WIP adding rename_execute_retire

Cargo.lock

@@ -388,7 +388,7 @@ checksum = "e8c02a5121d4ea3eb16a80748c74f5549a5665e4c21333c6098f283870fbdea6"
 [[package]]
 name = "fayalite"
 version = "0.3.0"
-source = "git+https://git.libre-chip.org/libre-chip/fayalite.git?branch=master#2aa41137d4eec592bd591fddb558bc78a52c635c"
+source = "git+https://git.libre-chip.org/libre-chip/fayalite.git?branch=master#80b92c7dd3adcad8487a3f8224846e381219bfa6"
 dependencies = [
  "base64",
  "bitvec",
@@ -417,7 +417,7 @@ dependencies = [
 [[package]]
 name = "fayalite-proc-macros"
 version = "0.3.0"
-source = "git+https://git.libre-chip.org/libre-chip/fayalite.git?branch=master#2aa41137d4eec592bd591fddb558bc78a52c635c"
+source = "git+https://git.libre-chip.org/libre-chip/fayalite.git?branch=master#80b92c7dd3adcad8487a3f8224846e381219bfa6"
 dependencies = [
  "fayalite-proc-macros-impl",
 ]
@@ -425,7 +425,7 @@ dependencies = [
 [[package]]
 name = "fayalite-proc-macros-impl"
 version = "0.3.0"
-source = "git+https://git.libre-chip.org/libre-chip/fayalite.git?branch=master#2aa41137d4eec592bd591fddb558bc78a52c635c"
+source = "git+https://git.libre-chip.org/libre-chip/fayalite.git?branch=master#80b92c7dd3adcad8487a3f8224846e381219bfa6"
 dependencies = [
  "base16ct 0.2.0",
  "num-bigint",
@@ -440,7 +440,7 @@ dependencies = [
 [[package]]
 name = "fayalite-visit-gen"
 version = "0.3.0"
-source = "git+https://git.libre-chip.org/libre-chip/fayalite.git?branch=master#2aa41137d4eec592bd591fddb558bc78a52c635c"
+source = "git+https://git.libre-chip.org/libre-chip/fayalite.git?branch=master#80b92c7dd3adcad8487a3f8224846e381219bfa6"
 dependencies = [
  "indexmap",
  "prettyplease",

crates/cpu/Cargo.toml

@@ -33,3 +33,6 @@ hex-literal.workspace = true
 regex = "1.12.2"
 sha2.workspace = true
 which.workspace = true
+
+[lints.rust]
+unexpected_cfgs = { level = "warn", check-cfg = ['cfg(todo)'] }

crates/cpu/src/config.rs

@@ -1,12 +1,6 @@
 // SPDX-License-Identifier: LGPL-3.0-or-later
 // See Notices.txt for copyright information
-use crate::{
-    instruction::{CONST_ZERO_UNIT_NUM, MOpTrait, PRegNum, RenamedMOp, UnitNum, UnitOutRegNum},
-    unit::{
-        UnitCancelInput, UnitKind, UnitOutputWrite,
-        unit_base::{UnitForwardingInfo, UnitToRegAlloc},
-    },
-};
+use crate::{instruction::CONST_ZERO_UNIT_NUM, unit::UnitKind};
 use fayalite::prelude::*;
 use serde::{Deserialize, Serialize};
 use std::num::NonZeroUsize;
@@ -101,55 +95,14 @@ impl CpuConfig {
     pub fn unit_num_width(&self) -> usize {
         UInt::range(CONST_ZERO_UNIT_NUM..self.non_const_unit_nums().end).width()
     }
-    pub fn unit_num(&self) -> UnitNum<DynSize> {
-        UnitNum[self.unit_num_width()]
-    }
-    pub fn unit_out_reg_num(&self) -> UnitOutRegNum<DynSize> {
-        UnitOutRegNum[self.out_reg_num_width]
-    }
-    pub fn p_reg_num(&self) -> PRegNum<DynSize, DynSize> {
-        PRegNum[self.unit_num_width()][self.out_reg_num_width]
-    }
     pub fn p_reg_num_width(&self) -> usize {
         self.unit_num_width() + self.out_reg_num_width
     }
-    pub fn renamed_mop_in_unit(&self) -> RenamedMOp<UnitOutRegNum<DynSize>, DynSize> {
-        RenamedMOp[self.unit_out_reg_num()][self.p_reg_num_width()]
-    }
-    pub fn unit_output_write(&self) -> UnitOutputWrite<DynSize> {
-        UnitOutputWrite[self.out_reg_num_width]
-    }
-    pub fn unit_output_writes(&self) -> Array<HdlOption<UnitOutputWrite<DynSize>>> {
-        Array[HdlOption[self.unit_output_write()]][self.non_const_unit_nums().len()]
-    }
-    pub fn unit_cancel_input(&self) -> UnitCancelInput<DynSize> {
-        UnitCancelInput[self.out_reg_num_width]
-    }
-    pub fn unit_forwarding_info(&self) -> UnitForwardingInfo<DynSize, DynSize, DynSize> {
-        UnitForwardingInfo[self.unit_num_width()][self.out_reg_num_width]
-            [self.non_const_unit_nums().len()]
-    }
     pub fn unit_max_in_flight(&self, unit_index: usize) -> NonZeroUsize {
         self.units[unit_index]
             .max_in_flight
             .unwrap_or(self.default_unit_max_in_flight)
     }
-    pub fn unit_to_reg_alloc<
-        MOp: Type + MOpTrait<DestReg = UnitOutRegNum<DynSize>, SrcRegWidth = DynSize>,
-        ExtraOut: Type,
-    >(
-        &self,
-        mop_ty: MOp,
-        extra_out_ty: ExtraOut,
-    ) -> UnitToRegAlloc<MOp, ExtraOut, DynSize, DynSize, DynSize> {
-        assert_eq!(
-            mop_ty.dest_reg_ty(),
-            self.unit_out_reg_num(),
-            "inconsistent types",
-        );
-        UnitToRegAlloc[mop_ty][extra_out_ty][self.unit_num_width()][self.out_reg_num_width]
-            [self.non_const_unit_nums().len()]
-    }
     pub fn fetch_width_in_bytes(&self) -> usize {
         1usize
             .checked_shl(self.log2_fetch_width_in_bytes.into())
@@ -188,6 +141,15 @@ impl CpuConfig {
     }
 }
 
+#[hdl(get(|c| c.out_reg_num_width))]
+pub type CpuConfigOutRegNumWidth<C: PhantomConstGet<CpuConfig>> = DynSize;
+
+#[hdl(get(|c| c.unit_num_width()))]
+pub type CpuConfigUnitNumWidth<C: PhantomConstGet<CpuConfig>> = DynSize;
+
+#[hdl(get(|c| c.non_const_unit_nums().len()))]
+pub type CpuConfigUnitCount<C: PhantomConstGet<CpuConfig>> = DynSize;
+
 #[hdl(get(|c| c.fetch_width.get()))]
 pub type CpuConfigFetchWidth<C: PhantomConstGet<CpuConfig>> = DynSize;
 

crates/cpu/src/fetch.rs

@@ -9,8 +9,9 @@ use crate::{
         CpuConfigMaxFetchesInFlight, PhantomConstCpuConfig,
     },
     main_memory_and_io::{
-        MemoryInterface, MemoryOperationErrorKind, MemoryOperationFinish,
-        MemoryOperationFinishKind, MemoryOperationKind, MemoryOperationStart,
+        CpuConfigFetchMemoryInterfaceConfig, MemoryInterface, MemoryInterfaceConfig,
+        MemoryOperationErrorKind, MemoryOperationFinish, MemoryOperationFinishKind,
+        MemoryOperationKind, MemoryOperationStart,
     },
     next_pc::{
         FETCH_BLOCK_ID_WIDTH, NextPcToFetchInterface, NextPcToFetchInterfaceInner, ResetStatus,
@@ -415,7 +416,9 @@ impl<C: PhantomConstCpuConfig> L1ICacheStateSim<C> {
         self.state_expr.ty().config
     }
     #[hdl]
-    fn try_start_memory_operation(&mut self) -> Option<SimValue<MemoryOperationStart<C>>> {
+    fn try_start_memory_operation(
+        &mut self,
+    ) -> Option<SimValue<MemoryOperationStart<CpuConfigFetchMemoryInterfaceConfig<C>>>> {
         let config = self.config();
         for cache_miss in &mut self.cache_misses {
             let Some(next_start_fetch_block) = CacheMiss::next_start_fetch_block(cache_miss) else {
@@ -430,7 +433,7 @@ impl<C: PhantomConstCpuConfig> L1ICacheStateSim<C> {
                 error: _, // handled by CacheMiss::next_start_fetch_block()
                 config: _,
             } = cache_miss;
-            let mem_op_ty = MemoryOperationStart[config];
+            let mem_op_ty = MemoryOperationStart[CpuConfigFetchMemoryInterfaceConfig[config]];
             let mut addr = SplitAddr[config].split_addr_sim(addr);
             #[hdl(sim)]
             let SplitAddr::<_> {
@@ -453,8 +456,8 @@ impl<C: PhantomConstCpuConfig> L1ICacheStateSim<C> {
                         mem_op_ty.write_data.len(),
                     ),
                     rw_mask: repeat(true, mem_op_ty.write_data.len()),
-                    fetch_block_id,
-                    config,
+                    op_id: SimValue::into_dyn_int(fetch_block_id.clone()),
+                    config: CpuConfigFetchMemoryInterfaceConfig[config],
                 },
             );
         }
@@ -497,7 +500,9 @@ impl<C: PhantomConstCpuConfig> L1ICacheStateSim<C> {
     fn do_memory_operation_finish<'a>(
         &mut self,
         ready_for_memory_operation_finish: ReadyForMemoryOperationFinish,
-        memory_operation_finish: impl ToSimValue<Type = MemoryOperationFinish<C>>,
+        memory_operation_finish: impl ToSimValue<
+            Type = MemoryOperationFinish<CpuConfigFetchMemoryInterfaceConfig<C>>,
+        >,
     ) -> Option<SimValue<WriteBackStep<C>>> {
         let config = self.config();
         let cache_miss = &mut self.cache_misses[ready_for_memory_operation_finish.cache_miss_index];
@@ -621,9 +626,7 @@ impl<C: PhantomConstCpuConfig> L1ICacheStateSim<C> {
     #[hdl]
     fn check_memory_next_fetch_block_ids(
         &self,
-        memory_next_fetch_block_ids: SimValue<
-            ArrayVec<UInt<{ FETCH_BLOCK_ID_WIDTH }>, CpuConfigMaxFetchesInFlight<C>>,
-        >,
+        memory_next_fetch_block_ids: SimValue<ArrayVec<UInt, DynSize>>,
     ) {
         let memory_next_fetch_block_ids = ArrayVec::elements_sim_ref(&memory_next_fetch_block_ids);
         let mut expected_memory_next_fetch_block_ids = Vec::new();
@@ -1071,8 +1074,8 @@ fn l1_i_cache_impl(config: PhantomConst<CpuConfig>) {
     #[hdl]
     let cd: ClockDomain = m.input();
     #[hdl]
-    let memory_interface: MemoryInterface<PhantomConst<CpuConfig>> =
-        m.output(MemoryInterface[config]);
+    let memory_interface: MemoryInterface<PhantomConst<MemoryInterfaceConfig>> =
+        m.output(MemoryInterface[CpuConfigFetchMemoryInterfaceConfig[config]]);
     #[hdl]
     let from_next_pc: NextPcToFetchInterface<PhantomConst<CpuConfig>> =
         m.input(NextPcToFetchInterface[config]);
@@ -1103,7 +1106,7 @@ fn l1_i_cache_impl(config: PhantomConst<CpuConfig>) {
     async fn run(
         mut sim: ExternModuleSimulationState,
         cd: Expr<ClockDomain>,
-        memory_interface: Expr<MemoryInterface<PhantomConst<CpuConfig>>>,
+        memory_interface: Expr<MemoryInterface<PhantomConst<MemoryInterfaceConfig>>>,
         from_next_pc: Expr<NextPcToFetchInterface<PhantomConst<CpuConfig>>>,
         to_decode_fetched: Expr<ReadyValid<FetchToDecodeInterfaceInner<PhantomConst<CpuConfig>>>>,
         to_decode_next_fetch_block_ids: Expr<
@@ -1196,9 +1199,8 @@ fn l1_i_cache_impl(config: PhantomConst<CpuConfig>) {
                 )
                 .await;
             #[hdl(sim)]
-            if let HdlSome(next_fetch_block_ids) = sim
-                .read_past(memory_interface.next_fetch_block_ids, cd.clk)
-                .await
+            if let HdlSome(next_fetch_block_ids) =
+                sim.read_past(memory_interface.next_op_ids, cd.clk).await
             {
                 state.check_memory_next_fetch_block_ids(next_fetch_block_ids);
             }
@@ -1359,7 +1361,7 @@ fn l1_i_cache_impl(config: PhantomConst<CpuConfig>) {
             state_for_debug,
         ),
          mut sim| async move {
-            let config = memory_interface.ty().config;
+            let config = from_next_pc.ty().config;
             let cache_line_ty = CacheLine[config];
             sim.write(i_cache_port.clk, false).await; // externally overridden with cd.clk, so just write a constant here
             sim.resettable(
@@ -1426,8 +1428,8 @@ pub fn l1_i_cache(config: PhantomConst<CpuConfig>) {
     #[hdl]
     let cd: ClockDomain = m.input();
     #[hdl]
-    let memory_interface: MemoryInterface<PhantomConst<CpuConfig>> =
-        m.output(MemoryInterface[config]);
+    let memory_interface: MemoryInterface<PhantomConst<MemoryInterfaceConfig>> =
+        m.output(MemoryInterface[CpuConfigFetchMemoryInterfaceConfig[config]]);
     #[hdl]
     let from_next_pc: NextPcToFetchInterface<PhantomConst<CpuConfig>> =
         m.input(NextPcToFetchInterface[config]);
@@ -1524,8 +1526,8 @@ pub fn fetch(config: PhantomConst<CpuConfig>) {
     #[hdl]
     let cd: ClockDomain = m.input();
     #[hdl]
-    let memory_interface: MemoryInterface<PhantomConst<CpuConfig>> =
-        m.output(MemoryInterface[config]);
+    let memory_interface: MemoryInterface<PhantomConst<MemoryInterfaceConfig>> =
+        m.output(MemoryInterface[CpuConfigFetchMemoryInterfaceConfig[config]]);
     #[hdl]
     let from_next_pc: NextPcToFetchInterface<PhantomConst<CpuConfig>> =
         m.input(NextPcToFetchInterface[config]);

crates/cpu/src/instruction.rs

@@ -1,6 +1,7 @@
 // SPDX-License-Identifier: LGPL-3.0-or-later
 // See Notices.txt for copyright information
 use crate::{
+    config::{CpuConfig, CpuConfigOutRegNumWidth, CpuConfigUnitNumWidth, PhantomConstCpuConfig},
     register::{PRegFlags, PRegFlagsViewTrait, PRegValue, ViewUnused},
     unit::UnitMOp,
     util::{Rotate, range_u32_len},
@@ -2595,19 +2596,21 @@ impl<DestReg: Type, SrcRegWidth: Size> MoveRegMOp<DestReg, SrcRegWidth> {
     }
 }
 
-#[hdl(cmp_eq)]
+#[hdl(cmp_eq, no_static)]
 /// there may be more than one unit of a given kind, so UnitNum is not the same as UnitKind.
 /// zero is used for built-in constants, such as the zero register
-pub struct UnitNum<Width: Size> {
-    pub adj_value: UIntType<Width>,
+pub struct UnitNum<C: PhantomConstGet<CpuConfig>> {
+    pub adj_value: UIntType<CpuConfigUnitNumWidth<C>>,
+    pub config: C,
 }
 
