mulimoen/SummationByParts
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

simplify RK logic

Browse Source
This commit is contained in:
Magnus Ulimoen
2019-08-09 18:28:40 +02:00
parent 3ab7e532c6
commit ceee1463dd
1 changed files with 48 additions and 142 deletions
Download Patch File Download Diff File Expand all files Collapse all files

190
src/lib.rs
View File

@@ -2,7 +2,7 @@ use ndarray::{Array2, Zip};
use wasm_bindgen::prelude::*;
mod operators;
use operators::{diffx, diffx_periodic, diffy, diffy_periodic, dissx, dissy};
use operators::diffx_periodic;
#[cfg(feature = "wee_alloc")]
#[global_allocator]
@@ -87,150 +87,62 @@ impl Universe {
pub fn advance(&self, fut: &mut Universe, dt: f32, work_buffers: Option<WorkBuffers>) {
assert_eq!(self.ex.shape(), fut.ex.shape());
let mut y_ex = self.ex.clone();
let mut y_hz = self.hz.clone();
let mut buffers = work_buffers
.unwrap_or_else(|| WorkBuffers::new(self.ex.shape()[1], self.ex.shape()[0]));
let mut k0_ex = &mut buffers.k1_ex;
let mut k0_hz = &mut buffers.k1_hz;
k0_ex.fill(0.0);
k0_hz.fill(0.0);
// u_x
diffx_periodic(y_ex.view(), k0_ex.view_mut());
diffx_periodic(y_hz.view(), k0_hz.view_mut());
let mut y = buffers.y;
let mut k = buffers.buf;
// y = y + 1/2*dt Au_x
y_ex.scaled_add(1.0 / 2.0 * dt, &k0_hz);
y_hz.scaled_add(1.0 / 2.0 * dt, &k0_ex);
let mut k1_ex = buffers.k1_ex;
let mut k1_hz = buffers.k1_hz;
k1_ex.fill(0.0);
k1_hz.fill(0.0);
diffx_periodic(y_ex.view(), k1_ex.view_mut());
diffx_periodic(y_hz.view(), k1_hz.view_mut());
for i in 0..4 {
// y = y0 + c*kn
y.0.assign(&self.ex);
y.1.assign(&self.hz);
match i {
0 => {}
1 => {
y.0.scaled_add(1.0 / 2.0 * dt, &k[i - 1].0);
y.1.scaled_add(1.0 / 2.0 * dt, &k[i - 1].1);
}
2 => {
y.0.scaled_add(1.0 / 2.0 * dt, &k[i - 1].0);
y.1.scaled_add(1.0 / 2.0 * dt, &k[i - 1].1);
}
3 => {
y.0.scaled_add(dt, &k[i - 1].0);
y.1.scaled_add(dt, &k[i - 1].1);
}
_ => {
unreachable!();
}
};
y_ex.assign(&self.ex);
y_hz.assign(&self.hz);
y_ex.scaled_add(1.0 / 2.0 * dt, &k1_hz);
y_hz.scaled_add(1.0 / 2.0 * dt, &k1_ex);
let mut k2_ex = buffers.k2_ex;
let mut k2_hz = buffers.k2_hz;
k2_ex.fill(0.0);
k2_hz.fill(0.0);
diffx_periodic(y_ex.view(), k2_ex.view_mut());
diffx_periodic(y_hz.view(), k2_hz.view_mut());
k[i].0.fill(0.0);
k[i].1.fill(0.0);
// y = y0
y_ex.assign(&self.ex);
y_hz.assign(&self.hz);
y_ex.scaled_add(1.0 / 2.0 * dt, &k2_hz);
y_hz.scaled_add(1.0 / 2.0 * dt, &k2_ex);
let mut k3_ex = buffers.k3_ex;
let mut k3_hz = buffers.k3_hz;
k3_ex.fill(0.0);
k3_hz.fill(0.0);
diffx_periodic(y_ex.view(), k3_ex.view_mut());
diffx_periodic(y_hz.view(), k3_hz.view_mut());
y_ex.assign(&self.ex);
y_hz.assign(&self.hz);
y_ex.scaled_add(dt, &k2_hz);
y_hz.scaled_add(dt, &k2_ex);
let mut k4_ex = buffers.k4_ex;
let mut k4_hz = buffers.k4_hz;
k4_ex.fill(0.0);
k4_hz.fill(0.0);
diffx_periodic(y_ex.view(), k4_ex.view_mut());
diffx_periodic(y_hz.view(), k4_hz.view_mut());
diffx_periodic(y.0.view(), k[i].1.view_mut());
diffx_periodic(y.1.view(), k[i].0.view_mut());
}
Zip::from(&mut fut.ex)
