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
//! This crate provides a fully written in Rust memory allocator for Vulkan, DirectX 12 and Metal.
//!
//! # [Windows-rs] and [winapi]
//!
//! `gpu-allocator` recently migrated from [winapi] to [windows-rs] but still provides convenient helpers to convert to and from [winapi] types, enabled when compiling with the `public-winapi` crate feature.
//!
//! [Windows-rs]: https://github.com/microsoft/windows-rs
//! [winapi]: https://github.com/retep998/winapi-rs
//!
//! # Setting up the Vulkan memory allocator
//!
//! ```no_run
//! # #[cfg(feature = "vulkan")]
//! # fn main() {
//! use gpu_allocator::vulkan::*;
//! # use ash::vk;
//! # let device = todo!();
//! # let instance = todo!();
//! # let physical_device = todo!();
//!
//! let mut allocator = Allocator::new(&AllocatorCreateDesc {
//!     instance,
//!     device,
//!     physical_device,
//!     debug_settings: Default::default(),
//!     buffer_device_address: true,  // Ideally, check the BufferDeviceAddressFeatures struct.
//!     allocation_sizes: Default::default(),
//! });
//! # }
//! # #[cfg(not(feature = "vulkan"))]
//! # fn main() {}
//! ```
//!
//! # Simple Vulkan allocation example
//!
//! ```no_run
//! # #[cfg(feature = "vulkan")]
//! # fn main() {
//! use gpu_allocator::vulkan::*;
//! use gpu_allocator::MemoryLocation;
//! # use ash::vk;
//! # let device = todo!();
//! # let instance = todo!();
//! # let physical_device = todo!();
//! # let mut allocator = Allocator::new(&AllocatorCreateDesc {
//! #     instance,
//! #     device,
//! #     physical_device,
//! #     debug_settings: Default::default(),
//! #     buffer_device_address: true,  // Ideally, check the BufferDeviceAddressFeatures struct.
//! #     allocation_sizes: Default::default(),
//! # }).unwrap();
//!
//! // Setup vulkan info
//! let vk_info = vk::BufferCreateInfo::default()
//!     .size(512)
//!     .usage(vk::BufferUsageFlags::STORAGE_BUFFER);
//!
//! let buffer = unsafe { device.create_buffer(&vk_info, None) }.unwrap();
//! let requirements = unsafe { device.get_buffer_memory_requirements(buffer) };
//!
//! let allocation = allocator
//!     .allocate(&AllocationCreateDesc {
//!         name: "Example allocation",
//!         requirements,
//!         location: MemoryLocation::CpuToGpu,
//!         linear: true, // Buffers are always linear
//!         allocation_scheme: AllocationScheme::GpuAllocatorManaged,
//!     }).unwrap();
//!
//! // Bind memory to the buffer
//! unsafe { device.bind_buffer_memory(buffer, allocation.memory(), allocation.offset()).unwrap() };
//!
//! // Cleanup
//! allocator.free(allocation).unwrap();
//! unsafe { device.destroy_buffer(buffer, None) };
//! # }
//! # #[cfg(not(feature = "vulkan"))]
//! # fn main() {}
//! ```
//!
//! # Setting up the D3D12 memory allocator
//!
//! ```no_run
//! # #[cfg(feature = "d3d12")]
//! # fn main() {
//! use gpu_allocator::d3d12::*;
//! # let device = todo!();
//!
//! let mut allocator = Allocator::new(&AllocatorCreateDesc {
//!     device: ID3D12DeviceVersion::Device(device),
//!     debug_settings: Default::default(),
//!     allocation_sizes: Default::default(),
//! });
//! # }
//! # #[cfg(not(feature = "d3d12"))]
//! # fn main() {}
//! ```
//!