-impl<Width: Size> UnitNum<Width> {
+impl<C: PhantomConstCpuConfig> UnitNum<C> {
     #[hdl]
     pub fn const_zero(self) -> Expr<Self> {
         #[hdl]
         UnitNum {
             adj_value: CONST_ZERO_UNIT_NUM.cast_to(self.adj_value),
+            config: self.config,
         }
     }
     #[hdl]
@@ -2615,6 +2618,7 @@ impl<Width: Size> UnitNum<Width> {
         #[hdl]
         UnitNum {
             adj_value: (index + 1).cast_to(self.adj_value),
+            config: self.config,
         }
     }
     pub fn is_index(expr: impl ToExpr<Type = Self>, index: usize) -> Expr<Bool> {
@@ -2622,7 +2626,9 @@ impl<Width: Size> UnitNum<Width> {
         expr.ty().from_index(index).adj_value.cmp_eq(expr.adj_value)
     }
     #[hdl]
-    pub fn as_index(expr: impl ToExpr<Type = Self>) -> Expr<HdlOption<UIntType<Width>>> {
+    pub fn as_index(
+        expr: impl ToExpr<Type = Self>,
+    ) -> Expr<HdlOption<UIntType<CpuConfigUnitNumWidth<C>>>> {
         let expr = expr.to_expr();
         #[hdl]
         let unit_index = wire(HdlOption[expr.ty().adj_value]);
@@ -2640,19 +2646,20 @@ impl<Width: Size> UnitNum<Width> {
 
 pub const CONST_ZERO_UNIT_NUM: usize = 0;
 
-#[hdl(cmp_eq)]
-pub struct UnitOutRegNum<Width: Size> {
-    pub value: UIntType<Width>,
+#[hdl(cmp_eq, no_static)]
+pub struct UnitOutRegNum<C: PhantomConstGet<CpuConfig>> {
+    pub value: UIntType<CpuConfigOutRegNumWidth<C>>,
+    pub config: C,
 }
 
-#[hdl(cmp_eq)]
+#[hdl(cmp_eq, no_static)]
 /// Physical Register Number -- registers in the CPU's backend
-pub struct PRegNum<UnitNumWidth: Size, OutRegNumWidth: Size> {
-    pub unit_num: UnitNum<UnitNumWidth>,
-    pub unit_out_reg: UnitOutRegNum<OutRegNumWidth>,
+pub struct PRegNum<C: PhantomConstGet<CpuConfig>> {
+    pub unit_num: UnitNum<C>,
+    pub unit_out_reg: UnitOutRegNum<C>,
 }
 
-impl<UnitNumWidth: Size, OutRegNumWidth: Size> PRegNum<UnitNumWidth, OutRegNumWidth> {
+impl<C: PhantomConstCpuConfig> PRegNum<C> {
     #[hdl]
     pub fn const_zero(self) -> Expr<Self> {
         #[hdl]
@@ -2661,6 +2668,7 @@ impl<UnitNumWidth: Size, OutRegNumWidth: Size> PRegNum<UnitNumWidth, OutRegNumWi
             unit_out_reg: #[hdl]
             UnitOutRegNum {
                 value: 0u8.cast_to(self.unit_out_reg.value),
+                config: self.unit_out_reg.config,
             },
         }
     }

crates/cpu/src/lib.rs

@@ -7,7 +7,9 @@ pub mod instruction;
 pub mod main_memory_and_io;
 pub mod next_pc;
 pub mod powerisa_instructions_xml;
+#[cfg(todo)]
 pub mod reg_alloc;
 pub mod register;
+pub mod rename_execute_retire;
 pub mod unit;
 pub mod util;

crates/cpu/src/main_memory_and_io/simple_uart.rs

@@ -0,0 +1,618 @@
+// SPDX-License-Identifier: LGPL-3.0-or-later
+// See Notices.txt for copyright information
+
+use std::num::{NonZeroU64, NonZeroUsize, Wrapping};
+
+use crate::{
+    main_memory_and_io::{
+        AddressRange, MemoryInterface, MemoryInterfaceConfig, MemoryOperationErrorKind,
+        MemoryOperationFinish, MemoryOperationFinishKind, MemoryOperationKind,
+        MemoryOperationStart,
+    },
+    util::array_vec::ArrayVec,
+};
+use fayalite::{
+    int::UIntInRange,
+    prelude::*,
+    util::ready_valid::{ReadyValid, queue},
+};
+
+const TICKS_PER_BAUD: usize = 16;
+const PRECISION_BITS: usize = 16;
+
+#[hdl_module]
+pub fn uart_clock_gen(
+    clock_input_properties: PhantomConst<peripherals::ClockInputProperties>,
+    baud_rate: f64,
+) {
+    #[hdl]
+    let cd: ClockDomain = m.input();
+    #[hdl]
+    let tick: Bool = m.output();
+
+    let divisor =
+        clock_input_properties.get().frequency.into_inner() / (baud_rate * TICKS_PER_BAUD as f64);
+    type NumeratorType = u128;
+    let numerator: NumeratorType = 1 << PRECISION_BITS;
+    let denominator: NumeratorType = (divisor * numerator as f64).round() as _;
+    let remainder_ty = UInt::range(0..denominator);
+
+    #[hdl]
+    let remainder_reg = reg_builder().clock_domain(cd).reset(remainder_ty.zero());
+
+    #[hdl]
+    let sum = wire((remainder_reg + numerator).ty());
+    connect_any(sum, remainder_reg + numerator);
+
+    #[hdl]
+    let tick_reg = reg_builder().clock_domain(cd).reset(false);
+    connect(tick_reg, false);
+
+    #[hdl]
+    let next_remainder = wire(remainder_ty);
+    connect(remainder_reg, next_remainder);
+
+    #[hdl]
+    if sum.cmp_ge(denominator) {
+        connect_any(next_remainder, sum - denominator);
+        connect(tick_reg, true);
+    } else {
+        connect(next_remainder, sum);
+    }
+
+    connect(tick, tick_reg);
+}
+
+#[hdl_module]
+pub fn transmitter() {
+    #[hdl]
+    let cd: ClockDomain = m.input();
+    #[hdl]
+    let tick: Bool = m.input();
+    #[hdl]
+    let tx: Bool = m.output();
+    #[hdl]
+    let input_byte: ReadyValid<UInt<8>> = m.input();
+
+    #[hdl]
+    let not_tx_reg: Bool = reg_builder().clock_domain(cd).reset(false);
+
+    connect(tx, !not_tx_reg);
+
+    #[hdl]
+    let input_buf_reg: HdlOption<UInt<8>> = reg_builder().clock_domain(cd).reset(HdlNone());
+
+    const BYTE_ON_WIRE_WIDTH: usize = 10;
+
+    #[hdl]
+    let shift_reg: Array<HdlOption<Bool>, _> = reg_builder()
+        .clock_domain(cd)
+        .reset([HdlNone(); BYTE_ON_WIRE_WIDTH]);
+
+    #[hdl]
+    if let HdlSome(v) = shift_reg[0] {
+        connect(not_tx_reg, !v);
+    } else {
+        connect(not_tx_reg, false);
+    }
+
+    #[hdl]
+    let tick_count_reg = reg_builder()
+        .clock_domain(cd)
+        .reset(0u8.cast_to_static::<UIntInRange<0, { TICKS_PER_BAUD }>>());
+
+    #[hdl]
+    let next_tick_count = wire(tick_count_reg.ty());
+
+    connect(tick_count_reg, next_tick_count);
+
+    #[hdl]
+    if !tick {
+        connect(next_tick_count, tick_count_reg);
+    } else if tick_count_reg.cmp_ge(TICKS_PER_BAUD - 1) {
+        connect(next_tick_count, 0u8.cast_to(next_tick_count.ty()));
+    } else {
+        connect(
+            next_tick_count,
+            (tick_count_reg.cast_to(UInt[tick_count_reg.ty().bit_width()]) + 1u8)
+                .cast_to(next_tick_count.ty()),
+        );
+    }
+
+    #[hdl]
+    if tick & tick_count_reg.cmp_eq(0u8) {
+        #[hdl]
+        if let HdlSome(_) = shift_reg[0] {
+            for v in shift_reg.windows(2) {
+                connect(v[0], v[1]);
+            }
+            connect(shift_reg.last().expect("known to be non-empty"), HdlNone());
+        }
+        #[hdl]
+        if let HdlNone = shift_reg[1] {
+            // when shift_reg[1] is HdlNone that means shift_reg would be completely empty in the next clock
+            // cycle, so instead fill it with the next byte if there is one.
+            #[hdl]
+            if let HdlSome(byte) = input_buf_reg {
+                connect(input_buf_reg, HdlNone());
+                connect(shift_reg, [HdlSome(true); _]);
+                connect(shift_reg[0], HdlSome(false));
+                for (i, v) in (*shift_reg)[1..][..8].iter().enumerate() {
+                    connect(v, HdlSome(byte[i]));
+                }
+            }
+        }
+    }
+
+    #[hdl]
+    if let HdlSome(_) = input_buf_reg {
+        connect(input_byte.ready, false);
+    } else {
+        connect(input_byte.ready, true);
+        connect(input_buf_reg, input_byte.data);
+    }
+}
+
+/// `rx_synchronized` is [`peripherals::Uart::rx`], but after being synchronized to `cd.clk`.
+/// `output_byte` is `HdlSome` for 1 clock cycle for every received byte.
+#[hdl_module]
+pub fn receiver_no_queue() {
+    #[hdl]
+    let cd: ClockDomain = m.input();
+    #[hdl]
+    let tick: Bool = m.input();
+    #[hdl]
+    let rx_synchronized: Bool = m.input();
+    #[hdl]
+    let output_byte: HdlOption<UInt<8>> = m.output();
+
+    const BITS_PER_BYTE: usize = 8;
+
+    #[hdl]
+    let output_byte_reg: HdlOption<UInt<{ BITS_PER_BYTE }>> =
+        reg_builder().clock_domain(cd).reset(HdlNone());
+
+    connect(output_byte, output_byte_reg);
+    connect(output_byte_reg, HdlNone()); // by default always reset to HdlNone()
+
+    #[hdl]
+    let tick_count_reg = reg_builder()
+        .clock_domain(cd)
+        .reset(0u8.cast_to(UInt::range(0..TICKS_PER_BAUD)));
+
+    #[hdl]
+    let shift_reg: UInt<{ BITS_PER_BYTE }> = reg_builder().clock_domain(cd).reset(0u8);
+
+    #[hdl]
+    enum State {
+        WaitingForStartBit,
+        StartBit,
+        DataBits(UIntInRange<0, { BITS_PER_BYTE }>),
+        StopBit,
+    }
+
+    #[hdl]
+    let state_reg = reg_builder()
+        .clock_domain(cd)
+        .reset(State.WaitingForStartBit());
+
+    #[hdl]
+    if tick {
+        #[hdl]
+        match state_reg {
+            State::WaitingForStartBit => {
+                connect(tick_count_reg, tick_count_reg.ty().zero());
+                #[hdl]
+                if !rx_synchronized {
+                    connect(state_reg, State.StartBit());
+                }
+            }
+            State::StartBit =>
+            {
+                #[hdl]
+                if tick_count_reg.cmp_eq(TICKS_PER_BAUD - 1) {
+                    connect(tick_count_reg, tick_count_reg.ty().zero());
+                    connect(state_reg, State.DataBits(0u8.cast_to(State.DataBits)));
+                } else {
+                    connect_any(tick_count_reg, tick_count_reg + 1u8);
+                }
+            }
+            State::DataBits(count) => {
+                #[hdl]
+                if tick_count_reg.cmp_eq(TICKS_PER_BAUD - 1) {
+                    connect(tick_count_reg, tick_count_reg.ty().zero());
+                    #[hdl]
+                    if count.cmp_eq(BITS_PER_BYTE - 1) {
+                        connect(state_reg, State.StopBit());
+                        connect(output_byte_reg, HdlSome(shift_reg));
+                    } else {
+                        connect(
+                            state_reg,
+                            State.DataBits((count.cast_to(UInt[8]) + 1u8).cast_to(State.DataBits)),
+                        );
+                    }
+                } else {
+                    connect_any(tick_count_reg, tick_count_reg + 1u8);
+                }
+                #[hdl]
+                if tick_count_reg.cmp_eq(TICKS_PER_BAUD / 2) {
+                    connect_any(
+                        shift_reg,
+                        (shift_reg >> 1) | (rx_synchronized.cast_to_static::<UInt<1>>() << 7),
+                    );
+                }
+            }
+            State::StopBit => {
+                #[hdl]
+                if tick_count_reg.cmp_eq(TICKS_PER_BAUD / 2) {
+                    // go to WaitingForStartBit in the middle of the stop bit so we can adjust if the sender is faster than us.
+                    connect(tick_count_reg, tick_count_reg.ty().zero());
+                    connect(state_reg, State.WaitingForStartBit());
+                } else {
+                    connect_any(tick_count_reg, tick_count_reg + 1u8);
+                }
+            }
+        }
+    }
+}
+
+#[hdl]
+pub enum ReceiverQueueStatus {
+    QueueEmpty,
+    QueueNotEmpty,
+    QueueAlmostFull,
+    QueueFull,
+    QueueOverflowed,
+}
+
+impl ReceiverQueueStatus {
+    pub fn sim_as_u8(this: impl ToSimValue<Type = Self>) -> u8 {
+        SimValue::bits(&this.into_sim_value())
+            .cast_to_static::<UInt<8>>()
+            .as_int()
+    }
+    #[hdl]
+    pub fn for_queue_len_sim(current_count: usize, queue_size: NonZeroUsize) -> SimValue<Self> {
+        if current_count == 0 {
+            #[hdl(sim)]
+            ReceiverQueueStatus.QueueEmpty()
+        } else if current_count == queue_size.get() {
+            #[hdl(sim)]
+            ReceiverQueueStatus.QueueFull()
+        } else if current_count == queue_size.get() - 1 {
+            #[hdl(sim)]
+            ReceiverQueueStatus.QueueAlmostFull()
+        } else if current_count > queue_size.get() {
+            #[hdl(sim)]
+            ReceiverQueueStatus.QueueOverflowed()
+        } else {
+            #[hdl(sim)]
+            ReceiverQueueStatus.QueueNotEmpty()
+        }
+    }
+    #[hdl]
+    pub fn for_queue_len_as_u8(current_count: usize, queue_size: NonZeroUsize) -> u8 {
+        Self::sim_as_u8(Self::for_queue_len_sim(current_count, queue_size))
+    }
+}
+
+/// `rx_synchronized` is [`peripherals::Uart::rx`], but after being synchronized to `cd.clk`.
+#[hdl_module]
+pub fn receiver(queue_capacity: NonZeroUsize) {
+    #[hdl]
+    let cd: ClockDomain = m.input();
+    #[hdl]
+    let tick: Bool = m.input();
+    #[hdl]
+    let rx_synchronized: Bool = m.input();
+    #[hdl]
+    let output_byte: ReadyValid<UInt<8>> = m.output();
+    #[hdl]
+    let queue_status: ReceiverQueueStatus = m.output();
+
+    #[hdl]
+    let queue = instance(queue(UInt::<8>::new_static(), queue_capacity, false, false));
+
+    connect(queue.cd, cd);
+    connect(output_byte, queue.out);
+
+    #[hdl]
+    let queue_overflowed_reg = reg_builder().clock_domain(cd).reset(false);
+
+    #[hdl]
+    if queue_overflowed_reg {
+        connect(queue_status, ReceiverQueueStatus.QueueOverflowed());
+    } else if queue.count.cmp_eq(0u8) {
+        connect(queue_status, ReceiverQueueStatus.QueueEmpty());
+    } else if queue.count.cmp_eq(queue_capacity) {
+        connect(queue_status, ReceiverQueueStatus.QueueFull());
+    } else if queue.count.cmp_eq(queue_capacity.get() - 1) {
+        connect(queue_status, ReceiverQueueStatus.QueueAlmostFull());
+    } else {
+        connect(queue_status, ReceiverQueueStatus.QueueNotEmpty());
+    }
+
+    #[hdl]
+    if ReadyValid::firing(output_byte) {
+        // clear overflow when a byte is read from the queue
+        connect(queue_overflowed_reg, false);
+    }
+
+    // we ignore queue.inp.ready other than noting an overflow when there was data but the queue wasn't ready
+    #[hdl]
+    if !queue.inp.ready {
+        #[hdl]
+        if let HdlSome(_) = queue.inp.data {
+            connect(queue_overflowed_reg, true);
+        }
+    }
+
+    #[hdl]
+    let inner = instance(receiver_no_queue());
+    connect(inner.cd, cd);
+    connect(inner.tick, tick);
+    connect(inner.rx_synchronized, rx_synchronized);
+    connect(queue.inp.data, inner.output_byte);
+}
+
+pub const SIMPLE_UART_RECEIVE_OFFSET: u64 = 0;
+pub const SIMPLE_UART_TRANSMIT_OFFSET: u64 = 0;
+pub const SIMPLE_UART_STATUS_OFFSET: u64 = 1;
+pub const SIMPLE_UART_LOG2_BUS_WIDTH: u8 = 1;
+pub const SIMPLE_UART_USED_SIZE: NonZeroU64 = NonZeroU64::new(2).expect("known non-zero");
+pub const SIMPLE_UART_ADDRESS_SIZE: NonZeroU64 = NonZeroU64::new(1 << 6).expect("known non-zero");
+
+#[hdl(no_static)]
+struct Operation<C: PhantomConstGet<MemoryInterfaceConfig>> {
+    start: MemoryOperationStart<C>,
+    finish: HdlOption<MemoryOperationFinish<C>>,
+}
+
+pub const fn simple_uart_memory_interface_config(
+    op_id_width: usize,
+    start_address: Wrapping<u64>,
+) -> MemoryInterfaceConfig {
+    assert!(
+        start_address
+            .0
+            .is_multiple_of(SIMPLE_UART_ADDRESS_SIZE.get()),
+        "start_address must be properly aligned"
+    );
+    MemoryInterfaceConfig {
+        log2_bus_width_in_bytes: SIMPLE_UART_LOG2_BUS_WIDTH,
+        queue_capacity: const { NonZeroUsize::new(1).unwrap() },
+        op_id_width,
+        address_range: AddressRange::Limited {
+            start: start_address,
+            size: SIMPLE_UART_ADDRESS_SIZE,
+        },
+    }
+}
+
+#[hdl_module]
+pub fn simple_uart(
+    config: PhantomConst<MemoryInterfaceConfig>,
+    clock_input_properties: PhantomConst<peripherals::ClockInputProperties>,
+    baud_rate: f64,
+    receiver_queue_size: NonZeroUsize,
+) {
+    let start_address = config.get().address_range.start();
+    assert_eq!(
+        *config.get(),
+        simple_uart_memory_interface_config(config.get().op_id_width, start_address),
+    );
+    #[hdl]
+    let cd: ClockDomain = m.input();
+    #[hdl]
+    let memory_interface: MemoryInterface<PhantomConst<MemoryInterfaceConfig>> =
+        m.input(MemoryInterface[config]);
+    #[hdl]
+    let uart: peripherals::Uart = m.output();
+
+    #[hdl]
+    let rx_sync_intermediate_reg: Bool = reg_builder().clock_domain(cd).reset(true);
+    annotate(rx_sync_intermediate_reg, DontTouchAnnotation);
+    #[hdl]
+    let rx_sync_final_reg: Bool = reg_builder().clock_domain(cd).reset(true);
+    annotate(rx_sync_final_reg, DontTouchAnnotation);
+    connect(rx_sync_intermediate_reg, uart.rx);
+    connect(rx_sync_final_reg, rx_sync_intermediate_reg);
+    #[hdl]
+    let rx_synchronized: Bool = wire();
+    annotate(rx_synchronized, DontTouchAnnotation);
+    connect(rx_synchronized, rx_sync_final_reg);
+
+    #[hdl]
+    let clk_gen = instance(uart_clock_gen(clock_input_properties, baud_rate));
+    connect(clk_gen.cd, cd);
+
+    #[hdl]
+    let transmitter = instance(transmitter());
+    connect(transmitter.cd, cd);
+    connect(transmitter.tick, clk_gen.tick);
+    connect(uart.tx, transmitter.tx);
+
+    #[hdl]
+    let receiver = instance(receiver(receiver_queue_size));
+    connect(receiver.cd, cd);
+    connect(receiver.tick, clk_gen.tick);
+    connect(receiver.rx_synchronized, rx_synchronized);
+
+    #[hdl]
+    let operation_reg = reg_builder()
+        .clock_domain(cd)
+        .reset(HdlOption[Operation[config]].HdlNone());
+
+    let next_op_ids_ty = memory_interface.ty().next_op_ids.HdlSome;
+    #[hdl]
+    if let HdlSome(operation) = operation_reg {
+        #[hdl]
+        let MemoryInterface::<_> {
+            start,
+            finish,
+            next_op_ids,
+            config: _,
+        } = memory_interface;
+        connect(start.ready, false);
+        connect(finish.data, operation.finish);
+        connect(
+            next_op_ids,
+            HdlSome(ArrayVec::from_parts_unchecked(
+                repeat(operation.start.op_id, next_op_ids_ty.capacity()),
+                1u8.cast_to(next_op_ids_ty.len_ty()),
+            )),
+        );
+        connect(transmitter.input_byte.data, HdlNone());
+        connect(receiver.output_byte.ready, false);
+        #[hdl]
+        if let HdlSome(_) = operation.finish {
+            #[hdl]
+            if finish.ready {
+                connect(operation_reg, operation_reg.ty().HdlNone());
+            }
+        } else {
+            #[hdl]
+            let MemoryOperationStart::<_> {
+                kind,
+                addr,
+                write_data,
+                rw_mask,
+                op_id: _,
+                config: _,
+            } = operation.start;
+            #[hdl]
+            let valid_addr = wire();
+            connect(valid_addr, true);
+            #[hdl]
+            let all_ready = wire();
+            connect(all_ready, true);
+            #[hdl]
+            let read_data = wire(write_data.ty());
+            for (byte_index, ((rw_mask, write_data), read_data)) in rw_mask
+                .into_iter()
+                .zip(write_data)
+                .zip(read_data)
+                .enumerate()
+            {
+                let byte_addr = (addr | byte_index).cast_to_static::<UInt<64>>();
+                connect(read_data, 0u8);
+                #[hdl]
+                if rw_mask {
+                    #[hdl]
+                    match kind {
+                        MemoryOperationKind::Read => {
+                            #[hdl]
+                            if byte_addr
+                                .cmp_eq(start_address.0.wrapping_add(SIMPLE_UART_RECEIVE_OFFSET))
+                            {
+                                connect(receiver.output_byte.ready, valid_addr);
+                                #[hdl]
+                                if let HdlSome(byte) = receiver.output_byte.data {
+                                    connect(read_data, byte);
+                                }
+                                // if there is no byte ready yet, we read a zero to avoid blocking the CPU on external inputs.
+                            } else if byte_addr
+                                .cmp_eq(start_address.0.wrapping_add(SIMPLE_UART_STATUS_OFFSET))
+                            {
+                                connect(
+                                    read_data,
+                                    receiver
+                                        .queue_status
+                                        .cast_to_bits()
+                                        .cast_to_static::<UInt<8>>(),
+                                );
+                            } else {
+                                connect(valid_addr, false);
+                            }
+                        }
+                        MemoryOperationKind::Write => {
+                            #[hdl]
+                            if byte_addr
+                                .cmp_eq(start_address.0.wrapping_add(SIMPLE_UART_TRANSMIT_OFFSET))
+                            {
+                                #[hdl]
+                                if !transmitter.input_byte.ready {
+                                    connect(all_ready, false);
+                                }
+                                #[hdl]
+                                if valid_addr {
+                                    connect(transmitter.input_byte.data, HdlSome(write_data));
+                                }
+                            } else {
+                                connect(valid_addr, false);
+                            }
+                        }
+                    }
+                }
+            }
+            #[hdl]
+            if !valid_addr {
+                connect(
+                    operation_reg,
+                    HdlSome(
+                        #[hdl]
+                        Operation::<_> {
+                            start: operation.start,
+                            finish: HdlSome(
+                                #[hdl]
+                                MemoryOperationFinish::<_> {
+                                    kind: MemoryOperationFinishKind
+                                        .Error(MemoryOperationErrorKind.Generic()),
+                                    read_data,
+                                    config,
+                                },
+                            ),
+                        },
+                    ),
+                );
+            } else if all_ready {
+                connect(
+                    operation_reg,
+                    HdlSome(
+                        #[hdl]
+                        Operation::<_> {
+                            start: operation.start,
+                            finish: HdlSome(
+                                #[hdl]
+                                MemoryOperationFinish::<_> {
+                                    kind: MemoryOperationFinishKind.Success(kind),
+                                    read_data,
+                                    config,
+                                },
+                            ),
+                        },
+                    ),
+                );
+            }
+        }
+    } else {
+        #[hdl]
+        let MemoryInterface::<_> {
+            start,
+            finish,
+            next_op_ids,
+            config: _,
+        } = memory_interface;
+        connect(start.ready, true);
+        connect(finish.data, finish.ty().data.HdlNone());
+        connect(
+            next_op_ids,
+            HdlSome(next_op_ids_ty.new_sim(next_op_ids_ty.element().zero())),
+        );
+        #[hdl]
+        if let HdlSome(start) = start.data {
+            connect(
+                operation_reg,
+                HdlSome(
+                    #[hdl]
+                    Operation::<_> {
+                        start,
+                        finish: operation_reg.ty().HdlSome.finish.HdlNone(),
+                    },
+                ),
+            );
+        }
+        connect(transmitter.input_byte.data, HdlNone());
+        connect(receiver.output_byte.ready, false);
+    }
+}

