cubecl_runtime/
runtime.rs

1use alloc::vec::Vec;
2use cubecl_common::device::{Device, DeviceId};
3use cubecl_ir::TargetProperties;
4use cubecl_zspace::{Shape, Strides};
5
6use crate::{client::Client, server::ServerStorage};
7
8pub use crate::description::{DeviceDescription, DeviceKind};
9pub use cubecl_monitoring::{
10    DevicePower, DeviceTemperature, DeviceUtilization, UtilizationUnavailable,
11};
12
13/// Runtime for the `CubeCL`.
14pub trait Runtime: Sized + Send + Sync + 'static + core::fmt::Debug + Clone {
15    /// The compute server used to run kernels and perform autotuning.
16    type Server: ServerStorage;
17    /// The device used to retrieve the compute client.
18    type Device: Device;
19
20    /// Retrieve the compute client from the runtime device, initializing the
21    /// server on first use.
22    fn client(device: &Self::Device) -> Client {
23        Client::load::<Self::Server>(device.to_id())
24    }
25
26    /// Whether a tensor with `shape` and `strides` can be read as is. If the result is false, the
27    /// tensor should be made contiguous before reading.
28    fn can_read_tensor(shape: &Shape, strides: &Strides) -> bool;
29
30    /// Returns the properties of the target hardware architecture.
31    fn target_properties() -> TargetProperties;
32
33    /// Returns all devices available under the provided type id.
34    fn enumerate_devices(type_id: u16) -> Vec<DeviceId>;
35    /// Returns all devices that can be handled by the runtime.
36    fn enumerate_all_devices() -> Vec<DeviceId> {
37        Self::enumerate_devices(0)
38    }
39
40    /// The devices of `device_id`'s own kind: its peers, and itself where the
41    /// machine has it.
42    ///
43    /// [`enumerate_devices`](Self::enumerate_devices) of its type, for a
44    /// runtime whose ids carry nothing more. One whose ids do answers from the
45    /// list that extra part names — a wgpu device pinned to a graphics API,
46    /// from that API's, the same adapter on another being another device.
47    fn enumerate_devices_like(device_id: DeviceId) -> Vec<DeviceId> {
48        Self::enumerate_devices(device_id.type_id)
49    }
50
51    /// Whether this machine has the device `device_id` names — and where it
52    /// does not, how many of that kind it has instead.
53    ///
54    /// What naming a device is checked against. The runtime's "you choose"
55    /// device names no hardware of its own, so it is there as soon as the
56    /// runtime found anything of its kind.
57    fn find_device(device_id: DeviceId) -> Result<(), usize> {
58        let of_kind = Self::enumerate_devices_like(device_id);
59
60        let found = of_kind.contains(&device_id)
61            || (device_id == Self::Device::default().to_id() && !of_kind.is_empty());
62
63        match found {
64            true => Ok(()),
65            false => Err(of_kind.len()),
66        }
67    }
68
69    /// Whether this machine has hardware worth choosing this runtime for.
70    ///
71    /// What picking a default device walks, so the question is not "could this
72    /// runtime run at all" but "would a caller who did not choose want it".
73    /// Enumerating a device is the answer for most runtimes; one that also
74    /// exposes a software fallback answers no on a machine that has only that,
75    /// leaving a native CPU runtime to claim it.
76    fn is_available() -> bool {
77        !Self::enumerate_all_devices().is_empty()
78    }
79
80    /// Description of the device (name, kind, memory, card, architecture, driver) read from the
81    /// driver without starting a server.
82    ///
83    /// Returns `None` when the machine has no such device, the device is the runtime's "you
84    /// choose" device (which names no hardware until a server picks some), or the runtime
85    /// describes nothing without a server.
86    fn description(_device: &Self::Device) -> Option<DeviceDescription> {
87        None
88    }
89
90    /// How busy `device` is, by the counter its card's driver or its platform keeps. Asking
91    /// creates the device's client, and must never wait on the work the device is running.
92    ///
93    /// A runtime that reads no counter for its devices says so.
94    fn utilization(_device: &Self::Device) -> Result<DeviceUtilization, UtilizationUnavailable> {
95        Err(UtilizationUnavailable::RuntimeReadsNoCounter)
96    }
97
98    /// The current temperature of `device`. Creates a client if needed.
99    ///
100    /// Returns an [`UtilizationUnavailable::RuntimeReadsNoCounter`]
101    /// error if the runtime doesn't read any counter.
102    fn temperature(_device: &Self::Device) -> Result<DeviceTemperature, UtilizationUnavailable> {
103        Err(UtilizationUnavailable::RuntimeReadsNoCounter)
104    }
105
106    /// The power usage of `device`, in Watt. Creates a client if needed.
107    ///
108    /// Returns an [`UtilizationUnavailable::RuntimeReadsNoCounter`]
109    /// error if the runtime doesn't read any counter.
110    fn power(_device: &Self::Device) -> Result<DevicePower, UtilizationUnavailable> {
111        Err(UtilizationUnavailable::RuntimeReadsNoCounter)
112    }
113}
cubecl_runtime/runtime.rs

cubecl_runtime/
runtime.rs
