Initial AVX-512 support

.github/workflows/ci.yml

@@ -3,12 +3,12 @@ env:
   # version like 1.70. Note that we only specify MAJOR.MINOR and not PATCH so that bugfixes still
   # come automatically. If the version specified here is no longer the latest stable version,
   # then please feel free to submit a PR that adjusts it along with the potential clippy fixes.
-  RUST_STABLE_VER: "1.88" # In quotes because otherwise (e.g.) 1.70 would be interpreted as 1.7
+  RUST_STABLE_VER: "1.89" # In quotes because otherwise (e.g.) 1.70 would be interpreted as 1.7
   # The purpose of checking with the minimum supported Rust toolchain is to detect its staleness.
   # If the compilation fails, then the version specified here needs to be bumped up to reality.
   # Be sure to also update the rust-version property in the workspace Cargo.toml file,
   # plus all the README.md files of the affected packages.
-  RUST_MIN_VER: "1.88"
+  RUST_MIN_VER: "1.89"
   # List of packages that will be checked with the minimum supported Rust version.
   # This should be limited to packages that are intended for publishing.
   RUST_MIN_VER_PKGS: "-p fearless_simd"

Cargo.toml

@@ -13,7 +13,7 @@ license = "Apache-2.0 OR MIT"
 repository = "https://github.com/linebender/fearless_simd"
 # Keep in sync with RUST_MIN_VER in .github/workflows/ci.yml, with the relevant README.md files
 # and with the MSRV in the `Unreleased` section of CHANGELOG.md.
-rust-version = "1.88"
+rust-version = "1.89"
 
 [workspace.lints]
 
@@ -47,7 +47,7 @@ clippy.debug_assert_with_mut_call = "warn"
 clippy.doc_markdown = "warn"
 clippy.fn_to_numeric_cast_any = "warn"
 clippy.infinite_loop = "warn"
-clippy.large_stack_arrays = "warn"
+clippy.large_stack_arrays = "allow"             # appears to be buggy as of 1.93, fixed in nightly. TODO: re-enable
 clippy.mismatching_type_param_order = "warn"
 clippy.missing_assert_message = "warn"
 clippy.missing_fields_in_debug = "warn"

README.md

@@ -59,7 +59,7 @@ It benefited from conversations with Luca Versari, though he is not responsible
 
 ## Minimum supported Rust Version (MSRV)
 
-This version of Fearless SIMD has been verified to compile with **Rust 1.88** and later.
+This version of Fearless SIMD has been verified to compile with **Rust 1.89** and later.
 
 Future versions of Fearless SIMD might increase the Rust version requirement.
 It will not be treated as a breaking change and as such can even happen with small patch releases.

fearless_simd/README.md

@@ -163,7 +163,7 @@ At least one of `std` and `libm` is required; `std` overrides `libm`.
 
 ## Minimum supported Rust Version (MSRV)
 
-This version of Fearless SIMD has been verified to compile with **Rust 1.88** and later.
+This version of Fearless SIMD has been verified to compile with **Rust 1.89** and later.
 
 Future versions of Fearless SIMD might increase the Rust version requirement.
 It will not be treated as a breaking change and as such can even happen with small patch releases.

fearless_simd/src/core_arch/x86/avx512.rs

@@ -0,0 +1,21 @@
+// Copyright 2025 the Fearless_SIMD Authors
+// SPDX-License-Identifier: Apache-2.0 OR MIT
+
+//! Access to AVX-512 intrinsics (Ice Lake feature set).
+
+/// A token for AVX-512 intrinsics (Ice Lake feature set) on `x86` and `x86_64`.
+#[derive(Clone, Copy, Debug)]
+pub struct Avx512 {
+    _private: (),
+}
+
+impl Avx512 {
+    /// Create a SIMD token.
+    ///
+    /// # Safety
+    ///
+    /// The required CPU features must be available.
+    pub const unsafe fn new_unchecked() -> Self {
+        Self { _private: () }
+    }
+}

fearless_simd/src/core_arch/x86/mod.rs

@@ -5,6 +5,7 @@
 
 mod avx;
 mod avx2;
+mod avx512;
 mod fma;
 mod sse;
 mod sse2;
@@ -15,6 +16,7 @@ mod ssse3;
 
 pub use avx::Avx;
 pub use avx2::Avx2;
+pub use avx512::Avx512;
 pub use fma::Fma;
 pub use sse::Sse;
 pub use sse2::Sse2;

fearless_simd/src/generated.rs

@@ -46,6 +46,8 @@
 
 #[cfg(any(target_arch = "x86", target_arch = "x86_64"))]
 mod avx2;
+#[cfg(any(target_arch = "x86", target_arch = "x86_64"))]
+mod avx512;
 mod fallback;
 #[cfg(target_arch = "aarch64")]
 mod neon;
@@ -59,6 +61,8 @@ mod wasm;
 
 #[cfg(any(target_arch = "x86", target_arch = "x86_64"))]
 pub use avx2::*;
+#[cfg(any(target_arch = "x86", target_arch = "x86_64"))]
+pub use avx512::*;
 pub use fallback::*;
 #[cfg(target_arch = "aarch64")]
 pub use neon::*;

fearless_simd/src/generated/avx2.rs

@@ -8645,6 +8645,13 @@ unsafe fn cross_block_alignr_one(
     };
     unsafe { dyn_alignr_256(hi_blocks, lo_blocks, intra_shift) }
 }
+#[doc = r" Concatenates `b` and `a` (each 1 x __m256i = 2 blocks) and extracts 2 blocks starting at byte offset"]
+#[doc = r" `shift_bytes`. Extracts from [b : a] (b in low bytes, a in high bytes), matching alignr semantics."]
+#[inline(always)]
+unsafe fn cross_block_alignr_256x1(a: __m256i, b: __m256i, shift_bytes: usize) -> __m256i {
+    let regs = [b, a];
+    unsafe { cross_block_alignr_one(&regs, 0, shift_bytes) }
+}
 #[doc = r" Concatenates `b` and `a` (each 2 x __m256i = 4 blocks) and extracts 4 blocks starting at byte offset"]
 #[doc = r" `shift_bytes`. Extracts from [b : a] (b in low bytes, a in high bytes), matching alignr semantics."]
 #[inline(always)]
@@ -8661,10 +8668,3 @@ unsafe fn cross_block_alignr_256x2(
         ]
     }
 }
-#[doc = r" Concatenates `b` and `a` (each 1 x __m256i = 2 blocks) and extracts 2 blocks starting at byte offset"]
-#[doc = r" `shift_bytes`. Extracts from [b : a] (b in low bytes, a in high bytes), matching alignr semantics."]
-#[inline(always)]
-unsafe fn cross_block_alignr_256x1(a: __m256i, b: __m256i, shift_bytes: usize) -> __m256i {
-    let regs = [b, a];
-    unsafe { cross_block_alignr_one(&regs, 0, shift_bytes) }
-}