1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335
use super::DeviceCopy;
use error::*;
use memory::malloc::{cuda_free_locked, cuda_malloc_locked};
use std::mem;
use std::ops;
use std::ptr;
use std::slice;
/// Fixed-size host-side buffer in page-locked memory.
///
/// See the [`module-level documentation`](../memory/index.html) for more details on page-locked
/// memory.
#[derive(Debug)]
pub struct LockedBuffer<T: DeviceCopy> {
buf: *mut T,
capacity: usize,
}
impl<T: DeviceCopy + Clone> LockedBuffer<T> {
/// Allocate a new page-locked buffer large enough to hold `size` `T`'s and initialized with
/// clones of `value`.
///
/// # Errors:
///
/// If the allocation fails, returns the error from CUDA. If `size` is large enough that
/// `size * mem::sizeof::<T>()` overflows usize, then returns InvalidMemoryAllocation.
///
/// # Examples:
///
/// ```
/// # let _context = rustacuda::quick_init().unwrap();
/// use rustacuda::memory::*;
/// let mut buffer = LockedBuffer::new(&0u64, 5).unwrap();
/// buffer[0] = 1;
/// ```
pub fn new(value: &T, size: usize) -> CudaResult<Self> {
unsafe {
let mut uninit = LockedBuffer::uninitialized(size)?;
for x in 0..size {
*uninit.get_unchecked_mut(x) = value.clone();
}
Ok(uninit)
}
}
/// Allocate a new page-locked buffer of the same size as `slice`, initialized with a clone of
/// the data in `slice`.
///
/// # Errors:
///
/// If the allocation fails, returns the error from CUDA.
///
/// # Examples:
///
/// ```
/// # let _context = rustacuda::quick_init().unwrap();
/// use rustacuda::memory::*;
/// let values = [0u64; 5];
/// let mut buffer = LockedBuffer::from_slice(&values).unwrap();
/// buffer[0] = 1;
/// ```
pub fn from_slice(slice: &[T]) -> CudaResult<Self> {
unsafe {
let mut uninit = LockedBuffer::uninitialized(slice.len())?;
for (i, x) in slice.iter().enumerate() {
*uninit.get_unchecked_mut(i) = x.clone();
}
Ok(uninit)
}
}
}
impl<T: DeviceCopy> LockedBuffer<T> {
/// Allocate a new page-locked buffer large enough to hold `size` `T`'s, but without
/// initializing the contents.
///
/// # Errors:
///
/// If the allocation fails, returns the error from CUDA. If `size` is large enough that
/// `size * mem::sizeof::<T>()` overflows usize, then returns InvalidMemoryAllocation.
///
/// # Safety:
///
/// The caller must ensure that the contents of the buffer are initialized before reading from
/// the buffer.
///
/// # Examples:
///
/// ```
/// # let _context = rustacuda::quick_init().unwrap();
/// use rustacuda::memory::*;
/// let mut buffer = unsafe { LockedBuffer::uninitialized(5).unwrap() };
/// for i in buffer.iter_mut() {
/// *i = 0u64;
/// }
/// ```
pub unsafe fn uninitialized(size: usize) -> CudaResult<Self> {
let bytes = size
.checked_mul(mem::size_of::<T>())
.ok_or(CudaError::InvalidMemoryAllocation)?;
let ptr: *mut T = if bytes > 0 {
cuda_malloc_locked(bytes)?
} else {
ptr::NonNull::dangling().as_ptr()
};
Ok(LockedBuffer {
buf: ptr as *mut T,
capacity: size,
})
}
/// Extracts a slice containing the entire buffer.
///
/// Equivalent to `&s[..]`.
///
/// # Examples:
///
/// ```
/// # let _context = rustacuda::quick_init().unwrap();
/// use rustacuda::memory::*;
/// let buffer = LockedBuffer::new(&0u64, 5).unwrap();
/// let sum : u64 = buffer.as_slice().iter().sum();
/// ```
pub fn as_slice(&self) -> &[T] {
self
}
/// Extracts a mutable slice of the entire buffer.
///
/// Equivalent to `&mut s[..]`.
///
/// # Examples:
///
/// ```
/// # let _context = rustacuda::quick_init().unwrap();
/// use rustacuda::memory::*;
/// let mut buffer = LockedBuffer::new(&0u64, 5).unwrap();
/// for i in buffer.as_mut_slice() {
/// *i = 12u64;
/// }
/// ```
pub fn as_mut_slice(&mut self) -> &mut [T] {
self
}
/// Creates a `LockedBuffer<T>` directly from the raw components of another locked buffer.
///
/// # Safety
///
/// This is highly unsafe, due to the number of invariants that aren't
/// checked:
///
/// * `ptr` needs to have been previously allocated via `LockedBuffer` or
/// [`cuda_malloc_locked`](fn.cuda_malloc_locked.html).
/// * `ptr`'s `T` needs to have the same size and alignment as it was allocated with.
/// * `capacity` needs to be the capacity that the pointer was allocated with.
///
/// Violating these may cause problems like corrupting the CUDA driver's
/// internal data structures.
