diff --git a/crates/cpu/src/unit.rs b/crates/cpu/src/unit.rs
index ce024a4..0c0c4e1 100644
--- a/crates/cpu/src/unit.rs
+++ b/crates/cpu/src/unit.rs
@@ -18,6 +18,7 @@ use serde::{Deserialize, Serialize};
 use std::ops::ControlFlow;
 
 pub mod alu_branch;
+pub mod load_store;
 
 macro_rules! all_units {
     (
@@ -333,7 +334,7 @@ all_units! {
         #[create_dyn_unit_fn = |config, unit_index, filter| todo!()]
         #[extract(transformed_move_mop, transformed_move_mop_sim, transformed_move_mop_sim_ref, transformed_move_mop_sim_mut)]
         TransformedMove(TransformedMoveOp),
-        #[create_dyn_unit_fn = |config, unit_index, filter| todo!()]
+        #[create_dyn_unit_fn = |config, unit_index, filter| load_store::LoadStore::new(config, unit_index, filter).to_dyn()]
         #[extract(load_store_mop, load_store_mop_sim, load_store_mop_sim_ref, load_store_mop_sim_mut)]
         LoadStore(LoadStoreMOp<DestReg, SrcReg>),
     }
diff --git a/crates/cpu/src/unit/load_store.rs b/crates/cpu/src/unit/load_store.rs
new file mode 100644
index 0000000..0b6bbd0
--- /dev/null
+++ b/crates/cpu/src/unit/load_store.rs
@@ -0,0 +1,157 @@
+// SPDX-License-Identifier: LGPL-3.0-or-later
+// See Notices.txt for copyright information
+
+use crate::{
+    config::CpuConfig,
+    instruction::LoadStoreMOp,
+    next_pc::CallStackOp,
+    register::PRegValue,
+    rename_execute_retire::{
+        ExecuteToUnitInterface, NextPcPredictorOp, RenamedMOp, UnitCausedCancel,
+        UnitFinishCauseCancel, UnitInputsReady, UnitOutputReady,
+    },
+    unit::{DynUnit, DynUnitWrapper, RenamedMOpFilter, UnitKind, UnitTrait},
+};
+use fayalite::{intern::Interned, prelude::*};
+
+#[hdl]
+async fn load_store_impl(
+    config: PhantomConst<CpuConfig>,
+    unit_index: usize,
+    filter: LoadStoreFilter,
+    cd: Expr<ClockDomain>,
+    from_execute: Expr<ExecuteToUnitInterface<PhantomConst<CpuConfig>>>,
+    mut sim: ExternModuleSimulationState,
+) {
+    todo!()
+}
+
+#[derive(Copy, Clone, PartialEq, Eq, Hash, Debug)]
+struct LoadStoreFilter {
+    load: bool,
+    store: bool,
+}
+
+impl LoadStoreFilter {
+    fn new(
+        from_execute: ExecuteToUnitInterface<PhantomConst<CpuConfig>>,
+        filter: &mut impl RenamedMOpFilter,
+    ) -> Self {
+        let LoadStoreMOp { Load, Store } = from_execute
+            .inputs_ready
+            .HdlSome
+            .mop
+            .mop
+            .inner_ty()
+            .LoadStore;
+        Self {
+            load: filter.should_include_ty(Load),
+            store: filter.should_include_ty(Store),
+        }
+    }
+}
+
+#[hdl_module(extern)]
+pub fn load_store(
+    config: PhantomConst<CpuConfig>,
+    unit_index: usize,
+    filter: &mut impl RenamedMOpFilter,
+) {
+    #[hdl]
+    let cd: ClockDomain = m.input();
+
+    #[hdl]
+    let from_execute: ExecuteToUnitInterface<PhantomConst<CpuConfig>> =
+        m.input(ExecuteToUnitInterface[config]);
+
+    let filter = LoadStoreFilter::new(from_execute.ty(), filter);
+
+    assert_eq!(config.get().units[unit_index].kind, UnitKind::LoadStore);
+
+    m.register_clock_for_past(cd.clk);
+
+    m.extern_module_simulation_fn(
+        (config, unit_index, filter, cd, from_execute),
+        async |(config, unit_index, filter, cd, from_execute), mut sim| {
+            sim.resettable(
+                cd,
+                async |mut sim| {
+                    #[hdl]
+                    let ExecuteToUnitInterface::<_> {
+                        global_state: _,
+                        enqueue,
+                        inputs_ready: _,
+                        is_no_longer_speculative: _,
+                        cant_cause_cancel,
+                        output_ready,
+                        finish_cause_cancel,
+                        unit_outputs_ready: _,
+                        cancel_all,
+                        config: _,
+                    } = from_execute;
+                    sim.write(enqueue.ready, false).await;
+                    sim.write(cant_cause_cancel, cant_cause_cancel.ty().HdlNone())
+                        .await;
+                    sim.write(output_ready, output_ready.ty().HdlNone()).await;
+                    sim.write(finish_cause_cancel, finish_cause_cancel.ty().HdlNone())
+                        .await;
+                    sim.write(cancel_all.ready, false).await;
+                },
+                async |sim, ()| {
+                    load_store_impl(config, unit_index, filter, cd, from_execute, sim).await
+                },
+            )
+            .await
+        },
+    );
+}
+
+#[derive(Debug, Copy, Clone, PartialEq, Eq, Hash)]
+pub struct LoadStore {
+    config: PhantomConst<CpuConfig>,
+    module: Interned<Module<load_store>>,
+}
+
+impl LoadStore {
+    pub fn new(
+        config: PhantomConst<CpuConfig>,
+        unit_index: usize,
+        filter: &mut impl RenamedMOpFilter,
+    ) -> Self {
+        Self {
+            config,
+            module: load_store(config, unit_index, filter),
+        }
+    }
+}
+
+impl UnitTrait for LoadStore {
+    type Type = load_store;
+
+    fn ty(&self) -> Self::Type {
+        self.module.io_ty()
+    }
+
+    fn unit_kind(&self) -> UnitKind {
+        UnitKind::LoadStore
+    }
+
+    fn module(&self) -> Interned<Module<Self::Type>> {
+        self.module
+    }
+
+    fn cd(&self, this: Expr<Self::Type>) -> Option<Expr<ClockDomain>> {
+        Some(this.cd)
+    }
+
+    fn from_execute(
+        &self,
+        this: Expr<Self::Type>,
+    ) -> Expr<ExecuteToUnitInterface<PhantomConst<CpuConfig>>> {
+        this.from_execute
+    }
+
+    fn to_dyn(&self) -> DynUnit {
+        DynUnitWrapper(*self).to_dyn()
+    }
+}