crates/cpu/src/rename_execute_retire.rs

@@ -0,0 +1,192 @@
+// SPDX-License-Identifier: LGPL-3.0-or-later
+// See Notices.txt for copyright information
+
+use std::collections::VecDeque;
+
+use crate::{
+    config::{CpuConfig, CpuConfigFetchWidth, CpuConfigRobSize, PhantomConstCpuConfig},
+    instruction::{MOp, MOpRegNum, PRegNum},
+    next_pc::{RetireToNextPcInterfaceInner, SimValueDefault},
+    util::array_vec::ArrayVec,
+};
+use fayalite::{
+    int::UIntInRangeInclusiveType, prelude::*, ty::OpaqueSimValue, util::ready_valid::ReadyValid,
+};
+
+#[hdl]
+/// A µOp along with the state needed for this instance of the µOp.
+pub struct MOpInstance<MOp> {
+    pub fetch_block_id: UInt<8>,
+    pub id: UInt<12>,
+    pub pc: UInt<64>,
+    /// initialized to 0 by decoder, overwritten by `next_pc()`
+    pub predicted_next_pc: UInt<64>,
+    pub size_in_bytes: UInt<4>,
+    /// `true` if this &micro;Op is the first &micro;Op in the ISA-level instruction.
+    /// In general, a single &micro;Op can't be cancelled by itself,
+    /// it needs to be cancelled along with all other &micro;Ops that
+    /// come from the same ISA-level instruction.
+    pub is_first_mop_in_insn: Bool,
+    pub mop: MOp,
+}
+
+#[hdl(no_static)]
+/// TODO: merge with [`crate::next_pc::PostDecodeOutputInterface`]
+pub struct PostDecodeOutputInterface<C: PhantomConstGet<CpuConfig>> {
+    pub insns: ArrayVec<MOpInstance<MOp>, CpuConfigFetchWidth<C>>,
+    #[hdl(flip)]
+    pub ready: UIntInRangeInclusiveType<ConstUsize<0>, CpuConfigFetchWidth<C>>,
+    /// tells the rename/execute/retire circuit to cancel all non-retired instructions
+    pub cancel: ReadyValid<()>,
+    pub config: C,
+}
+
+#[hdl(no_static)]
+/// handles updating speculative branch predictor state (e.g. branch histories)
+/// when instructions retire, as well as updating state when a
+/// branch instruction is mis-speculated.
+pub struct RetireToNextPcInterface<C: PhantomConstGet<CpuConfig>> {
+    pub inner: ReadyValid<RetireToNextPcInterfaceInner<C>>,
+    /// only for debugging
+    pub next_insns: HdlOption<ArrayVec<MOpInstance<MOp>, CpuConfigRobSize<C>>>,
+}
+
+#[hdl(no_static)]
+pub struct RenameExecuteRetireDebugState<C: PhantomConstGet<CpuConfig>> {
+    rename_table: Array<PRegNum<C>, { 1 << MOpRegNum::WIDTH }>,
+    retire_rename_table: Array<PRegNum<C>, { 1 << MOpRegNum::WIDTH }>,
+    rob: ArrayVec<RobEntryDebugState<C>, CpuConfigRobSize<C>>,
+}
+
+impl<C: PhantomConstCpuConfig> SimValueDefault for RenameExecuteRetireDebugState<C> {
+    fn sim_value_default(self) -> SimValue<Self> {
+        SimValue::from_opaque(
+            self,
+            OpaqueSimValue::from_bits(UInt::new(self.canonical().bit_width()).zero()),
+        )
+    }
+}
+
+#[derive(Debug, Clone)]
+struct RenameTable<C: PhantomConstCpuConfig> {
+    entries: Box<[SimValue<PRegNum<C>>; 1 << MOpRegNum::WIDTH]>,
+    config: C,
+}
+
+impl<C: PhantomConstCpuConfig> RenameTable<C> {
+    fn new(config: C) -> Self {
+        Self {
+            entries: vec![PRegNum[config].const_zero().into_sim_value(); 1 << MOpRegNum::WIDTH]
+                .try_into()
+                .expect("size is known to match"),
+            config,
+        }
+    }
+}
+
+#[hdl(no_static)]
+struct RobEntryDebugState<C: PhantomConstGet<CpuConfig>> {
+    config: C,
+}
+
+#[derive(Debug)]
+struct RobEntry<C: PhantomConstCpuConfig> {
+    config: C,
+}
+
+#[derive(Debug)]
+struct RenameExecuteRetireState<C: PhantomConstCpuConfig> {
+    rename_table: RenameTable<C>,
+    retire_rename_table: RenameTable<C>,
+    rob: VecDeque<RobEntry<C>>,
+}
+
+impl<C: PhantomConstCpuConfig> RenameExecuteRetireState<C> {
+    fn new(config: C) -> Self {
+        let rename_table = RenameTable::new(config);
+        Self {
+            rename_table: rename_table.clone(),
+            retire_rename_table: rename_table,
+            rob: VecDeque::with_capacity(CpuConfigRobSize[config]),
+        }
+    }
+    async fn write_for_debug(
+        &self,
+        sim: &mut ExternModuleSimulationState,
+        state_for_debug: Expr<RenameExecuteRetireDebugState<C>>,
+    ) {
+        // TODO
+    }
+    async fn write_to_next_pc_next_insns(
+        &self,
+        sim: &mut ExternModuleSimulationState,
+        next_insns: Expr<HdlOption<ArrayVec<MOpInstance<MOp>, CpuConfigRobSize<C>>>>,
+    ) {
+        // TODO
+    }
+}
+
+#[hdl]
+async fn rename_execute_retire_run(
+    mut sim: ExternModuleSimulationState,
+    cd: Expr<ClockDomain>,
+    from_post_decode: Expr<PostDecodeOutputInterface<PhantomConst<CpuConfig>>>,
+    to_next_pc: Expr<RetireToNextPcInterface<PhantomConst<CpuConfig>>>,
+    state_for_debug: Expr<RenameExecuteRetireDebugState<PhantomConst<CpuConfig>>>,
+    config: PhantomConst<CpuConfig>,
+) {
+    let mut state = RenameExecuteRetireState::new(config);
+    loop {
+        state
+            .write_to_next_pc_next_insns(&mut sim, to_next_pc.next_insns)
+            .await;
+        state.write_for_debug(&mut sim, state_for_debug).await;
+        sim.wait_for_clock_edge(cd.clk).await;
+        todo!("step state based on I/O");
+    }
+}
+
+#[hdl_module(extern)]
+pub fn rename_execute_retire(config: PhantomConst<CpuConfig>) {
+    #[hdl]
+    let cd: ClockDomain = m.input();
+    #[hdl]
+    let from_post_decode: PostDecodeOutputInterface<PhantomConst<CpuConfig>> =
+        m.input(PostDecodeOutputInterface[config]);
+    #[hdl]
+    let to_next_pc: RetireToNextPcInterface<PhantomConst<CpuConfig>> =
+        m.output(RetireToNextPcInterface[config]);
+    #[hdl]
+    let state_for_debug: RenameExecuteRetireDebugState<PhantomConst<CpuConfig>> =
+        m.output(RenameExecuteRetireDebugState[config]);
+    m.register_clock_for_past(cd.clk);
+    m.extern_module_simulation_fn(
+        (cd, from_post_decode, to_next_pc, state_for_debug, config),
+        |(cd, from_post_decode, to_next_pc, state_for_debug, config), mut sim| async move {
+            sim.write(state_for_debug, state_for_debug.ty().sim_value_default())
+                .await;
+            sim.resettable(
+                cd,
+                |mut sim: ExternModuleSimulationState| async move {
+                    sim.write(from_post_decode.ready, 0usize).await;
+                    sim.write(from_post_decode.cancel.ready, false).await;
+                    sim.write(to_next_pc.inner.data, to_next_pc.ty().inner.data.HdlNone())
+                        .await;
+                    sim.write(to_next_pc.next_insns, to_next_pc.ty().next_insns.HdlNone())
+                        .await;
+                },
+                |sim, ()| {
+                    rename_execute_retire_run(
+                        sim,
+                        cd,
+                        from_post_decode,
+                        to_next_pc,
+                        state_for_debug,
+                        config,
+                    )
+                },
+            )
+            .await;
+        },
+    );
+}

