auto merge of #6724 : thestinger/rust/swap_fast, r=thestinger

Passing higher alignment values gives the optimization passes more freedom since it can copy in larger chunks. This change results in rustc outputting the same post-optimization IR as clang for swaps and most copies excluding the lack of information about padding.

Code snippet:

```rust
#[inline(never)]
fn swap<T>(x: &mut T, y: &mut T) {
    util::swap(x, y);
}
```

Original IR (for `int`):

```llvm
define internal fastcc void @_ZN9swap_283417_a71830ca3ed2d65d3_00E(i64*, i64*) #1 {
static_allocas:
  %2 = icmp eq i64* %0, %1
  br i1 %2, label %_ZN4util9swap_283717_a71830ca3ed2d65d3_00E.exit, label %3

; <label>:3                                       ; preds = %static_allocas
  %4 = load i64* %0, align 1
  %5 = load i64* %1, align 1
  store i64 %5, i64* %0, align 1
  store i64 %4, i64* %1, align 1
  br label %_ZN4util9swap_283717_a71830ca3ed2d65d3_00E.exit

_ZN4util9swap_283717_a71830ca3ed2d65d3_00E.exit:  ; preds = %3, %static_allocas
  ret void
}
```

After #6710:

```llvm
define internal fastcc void @_ZN9swap_283017_a71830ca3ed2d65d3_00E(i64* nocapture, i64* nocapture) #1 {
static_allocas:
  %2 = load i64* %0, align 1
  %3 = load i64* %1, align 1
  store i64 %3, i64* %0, align 1
  store i64 %2, i64* %1, align 1
  ret void
}
```

After this change:

```llvm
define internal fastcc void @_ZN9swap_283017_a71830ca3ed2d65d3_00E(i64* nocapture, i64* nocapture) #1 {
static_allocas:
  %2 = load i64* %0, align 8
  %3 = load i64* %1, align 8
  store i64 %3, i64* %0, align 8
  store i64 %2, i64* %1, align 8
  ret void
}
```

Another example:

```rust
#[inline(never)]
fn set<T>(x: &mut T, y: T) {
    *x = y;
}
```

Before, with `(int, int)` (align 1):

```llvm
define internal fastcc void @_ZN8set_282517_8fa972e3f9e451983_00E({ i64, i64 }* nocapture, { i64, i64 }* nocapture) #1 {
static_allocas:
  %2 = bitcast { i64, i64 }* %1 to i8*
  %3 = bitcast { i64, i64 }* %0 to i8*
  tail call void @llvm.memcpy.p0i8.p0i8.i64(i8* %3, i8* %2, i64 16, i32 1, i1 false)
  ret void
}
```

After, with `(int, int)` (align 8):

```llvm
define internal fastcc void @_ZN8set_282617_8fa972e3f9e451983_00E({ i64, i64 }* nocapture, { i64, i64 }* nocapture) #1 {
static_allocas:
  %2 = bitcast { i64, i64 }* %1 to i8*
  %3 = bitcast { i64, i64 }* %0 to i8*
  tail call void @llvm.memcpy.p0i8.p0i8.i64(i8* %3, i8* %2, i64 16, i32 8, i1 false)
  ret void
}
```

1 files changed, 28 insertions, 5 deletions

diff --git a/src/libstd/unstable/intrinsics.rs b/src/libstd/unstable/intrinsics.rs
index 521708621fc..908c5e23ab0 100644
--- a/src/libstd/unstable/intrinsics.rs
+++ b/src/libstd/unstable/intrinsics.rs
@@ -16,7 +16,7 @@ The corresponding definitions are in librustc/middle/trans/foreign.rs.
 
 The atomic intrinsics provide common atomic operations on machine
 words, with multiple possible memory orderings. They obey the same
-semantics as C++0x. See the LLVM documentation on [[atomics]].
+semantics as C++11. See the LLVM documentation on [[atomics]].
 
 [atomics]: http://llvm.org/docs/Atomics.html
 
@@ -31,6 +31,7 @@ A quick refresher on memory ordering:
   with atomic types and is equivalent to Java's `volatile`.
 
 */
+
 #[abi = "rust-intrinsic"]
 pub extern "rust-intrinsic" {
 
@@ -127,18 +128,40 @@ pub extern "rust-intrinsic" {
     /// Get the address of the `__morestack` stack growth function.
     pub fn morestack_addr() -> *();
 
-    /// Equivalent to the `llvm.memcpy.p0i8.0i8.i32` intrinsic.
+    /// Equivalent to the `llvm.memcpy.p0i8.0i8.i32` intrinsic, with a size of
+    /// `count` * `size_of::<T>()` and an alignment of `min_align_of::<T>()`
     #[cfg(not(stage0))]
-    pub fn memcpy32(dst: *mut u8, src: *u8, size: u32);
-    /// Equivalent to the `llvm.memcpy.p0i8.0i8.i64` intrinsic.
+    pub fn memcpy32<T>(dst: *mut T, src: *T, count: u32);
+    /// Equivalent to the `llvm.memcpy.p0i8.0i8.i64` intrinsic, with a size of
+    /// `count` * `size_of::<T>()` and an alignment of `min_align_of::<T>()`
     #[cfg(not(stage0))]
-    pub fn memcpy64(dst: *mut u8, src: *u8, size: u64);
+    pub fn memcpy64<T>(dst: *mut T, src: *T, count: u64);
 
     /// Equivalent to the `llvm.memmove.p0i8.0i8.i32` intrinsic.
+    #[cfg(stage0)]
     pub fn memmove32(dst: *mut u8, src: *u8, size: u32);
     /// Equivalent to the `llvm.memmove.p0i8.0i8.i64` intrinsic.
+    #[cfg(stage0)]
     pub fn memmove64(dst: *mut u8, src: *u8, size: u64);
 
+    /// Equivalent to the `llvm.memmove.p0i8.0i8.i32` intrinsic, with a size of
+    /// `count` * `size_of::<T>()` and an alignment of `min_align_of::<T>()`
+    #[cfg(not(stage0))]
+    pub fn memmove32<T>(dst: *mut T, src: *T, count: u32);
+    /// Equivalent to the `llvm.memmove.p0i8.0i8.i64` intrinsic, with a size of
+    /// `count` * `size_of::<T>()` and an alignment of `min_align_of::<T>()`
+    #[cfg(not(stage0))]
+    pub fn memmove64<T>(dst: *mut T, src: *T, count: u64);
+
+    /// Equivalent to the `llvm.memset.p0i8.i32` intrinsic, with a size of
+    /// `count` * `size_of::<T>()` and an alignment of `min_align_of::<T>()`
+    #[cfg(not(stage0))]
+    pub fn memset32<T>(dst: *mut T, val: u8, count: u32);
+    /// Equivalent to the `llvm.memset.p0i8.i64` intrinsic, with a size of
+    /// `count` * `size_of::<T>()` and an alignment of `min_align_of::<T>()`
+    #[cfg(not(stage0))]
+    pub fn memset64<T>(dst: *mut T, val: u8, count: u64);
+
     pub fn sqrtf32(x: f32) -> f32;
     pub fn sqrtf64(x: f64) -> f64;