diff --git a/piet-gpu-hal/examples/dx12_toy.rs b/piet-gpu-hal/examples/dx12_toy.rs
index 63e3db4..23737bf 100644
--- a/piet-gpu-hal/examples/dx12_toy.rs
+++ b/piet-gpu-hal/examples/dx12_toy.rs
@@ -73,7 +73,7 @@ fn toy() -> Result<(), Error> {
         let pipeline = device.create_simple_compute_pipeline(SHADER_CODE, 1, 1)?;
         let ds = device.create_descriptor_set(&pipeline, &[&dev_buf], &[&img])?;
         let mut cmd_buf = device.create_cmd_buf()?;
-        let fence = device.create_fence(false)?;
+        let mut fence = device.create_fence(false)?;
         cmd_buf.begin();
         cmd_buf.copy_buffer(&buf, &dev_buf);
         cmd_buf.memory_barrier();
@@ -86,8 +86,8 @@ fn toy() -> Result<(), Error> {
         cmd_buf.finish_timestamps(&query_pool);
         cmd_buf.host_barrier();
         cmd_buf.finish();
-        device.run_cmd_bufs(&[&cmd_buf], &[], &[], Some(&fence))?;
-        device.wait_and_reset(&[&fence])?;
+        device.run_cmd_bufs(&[&cmd_buf], &[], &[], Some(&mut fence))?;
+        device.wait_and_reset(&[&mut fence])?;
         let mut readback: Vec<u32> = vec![0u32; 256];
         device.read_buffer(&buf, readback.as_mut_ptr() as *mut u8, 0, 1024)?;
         println!("{:?}", readback);
diff --git a/piet-gpu-hal/src/dx12.rs b/piet-gpu-hal/src/dx12.rs
index 0194e1e..637aa2f 100644
--- a/piet-gpu-hal/src/dx12.rs
+++ b/piet-gpu-hal/src/dx12.rs
@@ -348,7 +348,7 @@ impl crate::Device for Dx12Device {
         cmd_bufs: &[&Self::CmdBuf],
         _wait_semaphores: &[&Self::Semaphore],
         _signal_semaphores: &[&Self::Semaphore],
-        fence: Option<&Self::Fence>,
+        fence: Option<&mut Self::Fence>,
     ) -> Result<(), Error> {
         // TODO: handle semaphores
         let lists = cmd_bufs
@@ -405,7 +405,7 @@ impl crate::Device for Dx12Device {
         Ok(Fence { fence, event, val })
     }
 
-    unsafe fn wait_and_reset(&self, fences: &[&Self::Fence]) -> Result<(), Error> {
+    unsafe fn wait_and_reset(&self, fences: &[&mut Self::Fence]) -> Result<(), Error> {
         for fence in fences {
             // TODO: probably handle errors here.
             let _status = fence.event.wait(winapi::um::winbase::INFINITE);
diff --git a/piet-gpu-hal/src/hub.rs b/piet-gpu-hal/src/hub.rs
index 80c03cf..a53ef2f 100644
--- a/piet-gpu-hal/src/hub.rs
+++ b/piet-gpu-hal/src/hub.rs
@@ -124,9 +124,9 @@ impl Session {
             let mut i = 0;
             while i < pending.len() {
                 if let Ok(true) = self.0.device.get_fence_status(&pending[i].fence) {
-                    let item = pending.swap_remove(i);
+                    let mut item = pending.swap_remove(i);
                     // TODO: wait is superfluous, can just reset
-                    let _ = self.0.device.wait_and_reset(&[&item.fence]);
+                    let _ = self.0.device.wait_and_reset(vec![&mut item.fence]);
                     let mut pool = self.0.cmd_buf_pool.lock().unwrap();
                     pool.push((item.cmd_buf, item.fence));
                     std::mem::drop(item.resources);
@@ -143,7 +143,7 @@ impl Session {
 
     pub unsafe fn run_cmd_buf(
         &self,
-        cmd_buf: CmdBuf,
+        mut cmd_buf: CmdBuf,
         wait_semaphores: &[&Semaphore],
         signal_semaphores: &[&Semaphore],
     ) -> Result<SubmittedCmdBuf, Error> {
@@ -162,7 +162,7 @@ impl Session {
             &cmd_bufs,
             wait_semaphores,
             signal_semaphores,
-            Some(&cmd_buf.fence),
+            Some(&mut cmd_buf.fence),
         )?;
         Ok(SubmittedCmdBuf(
             Some(SubmittedCmdBufInner {
@@ -313,10 +313,10 @@ impl CmdBuf {
 