crates/cpu/src/unit.rs

@@ -2,7 +2,7 @@
 // See Notices.txt for copyright information
 
 use crate::{
-    config::CpuConfig,
+    config::{CpuConfig, PhantomConstCpuConfig},
     instruction::{
         AluBranchMOp, LoadStoreMOp, MOp, MOpDestReg, MOpInto, MOpRegNum, MOpTrait,
         MOpVariantVisitOps, MOpVariantVisitor, MOpVisitVariants, RenamedMOp, UnitOutRegNum,
@@ -48,7 +48,7 @@ macro_rules! all_units {
         }
 
         impl $UnitKind {
-            pub fn unit(self, config: &CpuConfig, unit_index: usize) -> DynUnit {
+            pub fn unit(self, config: PhantomConst<CpuConfig>, unit_index: usize) -> DynUnit {
                 match self {
                     $($UnitKind::$Unit => $create_dyn_unit_fn(config, unit_index),)*
                 }
@@ -277,9 +277,9 @@ pub struct UnitResultCompleted<ExtraOut> {
     pub extra_out: ExtraOut,
 }
 
-#[hdl(cmp_eq)]
-pub struct UnitOutputWrite<OutRegNumWidth: Size> {
-    pub which: UnitOutRegNum<OutRegNumWidth>,
+#[hdl(cmp_eq, no_static)]
+pub struct UnitOutputWrite<C: PhantomConstGet<CpuConfig>> {
+    pub which: UnitOutRegNum<C>,
     pub value: PRegValue,
 }
 
@@ -300,21 +300,21 @@ impl<ExtraOut: Type> UnitResult<ExtraOut> {
     }
 }
 
-#[hdl]
-pub struct UnitOutput<OutRegNumWidth: Size, ExtraOut> {
-    pub which: UnitOutRegNum<OutRegNumWidth>,
+#[hdl(no_static)]
+pub struct UnitOutput<C: PhantomConstGet<CpuConfig>, ExtraOut> {
+    pub which: UnitOutRegNum<C>,
     pub result: UnitResult<ExtraOut>,
 }
 
-impl<OutRegNumWidth: Size, ExtraOut: Type> UnitOutput<OutRegNumWidth, ExtraOut> {
+impl<C: PhantomConstCpuConfig, ExtraOut: Type> UnitOutput<C, ExtraOut> {
     pub fn extra_out_ty(self) -> ExtraOut {
         self.result.extra_out_ty()
     }
 }
 
-#[hdl(cmp_eq)]
-pub struct UnitCancelInput<OutRegNumWidth: Size> {
-    pub which: UnitOutRegNum<OutRegNumWidth>,
+#[hdl(cmp_eq, no_static)]
+pub struct UnitCancelInput<C: PhantomConstGet<CpuConfig>> {
+    pub which: UnitOutRegNum<C>,
 }
 
 pub trait UnitTrait:
@@ -332,7 +332,7 @@ pub trait UnitTrait:
 
     fn extract_mop(
         &self,
-        mop: Expr<RenamedMOp<UnitOutRegNum<DynSize>, DynSize>>,
+        mop: Expr<RenamedMOp<UnitOutRegNum<PhantomConst<CpuConfig>>, DynSize>>,
     ) -> Expr<HdlOption<Self::MOp>>;
 
     fn module(&self) -> Interned<Module<Self::Type>>;
@@ -340,7 +340,7 @@ pub trait UnitTrait:
     fn unit_to_reg_alloc(
         &self,
         this: Expr<Self::Type>,
-    ) -> Expr<UnitToRegAlloc<Self::MOp, Self::ExtraOut, DynSize, DynSize, DynSize>>;
+    ) -> Expr<UnitToRegAlloc<PhantomConst<CpuConfig>, Self::MOp, Self::ExtraOut>>;
 
     fn cd(&self, this: Expr<Self::Type>) -> Expr<ClockDomain>;
 
@@ -390,7 +390,7 @@ impl UnitTrait for DynUnit {
 
     fn extract_mop(
         &self,
-        mop: Expr<RenamedMOp<UnitOutRegNum<DynSize>, DynSize>>,
+        mop: Expr<RenamedMOp<UnitOutRegNum<PhantomConst<CpuConfig>>, DynSize>>,
     ) -> Expr<HdlOption<Self::MOp>> {
         self.unit.extract_mop(mop)
     }
@@ -402,7 +402,7 @@ impl UnitTrait for DynUnit {
     fn unit_to_reg_alloc(
         &self,
         this: Expr<Self::Type>,
-    ) -> Expr<UnitToRegAlloc<Self::MOp, Self::ExtraOut, DynSize, DynSize, DynSize>> {
+    ) -> Expr<UnitToRegAlloc<PhantomConst<CpuConfig>, Self::MOp, Self::ExtraOut>> {
         self.unit.unit_to_reg_alloc(this)
     }
 
@@ -445,7 +445,7 @@ impl<T: UnitTrait + Clone + std::hash::Hash + Eq> UnitTrait for DynUnitWrapper<T
 
     fn extract_mop(
         &self,
-        mop: Expr<RenamedMOp<UnitOutRegNum<DynSize>, DynSize>>,
+        mop: Expr<RenamedMOp<UnitOutRegNum<PhantomConst<CpuConfig>>, DynSize>>,
     ) -> Expr<HdlOption<Self::MOp>> {
         Expr::from_enum(Expr::as_enum(self.0.extract_mop(mop)))
     }
@@ -457,7 +457,7 @@ impl<T: UnitTrait + Clone + std::hash::Hash + Eq> UnitTrait for DynUnitWrapper<T
     fn unit_to_reg_alloc(
         &self,
         this: Expr<Self::Type>,
-    ) -> Expr<UnitToRegAlloc<Self::MOp, Self::ExtraOut, DynSize, DynSize, DynSize>> {
+    ) -> Expr<UnitToRegAlloc<PhantomConst<CpuConfig>, Self::MOp, Self::ExtraOut>> {
         Expr::from_bundle(Expr::as_bundle(
             self.0.unit_to_reg_alloc(Expr::from_bundle(this)),
         ))

crates/cpu/src/unit/alu_branch.rs

@@ -19,16 +19,13 @@ use crate::{
     },
 };
 use fayalite::{
-    intern::{Intern, Interned},
-    module::wire_with_loc,
-    prelude::*,
-    util::ready_valid::ReadyValid,
+    intern::Interned, module::wire_with_loc, prelude::*, util::ready_valid::ReadyValid,
 };
 use std::{collections::HashMap, ops::RangeTo};
 
 #[hdl]
 fn add_sub<SrcCount: KnownSize>(
-    mop: Expr<AddSubMOp<UnitOutRegNum<DynSize>, DynSize, SrcCount>>,
+    mop: Expr<AddSubMOp<UnitOutRegNum<PhantomConst<CpuConfig>>, DynSize, SrcCount>>,
     pc: Expr<UInt<64>>,
     flags_mode: Expr<FlagsMode>,
     src_values: Expr<Array<PRegValue, { COMMON_MOP_SRC_LEN }>>,
@@ -245,7 +242,7 @@ fn add_sub<SrcCount: KnownSize>(
 
 #[hdl]
 fn logical_flags(
-    mop: Expr<LogicalFlagsMOp<UnitOutRegNum<DynSize>, DynSize>>,
+    mop: Expr<LogicalFlagsMOp<UnitOutRegNum<PhantomConst<CpuConfig>>, DynSize>>,
     flags_mode: Expr<FlagsMode>,
     src_values: Expr<Array<PRegValue, { COMMON_MOP_SRC_LEN }>>,
 ) -> Expr<UnitResultCompleted<()>> {
@@ -259,7 +256,7 @@ fn logical_flags(
 
 #[hdl]
 fn logical(
-    mop: Expr<LogicalMOp<UnitOutRegNum<DynSize>, DynSize, ConstUsize<2>>>,
+    mop: Expr<LogicalMOp<UnitOutRegNum<PhantomConst<CpuConfig>>, DynSize, ConstUsize<2>>>,
     flags_mode: Expr<FlagsMode>,
     src_values: Expr<Array<PRegValue, { COMMON_MOP_SRC_LEN }>>,
 ) -> Expr<UnitResultCompleted<()>> {
@@ -273,7 +270,7 @@ fn logical(
 
 #[hdl]
 fn logical_i(
-    mop: Expr<LogicalMOp<UnitOutRegNum<DynSize>, DynSize, ConstUsize<1>>>,
+    mop: Expr<LogicalMOp<UnitOutRegNum<PhantomConst<CpuConfig>>, DynSize, ConstUsize<1>>>,
     flags_mode: Expr<FlagsMode>,
     src_values: Expr<Array<PRegValue, { COMMON_MOP_SRC_LEN }>>,
 ) -> Expr<UnitResultCompleted<()>> {
@@ -287,7 +284,7 @@ fn logical_i(
 
 #[hdl]
 fn shift_rotate(
-    mop: Expr<ShiftRotateMOp<UnitOutRegNum<DynSize>, DynSize>>,
+    mop: Expr<ShiftRotateMOp<UnitOutRegNum<PhantomConst<CpuConfig>>, DynSize>>,
     flags_mode: Expr<FlagsMode>,
     src_values: Expr<Array<PRegValue, { COMMON_MOP_SRC_LEN }>>,
 ) -> Expr<UnitResultCompleted<()>> {
@@ -301,7 +298,7 @@ fn shift_rotate(
 
 #[hdl]
 fn compare<SrcCount: KnownSize>(
-    mop: Expr<CompareMOp<UnitOutRegNum<DynSize>, DynSize, SrcCount>>,
+    mop: Expr<CompareMOp<UnitOutRegNum<PhantomConst<CpuConfig>>, DynSize, SrcCount>>,
     flags_mode: Expr<FlagsMode>,
     src_values: Expr<Array<PRegValue, { COMMON_MOP_SRC_LEN }>>,
 ) -> Expr<UnitResultCompleted<()>> {
@@ -315,7 +312,7 @@ fn compare<SrcCount: KnownSize>(
 
 #[hdl]
 fn branch<SrcCount: KnownSize>(
-    mop: Expr<BranchMOp<UnitOutRegNum<DynSize>, DynSize, SrcCount>>,
+    mop: Expr<BranchMOp<UnitOutRegNum<PhantomConst<CpuConfig>>, DynSize, SrcCount>>,
     pc: Expr<UInt<64>>,
     flags_mode: Expr<FlagsMode>,
     src_values: Expr<Array<PRegValue, { COMMON_MOP_SRC_LEN }>>,
@@ -330,7 +327,7 @@ fn branch<SrcCount: KnownSize>(
 
 #[hdl]
 fn read_special(
-    mop: Expr<ReadSpecialMOp<UnitOutRegNum<DynSize>, DynSize>>,
+    mop: Expr<ReadSpecialMOp<UnitOutRegNum<PhantomConst<CpuConfig>>, DynSize>>,
     pc: Expr<UInt<64>>,
     flags_mode: Expr<FlagsMode>,
     src_values: Expr<Array<PRegValue, { COMMON_MOP_SRC_LEN }>>,
@@ -344,20 +341,18 @@ fn read_special(
 }
 
 #[hdl_module]
-pub fn alu_branch(config: &CpuConfig, unit_index: usize) {
+pub fn alu_branch(config: PhantomConst<CpuConfig>, unit_index: usize) {
     #[hdl]
     let cd: ClockDomain = m.input();
     #[hdl]
     let unit_to_reg_alloc: UnitToRegAlloc<
-        AluBranchMOp<UnitOutRegNum<DynSize>, DynSize>,
+        PhantomConst<CpuConfig>,
+        AluBranchMOp<UnitOutRegNum<PhantomConst<CpuConfig>>, DynSize>,
         (),
-        DynSize,
-        DynSize,
-        DynSize,
-    > = m.output(config.unit_to_reg_alloc(
-        AluBranchMOp[config.unit_out_reg_num()][config.p_reg_num_width()],
-        (),
-    ));
+    > = m.output(
+        UnitToRegAlloc[config][AluBranchMOp[UnitOutRegNum[config]][config.get().p_reg_num_width()]]
+            [()],
+    );
     #[hdl]
     let global_state: GlobalState = m.input();
 
@@ -375,10 +370,11 @@ pub fn alu_branch(config: &CpuConfig, unit_index: usize) {
     #[hdl]
     if let HdlSome(execute_start) = ReadyValid::firing_data(unit_base.execute_start) {
         #[hdl]
-        let ExecuteStart::<_> {
+        let ExecuteStart::<_, _> {
             mop,
             pc,
             src_values,
+            config: _,
         } = execute_start;
         #[hdl]
         match mop {
@@ -580,14 +576,14 @@ pub fn alu_branch(config: &CpuConfig, unit_index: usize) {
 
 #[derive(Debug, Copy, Clone, PartialEq, Eq, Hash)]
 pub struct AluBranch {
-    config: Interned<CpuConfig>,
+    config: PhantomConst<CpuConfig>,
     module: Interned<Module<alu_branch>>,
 }
 
 impl AluBranch {
-    pub fn new(config: &CpuConfig, unit_index: usize) -> Self {
+    pub fn new(config: PhantomConst<CpuConfig>, unit_index: usize) -> Self {
         Self {
-            config: config.intern(),
+            config,
             module: alu_branch(config, unit_index),
         }
     }
@@ -596,7 +592,7 @@ impl AluBranch {
 impl UnitTrait for AluBranch {
     type Type = alu_branch;
     type ExtraOut = ();
-    type MOp = AluBranchMOp<UnitOutRegNum<DynSize>, DynSize>;
+    type MOp = AluBranchMOp<UnitOutRegNum<PhantomConst<CpuConfig>>, DynSize>;
 
     fn ty(&self) -> Self::Type {
         self.module.io_ty()
@@ -616,7 +612,7 @@ impl UnitTrait for AluBranch {
 
     fn extract_mop(
         &self,
-        mop: Expr<RenamedMOp<UnitOutRegNum<DynSize>, DynSize>>,
+        mop: Expr<RenamedMOp<UnitOutRegNum<PhantomConst<CpuConfig>>, DynSize>>,
     ) -> Expr<HdlOption<Self::MOp>> {
         UnitMOp::alu_branch_mop(mop)
     }
@@ -628,7 +624,7 @@ impl UnitTrait for AluBranch {
     fn unit_to_reg_alloc(
         &self,
         this: Expr<Self::Type>,
-    ) -> Expr<UnitToRegAlloc<Self::MOp, Self::ExtraOut, DynSize, DynSize, DynSize>> {
+    ) -> Expr<UnitToRegAlloc<PhantomConst<CpuConfig>, Self::MOp, Self::ExtraOut>> {
         this.unit_to_reg_alloc
     }
 

crates/cpu/src/unit/unit_base.rs

@@ -2,7 +2,7 @@
 // See Notices.txt for copyright information
 
 use crate::{
-    config::CpuConfig,
+    config::{CpuConfig, CpuConfigUnitCount, PhantomConstCpuConfig},
     instruction::{COMMON_MOP_SRC_LEN, MOpTrait, PRegNum, UnitNum, UnitOutRegNum},
     register::PRegValue,
     unit::{UnitCancelInput, UnitOutput, UnitOutputWrite},
@@ -15,13 +15,11 @@ use fayalite::{
     ty::StaticType,
     util::ready_valid::ReadyValid,
 };
-use std::marker::PhantomData;
 
