spirv: port all backends to use R for the stimulus

particularly, this patches over a difference where the gpu backend
expected the stimulus to be R, while the CPU thought it should be f32.
that would likely have revealed a crash if we had tested it with f64
(TODO).

crates/coremem/src/sim/spirv/cpu.rs

@@ -16,8 +16,8 @@ impl<R: Real, M: Material<R>> SimBackend<R, M> for CpuBackend {
         &self,
         num_steps: u32,
         meta: SimMeta<R>,
-        stim_e: &[Vec3<f32>],
-        stim_h: &[Vec3<f32>],
+        stim_e: &[Vec3<R>],
+        stim_h: &[Vec3<R>],
         mat: &[M],
         e: &mut [Vec3<R>],
         h: &mut [Vec3<R>],
@@ -32,7 +32,7 @@ impl<R: Real, M: Material<R>> SimBackend<R, M> for CpuBackend {
 
 fn step_e<R: Real, M: Material<R>>(
     meta: SimMeta<R>,
-    stim_e: &[Vec3<f32>],
+    stim_e: &[Vec3<R>],
     mat: &[M],
     e: &mut [Vec3<R>],
     h: &[Vec3<R>],
@@ -44,7 +44,7 @@ fn step_e<R: Real, M: Material<R>>(
 }
 fn step_h<R: Real, M: Material<R>>(
     meta: SimMeta<R>,
-    stim_h: &[Vec3<f32>],
+    stim_h: &[Vec3<R>],
     mat: &[M],
     e: &[Vec3<R>],
     h: &mut [Vec3<R>],
@@ -58,7 +58,7 @@ fn step_h<R: Real, M: Material<R>>(
 fn step_e_cell<R: Real, M: Material<R>>(
     idx: Vec3u,
     meta: SimMeta<R>,
-    stim_e: &[Vec3<f32>],
+    stim_e: &[Vec3<R>],
     mat: &[M],
     e: &mut [Vec3<R>],
     h: &[Vec3<R>],
@@ -78,7 +78,7 @@ fn step_e_cell<R: Real, M: Material<R>>(
     let update_state = StepEContext {
         inv_feature_size: meta.inv_feature_size,
         time_step: meta.time_step,
-        stim_e: stim_e.cast(),
+        stim_e,
         mat,
         in_h,
         in_e,
@@ -89,7 +89,7 @@ fn step_e_cell<R: Real, M: Material<R>>(
 fn step_h_cell<R: Real, M: Material<R>>(
     idx: Vec3u,
     meta: SimMeta<R>,
-    stim_h: &[Vec3<f32>],
+    stim_h: &[Vec3<R>],
     mat: &[M],
     e: &[Vec3<R>],
     h: &mut [Vec3<R>],
@@ -112,7 +112,7 @@ fn step_h_cell<R: Real, M: Material<R>>(
     let update_state = StepHContext {
         inv_feature_size: meta.inv_feature_size,
         time_step: meta.time_step,
-        stim_h: stim_h.cast(),
+        stim_h,
         mat,
         in_e,
         in_h,

crates/coremem/src/sim/spirv/gpu.rs

@@ -67,8 +67,8 @@ impl<R: Copy, M: Send + Sync + 'static> SimBackend<R, M> for WgpuBackend<R, M> {
         &self,
         num_steps: u32,
         meta: SimMeta<R>,
-        stim_cpu_e: &[Vec3<f32>],
-        stim_cpu_h: &[Vec3<f32>],
+        stim_cpu_e: &[Vec3<R>],
+        stim_cpu_h: &[Vec3<R>],
         mat: &[M],
         e: &mut [Vec3<R>],
         h: &mut [Vec3<R>],

crates/coremem/src/sim/spirv/mod.rs

@@ -22,8 +22,8 @@ pub trait SimBackend<R, M> {
         &self,
         num_steps: u32,
         meta: SimMeta<R>,
-        stim_e: &[Vec3<f32>],
-        stim_h: &[Vec3<f32>],
+        stim_e: &[Vec3<R>],
+        stim_h: &[Vec3<R>],
         mat: &[M],
         e: &mut [Vec3<R>],
         h: &mut [Vec3<R>],
@@ -227,7 +227,7 @@ where
     #[allow(unused)] // used for test
     fn apply_stimulus(&mut self, stim: &dyn AbstractStimulus) {
         trace!("apply_stimulus begin");
-        iterate_stim(stim, self.size(), self.feature_size(), self.time(), self.timestep(), |pos_idx, value_e, value_h| {
+        iterate_stim(stim, self.size(), self.feature_size(), self.time(), self.meta.time_step, |pos_idx, value_e, value_h| {
             let flat_idx = self.flat_index(pos_idx).unwrap();
             self.e[flat_idx] += value_e.cast();
             self.h[flat_idx] += value_h.cast();
@@ -236,13 +236,13 @@ where
     }
 
     fn eval_stimulus<S: AbstractStimulus>(&self, stim: &S)
-        -> (Array3<Vec3<f32>>, Array3<Vec3<f32>>)
+        -> (Array3<Vec3<R>>, Array3<Vec3<R>>)
     {
         trace!("eval_stimulus begin");
         let dim = self.size();
         let feature_size = self.feature_size();
         let t_sec = self.time();
-        let timestep = self.timestep();
+        let timestep = self.meta.time_step;
 
         // TODO(perf): do this in one loop!
         let e = ndarray::Zip::from(ndarray::indices(
@@ -251,7 +251,7 @@ where
             let pos_idx = Index::new(x as _, y as _, z as _);
             let pos_meters = pos_idx.to_meters(feature_size);
             let (density_e, _density_h) = stim.at(t_sec, pos_meters);
-            density_e * timestep
+            density_e.cast::<R>() * timestep
         });
         let h = ndarray::Zip::from(ndarray::indices(
             [dim.z() as usize, dim.y() as usize, dim.x() as usize]
@@ -259,16 +259,21 @@ where
             let pos_idx = Index::new(x as _, y as _, z as _);
             let pos_meters = pos_idx.to_meters(feature_size);
             let (_density_e, density_h) = stim.at(t_sec, pos_meters);
-            density_h * timestep
+            density_h.cast::<R>() * timestep
         });
         trace!("eval_stimulus end");
         (e, h)
     }
 }
 
-fn iterate_stim<S: AbstractStimulus + ?Sized, F: FnMut(Index, Vec3<f32>, Vec3<f32>)>(
-    stim: &S, dim: Index, feature_size: f32, t_sec: f32, timestep: f32, mut f: F
-) {
+fn iterate_stim<R, S, F>(
+    stim: &S, dim: Index, feature_size: f32, t_sec: f32, timestep: R, mut f: F
+)
+where
+    R: Real,
+    S: AbstractStimulus + ?Sized,
+    F: FnMut(Index, Vec3<R>, Vec3<R>),
+{
     // TODO: parallelize
     for z in 0..dim.z() {
         for y in 0..dim.y() {
@@ -276,7 +281,7 @@ fn iterate_stim<S: AbstractStimulus + ?Sized, F: FnMut(Index, Vec3<f32>, Vec3<f3
                 let pos_idx = Index::new(x, y, z);
                 let pos_meters = pos_idx.to_meters(feature_size);
                 let (density_e, density_h) = stim.at(t_sec, pos_meters);
-                let (value_e, value_h) = (density_e * timestep, density_h * timestep);
+                let (value_e, value_h) = (density_e.cast::<R>() * timestep, density_h.cast::<R>() * timestep);
                 f(pos_idx, value_e, value_h)
             }
         }