//! # Simple d3d12 allocation example
//!
//! ```no_run
//! # #[cfg(feature = "d3d12")]
//! # fn main() -> windows::core::Result<()> {
//! use gpu_allocator::d3d12::*;
//! use gpu_allocator::MemoryLocation;
//! # use windows::Win32::Graphics::{Dxgi, Direct3D12};
//! # let device = todo!();
//!
//! # let mut allocator = Allocator::new(&AllocatorCreateDesc {
//! #     device: ID3D12DeviceVersion::Device(device),
//! #     debug_settings: Default::default(),
//! #     allocation_sizes: Default::default(),
//! # }).unwrap();
//!
//! let buffer_desc = Direct3D12::D3D12_RESOURCE_DESC {
//!     Dimension: Direct3D12::D3D12_RESOURCE_DIMENSION_BUFFER,
//!     Alignment: 0,
//!     Width: 512,
//!     Height: 1,
//!     DepthOrArraySize: 1,
//!     MipLevels: 1,
//!     Format: Dxgi::Common::DXGI_FORMAT_UNKNOWN,
//!     SampleDesc: Dxgi::Common::DXGI_SAMPLE_DESC {
//!         Count: 1,
//!         Quality: 0,
//!     },
//!     Layout: Direct3D12::D3D12_TEXTURE_LAYOUT_ROW_MAJOR,
//!     Flags: Direct3D12::D3D12_RESOURCE_FLAG_NONE,
//! };
//! let allocation_desc = AllocationCreateDesc::from_d3d12_resource_desc(
//!     &allocator.device(),
//!     &buffer_desc,
//!     "Example allocation",
//!     MemoryLocation::GpuOnly,
//! );
//! let allocation = allocator.allocate(&allocation_desc).unwrap();
//! let mut resource: Option<Direct3D12::ID3D12Resource> = None;
//! let hr = unsafe {
//!     device.CreatePlacedResource(
//!         allocation.heap(),
//!         allocation.offset(),
//!         &buffer_desc,
//!         Direct3D12::D3D12_RESOURCE_STATE_COMMON,
//!         None,
//!         &mut resource,
//!     )
//! }?;
//!
//! // Cleanup
//! drop(resource);
//! allocator.free(allocation).unwrap();
//! # Ok(())
//! # }
//! # #[cfg(not(feature = "d3d12"))]
//! # fn main() {}
//! ```
//!
//! # Setting up the Metal memory allocator
//!
//! ```no_run
//! # #[cfg(feature = "metal")]
//! # fn main() {
//! # use std::sync::Arc;
//! use gpu_allocator::metal::*;
//!
//! # let device = Arc::new(metal::Device::system_default().unwrap());
//! let mut allocator = Allocator::new(&AllocatorCreateDesc {
//!     device: device.clone(),
//!     debug_settings: Default::default(),
//!     allocation_sizes: Default::default(),
//! });
//! # }
//! # #[cfg(not(feature = "metal"))]
//! # fn main() {}
//! ```
//!
//! # Simple Metal allocation example
//! ```no_run
//! # #[cfg(feature = "metal")]
//! # fn main() {
//! # use std::sync::Arc;
//! use gpu_allocator::metal::*;
//! use gpu_allocator::MemoryLocation;
//! # let device = Arc::new(metal::Device::system_default().unwrap());
//! # let mut allocator = Allocator::new(&AllocatorCreateDesc {
//! #     device: device.clone(),
//! #     debug_settings: Default::default(),
//! #     allocation_sizes: Default::default(),
//! # })
//! # .unwrap();
//!
//! let allocation_desc = AllocationCreateDesc::buffer(
//!     &device,
//!     "Example allocation",
//!     512, // size in bytes
//!     gpu_allocator::MemoryLocation::GpuOnly,
//! );
//! let allocation = allocator.allocate(&allocation_desc).unwrap();
//! let resource = allocation.make_buffer().unwrap();
//!
//! // Cleanup
//! drop(resource);
//! allocator.free(&allocation).unwrap();
//! # }
//! # #[cfg(not(feature = "metal"))]
//! # fn main() {}
//! ```