-#[hdl]
-pub struct UnitForwardingInfo<UnitNumWidth: Size, OutRegNumWidth: Size, UnitCount: Size> {
-    pub unit_output_writes: ArrayType<HdlOption<UnitOutputWrite<OutRegNumWidth>>, UnitCount>,
-    pub unit_reg_frees: ArrayType<HdlOption<UnitOutRegNum<OutRegNumWidth>>, UnitCount>,
-    pub _phantom: PhantomData<UnitNumWidth>,
+#[hdl(no_static)]
+pub struct UnitForwardingInfo<C: PhantomConstGet<CpuConfig>> {
+    pub unit_output_writes: ArrayType<HdlOption<UnitOutputWrite<C>>, CpuConfigUnitCount<C>>,
+    pub unit_reg_frees: ArrayType<HdlOption<UnitOutRegNum<C>>, CpuConfigUnitCount<C>>,
 }
 
 #[hdl]
@@ -30,26 +28,18 @@ pub struct UnitInput<MOp: Type> {
     pub pc: UInt<64>,
 }
 
-#[hdl]
-pub struct UnitToRegAlloc<
-    MOp: Type,
-    ExtraOut: Type,
-    UnitNumWidth: Size,
-    OutRegNumWidth: Size,
-    UnitCount: Size,
-> {
+#[hdl(no_static)]
+pub struct UnitToRegAlloc<C: PhantomConstGet<CpuConfig>, MOp: Type, ExtraOut: Type> {
     #[hdl(flip)]
-    pub unit_forwarding_info: UnitForwardingInfo<UnitNumWidth, OutRegNumWidth, UnitCount>,
+    pub unit_forwarding_info: UnitForwardingInfo<C>,
     #[hdl(flip)]
     pub input: ReadyValid<UnitInput<MOp>>,
     #[hdl(flip)]
-    pub cancel_input: HdlOption<UnitCancelInput<OutRegNumWidth>>,
-    pub output: HdlOption<UnitOutput<OutRegNumWidth, ExtraOut>>,
+    pub cancel_input: HdlOption<UnitCancelInput<C>>,
+    pub output: HdlOption<UnitOutput<C, ExtraOut>>,
 }
 
-impl<MOp: Type, ExtraOut: Type, UnitNumWidth: Size, OutRegNumWidth: Size, UnitCount: Size>
-    UnitToRegAlloc<MOp, ExtraOut, UnitNumWidth, OutRegNumWidth, UnitCount>
-{
+impl<C: PhantomConstCpuConfig, MOp: Type, ExtraOut: Type> UnitToRegAlloc<C, MOp, ExtraOut> {
     pub fn mop_ty(self) -> MOp {
         self.input.data.HdlSome.mop
     }
@@ -58,16 +48,20 @@ impl<MOp: Type, ExtraOut: Type, UnitNumWidth: Size, OutRegNumWidth: Size, UnitCo
     }
 }
 
-#[hdl]
-pub struct ExecuteStart<MOp: Type + MOpTrait<DestReg = UnitOutRegNum<DynSize>>> {
+#[hdl(no_static)]
+pub struct ExecuteStart<
+    C: PhantomConstGet<CpuConfig>,
+    MOp: Type + MOpTrait<DestReg = UnitOutRegNum<C>>,
+> {
     pub mop: MOp,
     pub pc: UInt<64>,
     pub src_values: Array<PRegValue, { COMMON_MOP_SRC_LEN }>,
+    pub config: C,
 }
 
-#[hdl]
-pub struct ExecuteEnd<OutRegNumWidth: Size, ExtraOut> {
-    pub unit_output: UnitOutput<OutRegNumWidth, ExtraOut>,
+#[hdl(no_static)]
+pub struct ExecuteEnd<C: PhantomConstGet<CpuConfig>, ExtraOut> {
+    pub unit_output: UnitOutput<C, ExtraOut>,
 }
 
 #[hdl]
@@ -240,10 +234,10 @@ impl InFlightOpsSummary<DynSize> {
 
 #[hdl_module]
 pub fn unit_base<
-    MOp: Type + MOpTrait<DestReg = UnitOutRegNum<DynSize>, SrcRegWidth = DynSize>,
+    MOp: Type + MOpTrait<DestReg = UnitOutRegNum<PhantomConst<CpuConfig>>, SrcRegWidth = DynSize>,
     ExtraOut: Type,
 >(
-    config: &CpuConfig,
+    config: PhantomConst<CpuConfig>,
     unit_index: usize,
     mop_ty: MOp,
     extra_out_ty: ExtraOut,
@@ -251,17 +245,18 @@ pub fn unit_base<
     #[hdl]
     let cd: ClockDomain = m.input();
     #[hdl]
-    let unit_to_reg_alloc: UnitToRegAlloc<MOp, ExtraOut, DynSize, DynSize, DynSize> =
-        m.output(config.unit_to_reg_alloc(mop_ty, extra_out_ty));
+    let unit_to_reg_alloc: UnitToRegAlloc<PhantomConst<CpuConfig>, MOp, ExtraOut> =
+        m.output(UnitToRegAlloc[config][mop_ty][extra_out_ty]);
     #[hdl]
-    let execute_start: ReadyValid<ExecuteStart<MOp>> = m.output(ReadyValid[ExecuteStart[mop_ty]]);
+    let execute_start: ReadyValid<ExecuteStart<PhantomConst<CpuConfig>, MOp>> =
+        m.output(ReadyValid[ExecuteStart[config][mop_ty]]);
     #[hdl]
-    let execute_end: HdlOption<ExecuteEnd<DynSize, ExtraOut>> =
-        m.input(HdlOption[ExecuteEnd[config.out_reg_num_width][extra_out_ty]]);
+    let execute_end: HdlOption<ExecuteEnd<PhantomConst<CpuConfig>, ExtraOut>> =
+        m.input(HdlOption[ExecuteEnd[config][extra_out_ty]]);
 
     connect(execute_start.data, execute_start.ty().data.HdlNone());
 
-    let max_in_flight = config.unit_max_in_flight(unit_index).get();
+    let max_in_flight = config.get().unit_max_in_flight(unit_index).get();
     let in_flight_op_ty = InFlightOp[mop_ty];
     #[hdl]
     let in_flight_ops = reg_builder()
@@ -279,16 +274,15 @@ pub fn unit_base<
     );
 
     #[hdl]
-    let UnitForwardingInfo::<_, _, _> {
+    let UnitForwardingInfo::<_> {
         unit_output_writes,
         unit_reg_frees,
-        _phantom: _,
     } = unit_to_reg_alloc.unit_forwarding_info;
     #[hdl]
     let read_src_regs = wire(mop_ty.src_regs_ty());
     connect(
         read_src_regs,
-        repeat(config.p_reg_num().const_zero().cast_to_bits(), ConstUsize),
+        repeat(PRegNum[config].const_zero().cast_to_bits(), ConstUsize),
     );
     #[hdl]
     let read_src_values = wire();
@@ -297,7 +291,7 @@ pub fn unit_base<
     let input_src_regs = wire(mop_ty.src_regs_ty());
     connect(
         input_src_regs,
-        repeat(config.p_reg_num().const_zero().cast_to_bits(), ConstUsize),
+        repeat(PRegNum[config].const_zero().cast_to_bits(), ConstUsize),
     );
     #[hdl]
     let input_src_regs_valid = wire();
@@ -309,7 +303,7 @@ pub fn unit_base<
                 Bool,
                 SourceLocation::caller(),
             );
-            mem.depth(1 << config.out_reg_num_width);
+            mem.depth(1 << config.get().out_reg_num_width);
             mem
         })
         .collect();
@@ -319,11 +313,11 @@ pub fn unit_base<
             PRegValue,
             SourceLocation::caller(),
         );
-        unit_output_regs.depth(1 << config.out_reg_num_width);
+        unit_output_regs.depth(1 << config.get().out_reg_num_width);
 
         for src_index in 0..COMMON_MOP_SRC_LEN {
             let read_port = unit_output_regs.new_read_port();
-            let p_reg_num = read_src_regs[src_index].cast_bits_to(config.p_reg_num());
+            let p_reg_num = read_src_regs[src_index].cast_bits_to(PRegNum[config]);
             connect_any(read_port.addr, p_reg_num.unit_out_reg.value);
             connect(read_port.en, false);
             connect(read_port.clk, cd.clk);
@@ -336,7 +330,7 @@ pub fn unit_base<
 
         for src_index in 0..COMMON_MOP_SRC_LEN {
             let read_port = unit_output_regs_valid[unit_index].new_read_port();
-            let p_reg_num = input_src_regs[src_index].cast_bits_to(config.p_reg_num());
+            let p_reg_num = input_src_regs[src_index].cast_bits_to(PRegNum[config]);
             connect_any(read_port.addr, p_reg_num.unit_out_reg.value);
             connect(read_port.en, false);
             connect(read_port.clk, cd.clk);
@@ -367,8 +361,8 @@ pub fn unit_base<
             connect_any(ready_write_port.addr, unit_output_write.which.value);
             connect(ready_write_port.en, true);
             let p_reg_num = #[hdl]
-            PRegNum::<_, _> {
-                unit_num: config.unit_num().from_index(unit_index),
+            PRegNum::<_> {
+                unit_num: UnitNum[config].from_index(unit_index),
                 unit_out_reg: unit_output_write.which,
             };
             for src_index in 0..COMMON_MOP_SRC_LEN {
@@ -399,10 +393,11 @@ pub fn unit_base<
                 execute_start.data,
                 HdlSome(
                     #[hdl]
-                    ExecuteStart::<_> {
+                    ExecuteStart::<_, _> {
                         mop: in_flight_op.mop,
                         pc: in_flight_op.pc,
                         src_values: read_src_values,
+                        config,
                     },
                 ),
             );
@@ -425,7 +420,7 @@ pub fn unit_base<
         let input_mop_src_regs = wire(mop_ty.src_regs_ty());
         connect(
             input_mop_src_regs,
-            repeat(config.p_reg_num().const_zero().cast_to_bits(), ConstUsize),
+            repeat(PRegNum[config].const_zero().cast_to_bits(), ConstUsize),
         );
         MOp::connect_src_regs(mop, input_mop_src_regs);
         let src_ready_flags = wire_with_loc(
@@ -497,7 +492,7 @@ pub fn unit_base<
             );
             connect(
                 src_regs,
-                repeat(config.p_reg_num().const_zero().cast_to_bits(), ConstUsize),
+                repeat(PRegNum[config].const_zero().cast_to_bits(), ConstUsize),
             );
             MOp::connect_src_regs(mop, src_regs);
 
@@ -521,8 +516,8 @@ pub fn unit_base<
                         value: _,
                     } = unit_output_write;
                     let p_reg_num = #[hdl]
-                    PRegNum::<_, _> {
-                        unit_num: config.unit_num().from_index(unit_index),
+                    PRegNum::<_> {
+                        unit_num: UnitNum[config].from_index(unit_index),
                         unit_out_reg,
                     };
                     for src_index in 0..COMMON_MOP_SRC_LEN {

crates/cpu/tests/fetch.rs

@@ -5,8 +5,9 @@ use cpu::{
     config::{CpuConfig, UnitConfig},
     fetch::{FetchToDecodeInterface, FetchToDecodeInterfaceInner, fetch},
     main_memory_and_io::{
-        MemoryInterface, MemoryOperationErrorKind, MemoryOperationFinish,
-        MemoryOperationFinishKind, MemoryOperationKind, MemoryOperationStart,
+        CpuConfigFetchMemoryInterfaceConfig, MemoryInterface, MemoryInterfaceConfig,
+        MemoryOperationErrorKind, MemoryOperationFinish, MemoryOperationFinishKind,
+        MemoryOperationKind, MemoryOperationStart,
     },
     next_pc::{FETCH_BLOCK_ID_WIDTH, NextPcToFetchInterface, NextPcToFetchInterfaceInner},
     unit::UnitKind,
@@ -92,8 +93,8 @@ fn mock_memory(config: PhantomConst<CpuConfig>) {
     #[hdl]
     let cd: ClockDomain = m.input();
     #[hdl]
-    let memory_interface: MemoryInterface<PhantomConst<CpuConfig>> =
-        m.input(MemoryInterface[config]);
+    let memory_interface: MemoryInterface<PhantomConst<MemoryInterfaceConfig>> =
+        m.input(MemoryInterface[CpuConfigFetchMemoryInterfaceConfig[config]]);
     #[hdl]
     let queue_debug: ArrayVec<MemoryQueueEntry, ConstUsize<{ MEMORY_QUEUE_SIZE }>> = m.output();
     m.register_clock_for_past(cd.clk);
@@ -109,13 +110,12 @@ fn mock_memory(config: PhantomConst<CpuConfig>) {
                     let MemoryInterface::<_> {
                         start,
                         finish,
-                        next_fetch_block_ids,
+                        next_op_ids,
                         config: _,
                     } = memory_interface;
                     sim.write(start.ready, false).await;
                     sim.write(finish.data, finish.ty().data.HdlNone()).await;
-                    sim.write(next_fetch_block_ids, next_fetch_block_ids.ty().HdlNone())
-                        .await;
+                    sim.write(next_op_ids, next_op_ids.ty().HdlNone()).await;
                     sim.write(
                         queue_debug,
                         queue_debug.ty().new_sim(MemoryQueueEntry.default_sim()),
@@ -130,7 +130,7 @@ fn mock_memory(config: PhantomConst<CpuConfig>) {
     #[hdl]
     async fn run_fn(
         cd: Expr<ClockDomain>,
-        memory_interface: Expr<MemoryInterface<PhantomConst<CpuConfig>>>,
+        memory_interface: Expr<MemoryInterface<PhantomConst<MemoryInterfaceConfig>>>,
         queue_debug: Expr<ArrayVec<MemoryQueueEntry, ConstUsize<{ MEMORY_QUEUE_SIZE }>>>,
         random: &RefCell<Random>,
         mut sim: ExternModuleSimulationState,
@@ -138,7 +138,7 @@ fn mock_memory(config: PhantomConst<CpuConfig>) {
         let mut random = random.borrow_mut();
         let config = memory_interface.config.ty();
         let finish_data_ty = memory_interface.finish.data.ty();
-        let next_fetch_block_ids_ty = memory_interface.next_fetch_block_ids.ty();
+        let next_op_ids_ty = memory_interface.next_op_ids.ty();
         let mut queue: VecDeque<SimValue<MemoryQueueEntry>> = VecDeque::new();
         loop {
             for entry in &mut queue {
@@ -156,12 +156,17 @@ fn mock_memory(config: PhantomConst<CpuConfig>) {
             )
             .await;
             sim.write(
-                memory_interface.next_fetch_block_ids,
+                memory_interface.next_op_ids,
                 #[hdl(sim)]
-                next_fetch_block_ids_ty.HdlSome(
-                    next_fetch_block_ids_ty
+                next_op_ids_ty.HdlSome(
+                    next_op_ids_ty
                         .HdlSome
-                        .from_iter_sim(0u8, queue.iter().map(|entry| &entry.fetch_block_id))
+                        .from_iter_sim(
+                            0u8,
+                            queue
+                                .iter()
+                                .map(|entry| SimValue::into_dyn_int(entry.fetch_block_id.clone())),
+                        )
                         .ok()
                         .expect("queue is known to be small enough"),
                 ),
@@ -235,7 +240,7 @@ fn mock_memory(config: PhantomConst<CpuConfig>) {
                         addr,
                         write_data: _,
                         rw_mask,
-                        fetch_block_id,
+                        op_id,
                         config: _,
                     } = memory_operation_start;
                     #[hdl(sim)]
@@ -250,7 +255,7 @@ fn mock_memory(config: PhantomConst<CpuConfig>) {
                     let entry = #[hdl(sim)]
                     MemoryQueueEntry {
                         addr,
-                        fetch_block_id,
+                        fetch_block_id: SimValue::from_dyn_int(op_id),
                         cycles_left: MemoryQueueEntry::get_next_delay(&mut random),
                     };
                     println!("mock memory start: {entry:#?}");

