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
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
use super::DeviceCopy;
use crate::error::*;
use crate::memory::DevicePointer;
use crate::memory::UnifiedPointer;
use std::mem;
use std::os::raw::c_void;
use std::ptr;

/// Unsafe wrapper around the `cuMemAlloc` function, which allocates some device memory and
/// returns a [`DevicePointer`](struct.DevicePointer.html) pointing to it. The memory is not cleared.
///
/// Note that `count` is in units of T; thus a `count` of 3 will allocate `3 * size_of::<T>()` bytes
/// of memory.
///
/// Memory buffers allocated using `cuda_malloc` must be freed using [`cuda_free`](fn.cuda_free.html).
///
/// # Errors
///
/// If allocating memory fails, returns the CUDA error value.
/// If the number of bytes to allocate is zero (either because count is zero or because T is a
/// zero-sized type), or if the size of the allocation would overflow a usize, returns InvalidValue.
///
/// # Safety
///
/// Since the allocated memory is not initialized, the caller must ensure that it is initialized
/// before copying it to the host in any way. Additionally, the caller must ensure that the memory
/// allocated is freed using cuda_free, or the memory will be leaked.
///
/// # Examples
///
/// ```
/// # let _context = rustacuda::quick_init().unwrap();
/// use rustacuda::memory::*;
/// unsafe {
///     // Allocate space for 5 u64s
///     let device_buffer = cuda_malloc::<u64>(5).unwrap();
///     cuda_free(device_buffer).unwrap();
/// }
/// ```
pub unsafe fn cuda_malloc<T>(count: usize) -> CudaResult<DevicePointer<T>> {
    let size = count.checked_mul(mem::size_of::<T>()).unwrap_or(0);
    if size == 0 {
        return Err(CudaError::InvalidMemoryAllocation);
    }

    let mut ptr: *mut c_void = ptr::null_mut();
    cuda_driver_sys::cuMemAlloc_v2(&mut ptr as *mut *mut c_void as *mut u64, size).to_result()?;
    let ptr = ptr as *mut T;
    Ok(DevicePointer::wrap(ptr as *mut T))
}

/// Unsafe wrapper around the `cuMemAllocManaged` function, which allocates some unified memory and
/// returns a [`UnifiedPointer`](struct.UnifiedPointer.html) pointing to it. The memory is not cleared.
///
/// Note that `count` is in units of T; thus a `count` of 3 will allocate `3 * size_of::<T>()` bytes
/// of memory.
///
/// Memory buffers allocated using `cuda_malloc_unified` must be freed using [`cuda_free`](fn.cuda_free.html).
///
/// # Errors
///
/// If allocating memory fails, returns the CUDA error value.
/// If the number of bytes to allocate is zero (either because count is zero or because T is a
/// zero-sized type), or if the size of the allocation would overflow a usize, returns InvalidValue.
///
/// # Safety
///
/// Since the allocated memory is not initialized, the caller must ensure that it is initialized
/// before reading from it in any way. Additionally, the caller must ensure that the memory
/// allocated is freed using cuda_free, or the memory will be leaked.
///
/// # Examples
///
/// ```
/// # let _context = rustacuda::quick_init().unwrap();
/// use rustacuda::memory::*;
/// unsafe {
///     // Allocate space for a u64
///     let mut unified_buffer = cuda_malloc_unified::<u64>(1).unwrap();
///     // Write to it
///     *unified_buffer.as_raw_mut() = 5u64;
///     cuda_free_unified(unified_buffer).unwrap();
/// }
/// ```
pub unsafe fn cuda_malloc_unified<T: DeviceCopy>(count: usize) -> CudaResult<UnifiedPointer<T>> {
    let size = count.checked_mul(mem::size_of::<T>()).unwrap_or(0);
    if size == 0 {
        return Err(CudaError::InvalidMemoryAllocation);
    }

    let mut ptr: *mut c_void = ptr::null_mut();
    cuda_driver_sys::cuMemAllocManaged(
        &mut ptr as *mut *mut c_void as *mut u64,
        size,
        cuda_driver_sys::CUmemAttach_flags_enum::CU_MEM_ATTACH_GLOBAL as u32,
    )
    .to_result()?;
    let ptr = ptr as *mut T;
    Ok(UnifiedPointer::wrap(ptr as *mut T))
}