///
/// The ownership of `ptr` is effectively transferred to the
/// `LockedBuffer<T>` which may then deallocate, reallocate or change the
/// contents of memory pointed to by the pointer at will. Ensure
/// that nothing else uses the pointer after calling this
/// function.
///
/// # Examples:
///
/// ```
/// # let _context = rustacuda::quick_init().unwrap();
/// use std::mem;
/// use rustacuda::memory::*;
///
/// let mut buffer = LockedBuffer::new(&0u64, 5).unwrap();
/// let ptr = buffer.as_mut_ptr();
/// let size = buffer.len();
///
/// mem::forget(buffer);
///
/// let buffer = unsafe { LockedBuffer::from_raw_parts(ptr, size) };
/// ```
pub unsafe fn from_raw_parts(ptr: *mut T, size: usize) -> LockedBuffer<T> {
LockedBuffer {
buf: ptr,
capacity: size,
}
}
/// Destroy a `LockedBuffer`, returning an error.
///
/// Deallocating page-locked memory can return errors from previous asynchronous work. This function
/// destroys the given buffer and returns the error and the un-destroyed buffer on failure.
///
/// # Example:
///
/// ```
/// # let _context = rustacuda::quick_init().unwrap();
/// use rustacuda::memory::*;
/// let x = LockedBuffer::new(&0u64, 5).unwrap();
/// match LockedBuffer::drop(x) {
/// Ok(()) => println!("Successfully destroyed"),
/// Err((e, buf)) => {
/// println!("Failed to destroy buffer: {:?}", e);
/// // Do something with buf
/// },
/// }
/// ```
pub fn drop(mut buf: LockedBuffer<T>) -> DropResult<LockedBuffer<T>> {
if buf.buf.is_null() {
return Ok(());
}
if buf.capacity > 0 && mem::size_of::<T>() > 0 {
let capacity = buf.capacity;
let ptr = mem::replace(&mut buf.buf, ptr::null_mut());
unsafe {
match cuda_free_locked(ptr) {
Ok(()) => {
mem::forget(buf);
Ok(())
}
Err(e) => Err((e, LockedBuffer::from_raw_parts(ptr, capacity))),
}
}
} else {
Ok(())
}
}
}
impl<T: DeviceCopy> AsRef<[T]> for LockedBuffer<T> {
fn as_ref(&self) -> &[T] {
self
}
}
impl<T: DeviceCopy> AsMut<[T]> for LockedBuffer<T> {
fn as_mut(&mut self) -> &mut [T] {
self
}
}
impl<T: DeviceCopy> ops::Deref for LockedBuffer<T> {
type Target = [T];
fn deref(&self) -> &[T] {
unsafe {
let p = self.buf;
slice::from_raw_parts(p, self.capacity)
}
}
}
impl<T: DeviceCopy> ops::DerefMut for LockedBuffer<T> {
fn deref_mut(&mut self) -> &mut [T] {
unsafe {
let ptr = self.buf;
slice::from_raw_parts_mut(ptr, self.capacity)
}
}
}
impl<T: DeviceCopy> Drop for LockedBuffer<T> {
fn drop(&mut self) {
if self.buf.is_null() {
return;
}
if self.capacity > 0 && mem::size_of::<T>() > 0 {
// No choice but to panic if this fails.
unsafe {
cuda_free_locked(self.buf).expect("Failed to deallocate CUDA page-locked memory.");
}
}
self.capacity = 0;
}
}
#[cfg(test)]
mod test {
use super::*;
use std::mem;
#[derive(Clone, Debug)]
struct ZeroSizedType;
unsafe impl ::memory::DeviceCopy for ZeroSizedType {}
#[test]
fn test_new() {
let _context = ::quick_init().unwrap();
let val = 0u64;
let mut buffer = LockedBuffer::new(&val, 5).unwrap();
buffer[0] = 1;
}
#[test]
fn test_from_slice() {
let _context = ::quick_init().unwrap();
let values = [0u64; 10];
let mut buffer = LockedBuffer::from_slice(&values).unwrap();
for i in buffer[0..3].iter_mut() {
*i = 10;
}
}
#[test]
fn from_raw_parts() {
let _context = ::quick_init().unwrap();
let mut buffer = LockedBuffer::new(&0u64, 5).unwrap();
buffer[2] = 1;
let ptr = buffer.as_mut_ptr();
let len = buffer.len();
mem::forget(buffer);
let buffer = unsafe { LockedBuffer::from_raw_parts(ptr, len) };
assert_eq!(&[0u64, 0, 1, 0, 0], buffer.as_slice());
drop(buffer);
}
#[test]
fn zero_length_buffer() {
let _context = ::quick_init().unwrap();
let buffer = LockedBuffer::new(&0u64, 0).unwrap();
drop(buffer);
}
#[test]
fn zero_size_type() {
let _context = ::quick_init().unwrap();
let buffer = LockedBuffer::new(&ZeroSizedType, 10).unwrap();
drop(buffer);
}
#[test]
fn overflows_usize() {
let _context = ::quick_init().unwrap();
let err = LockedBuffer::new(&0u64, ::std::usize::MAX - 1).unwrap_err();
assert_eq!(CudaError::InvalidMemoryAllocation, err);
}
}