crates/cpu/tests/main_memory_and_io.rs

@@ -0,0 +1,305 @@
+// SPDX-License-Identifier: LGPL-3.0-or-later
+// See Notices.txt for copyright information
+
+use cpu::{
+    main_memory_and_io::{
+        AddressRange, MemoryInterface, MemoryInterfaceConfig, MemoryOperationErrorKind,
+        MemoryOperationFinish, MemoryOperationFinishKind, MemoryOperationKind,
+        MemoryOperationStart, main_memory_and_io, simple_uart::SIMPLE_UART_TRANSMIT_OFFSET,
+    },
+    next_pc::FETCH_BLOCK_ID_WIDTH,
+};
+use fayalite::{prelude::*, sim::vcd::VcdWriterDecls, util::RcWriter};
+use std::num::{NonZeroU64, NonZeroUsize, Wrapping};
+
+const CLOCK_FREQUENCY: f64 = 2_000_000.0;
+
+const fn half_period(frequency: f64) -> SimDuration {
+    SimDuration::from_attos((SimDuration::from_secs(1).as_attos() as f64 / frequency / 2.0) as u128)
+}
+
+const CLOCK_HALF_PERIOD: SimDuration = half_period(CLOCK_FREQUENCY);
+
+fn clock_input_properties() -> PhantomConst<peripherals::ClockInputProperties> {
+    peripherals::ClockInput::new(CLOCK_FREQUENCY).properties
+}
+
+const SRAM_START: u64 = 0x1000;
+const SRAM_SIZE: NonZeroU64 = NonZeroU64::new(0x30).unwrap();
+const UART_START: u64 = 0x100000;
+const UART_BAUD_RATE: f64 = 115200.0;
+const UART_RECEIVER_QUEUE_SIZE: NonZeroUsize = NonZeroUsize::new(16).unwrap();
+const SRAM_INITIAL_CONTENTS: &[u8; SRAM_SIZE.get() as usize] =
+    b"This is a main memory and I/O test.             ";
+const SRAM_WRITING_CONTENTS: &[u8; SRAM_SIZE.get() as usize] =
+    b"Testing. Hello World. abcdefghijklmnopqrstuvwxyz";
+
+const CONFIG: MemoryInterfaceConfig =
+    MemoryInterfaceConfig::new(3, 8, FETCH_BLOCK_ID_WIDTH, AddressRange::Full);
+const BUS_WIDTH_IN_BYTES: usize = CONFIG.bus_width_in_bytes();
+
+fn config() -> PhantomConst<MemoryInterfaceConfig> {
+    PhantomConst::new_sized(CONFIG)
+}
+
+#[hdl_module]
+fn test_harness() {
+    let config = config();
+
+    #[hdl]
+    let cd: ClockDomain = m.input();
+    #[hdl]
+    let memory_interface: MemoryInterface<PhantomConst<MemoryInterfaceConfig>> =
+        m.input(MemoryInterface[config]);
+
+    #[hdl]
+    let dut = instance(main_memory_and_io(
+        config,
+        clock_input_properties(),
+        AddressRange::Limited {
+            start: Wrapping(SRAM_START),
+            size: SRAM_SIZE,
+        },
+        SRAM_INITIAL_CONTENTS,
+        UART_START,
+        UART_BAUD_RATE,
+        UART_RECEIVER_QUEUE_SIZE,
+    ));
+    connect(dut.cd, cd);
+    connect(dut.memory_interface, memory_interface);
+    connect(dut.uart.rx, dut.uart.tx); // loop back for testing
+}
+
+struct MainMemoryAndIoTester {
+    sim: Simulation<test_harness>,
+    next_op_id: u64,
+}
+
+impl MainMemoryAndIoTester {
+    fn reset(&mut self) {
+        let sim = &mut self.sim;
+        sim.write(
+            sim.io().memory_interface.start.data,
+            sim.io().ty().memory_interface.start.data.HdlNone(),
+        );
+        sim.write(sim.io().memory_interface.finish.ready, false);
+        sim.write_clock(sim.io().cd.clk, false);
+        sim.write_reset(sim.io().cd.rst, true);
+        sim.advance_time(CLOCK_HALF_PERIOD);
+        sim.write_clock(sim.io().cd.clk, true);
+        sim.advance_time(CLOCK_HALF_PERIOD);
+        sim.write_clock(sim.io().cd.clk, false);
+        sim.write_reset(sim.io().cd.rst, false);
+    }
+    fn wait(&mut self, cycles: u64) {
+        let sim = &mut self.sim;
+        sim.write(
+            sim.io().memory_interface.start.data,
+            sim.io().ty().memory_interface.start.data.HdlNone(),
+        );
+        sim.write(sim.io().memory_interface.finish.ready, false);
+        for _ in 0..cycles {
+            sim.advance_time(CLOCK_HALF_PERIOD);
+            sim.write_clock(sim.io().cd.clk, true);
+            sim.advance_time(CLOCK_HALF_PERIOD);
+            sim.write_clock(sim.io().cd.clk, false);
+        }
+    }
+    #[track_caller]
+    #[hdl]
+    fn try_rw_op(
+        &mut self,
+        addr: u64,
+        write: Option<[u8; BUS_WIDTH_IN_BYTES]>,
+        rw_mask: [bool; BUS_WIDTH_IN_BYTES],
+        max_wait: u64,
+    ) -> Result<[u8; BUS_WIDTH_IN_BYTES], SimValue<MemoryOperationErrorKind>> {
+        let config = config();
+        let sim = &mut self.sim;
+        let op_id = self.next_op_id.cast_to(UInt[FETCH_BLOCK_ID_WIDTH]);
+        self.next_op_id += 1;
+        sim.write(
+            sim.io().memory_interface.start.data,
+            #[hdl(sim)]
+            (sim.io().ty().memory_interface.start.data).HdlSome(match write {
+                Some(write) =>
+                {
+                    #[hdl(sim)]
+                    MemoryOperationStart::<_> {
+                        kind: #[hdl(sim)]
+                        MemoryOperationKind.Write(),
+                        addr,
+                        write_data: &write[..],
+                        rw_mask: &rw_mask[..],
+                        op_id,
+                        config,
+                    }
+                }
+                None =>
+                {
+                    #[hdl(sim)]
+                    MemoryOperationStart::<_> {
+                        kind: #[hdl(sim)]
+                        MemoryOperationKind.Read(),
+                        addr,
+                        write_data: &[0u8; BUS_WIDTH_IN_BYTES][..],
+                        rw_mask: &rw_mask[..],
+                        op_id,
+                        config,
+                    }
+                }
+            }),
+        );
+        sim.write(sim.io().memory_interface.finish.ready, true);
+        let mut cleared_start = false;
+        for _ in 0..max_wait {
+            sim.advance_time(CLOCK_HALF_PERIOD);
+            let start_ready = sim.read_bool(sim.io().memory_interface.start.ready);
+            let finish_data = sim.read(sim.io().memory_interface.finish.data);
+            sim.write_clock(sim.io().cd.clk, true);
+            sim.advance_time(CLOCK_HALF_PERIOD);
+            sim.write_clock(sim.io().cd.clk, false);
+            if start_ready && !cleared_start {
+                cleared_start = true;
+                sim.write(
+                    sim.io().memory_interface.start.data,
+                    sim.io().ty().memory_interface.start.data.HdlNone(),
+                );
+            }
+            #[hdl(sim)]
+            if let HdlSome(finish) = finish_data {
+                #[hdl(sim)]
+                let MemoryOperationFinish::<_> {
+                    kind,
+                    read_data,
+                    config: _,
+                } = finish;
+                #[hdl(sim)]
+                match kind {
+                    MemoryOperationFinishKind::Success(_) => {
+                        return Ok(std::array::from_fn(|i| read_data[i].as_int()));
+                    }
+                    MemoryOperationFinishKind::Error(e) => {
+                        return Err(e);
+                    }
+                };
+            }
+        }
+        panic!("waited too many cycles");
+    }
+    #[track_caller]
+    fn try_read<const N: usize>(
+        &mut self,
+        addr: u64,
+        max_wait: u64,
+    ) -> Result<[u8; N], SimValue<MemoryOperationErrorKind>> {
+        const { assert!(N <= BUS_WIDTH_IN_BYTES) }
+        let bus_addr = addr - addr % BUS_WIDTH_IN_BYTES as u64;
+        let mut rw_mask = [false; BUS_WIDTH_IN_BYTES];
+        let start = (addr % BUS_WIDTH_IN_BYTES as u64) as usize;
+        rw_mask[start..start + N].fill(true);
+        self.try_rw_op(bus_addr, None, rw_mask, max_wait)
+            .map(|v| std::array::from_fn(|i| v[i + start]))
+    }
+    #[track_caller]
+    fn read<const N: usize>(&mut self, addr: u64, max_wait: u64) -> [u8; N] {
+        self.try_read(addr, max_wait).expect("read returned error")
+    }
+    #[track_caller]
+    fn try_write<const N: usize>(
+        &mut self,
+        addr: u64,
+        data: [u8; N],
+        max_wait: u64,
+    ) -> Result<(), SimValue<MemoryOperationErrorKind>> {
+        const { assert!(N <= BUS_WIDTH_IN_BYTES) }
+        let bus_addr = addr - addr % BUS_WIDTH_IN_BYTES as u64;
+        let mut rw_mask = [false; BUS_WIDTH_IN_BYTES];
+        let mut write_data = [0u8; BUS_WIDTH_IN_BYTES];
+        let start = (addr % BUS_WIDTH_IN_BYTES as u64) as usize;
+        rw_mask[start..start + N].fill(true);
+        write_data[start..start + N].copy_from_slice(&data);
+        self.try_rw_op(bus_addr, Some(write_data), rw_mask, max_wait)?;
+        Ok(())
+    }
+    #[track_caller]
+    fn write<const N: usize>(&mut self, addr: u64, data: [u8; N], max_wait: u64) {
+        self.try_write(addr, data, max_wait)
+            .expect("read returned error")
+    }
+}
+
+#[test]
+#[hdl]
+fn test_main_memory_and_io() {
+    let _n = SourceLocation::normalize_files_for_tests();
+    let mut sim = Simulation::new(test_harness());
+    let writer = RcWriter::default();
+    sim.add_trace_writer(VcdWriterDecls::new(writer.clone()));
+    struct DumpVcdOnDrop {
+        writer: Option<RcWriter>,
+    }
+    impl Drop for DumpVcdOnDrop {
+        fn drop(&mut self) {
+            if let Some(mut writer) = self.writer.take() {
+                let vcd = String::from_utf8(writer.take()).unwrap();
+                println!("####### VCD:\n{vcd}\n#######");
+            }
+        }
+    }
+    let mut writer = DumpVcdOnDrop {
+        writer: Some(writer),
+    };
+    let mut tester = MainMemoryAndIoTester { sim, next_op_id: 0 };
+    tester.reset();
+
+    const UART_MESSAGE: &str = "Test.";
+    for b in UART_MESSAGE.bytes() {
+        tester.write(UART_START + SIMPLE_UART_TRANSMIT_OFFSET, [b], 200);
+    }
+
+    for (i, expected_chunk) in SRAM_INITIAL_CONTENTS
+        .as_chunks::<BUS_WIDTH_IN_BYTES>()
+        .0
+        .iter()
+        .enumerate()
+    {
+        let addr = (i as u64 * BUS_WIDTH_IN_BYTES as u64) + SRAM_START;
+        let chunk = tester.read::<BUS_WIDTH_IN_BYTES>(addr, 10);
+        assert_eq!(chunk, *expected_chunk);
+    }
+
+    for (i, chunk) in SRAM_WRITING_CONTENTS
+        .as_chunks::<BUS_WIDTH_IN_BYTES>()
+        .0
+        .iter()
+        .enumerate()
+    {
+        let addr = (i as u64 * BUS_WIDTH_IN_BYTES as u64) + SRAM_START;
+        tester.write(addr, *chunk, 10);
+    }
+
+    for (i, expected_chunk) in SRAM_WRITING_CONTENTS
+        .as_chunks::<BUS_WIDTH_IN_BYTES>()
+        .0
+        .iter()
+        .enumerate()
+    {
+        let addr = (i as u64 * BUS_WIDTH_IN_BYTES as u64) + SRAM_START;
+        let chunk = tester.read::<BUS_WIDTH_IN_BYTES>(addr, 10);
+        assert_eq!(chunk, *expected_chunk);
+    }
+
+    tester.wait(400);
+
+    for expected_byte in UART_MESSAGE.bytes() {
+        let [byte] = tester.read(UART_START + SIMPLE_UART_TRANSMIT_OFFSET, 10);
+        assert_eq!(byte, expected_byte);
+    }
+
+    let vcd = String::from_utf8(writer.writer.take().unwrap().take()).unwrap();
+    println!("####### VCD:\n{vcd}\n#######");
+    if vcd != include_str!("expected/main_memory_and_io.vcd") {
+        panic!();
+    }
+}