 impl SubmittedCmdBuf {
     pub fn wait(mut self) -> Result<(), Error> {
-        let item = self.0.take().unwrap();
+        let mut item = self.0.take().unwrap();
         if let Some(session) = Weak::upgrade(&self.1) {
             unsafe {
-                session.device.wait_and_reset(&[&item.fence])?;
+                session.device.wait_and_reset(vec![&mut item.fence])?;
             }
             session
                 .cmd_buf_pool
diff --git a/piet-gpu-hal/src/lib.rs b/piet-gpu-hal/src/lib.rs
index 51ad9e4..e9c62fd 100644
--- a/piet-gpu-hal/src/lib.rs
+++ b/piet-gpu-hal/src/lib.rs
@@ -196,7 +196,7 @@ pub trait Device: Sized {
         cmd_buf: &[&Self::CmdBuf],
         wait_semaphores: &[&Self::Semaphore],
         signal_semaphores: &[&Self::Semaphore],
-        fence: Option<&Self::Fence>,
+        fence: Option<&mut Self::Fence>,
     ) -> Result<(), Error>;
 
     /// Copy data from the buffer to memory.
@@ -233,7 +233,7 @@ pub trait Device: Sized {
 
     unsafe fn create_semaphore(&self) -> Result<Self::Semaphore, Error>;
     unsafe fn create_fence(&self, signaled: bool) -> Result<Self::Fence, Error>;
-    unsafe fn wait_and_reset(&self, fences: &[&Self::Fence]) -> Result<(), Error>;
+    unsafe fn wait_and_reset(&self, fences: &[&mut Self::Fence]) -> Result<(), Error>;
     unsafe fn get_fence_status(&self, fence: &Self::Fence) -> Result<bool, Error>;
 
     unsafe fn create_sampler(&self, params: SamplerParams) -> Result<Self::Sampler, Error>;
diff --git a/piet-gpu-hal/src/macros.rs b/piet-gpu-hal/src/macros.rs
index 98178e2..1810a3c 100644
--- a/piet-gpu-hal/src/macros.rs
+++ b/piet-gpu-hal/src/macros.rs
@@ -58,6 +58,16 @@ macro_rules! mux_enum {
                     }
                 }
             }
+            $crate::mux_cfg! {
+                #[cfg(vk)]
+                #[allow(unused)]
+                fn vk_mut(&mut self) -> &mut $vk {
+                    match self {
+                        $name::Vk(x) => x,
+                        _ => panic!("downcast error")
+                    }
+                }
+            }
 
             $crate::mux_cfg! {
                 #[cfg(dx12)]
@@ -69,6 +79,16 @@ macro_rules! mux_enum {
                     }
                 }
             }
+            $crate::mux_cfg! {
+                #[cfg(dx12)]
+                #[allow(unused)]
+                fn dx12_mut(&mut self) -> &mut $dx12 {
+                    match self {
+                        $name::Dx12(x) => x,
+                        _ => panic!("downcast error")
+                    }
+                }
+            }
 
             $crate::mux_cfg! {
                 #[cfg(mtl)]
@@ -76,7 +96,15 @@ macro_rules! mux_enum {
                 fn mtl(&self) -> &$mtl {
                     match self {
                         $name::Mtl(x) => x,
-                        _ => panic!("downcast error")
+                    }
+                }
+            }
+            $crate::mux_cfg! {
+                #[cfg(mtl)]
+                #[allow(unused)]
+                fn mtl_mut(&mut self) -> &mut $mtl {
+                    match self {
+                        $name::Mtl(x) => x,
                     }
                 }
             }
diff --git a/piet-gpu-hal/src/metal.rs b/piet-gpu-hal/src/metal.rs
index f7a97d6..9818c75 100644
--- a/piet-gpu-hal/src/metal.rs
+++ b/piet-gpu-hal/src/metal.rs
@@ -200,7 +200,7 @@ impl crate::Device for MtlDevice {
         cmd_bufs: &[&Self::CmdBuf],
         wait_semaphores: &[&Self::Semaphore],
         signal_semaphores: &[&Self::Semaphore],
-        fence: Option<&Self::Fence>,
+        fence: Option<&mut Self::Fence>,
     ) -> Result<(), Error> {
         todo!()
     }
@@ -251,7 +251,7 @@ impl crate::Device for MtlDevice {
         todo!()
     }
 
-    unsafe fn wait_and_reset(&self, fences: &[&Self::Fence]) -> Result<(), Error> {
+    unsafe fn wait_and_reset(&self, fences: &[&mut Self::Fence]) -> Result<(), Error> {
         todo!()
     }
 
diff --git a/piet-gpu-hal/src/mux.rs b/piet-gpu-hal/src/mux.rs
index 9e13a87..90be57a 100644
--- a/piet-gpu-hal/src/mux.rs
+++ b/piet-gpu-hal/src/mux.rs
@@ -228,31 +228,29 @@ impl Device {
         }
     }
 
