switch to ensure_simd and update readme

Author: RagnarGrootKoerkamp

CHANGELOG.md

@@ -1,5 +1,8 @@
 # Changelog
 
+## Upcoming
+- switch to `ensure_simd`
+
 ## 4.2.0
 - Perf: improve NEON codegen by avoiding `u32x8::splat`.
 - Perf: reuse thread-local allocations between invocations

Cargo.lock

@@ -13,6 +13,7 @@ description = "Constructing and iterating packed DNA sequences using SIMD"
 wide = "0.7"
 mem_dbg = "0.3"
 cfg-if = "1.0"
+ensure_simd = "0.1.0"
 
 # Optional generate random sequences for testing. (Default enabled.)
 rand = { version = "0.9", features = ["small_rng"], optional = true }
@@ -28,4 +29,4 @@ pyo3 = { version = "0.25", features = ["extension-module"], optional = true }
 default = ["rand"]
 
 # Hides the `simd` warnings when neither AVX2 nor NEON is detected.
-scalar = []
+scalar = ["ensure_simd/scalar"]

Cargo.toml

@@ -21,18 +21,18 @@ crate was developed:
   Ragnar Groot Koerkamp, Igor Martayan.
   SEA 2025 [https://doi.org/10.4230/LIPIcs.SEA.2025.20](doi.org/10.4230/LIPIcs.SEA.2025.20)
 
-## Requirements
+## Installation
 
-This library supports AVX2 and NEON instruction sets.
-Make sure to set `RUSTFLAGS="-C target-cpu=native"` when compiling to use the instruction sets available on your architecture.
+This library requires AVX2 or NEON instruction sets, which, on x64, requires
+either `target-cpu=native` or `target-cpu=x86-64-v3`.
+See [this README](https://github.com/ragnargrootkoerkamp/ensure_simd) for details and [this
+blog](https://curiouscoding.nl/posts/distributing-rust-simd-binaries/) for background.
+The same restrictions apply when using packed-seq in a larger project.
 
 ``` sh
 RUSTFLAGS="-C target-cpu=native" cargo run --release
 ```
 
-Enable the `-F scalar` feature flag to fall back to a scalar implementation with
-reduced performance.
-
 ## Usage example
 
 Full documentation can be found on [docs.rs](https://docs.rs/packed-seq).

README.md

@@ -105,17 +105,6 @@
 //! - `epserde` enables `derive(epserde::Epserde)` for `PackedSeqVec` and `AsciiSeqVec`, and adds its `SerializeInner` and `DeserializeInner` traits to `SeqVec`.
 //! - `pyo3` enables `derive(pyo3::pyclass)` for `PackedSeqVec` and `AsciiSeqVec`.
 
-#[cfg(not(any(
-    doc,
-    debug_assertions,
-    target_feature = "avx2",
-    target_feature = "neon",
-    feature = "scalar"
-)))]
-compile_error!(
-    "Packed-seq uses AVX2 or NEON SIMD instructions. Compile using `-C target-cpu=native` to get the expected performance. Silence this error using the `scalar` feature."
-);
-
 /// Functions with architecture-specific implementations.
 pub mod intrinsics {
     mod transpose;