crates/cpu/tests/memory_interface.rs

@@ -0,0 +1,750 @@
+// SPDX-License-Identifier: LGPL-3.0-or-later
+// See Notices.txt for copyright information
+
+use cpu::{
+    main_memory_and_io::{
+        AddressRange, MemoryInterface, MemoryInterfaceConfig, MemoryOperationErrorKind,
+        MemoryOperationFinish, MemoryOperationFinishKind, MemoryOperationKind,
+        MemoryOperationStart, memory_interface_adapter_no_split,
+    },
+    next_pc::FETCH_BLOCK_ID_WIDTH,
+    util::array_vec::ArrayVec,
+};
+use fayalite::{
+    bundle::{BundleField, BundleType},
+    intern::{Intern, Interned},
+    module::instance_with_loc,
+    prelude::*,
+    sim::vcd::VcdWriterDecls,
+    util::RcWriter,
+};
+use std::{cell::Cell, collections::VecDeque, fmt};
+
+fn get_next_delay(delay_sequence_index: &Cell<u64>) -> u8 {
+    let index = delay_sequence_index.get();
+    delay_sequence_index.set(delay_sequence_index.get().wrapping_add(1));
+    // make a pseudo-random number deterministically based on index
+    let random = index
+        .wrapping_add(1)
+        .wrapping_mul(0x8c16a62518f86883) // random prime
+        .rotate_left(32)
+        .wrapping_mul(0xf807b7df2082353d) // random prime
+        .rotate_right(60);
+    const DELAYS: &[u8; 0x20] = &[
+        0, 0, 0, 0, 0, 0, 0, 0, //
+        1, 1, 1, 1, 1, 1, 1, 1, //
+        2, 2, 2, 2, 2, 2, 2, 2, //
+        3, 3, 3, 3, 4, 5, 6, 20, //
+    ];
+    DELAYS[(random & 0x1F) as usize]
+}
+
+#[hdl_module(extern)]
+fn mock_memory(memory: Memory) {
+    let Memory { config, contents } = memory;
+    #[hdl]
+    let cd: ClockDomain = m.input();
+    #[hdl]
+    let input_interface: MemoryInterface<PhantomConst<MemoryInterfaceConfig>> =
+        m.input(MemoryInterface[config]);
+    m.register_clock_for_past(cd.clk);
+    #[hdl]
+    async fn run(
+        cd: Expr<ClockDomain>,
+        input_interface: Expr<MemoryInterface<PhantomConst<MemoryInterfaceConfig>>>,
+        config: PhantomConst<MemoryInterfaceConfig>,
+        contents: Interned<str>,
+        delay_sequence_index: &Cell<u64>,
+        mut sim: ExternModuleSimulationState,
+    ) {
+        #[derive(Debug)]
+        struct Op {
+            cycles_left: u8,
+            op_id: SimValue<UInt>,
+            finish: SimValue<MemoryOperationFinish<PhantomConst<MemoryInterfaceConfig>>>,
+        }
+        let mut ops = VecDeque::<Op>::with_capacity(config.get().queue_capacity.get());
+        let finish_ty = input_interface.ty().finish.data.HdlSome;
+        loop {
+            for op in &mut ops {
+                op.cycles_left = op.cycles_left.saturating_sub(1);
+            }
+            let next_op_ids_ty = input_interface.ty().next_op_ids.HdlSome;
+            sim.write(
+                input_interface.next_op_ids,
+                #[hdl(sim)]
+                (input_interface.ty().next_op_ids).HdlSome(
+                    next_op_ids_ty
+                        .from_iter_sim(
+                            next_op_ids_ty.element().zero(),
+                            ops.iter().map(|op| &op.op_id),
+                        )
+                        .expect("known to fit"),
+                ),
+            )
+            .await;
+            if let Some(Op {
+                cycles_left: 0,
+                op_id: _,
+                finish,
+            }) = ops.front()
+            {
+                sim.write(
+                    input_interface.finish.data,
+                    #[hdl(sim)]
+                    (input_interface.ty().finish.data).HdlSome(finish),
+                )
+                .await;
+            } else {
+                sim.write(
+                    input_interface.finish.data,
+                    #[hdl(sim)]
+                    (input_interface.ty().finish.data).HdlNone(),
+                )
+                .await;
+            }
+            sim.write_bool(
+                input_interface.start.ready,
+                ops.len() < config.get().queue_capacity.get(),
+            )
+            .await;
+            sim.wait_for_clock_edge(cd.clk).await;
+            if sim
+                .read_past_bool(input_interface.finish.ready, cd.clk)
+                .await
+            {
+                dbg!(ops.pop_front_if(|op| op.cycles_left == 0));
+            }
+            if sim
+                .read_past_bool(input_interface.start.ready, cd.clk)
+                .await
+            {
+                #[hdl(sim)]
+                if let HdlSome(start) = sim.read_past(input_interface.start.data, cd.clk).await {
+                    #[hdl(sim)]
+                    let MemoryOperationStart::<_> {
+                        kind,
+                        addr,
+                        write_data,
+                        rw_mask,
+                        op_id,
+                        config: _,
+                    } = start;
+                    let mut error = false;
+                    let mut read_data = vec![0u8; finish_ty.read_data.len()];
+                    #[hdl(sim)]
+                    match &kind {
+                        MemoryOperationKind::Read => {
+                            for (i, v) in read_data.iter_mut().enumerate() {
+                                if *rw_mask[i] {
+                                    let addr = addr.as_int().wrapping_add(i as u64);
+                                    let offset =
+                                        addr.wrapping_sub(config.get().address_range.start().0);
+                                    if !config.get().address_range.contains(addr)
+                                        || offset >= contents.len() as u64
+                                    {
+                                        error = true;
+                                        break;
+                                    }
+                                    *v = contents.as_bytes()[offset as usize];
+                                    println!(
+                                        "reading byte at {addr:#x} (offset={offset:#x}) -> {v:#x} (contents={contents:?})",
+                                    );
+                                }
+                            }
+                            if !error {
+                                println!(
+                                    "read chunk at {addr:#x}: {:#x?}",
+                                    std::fmt::from_fn(|f| f
+                                        .debug_map()
+                                        .entries(
+                                            read_data
+                                                .iter()
+                                                .enumerate()
+                                                .filter_map(|(i, &v)| rw_mask[i].then(|| (i, v)))
+                                        )
+                                        .finish()),
+                                    addr = addr.as_int(),
+                                );
+                            }
+                        }
+                        MemoryOperationKind::Write => {
+                            todo!("write {write_data:?}");
+                        }
+                    }
+                    let finish_kind = if error {
+                        println!("error at: {addr} config: {config:?}");
+                        read_data.fill(0);
+                        #[hdl(sim)]
+                        MemoryOperationFinishKind.Error(
+                            #[hdl(sim)]
+                            MemoryOperationErrorKind.Generic(),
+                        )
+                    } else {
+                        #[hdl(sim)]
+                        MemoryOperationFinishKind.Success(kind)
+                    };
+                    ops.push_back(Op {
+                        cycles_left: get_next_delay(delay_sequence_index),
+                        op_id,
+                        finish: #[hdl(sim)]
+                        MemoryOperationFinish::<_> {
+                            kind: finish_kind,
+                            read_data,
+                            config,
+                        },
+                    });
+                }
+            }
+        }
+    }
+    m.extern_module_simulation_fn(
+        (cd, input_interface, config, contents),
+        async |(cd, input_interface, config, contents), mut sim| {
+            // intentionally have a different sequence each time we're reset
+            let delay_sequence_index = Cell::new(0);
+            sim.resettable(
+                cd,
+                async |mut sim| {
+                    sim.write(
+                        input_interface.next_op_ids,
+                        #[hdl(sim)]
+                        (input_interface.ty().next_op_ids).HdlNone(),
+                    )
+                    .await;
+                    sim.write_bool(input_interface.start.ready, false).await;
+                    sim.write(
+                        input_interface.finish.data,
+                        #[hdl(sim)]
+                        (input_interface.ty().finish.data).HdlNone(),
+                    )
+                    .await;
+                },
+                |sim, ()| {
+                    run(
+                        cd,
+                        input_interface,
+                        config,
+                        contents,
+                        &delay_sequence_index,
+                        sim,
+                    )
+                },
+            )
+            .await
+        },
+    );
+}
+
+#[derive(Debug, Clone, Copy, PartialEq, Eq, Hash)]
+struct Memory {
+    contents: Interned<str>,
+    config: PhantomConst<MemoryInterfaceConfig>,
+}
+
+impl Memory {
+    fn new(
+        contents: impl AsRef<str>,
+        log2_bus_width_in_bytes: u8,
+        address_range: AddressRange,
+    ) -> Self {
+        Self {
+            contents: contents.as_ref().intern(),
+            config: PhantomConst::new_sized(MemoryInterfaceConfig::new(
+                log2_bus_width_in_bytes,
+                8,
+                FETCH_BLOCK_ID_WIDTH,
+                address_range,
+            )),
+        }
+    }
+}
+
+#[hdl_module(extern)]
+fn mock_cpu(memories: Interned<[Memory]>) {
+    const LOG2_BUS_WIDTH: u8 = 3;
+    const BUS_WIDTH: usize = 1 << LOG2_BUS_WIDTH;
+    let config = PhantomConst::new_sized(MemoryInterfaceConfig::new(
+        LOG2_BUS_WIDTH,
+        8,
+        FETCH_BLOCK_ID_WIDTH,
+        AddressRange::Full,
+    ));
+    #[hdl]
+    let cd: ClockDomain = m.input();
+    #[hdl]
+    let output_interface: MemoryInterface<PhantomConst<MemoryInterfaceConfig>> =
+        m.output(MemoryInterface[config]);
+    #[hdl]
+    let finished: Bool = m.output();
+    m.register_clock_for_past(cd.clk);
+    #[derive(PartialEq, Clone, Copy)]
+    struct Op {
+        addr: u64,
+        read_mask: [bool; BUS_WIDTH],
+    }
+    impl fmt::Debug for Op {
+        fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+            let Self { addr, read_mask } = self;
+            f.debug_struct("Op")
+                .field("addr", &fmt::from_fn(|f| write!(f, "{addr:#x}")))
+                .field("read_mask", read_mask)
+                .finish()
+        }
+    }
+    #[hdl]
+    async fn generator(
+        cd: Expr<ClockDomain>,
+        output_interface: Expr<MemoryInterface<PhantomConst<MemoryInterfaceConfig>>>,
+        config: PhantomConst<MemoryInterfaceConfig>,
+        sequence: &[Op],
+        delay_sequence_index: &Cell<u64>,
+        mut sim: ExternModuleSimulationState,
+    ) {
+        println!("generator: start");
+        let start_ty = MemoryOperationStart[config];
+        for (op_index, op) in sequence.iter().enumerate() {
+            sim.write(
+                output_interface.start.data,
+                #[hdl(sim)]
+                (output_interface.ty().start.data).HdlNone(),
+            )
+            .await;
+            let delay = get_next_delay(delay_sequence_index);
+            println!("generator: delay by {delay}");
+            for i in 0..delay {
+                println!("generator: delay cycle {i}");
+                sim.wait_for_clock_edge(cd.clk).await;
+            }
+            println!("generator: generated {op:?}");
+            sim.write(
+                output_interface.start.data,
+                #[hdl(sim)]
+                (output_interface.ty().start.data).HdlSome(
+                    #[hdl(sim)]
+                    MemoryOperationStart::<_> {
+                        kind: #[hdl(sim)]
+                        MemoryOperationKind.Read(),
+                        addr: op.addr,
+                        write_data: &[0u8; BUS_WIDTH][..],
+                        rw_mask: &op.read_mask[..],
+                        op_id: op_index.cast_to(start_ty.op_id),
+                        config,
+                    },
+                ),
+            )
+            .await;
+            sim.wait_for_clock_edge(cd.clk).await;
+            while !sim
+                .read_past_bool(output_interface.start.ready, cd.clk)
+                .await
+            {
+                println!("generator: start.ready was false");
+                sim.wait_for_clock_edge(cd.clk).await;
+            }
+        }
+        sim.write(
+            output_interface.start.data,
+            #[hdl(sim)]
+            (output_interface.ty().start.data).HdlNone(),
+        )
+        .await;
+    }
+    #[hdl]
+    async fn checker(
+        cd: Expr<ClockDomain>,
+        output_interface: Expr<MemoryInterface<PhantomConst<MemoryInterfaceConfig>>>,
+        config: PhantomConst<MemoryInterfaceConfig>,
+        sequence: &[Op],
+        memories: Interned<[Memory]>,
+        delay_sequence_index: &Cell<u64>,
+        sim: &mut ExternModuleSimulationState,
+    ) {
+        println!("checker: start");
+        for op in sequence {
+            sim.write_bool(output_interface.finish.ready, false).await;
+            let delay = get_next_delay(delay_sequence_index);
+            println!("checker: delay by {delay}");
+            for i in 0..delay {
+                println!("checker: delay cycle {i}");
+                sim.wait_for_clock_edge(cd.clk).await;
+            }
+            sim.write_bool(output_interface.finish.ready, true).await;
+            sim.wait_for_clock_edge(cd.clk).await;
+            let mut finish = loop {
+                #[hdl(sim)]
+                if let HdlSome(finish) = sim.read_past(output_interface.finish.data, cd.clk).await {
+                    break finish;
+                }
+                println!("checker: finish.data was HdlNone");
+                sim.wait_for_clock_edge(cd.clk).await;
+            };
+            println!("checker: checking for {op:?}");
+            let finish_unmasked = finish.clone();
+            for (v, &mask) in finish.read_data.iter_mut().zip(&op.read_mask) {
+                if !mask {
+                    *v = 0u8.to_sim_value(); // ignore outputs for ignored bytes
+                }
+            }
+            let mut expected_finish = memories
+                .iter()
+                .find(|m| dbg!(dbg!(m.config.get().address_range).contains(op.addr)))
+                .and_then(
+                    |&Memory {
+                         config: memory_config,
+                         contents,
+                     }|
+                     -> Option<_> {
+                        let mut read_data = [0u8; BUS_WIDTH];
+                        let mut first_enabled = None;
+                        let mut last_enabled = None;
+                        for (i, &mask) in op.read_mask.iter().enumerate() {
+                            if mask {
+                                first_enabled.get_or_insert(i);
+                                last_enabled = Some(i);
+                                read_data[i] = *dbg!(
+                                    dbg!(contents).as_bytes().get(
+                                        dbg!(usize::try_from(
+                                            op.addr.wrapping_add(i as u64).wrapping_sub(
+                                                memory_config.get().address_range.start().0,
+                                            ),
+                                        ))
+                                        .ok()?,
+                                    )
+                                )?;
+                            }
+                        }
+                        if let (Some(first_enabled), Some(last_enabled)) =
+                            (first_enabled, last_enabled)
+                        {
+                            let log2_bus_width_in_bytes =
+                                memory_config.get().log2_bus_width_in_bytes;
+                            if first_enabled >> log2_bus_width_in_bytes
+                                != last_enabled >> log2_bus_width_in_bytes
+                            {
+                                // this operation requires more than one operation at the final memory,
+                                // so it gets turned into an error since we're using
+                                // memory_interface_adapter_no_split
+                                return None;
+                            }
+                        }
+                        Some(
+                            #[hdl(sim)]
+                            MemoryOperationFinish::<_> {
+                                kind: #[hdl(sim)]
+                                MemoryOperationFinishKind.Success(
+                                    #[hdl(sim)]
+                                    MemoryOperationKind.Read(),
+                                ),
+                                read_data: &read_data[..],
+                                config,
+                            },
+                        )
+                    },
+                )
+                .unwrap_or_else(|| {
+                    #[hdl(sim)]
+                    MemoryOperationFinish::<_> {
+                        kind: #[hdl(sim)]
+                        MemoryOperationFinishKind.Error(
+                            #[hdl(sim)]
+                            MemoryOperationErrorKind.Generic(),
+                        ),
+                        read_data: &[0u8; BUS_WIDTH][..],
+                        config,
+                    }
+                });
+            // make SimValue fill in the enum padding so they format the same
+            SimValue::bits_mut(&mut finish);
+            SimValue::bits_mut(&mut expected_finish);
+            assert!(
+                format!("{finish:#?}") == format!("{expected_finish:#?}"),
+                "op={op:#?}\nexpected_finish={expected_finish:#?}\n\
+                finish={finish:#?}\nfinish_unmasked={finish_unmasked:#?}"
+            );
+        }
+    }
+    m.extern_module_simulation_fn(
+        (cd, output_interface, finished, config, memories),
+        async |(cd, output_interface, finished, config, memories), mut sim| {
+            sim.write_bool(finished, false).await;
+            sim.write_bool(output_interface.finish.ready, false).await;
+            sim.write(
+                output_interface.start.data,
+                #[hdl(sim)]
+                (output_interface.ty().start.data).HdlNone(),
+            )
+            .await;
+
+            // intentionally have a different sequence each time we're reset
+            let generator_delay_sequence_index = Cell::new(1 << 63);
+            let checker_delay_sequence_index = Cell::new(1 << 62);
+            let mut sequence = Vec::new();
+            sequence.push(Op {
+                addr: !0 << 16,
+                read_mask: [true; _],
+            });
+            for (i, memory) in memories.iter().enumerate() {
+                assert!(
+                    memory
+                        .config
+                        .get()
+                        .address_range
+                        .start()
+                        .0
+                        .is_multiple_of(BUS_WIDTH as u64)
+                );
+                assert!(
+                    memory
+                        .config
+                        .get()
+                        .address_range
+                        .last()
+                        .wrapping_add(1)
+                        .is_multiple_of(BUS_WIDTH as u64)
+                );
+                if i == 0 {
+                    for log2_read_size in 0..=LOG2_BUS_WIDTH {
+                        let read_size = 1 << log2_read_size;
+                        for offset in (0..BUS_WIDTH).step_by(read_size) {
+                            sequence.push(Op {
+                                addr: memory.config.get().address_range.start().0,
+                                read_mask: std::array::from_fn(|byte_index| {
+                                    byte_index
+                                        .checked_sub(offset)
+                                        .is_some_and(|v| v < read_size)
+                                }),
+                            });
+                        }
+                    }
+                }
+                for (addr, chunk) in (memory.config.get().address_range.start().0..)
+                    .step_by(BUS_WIDTH)
+                    .zip(memory.contents.as_bytes().chunks(BUS_WIDTH))
+                {
+                    let size = memory.config.get().bus_width_in_bytes();
+                    for offset in (0..BUS_WIDTH).step_by(size) {
+                        let mut op = Op {
+                            addr,
+                            read_mask: std::array::from_fn(|i| i >= offset && i < offset + size),
+                        };
+                        op.read_mask[chunk.len()..].fill(false);
+                        if op.read_mask.contains(&true) && sequence.last() != Some(&op) {
+                            sequence.push(op);
+                        }
+                    }
+                }
+            }
+            sequence.extend_from_within(..);
+            sequence.extend_from_within(..);
+            sim.fork_join_scope(async |scope, mut sim| {
+                scope.spawn_detached(async |_scope, mut sim| {
+                    sim.resettable(
+                        cd,
+                        async |mut sim| {
+                            sim.write(
+                                output_interface.start.data,
+                                #[hdl(sim)]
+                                (output_interface.ty().start.data).HdlNone(),
+                            )
+                            .await;
+                        },
+                        |sim, ()| {
+                            generator(
+                                cd,
+                                output_interface,
+                                config,
+                                &sequence,
+                                &generator_delay_sequence_index,
+                                sim,
+                            )
+                        },
+                    )
+                    .await
+                });
+                scope.spawn_detached(async |_scope, mut sim| {
+                    sim.resettable(
+                        cd,
+                        async |_| {},
+                        async |mut sim, ()| {
+                            let mut expected_range = 0u32..0u32;
+                            loop {
+                                sim.wait_for_clock_edge(cd.clk).await;
+                                let next_op_ids =
+                                    sim.read_past(output_interface.next_op_ids, cd.clk).await;
+                                #[hdl(sim)]
+                                if let HdlSome(next_op_ids) = next_op_ids {
+                                    let op_id_ty = next_op_ids.ty().element();
+                                    let capacity = next_op_ids.ty().capacity();
+                                    let next_op_ids = ArrayVec::elements_sim_ref(&next_op_ids);
+                                    let expected_next_op_ids = Vec::from_iter(
+                                        expected_range
+                                            .clone()
+                                            .map(|i| i.cast_to(op_id_ty).into_sim_value()),
+                                    );
+                                    assert_eq!(
+                                        next_op_ids,
+                                        expected_next_op_ids,
+                                        "{:#x}..{:#x} ({expected_range:?}){}",
+                                        expected_range.start,
+                                        expected_range.end,
+                                        if expected_range.len() > capacity {
+                                            "\nexpected_range is bigger than capacity"
+                                        } else {
+                                            ""
+                                        }
+                                    );
+                                }
+                                #[hdl(sim)]
+                                if let HdlSome(_) =
+                                    sim.read_past(output_interface.start.data, cd.clk).await
+                                {
+                                    if sim
+                                        .read_past_bool(output_interface.start.ready, cd.clk)
+                                        .await
+                                    {
+                                        expected_range.end += 1;
+                                    }
+                                }
+                                #[hdl(sim)]
+                                if let HdlSome(_) =
+                                    sim.read_past(output_interface.finish.data, cd.clk).await
+                                {
+                                    if sim
+                                        .read_past_bool(output_interface.finish.ready, cd.clk)
+                                        .await
+                                    {
+                                        expected_range.start += 1;
+                                    }
+                                }
+                            }
+                        },
+                    )
+                    .await;
+                });
+                sim.resettable(
+                    cd,
+                    async |mut sim| {
+                        sim.write_bool(finished, false).await;
+                        sim.write_bool(output_interface.finish.ready, false).await;
+                    },
+                    async |mut sim, ()| {
+                        checker(
+                            cd,
+                            output_interface,
+                            config,
+                            &sequence,
+                            memories,
+                            &checker_delay_sequence_index,
+                            &mut sim,
+                        )
+                        .await;
+                        sim.write_bool(finished, true).await;
+                        loop {
+                            sim.write_bool(output_interface.finish.ready, true).await;
+                            sim.wait_for_clock_edge(cd.clk).await;
+                            #[hdl(sim)]
+                            if let HdlSome(finish) =
+                                sim.read_past(output_interface.finish.data, cd.clk).await
+                            {
+                                panic!("spurious finished transaction: {finish:#?}");
+                            }
+                        }
+                    },
+                )
+                .await
+            })
+            .await;
+        },
+    );
+}
+
+#[hdl_module]
+fn memory_interface_adapter_no_split_dut(memories: Interned<[Memory]>) {
+    #[hdl]
+    let cd: ClockDomain = m.input();
+    #[hdl]
+    let finished: Bool = m.output();
+    #[hdl]
+    let mock_cpu = instance(mock_cpu(memories));
+    connect(mock_cpu.cd, cd);
+    connect(finished, mock_cpu.finished);
+    let (fields, inputs): (Vec<_>, Vec<_>) = memories
+        .iter()
+        .enumerate()
+        .map(|(index, &memory)| {
+            let mock_mem = instance_with_loc(
+                &format!("mock_mem_{index}"),
+                mock_memory(memory),
+                SourceLocation::caller(),
+            );
+            connect(mock_mem.cd, cd);
+            (
+                BundleField {
+                    name: format!("{index}").intern_deref(),
+                    flipped: false,
+                    ty: MemoryInterface[memory.config].canonical(),
+                },
+                mock_mem.input_interface,
+            )
+        })
+        .unzip();
+    let bundle_ty = Bundle::new(fields.intern_deref());
+    #[hdl]
+    let adapter = instance(memory_interface_adapter_no_split(
+        mock_cpu.ty().output_interface.config,
+        bundle_ty,
+    ));
+    connect(adapter.cd, cd);
+    connect(adapter.input_interface, mock_cpu.output_interface);
+    for (field, input) in bundle_ty.fields().into_iter().zip(inputs) {
+        connect(input, Expr::field(adapter.output_interfaces, &field.name));
+    }
+}
+
+#[test]
+#[hdl]
+fn test_memory_interface_adapter_no_split() {
+    let _n = SourceLocation::normalize_files_for_tests();
+    let memories = vec![
+        Memory::new("Testing", 3, AddressRange::from_range(0x1000..0x2000)),
+        Memory::new("Memory2.", 2, AddressRange::from_range(0x2000..0x2010)),
+        Memory::new("Contents Test", 0, AddressRange::from_range(0x3000..0x3100)),
+    ]
+    .intern_deref();
+    let m = memory_interface_adapter_no_split_dut(memories);
+    let mut sim = Simulation::new(m);
+    let writer = RcWriter::default();
+    sim.add_trace_writer(VcdWriterDecls::new(writer.clone()));
+    struct DumpVcdOnDrop {
+        writer: Option<RcWriter>,
+    }
+    impl Drop for DumpVcdOnDrop {
+        fn drop(&mut self) {
+            if let Some(mut writer) = self.writer.take() {
+                let vcd = String::from_utf8(writer.take()).unwrap();
+                println!("####### VCD:\n{vcd}\n#######");
+            }
+        }
+    }
+    let mut writer = DumpVcdOnDrop {
+        writer: Some(writer),
+    };
+    sim.write_clock(sim.io().cd.clk, false);
+    sim.write_reset(sim.io().cd.rst, true);
+    for cycle in 0..500 {
+        println!("cycle: {cycle}");
+        sim.advance_time(SimDuration::from_nanos(500));
+        sim.write_clock(sim.io().cd.clk, true);
+        sim.advance_time(SimDuration::from_nanos(500));
+        sim.write_clock(sim.io().cd.clk, false);
+        sim.write_reset(sim.io().cd.rst, false);
+    }
+    sim.advance_time(SimDuration::from_nanos(500));
+    assert!(sim.read_bool(sim.io().finished));
+    let vcd = String::from_utf8(writer.writer.take().unwrap().take()).unwrap();
+    println!("####### VCD:\n{vcd}\n#######");
+    if vcd != include_str!("expected/memory_interface_adapter_no_split.vcd") {
+        panic!();
+    }
+}

