logo
Expand description

This crate provides macros for runtime CPU feature detection. It’s intended as a stopgap until Rust RFC 2725 adding first-class target feature detection macros to libcore is implemented.

Supported target architectures:

  • aarch64: Linux and macOS/M4 only (ARM64 does not support OS-independent feature detection)
    • Target features: aes, sha2, sha3
  • x86/x86_64: OS independent and no_std-friendly
    • Target features: adx, aes, avx, avx2, bmi1, bmi2, fma, mmx, pclmulqdq, popcnt, rdrand, rdseed, sgx, sha, sse, sse2, sse3, sse4.1, sse4.2, ssse3

If you would like detection support for a target feature which is not on this list, please open a GitHub issue.

Example

// This macro creates `cpuid_aes_sha` module
cpufeatures::new!(cpuid_aes_sha, "aes", "sha");

// `token` is a Zero Sized Type (ZST) value, which guarantees
// that underlying static storage got properly initialized,
// which allows to omit initialization branch
let token: cpuid_aes_sha::InitToken = cpuid_aes_sha::init();

if token.get() {
    println!("CPU supports both SHA and AES extensions");
} else {
    println!("SHA and AES extensions are not supported");
}

// If stored value needed only once you can get stored value
// omitting the token
let val = cpuid_aes_sha::get();
assert_eq!(val, token.get());

// Additionally you can get both token and value
let (token, val) = cpuid_aes_sha::init_get();
assert_eq!(val, token.get());

Note that if all tested target features are enabled via compiler options (e.g. by using RUSTFLAGS), the get method will always return true and init will not use CPUID instruction. Such behavior allows compiler to completely eliminate fallback code.

After first call macro caches result and returns it in subsequent calls, thus runtime overhead for them is minimal.

Macros

Create module with CPU feature detection code.