use crate::device::DeviceCapabilities;
use crate::error::Result;
use crate::tensor::Tensor;
pub struct BesselJ0 {
input: Tensor,
}
impl BesselJ0 {
#[must_use]
pub fn new(input: Tensor) -> Self {
Self { input }
}
fn wgsl_shader() -> &'static str {
include_str!("../shaders/special/bessel_j0.wgsl")
}
pub fn execute(self) -> Result<Tensor> {
let device = self.input.device();
let size: usize = self.input.shape().iter().product();
let output_buffer = device.create_buffer_f32(size)?;
#[repr(C)]
#[derive(Copy, Clone, bytemuck::Pod, bytemuck::Zeroable)]
struct Metadata {
size: u32,
}
let metadata = Metadata { size: size as u32 };
let metadata_buffer = device
.device
.create_buffer_init(&wgpu::util::BufferInitDescriptor {
label: Some("BESSEL_J0 Metadata"),
contents: bytemuck::cast_slice(&[metadata]),
usage: wgpu::BufferUsages::UNIFORM,
});
let bind_group_layout =
device
.device
.create_bind_group_layout(&wgpu::BindGroupLayoutDescriptor {
label: Some("BESSEL_J0 Bind Group Layout"),
entries: &[
wgpu::BindGroupLayoutEntry {
binding: 0,
visibility: wgpu::ShaderStages::COMPUTE,
ty: wgpu::BindingType::Buffer {
ty: wgpu::BufferBindingType::Storage { read_only: true },
has_dynamic_offset: false,
min_binding_size: None,
},
count: None,
},
wgpu::BindGroupLayoutEntry {
binding: 1,
visibility: wgpu::ShaderStages::COMPUTE,
ty: wgpu::BindingType::Buffer {
ty: wgpu::BufferBindingType::Storage { read_only: false },
has_dynamic_offset: false,
min_binding_size: None,
},
count: None,
},
wgpu::BindGroupLayoutEntry {
binding: 2,
visibility: wgpu::ShaderStages::COMPUTE,
ty: wgpu::BindingType::Buffer {
ty: wgpu::BufferBindingType::Uniform,
has_dynamic_offset: false,
min_binding_size: None,
},
count: None,
},
],
});
let bind_group = device.device.create_bind_group(&wgpu::BindGroupDescriptor {
label: Some("BESSEL_J0 Bind Group"),
layout: &bind_group_layout,
entries: &[
wgpu::BindGroupEntry {
binding: 0,
resource: self.input.buffer().as_entire_binding(),
},
wgpu::BindGroupEntry {
binding: 1,
resource: output_buffer.as_entire_binding(),
},
wgpu::BindGroupEntry {
binding: 2,
resource: metadata_buffer.as_entire_binding(),
},
],
});
let shader = device.compile_shader(Self::wgsl_shader(), Some("BESSEL_J0"));
let pipeline_layout =
device
.device
.create_pipeline_layout(&wgpu::PipelineLayoutDescriptor {
label: Some("BESSEL_J0 Pipeline Layout"),
bind_group_layouts: &[&bind_group_layout],
immediate_size: 0,
});
let pipeline = device
.device
.create_compute_pipeline(&wgpu::ComputePipelineDescriptor {
label: Some("BESSEL_J0 Pipeline"),
layout: Some(&pipeline_layout),
module: &shader,
entry_point: Some("main"),
cache: None,
compilation_options: Default::default(),
});
let mut encoder = device.create_encoder_guarded(&wgpu::CommandEncoderDescriptor {
label: Some("BESSEL_J0 Encoder"),
});
{
let mut pass = encoder.begin_compute_pass(&wgpu::ComputePassDescriptor {
label: Some("BESSEL_J0 Pass"),
timestamp_writes: None,
});
pass.set_pipeline(&pipeline);
pass.set_bind_group(0, Some(&bind_group), &[]);
let caps = DeviceCapabilities::from_device(device);
let workgroups = caps.dispatch_1d(size as u32);
pass.dispatch_workgroups(workgroups, 1, 1);
}
device.submit_commands(Some(encoder.finish()));
Ok(Tensor::from_buffer(
output_buffer,
self.input.shape().to_vec(),
device.clone(),
))
}
}
impl Tensor {
pub fn bessel_j0(self) -> Result<Self> {
BesselJ0::new(self).execute()
}
}
#[cfg(test)]
mod tests {
use super::*;
#[tokio::test]
async fn test_bessel_j0() {
let device = crate::device::test_pool::get_test_device().await;
let data = vec![0.0, 1.0, 2.0, 5.0, 8.0];
let input = Tensor::new(data, vec![5], device);
let output = input.bessel_j0().unwrap();
let result = output.to_vec().unwrap();
assert!(
(result[0] - 1.0).abs() < 1e-5,
"J₀(0) should be 1, got {}",
result[0]
);
assert!(result.iter().all(|&x| x.is_finite()));
}
}