runmat-accelerate 0.4.5

Pluggable GPU acceleration layer for RunMat (CUDA, ROCm, Metal, Vulkan/Spir-V)
Documentation 
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

use super::common::submit;

pub fn warmup_noop(device: &wgpu::Device, queue: &wgpu::Queue, pipeline: &wgpu::ComputePipeline) {
    let mut enc = device.create_command_encoder(&wgpu::CommandEncoderDescriptor {
        label: Some("runmat-noop-elementwise"),
    });
    {
        let mut pass = enc.begin_compute_pass(&wgpu::ComputePassDescriptor {
            label: Some("runmat-noop-pass-elementwise"),
            timestamp_writes: None,
        });
        pass.set_pipeline(pipeline);
    }
    submit(device, queue, enc);
}

pub fn run(
    device: &wgpu::Device,
    queue: &wgpu::Queue,
    pipeline: &wgpu::ComputePipeline,
    bind_group: &wgpu::BindGroup,
    workgroups_x: u32,
) {
    let mut enc = device.create_command_encoder(&wgpu::CommandEncoderDescriptor {
        label: Some("runmat-elementwise-encoder"),
    });
    {
        let mut pass = enc.begin_compute_pass(&wgpu::ComputePassDescriptor {
            label: Some("runmat-elementwise-pass"),
            timestamp_writes: None,
        });
        pass.set_pipeline(pipeline);
        pass.set_bind_group(0, bind_group, &[]);
        if workgroups_x > 0 {
            pass.dispatch_workgroups(workgroups_x, 1, 1);
        }
    }
    submit(device, queue, enc);
}