mod result;
pub use result::*;

pub(crate) mod allocator;

pub use allocator::{AllocationReport, AllocatorReport, MemoryBlockReport};

#[cfg(feature = "visualizer")]
pub mod visualizer;

#[cfg(feature = "vulkan")]
pub mod vulkan;

#[cfg(all(windows, feature = "d3d12"))]
pub mod d3d12;

#[cfg(all(any(target_os = "macos", target_os = "ios"), feature = "metal"))]
pub mod metal;

#[derive(Clone, Copy, Debug, PartialEq, Eq, Hash)]
pub enum MemoryLocation {
    /// The allocated resource is stored at an unknown memory location; let the driver decide what's the best location
    Unknown,
    /// Store the allocation in GPU only accessible memory - typically this is the faster GPU resource and this should be
    /// where most of the allocations live.
    GpuOnly,
    /// Memory useful for uploading data to the GPU and potentially for constant buffers
    CpuToGpu,
    /// Memory useful for CPU readback of data
    GpuToCpu,
}

#[derive(Copy, Clone, Debug)]
pub struct AllocatorDebugSettings {
    /// Logs out debugging information about the various heaps the current device has on startup
    pub log_memory_information: bool,
    /// Logs out all memory leaks on shutdown with log level Warn
    pub log_leaks_on_shutdown: bool,
    /// Stores a copy of the full backtrace for every allocation made, this makes it easier to debug leaks
    /// or other memory allocations, but storing stack traces has a RAM overhead so should be disabled
    /// in shipping applications.
    pub store_stack_traces: bool,
    /// Log out every allocation as it's being made with log level Debug, rather spammy so off by default
    pub log_allocations: bool,
    /// Log out every free that is being called with log level Debug, rather spammy so off by default
    pub log_frees: bool,
    /// Log out stack traces when either `log_allocations` or `log_frees` is enabled.
    pub log_stack_traces: bool,
}

impl Default for AllocatorDebugSettings {
    fn default() -> Self {
        Self {
            log_memory_information: false,
            log_leaks_on_shutdown: true,
            store_stack_traces: false,
            log_allocations: false,
            log_frees: false,
            log_stack_traces: false,
        }
    }
}

/// The sizes of the memory blocks that the allocator will create.
///
/// Useful for tuning the allocator to your application's needs. For example most games will be fine with the default
/// values, but eg. an app might want to use smaller block sizes to reduce the amount of memory used.
///
/// Clamped between 4MB and 256MB, and rounds up to the nearest multiple of 4MB for alignment reasons.
#[derive(Clone, Copy, Debug)]
pub struct AllocationSizes {
    /// The size of the memory blocks that will be created for the GPU only memory type.
    ///
    /// Defaults to 256MB.
    device_memblock_size: u64,
    /// The size of the memory blocks that will be created for the CPU visible memory types.
    ///
    /// Defaults to 64MB.
    host_memblock_size: u64,
}

impl AllocationSizes {
    pub fn new(device_memblock_size: u64, host_memblock_size: u64) -> Self {
        const FOUR_MB: u64 = 4 * 1024 * 1024;
        const TWO_HUNDRED_AND_FIFTY_SIX_MB: u64 = 256 * 1024 * 1024;

        let mut device_memblock_size =
            device_memblock_size.clamp(FOUR_MB, TWO_HUNDRED_AND_FIFTY_SIX_MB);
        let mut host_memblock_size =
            host_memblock_size.clamp(FOUR_MB, TWO_HUNDRED_AND_FIFTY_SIX_MB);

        if device_memblock_size % FOUR_MB != 0 {
            let val = device_memblock_size / FOUR_MB + 1;
            device_memblock_size = val * FOUR_MB;
            log::warn!(
                "Device memory block size must be a multiple of 4MB, clamping to {}MB",
                device_memblock_size / 1024 / 1024
            )
        }

        if host_memblock_size % FOUR_MB != 0 {
            let val = host_memblock_size / FOUR_MB + 1;
            host_memblock_size = val * FOUR_MB;
            log::warn!(
                "Host memory block size must be a multiple of 4MB, clamping to {}MB",
                host_memblock_size / 1024 / 1024
            )
        }

        Self {
            device_memblock_size,
            host_memblock_size,
        }
    }
}

impl Default for AllocationSizes {
    fn default() -> Self {
        Self {
            device_memblock_size: 256 * 1024 * 1024,
            host_memblock_size: 64 * 1024 * 1024,
        }
    }
}