Skip to main content

launch/
launch.rs

1#[macro_use]
2extern crate rustacuda;
3
4use rustacuda::prelude::*;
5use std::error::Error;
6use std::ffi::CString;
7
8fn main() -> Result<(), Box<dyn Error>> {
9    // Set up the context, load the module, and create a stream to run kernels in.
10    rustacuda::init(CudaFlags::empty())?;
11    let device = Device::get_device(0)?;
12    let _ctx = Context::create_and_push(ContextFlags::MAP_HOST | ContextFlags::SCHED_AUTO, device)?;
13
14    let ptx = CString::new(include_str!("../resources/add.ptx"))?;
15    let module = Module::load_from_string(&ptx)?;
16    let stream = Stream::new(StreamFlags::NON_BLOCKING, None)?;
17
18    // Create buffers for data
19    let mut in_x = DeviceBuffer::from_slice(&[1.0f32; 10])?;
20    let mut in_y = DeviceBuffer::from_slice(&[2.0f32; 10])?;
21    let mut out_1 = DeviceBuffer::from_slice(&[0.0f32; 10])?;
22    let mut out_2 = DeviceBuffer::from_slice(&[0.0f32; 10])?;
23
24    // This kernel adds each element in `in_x` and `in_y` and writes the result into `out`.
25    unsafe {
26        // Launch the kernel with one block of one thread, no dynamic shared memory on `stream`.
27        let result = launch!(module.sum<<<1, 1, 0, stream>>>(
28            in_x.as_device_ptr(),
29            in_y.as_device_ptr(),
30            out_1.as_device_ptr(),
31            out_1.len()
32        ));
33        result?;
34
35        // Launch the kernel again using the `function` form:
36        let function_name = CString::new("sum")?;
37        let sum = module.get_function(&function_name)?;
38        // Launch with 1x1x1 (1) blocks of 10x1x1 (10) threads, to show that you can use tuples to
39        // configure grid and block size.
40        let result = launch!(sum<<<(1, 1, 1), (10, 1, 1), 0, stream>>>(
41            in_x.as_device_ptr(),
42            in_y.as_device_ptr(),
43            out_2.as_device_ptr(),
44            out_2.len()
45        ));
46        result?;
47    }
48
49    // Kernel launches are asynchronous, so we wait for the kernels to finish executing.
50    stream.synchronize()?;
51
52    // Copy the results back to host memory
53    let mut out_host = [0.0f32; 20];
54    out_1.copy_to(&mut out_host[0..10])?;
55    out_2.copy_to(&mut out_host[10..20])?;
56
57    for x in out_host.iter() {
58        assert_eq!(3.0 as u32, *x as u32);
59    }
60
61    println!("Launched kernel successfully.");
62    Ok(())
63}