diff options
| author | Björn Steinbrink <bsteinbr@gmail.com> | 2015-05-15 15:20:42 +0200 |
|---|---|---|
| committer | Björn Steinbrink <bsteinbr@gmail.com> | 2015-05-15 15:30:22 +0200 |
| commit | eeeb2cc0dffc016582f020c0a9e6d9f9fc751397 (patch) | |
| tree | 0218029e9937bf8ad03f8a8071ae2b6dcb1c6d2f /src/libcore | |
| parent | 579e31929feff51dcaf8d444648eff8de735f91a (diff) | |
| download | rust-eeeb2cc0dffc016582f020c0a9e6d9f9fc751397.tar.gz rust-eeeb2cc0dffc016582f020c0a9e6d9f9fc751397.zip | |
Allow for better optimizations of iterators for zero-sized types
Using regular pointer arithmetic to iterate collections of zero-sized types doesn't work, because we'd get the same pointer all the time. Our current solution is to convert the pointer to an integer, add an offset and then convert back, but this inhibits certain optimizations. What we should do instead is to convert the pointer to one that points to an i8*, and then use a LLVM GEP instructions without the inbounds flag to perform the pointer arithmetic. This allows to generate pointers that point outside allocated objects without causing UB (as long as you don't dereference them), and it wraps around using two's complement, i.e. it behaves exactly like the wrapping_* operations we're currently using, with the added benefit of LLVM being able to better optimize the resulting IR.
Diffstat (limited to 'src/libcore')
| -rw-r--r-- | src/libcore/intrinsics.rs | 14 | ||||
| -rw-r--r-- | src/libcore/slice.rs | 106 |
2 files changed, 75 insertions, 45 deletions
diff --git a/src/libcore/intrinsics.rs b/src/libcore/intrinsics.rs index d94b8884112..fa432e311eb 100644 --- a/src/libcore/intrinsics.rs +++ b/src/libcore/intrinsics.rs @@ -283,6 +283,20 @@ extern "rust-intrinsic" { /// returned value will result in undefined behavior. pub fn offset<T>(dst: *const T, offset: isize) -> *const T; + /// Calculates the offset from a pointer, potentially wrapping. + /// + /// This is implemented as an intrinsic to avoid converting to and from an + /// integer, since the conversion inhibits certain optimizations. + /// + /// # Safety + /// + /// Unlike the `offset` intrinsic, this intrinsic does not restrict the + /// resulting pointer to point into or one byte past the end of an allocated + /// object, and it wraps with two's complement arithmetic. The resulting + /// value is not necessarily valid to be used to actually access memory. + #[cfg(not(stage0))] + pub fn arith_offset<T>(dst: *const T, offset: isize) -> *const T; + /// Copies `count * size_of<T>` bytes from `src` to `dst`. The source /// and destination may *not* overlap. /// diff --git a/src/libcore/slice.rs b/src/libcore/slice.rs index 9db1ceddf0d..346afdc963d 100644 --- a/src/libcore/slice.rs +++ b/src/libcore/slice.rs @@ -124,6 +124,43 @@ pub trait SliceExt { fn clone_from_slice(&mut self, &[Self::Item]) -> usize where Self::Item: Clone; } +// Use macros to be generic over const/mut +#[cfg(stage0)] +macro_rules! slice_offset { + ($ptr:expr, $by:expr) => {{ + let ptr = $ptr; + if size_from_ptr(ptr) == 0 { + transmute((ptr as isize).wrapping_add($by)) + } else { + ptr.offset($by) + } + }}; +} + +#[cfg(not(stage0))] +macro_rules! slice_offset { + ($ptr:expr, $by:expr) => {{ + let ptr = $ptr; + if size_from_ptr(ptr) == 0 { + ::intrinsics::arith_offset(ptr as *mut i8, $by) as *mut _ + } else { + ptr.offset($by) + } + }}; +} + +macro_rules! slice_ref { + ($ptr:expr) => {{ + let ptr = $ptr; + if size_from_ptr(ptr) == 0 { + // Use a non-null pointer value + &mut *(1 as *mut _) + } else { + transmute(ptr) + } + }}; +} + #[unstable(feature = "core")] impl<T> SliceExt for [T] { type Item = T; @@ -136,16 +173,18 @@ impl<T> SliceExt for [T] { #[inline] fn iter<'a>(&'a self) -> Iter<'a, T> { unsafe { - let p = self.as_ptr(); - assume(!p.is_null()); - if mem::size_of::<T>() == 0 { - Iter {ptr: p, - end: ((p as usize).wrapping_add(self.len())) as *const T, - _marker: marker::PhantomData} + let p = if mem::size_of::<T>() == 0 { + 1 as *const _ } else { - Iter {ptr: p, - end: p.offset(self.len() as isize), - _marker: marker::PhantomData} + let p = self.as_ptr(); + assume(!p.is_null()); + p + }; + + Iter { + ptr: p, + end: slice_offset!(p, self.len() as isize), + _marker: marker::PhantomData } } } @@ -273,16 +312,18 @@ impl<T> SliceExt for [T] { #[inline] fn iter_mut<'a>(&'a mut self) -> IterMut<'a, T> { unsafe { - let p = self.as_mut_ptr(); - assume(!p.is_null()); - if mem::size_of::<T>() == 0 { - IterMut {ptr: p, - end: ((p as usize).wrapping_add(self.len())) as *mut T, - _marker: marker::PhantomData} + let p = if mem::size_of::<T>() == 0 { + 1 as *mut _ } else { - IterMut {ptr: p, - end: p.offset(self.len() as isize), - _marker: marker::PhantomData} + let p = self.as_mut_ptr(); + assume(!p.is_null()); + p + }; + + IterMut { + ptr: p, + end: slice_offset!(p, self.len() as isize), + _marker: marker::PhantomData } } } @@ -630,31 +671,6 @@ fn size_from_ptr<T>(_: *const T) -> usize { mem::size_of::<T>() } - -// Use macros to be generic over const/mut -macro_rules! slice_offset { - ($ptr:expr, $by:expr) => {{ - let ptr = $ptr; - if size_from_ptr(ptr) == 0 { - transmute((ptr as isize).wrapping_add($by)) - } else { - ptr.offset($by) - } - }}; -} - -macro_rules! slice_ref { - ($ptr:expr) => {{ - let ptr = $ptr; - if size_from_ptr(ptr) == 0 { - // Use a non-null pointer value - &mut *(1 as *mut _) - } else { - transmute(ptr) - } - }}; -} - // The shared definition of the `Iter` and `IterMut` iterators macro_rules! iterator { (struct $name:ident -> $ptr:ty, $elem:ty) => { @@ -781,7 +797,7 @@ impl<'a, T> Iter<'a, T> { match self.as_slice().get(n) { Some(elem_ref) => unsafe { self.ptr = slice_offset!(self.ptr, (n as isize).wrapping_add(1)); - Some(slice_ref!(elem_ref)) + Some(elem_ref) }, None => { self.ptr = self.end; @@ -849,7 +865,7 @@ impl<'a, T> IterMut<'a, T> { match make_mut_slice!(self.ptr, self.end).get_mut(n) { Some(elem_ref) => unsafe { self.ptr = slice_offset!(self.ptr, (n as isize).wrapping_add(1)); - Some(slice_ref!(elem_ref)) + Some(elem_ref) }, None => { self.ptr = self.end; |