/// Free memory allocated with [`cuda_malloc`](fn.cuda_malloc.html).
///
/// # Errors
///
/// If freeing memory fails, returns the CUDA error value. If the given pointer is null, returns
/// InvalidValue.
///
/// # Safety
///
/// The given pointer must have been allocated with `cuda_malloc`, or null.
/// The caller is responsible for ensuring that no other pointers to the deallocated buffer exist.
///
/// # Examples
///
/// ```
/// # let _context = rustacuda::quick_init().unwrap();
/// use rustacuda::memory::*;
/// unsafe {
///     let device_buffer = cuda_malloc::<u64>(5).unwrap();
///     // Free allocated memory.
///     cuda_free(device_buffer).unwrap();
/// }
/// ```
pub unsafe fn cuda_free<T>(mut p: DevicePointer<T>) -> CudaResult<()> {
    let ptr = p.as_raw_mut();
    if ptr.is_null() {
        return Err(CudaError::InvalidMemoryAllocation);
    }

    cuda_driver_sys::cuMemFree_v2(ptr as u64).to_result()?;
    Ok(())
}

/// Free memory allocated with [`cuda_malloc_unified`](fn.cuda_malloc_unified.html).
///
/// # Errors
///
/// If freeing memory fails, returns the CUDA error value. If the given pointer is null, returns
/// InvalidValue.
///
/// # Safety
///
/// The given pointer must have been allocated with `cuda_malloc_unified`, or null.
/// The caller is responsible for ensuring that no other pointers to the deallocated buffer exist.
///
/// # Examples
///
/// ```
/// # let _context = rustacuda::quick_init().unwrap();
/// use rustacuda::memory::*;
/// unsafe {
///     let unified_buffer = cuda_malloc_unified::<u64>(5).unwrap();
///     // Free allocated memory.
///     cuda_free_unified(unified_buffer).unwrap();
/// }
/// ```
pub unsafe fn cuda_free_unified<T: DeviceCopy>(mut p: UnifiedPointer<T>) -> CudaResult<()> {
    let ptr = p.as_raw_mut();
    if ptr.is_null() {
        return Err(CudaError::InvalidMemoryAllocation);
    }

    cuda_driver_sys::cuMemFree_v2(ptr as u64).to_result()?;
    Ok(())
}

/// Unsafe wrapper around the `cuMemAllocHost` function, which allocates some page-locked host memory
/// and returns a raw pointer pointing to it. The memory is not cleared.
///
/// Note that `count` is in units of T; thus a `count` of 3 will allocate `3 * size_of::<T>()` bytes
/// of memory.
///
/// Memory buffers allocated using `cuda_malloc_locked` must be freed using [`cuda_free_locked`](fn.cuda_free_locked.html).
///
/// # Errors
///
/// If allocating memory fails, returns the CUDA error value.
/// If the number of bytes to allocate is zero (either because count is zero or because T is a
/// zero-sized type), or if the size of the allocation would overflow a usize, returns InvalidValue.
///
/// # Safety
///
/// Since the allocated memory is not initialized, the caller must ensure that it is initialized
/// before reading from it in any way. Additionally, the caller must ensure that the memory
/// allocated is freed using `cuda_free_locked`, or the memory will be leaked.
///
/// # Examples
///
/// ```
/// # let _context = rustacuda::quick_init().unwrap();
/// use rustacuda::memory::*;
/// unsafe {
///     // Allocate space for 5 u64s
///     let locked_buffer = cuda_malloc_locked::<u64>(5).unwrap();
///     cuda_free_locked(locked_buffer).unwrap();
/// }
/// ```
pub unsafe fn cuda_malloc_locked<T>(count: usize) -> CudaResult<*mut T> {
    let size = count.checked_mul(mem::size_of::<T>()).unwrap_or(0);
    if size == 0 {
        return Err(CudaError::InvalidMemoryAllocation);
    }

    let mut ptr: *mut c_void = ptr::null_mut();
    cuda_driver_sys::cuMemAllocHost_v2(&mut ptr as *mut *mut c_void, size).to_result()?;
    let ptr = ptr as *mut T;
    Ok(ptr as *mut T)
}

/// Free page-locked memory allocated with [`cuda_malloc_host`](fn.cuda_malloc_host.html).
///
/// # Errors
///
/// If freeing memory fails, returns the CUDA error value. If the given pointer is null, returns
/// InvalidValue.
///
/// # Safety
///
/// The given pointer must have been allocated with `cuda_malloc_locked`, or null.
/// The caller is responsible for ensuring that no other pointers to the deallocated buffer exist.
///
/// # Examples
///
/// ```
/// # let _context = rustacuda::quick_init().unwrap();
/// use rustacuda::memory::*;
/// unsafe {
///     let locked_buffer = cuda_malloc_locked::<u64>(5).unwrap();
///     // Free allocated memory
///     cuda_free_locked(locked_buffer).unwrap();
/// }
/// ```
pub unsafe fn cuda_free_locked<T>(ptr: *mut T) -> CudaResult<()> {
    if ptr.is_null() {
        return Err(CudaError::InvalidMemoryAllocation);
    }

    cuda_driver_sys::cuMemFreeHost(ptr as *mut c_void).to_result()?;
    Ok(())
}

#[cfg(test)]
mod test {
    use super::*;

