mem: Add REP MOVSB/STOSB implementations

The assembly generated seems correct:
    https://rust.godbolt.org/z/GGnec8

Signed-off-by: Joe Richey <[email protected]>

Author: josephlr

src/lib.rs

@@ -1,6 +1,7 @@
 #![cfg_attr(feature = "compiler-builtins", compiler_builtins)]
 #![cfg_attr(feature = "asm", feature(asm))]
 #![feature(abi_unadjusted)]
+#![feature(asm)]
 #![feature(llvm_asm)]
 #![feature(global_asm)]
 #![feature(cfg_target_has_atomic)]

src/mem/mod.rs

@@ -10,6 +10,7 @@ use core::mem;
 use core::ops::{BitOr, Shl};
 
 // memcpy/memmove/memset have optimized implementations on some architectures
+#[cfg_attr(target_arch = "x86_64", path = "x86_64.rs")]
 mod memcpy;
 pub use self::memcpy::*;
 

src/mem/x86_64.rs

@@ -0,0 +1,46 @@
+use super::c_int;
+
+#[cfg_attr(all(feature = "mem", not(feature = "mangled-names")), no_mangle)]
+pub unsafe extern "C" fn memcpy(dest: *mut u8, src: *const u8, count: usize) -> *mut u8 {
+    asm!(
+        "rep movsb [rdi], [rsi]",
+        inout("rcx") count => _,
+        inout("rdi") dest => _,
+        inout("rsi") src => _,
+        options(nostack, preserves_flags)
+    );
+    dest
+}
+
+#[cfg_attr(all(feature = "mem", not(feature = "mangled-names")), no_mangle)]
+pub unsafe extern "C" fn memmove(dest: *mut u8, src: *const u8, count: usize) -> *mut u8 {
+    let delta = dest as usize - src as usize;
+    if delta >= count {
+        // We can copy forwards because either dest is far enough ahead of src,
+        // or src is ahead of dest (and delta overflowed).
+        return self::memcpy(dest, src, count);
+    }
+    // copy backwards
+    asm!(
+        "std",
+        "rep movsb [rdi], [rsi]",
+        "cld",
+        inout("rcx") count => _,
+        inout("rdi") dest.add(count).sub(1) => _,
+        inout("rsi") src.add(count).sub(1) => _,
+        options(nostack, preserves_flags)
+    );
+    dest
+}
+
+#[cfg_attr(all(feature = "mem", not(feature = "mangled-names")), no_mangle)]
+pub unsafe extern "C" fn memset(dest: *mut u8, c: c_int, count: usize) -> *mut u8 {
+    asm!(
+        "rep stosb [rdi], al",
+        inout("rcx") count => _,
+        inout("rdi") dest => _,
+        in("al") c as u8,
+        options(nostack, preserves_flags)
+    );
+    dest
+}