diff --git a/crates/coremem/src/sim/spirv/gpu.rs b/crates/coremem/src/sim/spirv/gpu.rs
index dcb7849..f0f2481 100644
--- a/crates/coremem/src/sim/spirv/gpu.rs
+++ b/crates/coremem/src/sim/spirv/gpu.rs
@@ -2,12 +2,12 @@ use futures::FutureExt as _;
 use log::{info, warn};
 use ndarray::Array3;
 use std::borrow::Cow;
+use std::marker::PhantomData;
 use std::num::NonZeroU64;
 use wgpu;
 use wgpu::util::DeviceExt as _;
 
 use coremem_types::vec::Vec3;
-use coremem_types::mat::{FullyGenericMaterial, Material};
 use coremem_types::step::SimMeta;
 
 pub fn entry_points<L: 'static>() -> Option<(&'static str, &'static str)>
@@ -15,16 +15,17 @@ pub fn entry_points<L: 'static>() -> Option<(&'static str, &'static str)>
     spirv_backend::entry_points::<L>().into()
 }
 
-pub(super) struct WgpuData {
+pub(super) struct WgpuData<M: 'static> {
     step_bind_group_layout: wgpu::BindGroupLayout,
     step_e_pipeline: wgpu::ComputePipeline,
     step_h_pipeline: wgpu::ComputePipeline,
     device: wgpu::Device,
     queue: wgpu::Queue,
+    mat: PhantomData<&'static M>,
 }
 
-impl WgpuData {
-    pub fn new<M: 'static>(volume: u64) -> Self
+impl<M: 'static> WgpuData<M> {
+    pub fn new(volume: u64) -> Self
     {
         info!("WgpuData::new({})", volume);
         use std::mem::size_of;
@@ -51,19 +52,20 @@ impl WgpuData {
             step_e_pipeline,
             device,
             queue,
+            mat: PhantomData,
         }
     }
 }
 
-impl Default for WgpuData {
+impl<M: 'static> Default for WgpuData<M> {
     fn default() -> Self {
-        Self::new::<FullyGenericMaterial<f32>>(0)
+        Self::new(0)
     }
 }
 
 
-impl WgpuData {
-    pub(super) fn step_spirv<M>(
+impl<M: Send + Sync + 'static> WgpuData<M> {
+    pub(super) fn step_spirv(
         &self,
         num_steps: u32,
         meta: SimMeta<f32>,
@@ -74,9 +76,7 @@ impl WgpuData {
         e: &mut [Vec3<f32>],
         h: &mut [Vec3<f32>],
         m: &mut [Vec3<f32>],
-    )
-    where M: Send + Sync + Material<f32> + 'static
-    {
+    ) {
         let field_bytes = meta.dim.product_sum() as usize * std::mem::size_of::<Vec3<f32>>();
 
         let device = &self.device;
diff --git a/crates/coremem/src/sim/spirv/mod.rs b/crates/coremem/src/sim/spirv/mod.rs
index f3e11d4..ec67ff6 100644
--- a/crates/coremem/src/sim/spirv/mod.rs
+++ b/crates/coremem/src/sim/spirv/mod.rs
@@ -17,6 +17,7 @@ use gpu::WgpuData;
 /// Wrapper around an inner state object which offloads stepping onto a spirv backend (e.g. GPU).
 #[derive(Clone, Default, Serialize, Deserialize)]
 pub struct SpirvSim<M=FullyGenericMaterial<f32>>
+where M: 'static
 {
     // TODO: make this generic over R
     meta: SimMeta<f32>,
@@ -29,7 +30,7 @@ pub struct SpirvSim<M=FullyGenericMaterial<f32>>
     // XXX not confident that wgpu is actually properly synchronized for us to omit a Mutex here
     // though.
     #[serde(skip)]
-    wgpu: Option<Arc<WgpuData>>,
+    wgpu: Option<Arc<WgpuData<M>>>,
 }
 
 impl<M> MaterialSim for SpirvSim<M>
@@ -94,7 +95,7 @@ where
     fn step_multiple<S: AbstractStimulus>(&mut self, num_steps: u32, stim: &S) {
         let vol = self.size().volume();
         self.wgpu.get_or_insert_with(
-            || Arc::new(WgpuData::new::<M>(vol))
+            || Arc::new(WgpuData::new(vol))
         );
 
         let (stim_e, stim_h) = self.eval_stimulus(stim);
@@ -132,14 +133,14 @@ impl<M> SpirvSim<M>
     where M: Default + 'static
 {
     pub fn new(size: Index, feature_size: f32) -> Self {
-        Self::new_with_wgpu_handle(size, feature_size, Some(Arc::new(WgpuData::new::<M>(size.volume()))))
+        Self::new_with_wgpu_handle(size, feature_size, Some(Arc::new(WgpuData::new(size.volume()))))
     }
 
     pub fn new_no_wgpu(size: Index, feature_size: f32) -> Self {
         Self::new_with_wgpu_handle(size, feature_size, None)
     }
 
-    fn new_with_wgpu_handle(size: Index, feature_size: f32, wgpu: Option<Arc<WgpuData>>) -> Self {
+    fn new_with_wgpu_handle(size: Index, feature_size: f32, wgpu: Option<Arc<WgpuData<M>>>) -> Self {
         info!("SpirvSim::new({:?}, {})", size, feature_size);
         let flat_size = size.volume() as usize;
         if flat_size * std::mem::size_of::<M>() >= 0x40000000 {