solvr 0.2.0

Advanced computing library for real-world problem solving - optimization, differential equations, interpolation, statistics, and more
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35

//! CUDA implementation of median filter algorithms.
//!
//! Median filtering is CPU-only because computing median requires sorting/selection
//! which doesn't parallelize efficiently on GPU for small sliding windows.

use crate::signal::traits::medfilt::MedianFilterAlgorithms;
use numr::error::{Error, Result};
use numr::runtime::cuda::{CudaClient, CudaRuntime};
use numr::tensor::Tensor;

impl MedianFilterAlgorithms<CudaRuntime> for CudaClient {
    fn medfilt(
        &self,
        _x: &Tensor<CudaRuntime>,
        _kernel_size: usize,
    ) -> Result<Tensor<CudaRuntime>> {
        Err(Error::backend_limitation(
            "cuda",
            "medfilt",
            "Median filtering is CPU-only due to sorting requirements. Transfer data to CPU first.",
        ))
    }

    fn medfilt2d(
        &self,
        _x: &Tensor<CudaRuntime>,
        _kernel_size: (usize, usize),
    ) -> Result<Tensor<CudaRuntime>> {
        Err(Error::backend_limitation(
            "cuda",
            "medfilt2d",
            "Median filtering is CPU-only due to sorting requirements. Transfer data to CPU first.",
        ))
    }
}