use crate::device::{DeviceCapabilities, WorkloadType};
use crate::error::Result;
use crate::tensor::Tensor;
pub struct Digamma {
input: Tensor,
}
impl Digamma {
#[must_use]
pub fn new(input: Tensor) -> Self {
Self { input }
}
fn wgsl_shader() -> &'static str {
include_str!("../shaders/special/digamma.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 Params {
size: u32,
}
let params = Params { size: size as u32 };
let params_buffer = device
.device
.create_buffer_init(&wgpu::util::BufferInitDescriptor {
label: Some("Digamma Params"),
contents: bytemuck::cast_slice(&[params]),
usage: wgpu::BufferUsages::UNIFORM,
});
let bind_group_layout =
device
.device
.create_bind_group_layout(&wgpu::BindGroupLayoutDescriptor {
label: Some("Digamma 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("Digamma 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: params_buffer.as_entire_binding(),
},
],
});
let shader = device.compile_shader(Self::wgsl_shader(), Some("Digamma"));
let pipeline_layout =
device
.device
.create_pipeline_layout(&wgpu::PipelineLayoutDescriptor {
label: Some("Digamma Pipeline Layout"),
bind_group_layouts: &[&bind_group_layout],
immediate_size: 0,
});
let pipeline = device
.device
.create_compute_pipeline(&wgpu::ComputePipelineDescriptor {
label: Some("Digamma 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("Digamma Encoder"),
});
{
let mut pass = encoder.begin_compute_pass(&wgpu::ComputePassDescriptor {
label: Some("Digamma Pass"),
timestamp_writes: None,
});
pass.set_pipeline(&pipeline);
pass.set_bind_group(0, Some(&bind_group), &[]);
let caps = DeviceCapabilities::from_device(device);
let optimal_wg_size = caps.optimal_workgroup_size(WorkloadType::ElementWise);
let workgroups = (size as u32).div_ceil(optimal_wg_size);
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 digamma(self) -> Result<Self> {
Digamma::new(self).execute()
}
}
#[cfg(test)]
mod tests {
use super::*;
const EULER_MASCHERONI: f32 = 0.577_215_7;
#[tokio::test]
async fn test_digamma_at_1() {
let device = crate::device::test_pool::get_test_device().await;
let input = Tensor::new(vec![1.0], vec![1], device);
let output = input.digamma().unwrap();
let result = output.to_vec().unwrap();
let expected = -EULER_MASCHERONI;
assert!(
(result[0] - expected).abs() < 0.001f32,
"ψ(1) = {}, expected {}",
result[0],
expected
);
}
#[tokio::test]
async fn test_digamma_at_2() {
let device = crate::device::test_pool::get_test_device().await;
let input = Tensor::new(vec![2.0], vec![1], device);
let output = input.digamma().unwrap();
let result = output.to_vec().unwrap();
let expected = 1.0 - EULER_MASCHERONI;
assert!(
(result[0] - expected).abs() < 0.001f32,
"ψ(2) = {}, expected {}",
result[0],
expected
);
}
#[tokio::test]
async fn test_digamma_large() {
let device = crate::device::test_pool::get_test_device().await;
let input = Tensor::new(vec![100.0], vec![1], device);
let output = input.digamma().unwrap();
let result = output.to_vec().unwrap();
let expected = 100.0_f32.ln() - 0.5 / 100.0;
assert!(
(result[0] - expected).abs() < 0.01,
"ψ(100) = {}, expected ~{}",
result[0],
expected
);
}
}