.and(&self.ex)
.and(&k1_hz)
.and(&k2_hz)
.and(&k3_hz)
.and(&k4_hz)
.and(&k[0].0)
.and(&k[1].0)
.and(&k[2].0)
.and(&k[3].0)
.apply(|y1, &y0, &k1, &k2, &k3, &k4| {
*y1 = y0 + dt / 6.0 * (k1 + 2.0 * k2 + 2.0 * k3 + k4)
});
Zip::from(&mut fut.hz)
.and(&self.hz)
.and(&k1_ex)
.and(&k2_ex)
.and(&k3_ex)
.and(&k4_ex)
.and(&k[0].1)
.and(&k[1].1)
.and(&k[2].1)
.and(&k[3].1)
.apply(|y1, &y0, &k1, &k2, &k3, &k4| {
*y1 = y0 + dt / 6.0 * (k1 + 2.0 * k2 + 2.0 * k3 + k4)
});
}
pub fn advance_upwind(&self, fut: &mut Universe, dt: f32) {
assert_eq!(self.ex.dim(), fut.ex.dim());
let mut ad = Array2::zeros(fut.ex.dim());
let mut y = self.ex.clone();
// y.assign(&self.ex)
// ad.fill(0.0);
let mut k0 = Array2::zeros(self.ex.dim());
diffx(y.view(), k0.view_mut());
diffy(y.view(), k0.view_mut());
dissx(y.view(), ad.view_mut());
dissy(y.view(), ad.view_mut());
k0.scaled_add(-0.5, &ad);
// y.assign(&self.ex);
ad.fill(0.0);
y.scaled_add(-1.0 / 2.0 * dt, &k0);
let mut k1 = Array2::zeros(self.ex.dim());
diffx(y.view(), k1.view_mut());
diffy(y.view(), k1.view_mut());
dissx(y.view(), ad.view_mut());
dissy(y.view(), ad.view_mut());
k1.scaled_add(-0.5, &ad);
y.assign(&self.ex);
ad.fill(0.0);
y.scaled_add(-1.0 / 2.0 * dt, &k1);
let mut k2 = Array2::zeros(self.ex.dim());
diffx(y.view(), k2.view_mut());
diffy(y.view(), k2.view_mut());
dissx(y.view(), ad.view_mut());
dissy(y.view(), ad.view_mut());
k2.scaled_add(-0.5, &ad);
y.assign(&self.ex);
ad.fill(0.0);
y.scaled_add(-1.0 / 2.0 * dt, &k2);
let mut k3 = Array2::zeros(self.ex.dim());
diffx(y.view(), k3.view_mut());
diffy(y.view(), k3.view_mut());
dissx(y.view(), ad.view_mut());
dissy(y.view(), ad.view_mut());
k3.scaled_add(-0.5, &ad);
y.assign(&self.ex);
ad.fill(0.0);
y.scaled_add(-dt, &k2);
let mut k4 = Array2::zeros(self.ex.dim());
diffx(y.view(), k4.view_mut());
diffy(y.view(), k4.view_mut());
dissx(y.view(), ad.view_mut());
dissy(y.view(), ad.view_mut());
k4.scaled_add(-0.5, &ad);
Zip::from(&mut fut.ex)
.and(&self.ex)
.and(&k1)
.and(&k2)
.and(&k3)
.and(&k4)
.apply(|y1, &y0, &k1, &k2, &k3, &k4| {
*y1 = y0 + -dt / 6.0 * (k1 + 2.0 * k2 + 2.0 * k3 + k4)
});
}
pub fn get_ptr(&mut self) -> *mut u8 {
self.ex.as_mut_ptr() as *mut u8
}
@@ -238,28 +150,22 @@ impl Universe {
#[wasm_bindgen]
pub struct WorkBuffers {
k1_ex: Array2<f32>,
k1_hz: Array2<f32>,
k2_ex: Array2<f32>,
k2_hz: Array2<f32>,
k3_ex: Array2<f32>,
k3_hz: Array2<f32>,
k4_ex: Array2<f32>,
k4_hz: Array2<f32>,
y: (Array2<f32>, Array2<f32>),
buf: [(Array2<f32>, Array2<f32>); 4],
}
#[wasm_bindgen]
impl WorkBuffers {
pub fn new(nx: usize, ny: usize) -> Self {
let arr = Array2::zeros((ny, nx));
Self {
k1_ex: Array2::zeros((ny, nx)),
k1_hz: Array2::zeros((ny, nx)),
k2_ex: Array2::zeros((ny, nx)),
k2_hz: Array2::zeros((ny, nx)),
k3_ex: Array2::zeros((ny, nx)),
k3_hz: Array2::zeros((ny, nx)),
k4_ex: Array2::zeros((ny, nx)),
k4_hz: Array2::zeros((ny, nx)),
y: (arr.clone(), arr.clone()),
buf: [
(arr.clone(), arr.clone()),
(arr.clone(), arr.clone()),
(arr.clone(), arr.clone()),
(arr.clone(), arr),
],
}
}
}