use crate::device::{DeviceCapabilities, WorkloadType};
use crate::error::Result;
use crate::tensor::Tensor;
pub struct NormPpf {
input: Tensor,
mu: f32,
sigma: f32,
}
impl NormPpf {
#[must_use]
pub fn standard(input: Tensor) -> Self {
Self {
input,
mu: 0.0,
sigma: 1.0,
}
}
#[must_use]
pub fn general(input: Tensor, mu: f32, sigma: f32) -> Self {
Self { input, mu, sigma }
}
fn wgsl_shader() -> &'static str {
static SHADER: std::sync::LazyLock<String> = std::sync::LazyLock::new(|| {
include_str!("../shaders/special/norm_ppf_f64.wgsl").to_string()
});
&SHADER
}
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,
mu: f32,
sigma: f32,
}
let params = Params {
size: size as u32,
mu: self.mu,
sigma: self.sigma,
};
let params_buffer = device
.device
.create_buffer_init(&wgpu::util::BufferInitDescriptor {
label: Some("NormPpf Params"),
contents: bytemuck::cast_slice(&[params]),
usage: wgpu::BufferUsages::UNIFORM,
});
let bind_group_layout =
device
.device
.create_bind_group_layout(&wgpu::BindGroupLayoutDescriptor {
label: Some("NormPpf 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("NormPpf 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("NormPpf"));
let pipeline_layout =
device
.device
.create_pipeline_layout(&wgpu::PipelineLayoutDescriptor {
label: Some("NormPpf Pipeline Layout"),
bind_group_layouts: &[&bind_group_layout],
immediate_size: 0,
});
let pipeline = device
.device
.create_compute_pipeline(&wgpu::ComputePipelineDescriptor {
label: Some("NormPpf 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("NormPpf Encoder"),
});
{
let mut pass = encoder.begin_compute_pass(&wgpu::ComputePassDescriptor {
label: Some("NormPpf 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 norm_ppf(self) -> Result<Self> {
NormPpf::standard(self).execute()
}
pub fn norm_ppf_params(self, mu: f32, sigma: f32) -> Result<Self> {
NormPpf::general(self, mu, sigma).execute()
}
}
#[cfg(test)]
mod tests {
use super::*;
#[tokio::test]
async fn test_norm_ppf_median() {
let device = crate::device::test_pool::get_test_device().await;
let input = Tensor::new(vec![0.5], vec![1], device);
let output = input.norm_ppf().unwrap();
let result = output.to_vec().unwrap();
assert!(
result[0].abs() < 0.001,
"Φ⁻¹(0.5) = {}, expected 0",
result[0]
);
}
#[tokio::test]
async fn test_norm_ppf_quartiles() {
let device = crate::device::test_pool::get_test_device().await;
let input = Tensor::new(vec![0.25, 0.75], vec![2], device);
let output = input.norm_ppf().unwrap();
let result = output.to_vec().unwrap();
assert!(
(result[0] - (-0.6745)).abs() < 0.01,
"Φ⁻¹(0.25) = {}, expected -0.6745",
result[0]
);
assert!(
(result[1] - 0.6745).abs() < 0.01,
"Φ⁻¹(0.75) = {}, expected 0.6745",
result[1]
);
}
#[tokio::test]
async fn test_norm_ppf_critical() {
let device = crate::device::test_pool::get_test_device().await;
let input = Tensor::new(vec![0.025, 0.975], vec![2], device);
let output = input.norm_ppf().unwrap();
let result = output.to_vec().unwrap();
assert!(
(result[0] - (-1.96)).abs() < 0.02,
"Φ⁻¹(0.025) = {}, expected -1.96",
result[0]
);
assert!(
(result[1] - 1.96).abs() < 0.02,
"Φ⁻¹(0.975) = {}, expected 1.96",
result[1]
);
}
#[tokio::test]
async fn test_norm_ppf_general() {
let device = crate::device::test_pool::get_test_device().await;
let input = Tensor::new(vec![0.5], vec![1], device);
let output = input.norm_ppf_params(10.0, 2.0).unwrap();
let result = output.to_vec().unwrap();
assert!(
(result[0] - 10.0).abs() < 0.01,
"Φ⁻¹(0.5; 10, 2) = {}, expected 10",
result[0]
);
}
}