When SIMD is enabled, use the SSE2 code for x86

When enabled, makes BLAKE2b ~20% faster and BLAKE2s ~60% faster on
32-bit x86.

Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "blake2-rfc"
-version = "0.2.4"
+version = "0.2.5"
 authors = ["Cesar Eduardo Barros <cesarb@cesarb.eti.br>"]
 description = "A pure Rust implementation of BLAKE2 based on the draft RFC."
 repository = "https://github.com/cesarb/blake2-rfc"

src/simd/mod.rs

@@ -26,17 +26,14 @@
 
 #![allow(non_camel_case_types)]
 
-#[cfg(not(feature = "simd"))]
-mod fallback;
-#[cfg(not(feature = "simd"))]
-pub use self::fallback::*;
-
-#[cfg(all(feature = "simd", target_arch = "x86_64"))]
+#[cfg(all(feature = "simd", any(target_arch = "x86", target_arch = "x86_64")))]
 mod sse2;
-#[cfg(all(feature = "simd", target_arch = "x86_64"))]
+#[cfg(all(feature = "simd", any(target_arch = "x86", target_arch = "x86_64")))]
 pub use self::sse2::*;
 
-#[cfg(all(feature = "simd", not(any(target_arch = "x86_64"))))]
+#[cfg(not(all(feature = "simd", any(target_arch = "x86",
+                                    target_arch = "x86_64"))))]
 mod fallback;
-#[cfg(all(feature = "simd", not(any(target_arch = "x86_64"))))]
+#[cfg(not(all(feature = "simd", any(target_arch = "x86",
+                                    target_arch = "x86_64"))))]
 pub use self::fallback::*;