x4) -> Self {
+    pick! {
+      if #[cfg(target_feature="sse2")] {
+        Self { sse: convert_to_m128d_from_lower2_i32_m128i(v.sse) }
+      } else if #[cfg(target_feature="simd128")] {
+        Self { simd: f64x2_convert_low_i32x4(v.simd)}
+      } else if #[cfg(all(target_feature="neon",target_arch="aarch64"))] {
+        Self { neon: unsafe { vcvtq_f64_s64(vmovl_s32(vget_low_s32(v.neon))) }}
+      } else {
+        Self { arr: [
+            v.as_array_ref()[0] as f64,
+            v.as_array_ref()[1] as f64,
+        ]}
+      }
+    }
+  }
+}
+
+impl From<i32x4> for f64x2 {
+  /// Converts the lower two `i32` lanes to two `f64` lanes (and dropping the higher two `i32` lanes)
+  #[inline]
+  fn from(v: i32x4) -> Self {
+    Self::from_i32x4_lower2(v)
+  }
 }
 
 impl Not for f64x2 {

src/f64x4_.rs

@@ -4,11 +4,11 @@ pick! {
   if #[cfg(target_feature="avx")] {
     #[derive(Default, Clone, Copy, PartialEq)]
     #[repr(C, align(32))]
-    pub struct f64x4 { avx: m256d }
+    pub struct f64x4 { pub(crate) avx: m256d }
   } else {
     #[derive(Default, Clone, Copy, PartialEq)]
     #[repr(C, align(32))]
-    pub struct f64x4 { a : f64x2, b : f64x2 }
+    pub struct f64x4 { pub(crate) a: f64x2, pub(crate) b: f64x2 }
   }
 }
 
@@ -1174,7 +1174,7 @@ impl f64x4 {
     cast::<_, f64x4>(c)
   }
 
-  /// Calculate the exponent of a packed f64x4
+  /// Calculate the exponent of a packed `f64x4`
   #[inline]
   #[must_use]
   #[allow(non_upper_case_globals)]
@@ -1473,6 +1473,29 @@ impl f64x4 {
   pub fn as_array_mut(&mut self) -> &mut [f64; 4] {
     cast_mut(self)
   }
+
+  #[inline]
+  pub fn from_i32x4(v: i32x4) -> Self {
+    pick! {
+      if #[cfg(target_feature="avx")] {
+        Self { avx: convert_to_m256d_from_i32_m128i(v.sse) }
+      } else {
+        Self::new([
+          v.as_array_ref()[0] as f64,
+          v.as_array_ref()[1] as f64,
+          v.as_array_ref()[2] as f64,
+          v.as_array_ref()[3] as f64,
+        ])
+      }
+    }
+  }
+}
+
+impl From<i32x4> for f64x4 {
+  #[inline]
+  fn from(v: i32x4) -> Self {
+    Self::from_i32x4(v)
+  }
 }
 
 impl Not for f64x4 {