    #[derive(Clone, Debug)]
    struct ZeroSizedType;
    unsafe impl DeviceCopy for ZeroSizedType {}

    #[test]
    fn test_cuda_malloc() {
        let _context = crate::quick_init().unwrap();
        unsafe {
            let device_mem = cuda_malloc::<u64>(1).unwrap();
            assert!(!device_mem.is_null());
            cuda_free(device_mem).unwrap();
        }
    }

    #[test]
    fn test_cuda_malloc_zero_bytes() {
        let _context = crate::quick_init().unwrap();
        unsafe {
            assert_eq!(
                CudaError::InvalidMemoryAllocation,
                cuda_malloc::<u64>(0).unwrap_err()
            );
        }
    }

    #[test]
    fn test_cuda_malloc_zero_sized() {
        let _context = crate::quick_init().unwrap();
        unsafe {
            assert_eq!(
                CudaError::InvalidMemoryAllocation,
                cuda_malloc::<ZeroSizedType>(10).unwrap_err()
            );
        }
    }

    #[test]
    fn test_cuda_alloc_overflow() {
        let _context = crate::quick_init().unwrap();
        unsafe {
            assert_eq!(
                CudaError::InvalidMemoryAllocation,
                cuda_malloc::<u64>(::std::usize::MAX - 1).unwrap_err()
            );
        }
    }

    #[test]
    fn test_cuda_malloc_unified() {
        let _context = crate::quick_init().unwrap();
        unsafe {
            let mut unified = cuda_malloc_unified::<u64>(1).unwrap();
            assert!(!unified.is_null());

            // Write to the allocated memory
            *unified.as_raw_mut() = 64;

            cuda_free_unified(unified).unwrap();
        }
    }

    #[test]
    fn test_cuda_malloc_unified_zero_bytes() {
        let _context = crate::quick_init().unwrap();
        unsafe {
            assert_eq!(
                CudaError::InvalidMemoryAllocation,
                cuda_malloc_unified::<u64>(0).unwrap_err()
            );
        }
    }

    #[test]
    fn test_cuda_malloc_unified_zero_sized() {
        let _context = crate::quick_init().unwrap();
        unsafe {
            assert_eq!(
                CudaError::InvalidMemoryAllocation,
                cuda_malloc_unified::<ZeroSizedType>(10).unwrap_err()
            );
        }
    }

    #[test]
    fn test_cuda_malloc_unified_overflow() {
        let _context = crate::quick_init().unwrap();
        unsafe {
            assert_eq!(
                CudaError::InvalidMemoryAllocation,
                cuda_malloc_unified::<u64>(::std::usize::MAX - 1).unwrap_err()
            );
        }
    }

    #[test]
    fn test_cuda_free_null() {
        let _context = crate::quick_init().unwrap();
        let null = ::std::ptr::null_mut::<u64>();
        unsafe {
            assert_eq!(
                CudaError::InvalidMemoryAllocation,
                cuda_free(DevicePointer::wrap(null)).unwrap_err()
            );
        }
    }

    #[test]
    fn test_cuda_malloc_locked() {
        let _context = crate::quick_init().unwrap();
        unsafe {
            let locked = cuda_malloc_locked::<u64>(1).unwrap();
            assert!(!locked.is_null());

            // Write to the allocated memory
            *locked = 64;

            cuda_free_locked(locked).unwrap();
        }
    }

    #[test]
    fn test_cuda_malloc_locked_zero_bytes() {
        let _context = crate::quick_init().unwrap();
        unsafe {
            assert_eq!(
                CudaError::InvalidMemoryAllocation,
                cuda_malloc_locked::<u64>(0).unwrap_err()
            );
        }
    }

    #[test]
    fn test_cuda_malloc_locked_zero_sized() {
        let _context = crate::quick_init().unwrap();
        unsafe {
            assert_eq!(
                CudaError::InvalidMemoryAllocation,
                cuda_malloc_locked::<ZeroSizedType>(10).unwrap_err()
            );
        }
    }

    #[test]
    fn test_cuda_malloc_locked_overflow() {
        let _context = crate::quick_init().unwrap();
        unsafe {
            assert_eq!(
                CudaError::InvalidMemoryAllocation,
                cuda_malloc_locked::<u64>(::std::usize::MAX - 1).unwrap_err()
            );
        }
    }

    #[test]
    fn test_cuda_free_locked_null() {
        let _context = crate::quick_init().unwrap();
        unsafe {
            assert_eq!(
                CudaError::InvalidMemoryAllocation,
                cuda_free_locked(::std::ptr::null_mut::<u64>()).unwrap_err()
            );
        }
    }
}