-    pub unsafe fn wait_and_reset(&self, fences: &[&Fence]) -> Result<(), Error> {
+    // Consider changing Vec to iterator (as is done in gfx-hal)
+    pub unsafe fn wait_and_reset(&self, fences: Vec<&mut Fence>) -> Result<(), Error> {
         mux_match! { self;
             Device::Vk(d) => {
-                let fences = fences
-                    .iter()
-                    .copied()
-                    .map(Fence::vk)
+                let mut fences = fences
+                    .into_iter()
+                    .map(|f| f.vk_mut())
                     .collect::<SmallVec<[_; 4]>>();
-                d.wait_and_reset(&*fences)
+                d.wait_and_reset(&mut fences)
             }
             Device::Dx12(d) => {
-                let fences = fences
-                    .iter()
-                    .copied()
-                    .map(Fence::dx12)
+                let mut fences = fences
+                    .into_iter()
+                    .map(|f| f.dx12_mut())
                     .collect::<SmallVec<[_; 4]>>();
-                d.wait_and_reset(&*fences)
+                d.wait_and_reset(&mut fences)
             }
             Device::Mtl(d) => {
-                let fences = fences
-                    .iter()
-                    .copied()
-                    .map(Fence::mtl)
+                let mut fences = fences
+                    .into_iter()
+                    .map(|f| f.mtl_mut())
                     .collect::<SmallVec<[_; 4]>>();
-                d.wait_and_reset(&*fences)
+                d.wait_and_reset(&mut fences)
             }
         }
     }
@@ -318,7 +316,7 @@ impl Device {
         cmd_bufs: &[&CmdBuf],
         wait_semaphores: &[&Semaphore],
         signal_semaphores: &[&Semaphore],
-        fence: Option<&Fence>,
+        fence: Option<&mut Fence>,
     ) -> Result<(), Error> {
         mux_match! { self;
             Device::Vk(d) => d.run_cmd_bufs(
@@ -336,7 +334,7 @@ impl Device {
                     .copied()
                     .map(Semaphore::vk)
                     .collect::<SmallVec<[_; 4]>>(),
-                fence.map(Fence::vk),
+                fence.map(Fence::vk_mut),
             ),
             Device::Dx12(d) => d.run_cmd_bufs(
                 &cmd_bufs
@@ -353,7 +351,7 @@ impl Device {
                     .copied()
                     .map(Semaphore::dx12)
                     .collect::<SmallVec<[_; 4]>>(),
-                fence.map(Fence::dx12),
+                fence.map(Fence::dx12_mut_mut),
             ),
             Device::Mtl(d) => d.run_cmd_bufs(
                 &cmd_bufs
@@ -370,7 +368,7 @@ impl Device {
                     .copied()
                     .map(Semaphore::mtl)
                     .collect::<SmallVec<[_; 4]>>(),
-                fence.map(Fence::mtl),
+                fence.map(Fence::mtl_mut),
             ),
         }
     }
diff --git a/piet-gpu-hal/src/vulkan.rs b/piet-gpu-hal/src/vulkan.rs
index 619c6bd..4d73c2f 100644
--- a/piet-gpu-hal/src/vulkan.rs
+++ b/piet-gpu-hal/src/vulkan.rs
@@ -619,12 +619,11 @@ impl crate::Device for VkDevice {
         Ok(device.create_semaphore(&vk::SemaphoreCreateInfo::default(), None)?)
     }
 
-    unsafe fn wait_and_reset(&self, fences: &[&Self::Fence]) -> Result<(), Error> {
+    unsafe fn wait_and_reset(&self, fences: &[&mut Self::Fence]) -> Result<(), Error> {
         let device = &self.device.device;
         let fences = fences
             .iter()
-            .copied()
-            .copied()
+            .map(|f| **f)
             .collect::<SmallVec<[_; 4]>>();
         device.wait_for_fences(&fences, true, !0)?;
         device.reset_fences(&fences)?;
@@ -718,7 +717,7 @@ impl crate::Device for VkDevice {
         cmd_bufs: &[&CmdBuf],
         wait_semaphores: &[&Self::Semaphore],
         signal_semaphores: &[&Self::Semaphore],
-        fence: Option<&Self::Fence>,
+        fence: Option<&mut Self::Fence>,
     ) -> Result<(), Error> {
         let device = &self.device.device;