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
use objc2::{extern_protocol, runtime::ProtocolObject};
use crate::*;
extern_protocol!(
/// Encodes commands for dispatching machine learning networks on Apple silicon.
///
/// See also [Apple's documentation](https://developer.apple.com/documentation/metal/mtl4machinelearningcommandencoder?language=objc)
pub unsafe trait MTL4MachineLearningCommandEncoder: MTL4CommandEncoder {
/// Configures the encoder with a machine learning pipeline state instance.
///
/// The pipeline state instance affects all subsequent Machine Learning commands.
///
/// - Parameters:
/// - pipelineState: A Machine Learning pipeline state instance.
#[unsafe(method(setPipelineState:))]
#[unsafe(method_family = none)]
fn set_pipeline_state(
&self,
pipeline_state: &ProtocolObject<dyn MTL4MachineLearningPipelineState>,
);
/// Sets an argument table for the command encoder's machine learning shader stage.
///
/// The argument table provides inputs to all subsequent Machine Learning dispatches.
/// - Parameters:
/// - argumentTable: An argument table to set on the command encoder's Machine Learning stage.
#[unsafe(method(setArgumentTable:))]
#[unsafe(method_family = none)]
fn set_argument_table(
&self,
argument_table: &ProtocolObject<dyn MTL4ArgumentTable>,
);
/// Dispatches a machine learning network using the current pipeline state and argument table.
///
/// This method takes a parameter consisting of a `MTLHeap` that Metal can use to allocate intermediate tensors.
/// You can query the minimum size Metal requires for this heap by calling
/// ``MTL4MachineLearningPipelineState/intermediatesHeapSize``.
///
/// - Parameters:
/// - heap: a heap that Metal can use to allocate intermediate tensors.
#[unsafe(method(dispatchNetworkWithIntermediatesHeap:))]
#[unsafe(method_family = none)]
fn dispatch_network_with_intermediates_heap(
&self,
heap: &ProtocolObject<dyn MTLHeap>,
);
}
);