Macro offload 

macro_rules! offload {
    (
        @ munch[kernel = $val:expr $(, $($rest:tt)*)?]; kernel = NONE;
        workgroup_dim = $w:tt; thread_dim = $t:tt; dyn_cache = $d:tt; device =
        $device:tt; args = $a:tt
    ) => { ... };
    (
        @ munch[kernel = $val:expr $(, $($rest:tt)*)?]; kernel = (SOME
        $old:expr); workgroup_dim = $w:tt; thread_dim = $t:tt; dyn_cache =
        $d:tt; device = $device:tt; args = $a:tt
    ) => { ... };
    (
        @ munch[workgroup_dim = $val:expr $(, $($rest:tt)*)?]; kernel = $k:tt;
        workgroup_dim = ([1, 1, 1]); thread_dim = $t:tt; dyn_cache = $d:tt;
        device = $device:tt; args = $a:tt
    ) => { ... };
    (
        @ munch[workgroup_dim = $val:expr $(, $($rest:tt)*)?]; kernel = $k:tt;
        workgroup_dim = (SOME $old:expr); thread_dim = $t:tt; dyn_cache =
        $d:tt; device = $device:tt; args = $a:tt
    ) => { ... };
    (
        @ munch[thread_dim = $val:expr $(, $($rest:tt)*)?]; kernel = $k:tt;
        workgroup_dim = $w:tt; thread_dim = ([1, 1, 1]); dyn_cache = $d:tt;
        device = $device:tt; args = $a:tt
    ) => { ... };
    (
        @ munch[thread_dim = $val:expr $(, $($rest:tt)*)?]; kernel = $k:tt;
        workgroup_dim = $w:tt; thread_dim = (SOME $old:expr); dyn_cache =
        $d:tt; device = $device:tt; args = $a:tt
    ) => { ... };
    (
        @ munch[dyn_cache = $val:expr $(, $($rest:tt)*)?]; kernel = $k:tt;
        workgroup_dim = $w:tt; thread_dim = $t:tt; dyn_cache = (0); device =
        $device:tt; args = $a:tt
    ) => { ... };
    (
        @ munch[dyn_cache = $val:expr $(, $($rest:tt)*)?]; kernel = $k:tt;
        workgroup_dim = $w:tt; thread_dim = $t:tt; dyn_cache = (SOME
        $old:expr); device = $device:tt; args = $a:tt
    ) => { ... };
    (
        @ munch[device = $val:expr $(, $($rest:tt)*)?]; kernel = $k:tt;
        workgroup_dim = $w:tt; thread_dim = $t:tt; dyn_cache = $d:tt; device =
        NONE; args = $a:tt
    ) => { ... };
    (
        @ munch[device = $val:expr $(, $($rest:tt)*)?]; kernel = $k:tt;
        workgroup_dim = $w:tt; thread_dim = $t:tt; dyn_cache = $d:tt; device =
        (SOME $old:expr); args = $a:tt
    ) => { ... };
    (
        @ munch[args = ($($arg:expr),* $(,)?) $(, $($rest:tt)*)?]; kernel =
        $k:tt; workgroup_dim = $w:tt; thread_dim = $t:tt; dyn_cache = $d:tt;
        device = $device:tt; args = NONE
    ) => { ... };
    (
        @ munch[args = $val:expr $(, $($rest:tt)*)?]; kernel = $k:tt;
        workgroup_dim = $w:tt; thread_dim = $t:tt; dyn_cache = $d:tt; device =
        $device:tt; args = NONE
    ) => { ... };
    (
        @ munch[args = $val:expr $(, $($rest:tt)*)?]; kernel = $k:tt;
        workgroup_dim = $w:tt; thread_dim = $t:tt; dyn_cache = $d:tt; device =
        $device:tt; args = (SOME $old:expr)
    ) => { ... };
    (
        @ munch[args = $val:expr $(, $($rest:tt)*)?]; kernel = $k:tt;
        workgroup_dim = $w:tt; thread_dim = $t:tt; dyn_cache = $d:tt; device =
        $device:tt; args = (SOME_TUPLE $old:tt)
    ) => { ... };
    (
        @ munch[$invalid:ident = $val:expr $(, $($rest:tt)*)?]; kernel =
        $k:tt; workgroup_dim = $w:tt; thread_dim = $t:tt; dyn_cache = $d:tt;
        device = $device:tt; args = $a:tt
    ) => { ... };
    (
        @ munch[]; kernel = NONE; workgroup_dim = $w:tt; thread_dim = $t:tt;
        dyn_cache = $d:tt; device = $device:tt; args = $a:tt
    ) => { ... };
    (
        @ munch[]; kernel = $k:tt; workgroup_dim = $w:tt; thread_dim = $t:tt;
        dyn_cache = $d:tt; device = $device:tt; args = NONE
    ) => { ... };
    (
        @ munch[]; kernel = (SOME $kernel:expr); workgroup_dim = $w:tt;
        thread_dim = $t:tt; dyn_cache = $d:tt; device = $device:tt; args =
        (SOME_TUPLE($($arg:expr),* $(,)?))
    ) => { ... };
    (
        @ munch[]; kernel = (SOME $kernel:expr); workgroup_dim = $w:tt;
        thread_dim = $t:tt; dyn_cache = $d:tt; device = $device:tt; args =
        (SOME $args:expr)
    ) => { ... };
    (@ value(SOME $val:expr)) => { ... };
    (@ value($val:expr)) => { ... };
    (@ device NONE) => { ... };
    (@ device(SOME $val:expr)) => { ... };
    ($($tt:tt)*) => { ... };
}

🔬This is a nightly-only experimental API. (gpu_offload #131513)

Launches a kernel on an offload device (e.g., a GPU).

This macro is an interface over the offload intrinsic. The kernel itself must be defined using the offload_kernel macro.

The following named arguments are accepted:

• kernel: The kernel function to offload. Must be a function item. (required)
• args: A tuple of arguments forwarded to kernel. (required)
• workgroup_dim: A 3D size specifying the number of workgroups to launch. Defaults to [1, 1, 1].
• thread_dim: A 3D size specifying the number of threads per workgroup. Defaults to [1, 1, 1].
• dyn_cache: The amount of dynamic shared memory, in bytes, to allocate for the kernel. Defaults to 0.
• device: The index of the device to offload to. Must be >= 0. If omitted, the default device is used. Use crate::intrinsics::offload_get_num_devices to discover which device ids are valid.

Each argument may only be specified once.

Examples

let mut x = [0.0f64; 256];
core::offload::offload! {
    kernel = kernel,
    workgroup_dim = [256, 1, 1],
    args = (&mut x as *mut [f64; 256],),
}