crates/cpu/tests/simple_uart.rs

@@ -0,0 +1,980 @@
+// SPDX-License-Identifier: LGPL-3.0-or-later
+// See Notices.txt for copyright information
+
+use cpu::{
+    main_memory_and_io::{
+        MemoryInterface, MemoryInterfaceConfig, MemoryOperationErrorKind, MemoryOperationFinish,
+        MemoryOperationFinishKind, MemoryOperationKind, MemoryOperationStart,
+        simple_uart::{
+            ReceiverQueueStatus, SIMPLE_UART_RECEIVE_OFFSET, SIMPLE_UART_STATUS_OFFSET,
+            SIMPLE_UART_TRANSMIT_OFFSET, SIMPLE_UART_USED_SIZE, receiver, receiver_no_queue,
+            simple_uart, simple_uart_memory_interface_config, transmitter, uart_clock_gen,
+        },
+    },
+    next_pc::FETCH_BLOCK_ID_WIDTH,
+    util::array_vec::ArrayVec,
+};
+use fayalite::{
+    bundle::BundleType,
+    intern::{Intern, Interned},
+    prelude::*,
+    sim::vcd::VcdWriterDecls,
+    util::{RcWriter, ready_valid::ReadyValid},
+};
+use std::{
+    borrow::BorrowMut,
+    marker::PhantomData,
+    num::{NonZeroUsize, Wrapping},
+};
+
+const fn half_period(frequency: f64) -> SimDuration {
+    SimDuration::from_attos((SimDuration::from_secs(1).as_attos() as f64 / frequency / 2.0) as u128)
+}
+
+const CLOCK_FREQUENCY: f64 = 200_000.0;
+const CLOCK_HALF_PERIOD: SimDuration = half_period(CLOCK_FREQUENCY);
+const BAUD_RATE: f64 = 9600.0;
+const BAUD_HALF_PERIOD: SimDuration = half_period(BAUD_RATE);
+const BAUD_PERIOD: SimDuration = BAUD_HALF_PERIOD.saturating_add(BAUD_HALF_PERIOD);
+
+const fn baud_period_slow_percentage(percent: i128) -> SimDuration {
+    let v = BAUD_PERIOD.as_attos();
+    let v = v.strict_add_signed(v as i128 * percent / 100);
+    SimDuration::from_attos(v)
+}
+
+/// 3% slow BAUD_PERIOD
+const BAUD_PERIOD_SLOW: SimDuration = baud_period_slow_percentage(3);
+/// 3% fast BAUD_PERIOD
+const BAUD_PERIOD_FAST: SimDuration = baud_period_slow_percentage(-3);
+
+const _: () = {
+    assert!(BAUD_HALF_PERIOD.as_attos() < BAUD_PERIOD_FAST.as_attos());
+    assert!(BAUD_PERIOD_FAST.as_attos() < BAUD_PERIOD.as_attos());
+    assert!(BAUD_PERIOD.as_attos() < BAUD_PERIOD_SLOW.as_attos());
+};
+
+fn clock_input_properties() -> PhantomConst<peripherals::ClockInputProperties> {
+    peripherals::ClockInput::new(CLOCK_FREQUENCY).properties
+}
+
+#[hdl_module(extern)]
+fn reference_baud_rate_clk_gen() {
+    #[hdl]
+    let reference_baud_rate_clk: Clock = m.output();
+    m.extern_module_simulation_fn(
+        reference_baud_rate_clk,
+        async |reference_baud_rate_clk, mut sim| {
+            loop {
+                sim.write_clock(reference_baud_rate_clk, false).await;
+                sim.advance_time(BAUD_HALF_PERIOD).await;
+                sim.write_clock(reference_baud_rate_clk, true).await;
+                sim.advance_time(BAUD_HALF_PERIOD).await;
+            }
+        },
+    );
+}
+
+#[hdl_module(extern)]
+fn sim_receiver() {
+    #[hdl]
+    let tx: Bool = m.input();
+    #[hdl]
+    let text_out: SimOnly<String> = m.output();
+    m.extern_module_simulation_fn((tx, text_out), async |(tx, text_out), mut sim| {
+        let mut text = SimOnlyValue::new(String::new());
+        sim.write(text_out, &text).await;
+        loop {
+            // wait for the starting edge of the start bit
+            while sim.read_bool(tx).await {
+                sim.wait_for_changes([tx], None).await;
+            }
+            // wait till the middle of the start bit
+            sim.advance_time(BAUD_HALF_PERIOD).await;
+            let mut byte = 0u8;
+            for i in 0..u8::BITS {
+                // wait till the middle of the next data bit
+                sim.advance_time(BAUD_PERIOD).await;
+                if sim.read_bool(tx).await {
+                    byte |= 1 << i;
+                }
+            }
+            assert!(byte.is_ascii());
+            text.push(byte as char);
+            sim.write(text_out, &text).await;
+            // wait till the middle of the stop bit
+            sim.advance_time(BAUD_PERIOD).await;
+            // we're in the middle of the stop bit, the stop bit is a one so we can just loop and wait for
+            // the next transition to zero which is the next start bit.
+        }
+    });
+}
+
+#[hdl_module]
+fn transmitter_and_uart_clock_gen() {
+    #[hdl]
+    let cd: ClockDomain = m.input();
+    #[hdl]
+    let tx: Bool = m.output();
+    #[hdl]
+    let input_byte: ReadyValid<UInt<8>> = m.input();
+    #[hdl]
+    let reference_baud_rate_clk: Clock = m.output();
+    #[hdl]
+    let text_out: SimOnly<String> = m.output();
+
+    #[hdl]
+    let sim_receiver = instance(sim_receiver());
+    connect(sim_receiver.tx, tx);
+    connect(text_out, sim_receiver.text_out);
+
+    #[hdl]
+    let reference_baud_rate_clk_gen = instance(reference_baud_rate_clk_gen());
+    connect(
+        reference_baud_rate_clk,
+        reference_baud_rate_clk_gen.reference_baud_rate_clk,
+    );
+
+    #[hdl]
+    let transmitter = instance(transmitter());
+    connect(transmitter.cd, cd);
+    connect(transmitter.input_byte, input_byte);
+    connect(tx, transmitter.tx);
+    #[hdl]
+    let clock_gen = instance(uart_clock_gen(clock_input_properties(), BAUD_RATE));
+    connect(clock_gen.cd, cd);
+    connect(transmitter.tick, clock_gen.tick);
+}
+
+#[test]
+#[hdl]
+fn test_transmitter_and_uart_clock_gen() {
+    let _n = SourceLocation::normalize_files_for_tests();
+    let m = transmitter_and_uart_clock_gen();
+    let mut sim = Simulation::new(m);
+    let writer = RcWriter::default();
+    sim.add_trace_writer(VcdWriterDecls::new(writer.clone()));
+    struct DumpVcdOnDrop {
+        writer: Option<RcWriter>,
+    }
+    impl Drop for DumpVcdOnDrop {
+        fn drop(&mut self) {
+            if let Some(mut writer) = self.writer.take() {
+                let vcd = String::from_utf8(writer.take()).unwrap();
+                println!("####### VCD:\n{vcd}\n#######");
+            }
+        }
+    }
+    let mut writer = DumpVcdOnDrop {
+        writer: Some(writer),
+    };
+    sim.write_clock(sim.io().cd.clk, false);
+    sim.write_reset(sim.io().cd.rst, true);
+    let text = "Hi!\n";
+    let mut bytes = text.as_bytes().iter();
+    for cycle in 0..1000 {
+        let input_byte = match bytes.clone().next() {
+            Some(b) if cycle > 10 =>
+            {
+                #[hdl(sim)]
+                HdlSome(b)
+            }
+            _ =>
+            {
+                #[hdl(sim)]
+                HdlNone()
+            }
+        };
+        sim.write(sim.io().input_byte.data, &input_byte);
+        sim.advance_time(CLOCK_HALF_PERIOD);
+        sim.write_clock(sim.io().cd.clk, true);
+        sim.advance_time(CLOCK_HALF_PERIOD);
+        if sim.read_bool(sim.io().input_byte.ready) {
+            #[hdl(sim)]
+            if let HdlSome(_) = input_byte {
+                bytes.next();
+            }
+        }
+        sim.write_clock(sim.io().cd.clk, false);
+        sim.write_reset(sim.io().cd.rst, false);
+    }
+    assert_eq!(sim.read(sim.io().text_out).as_str(), text);
+    assert!(bytes.as_slice().is_empty());
+    let vcd = String::from_utf8(writer.writer.take().unwrap().take()).unwrap();
+    println!("####### VCD:\n{vcd}\n#######");
+    if vcd != include_str!("expected/transmitter_and_uart_clock_gen.vcd") {
+        panic!();
+    }
+}
+
+#[hdl_module(extern)]
+fn receiver_no_queue_test_cases(text: Interned<str>) {
+    #[hdl]
+    let rx: Bool = m.output();
+    #[hdl]
+    let cur_byte: UInt<8> = m.output();
+    m.extern_module_simulation_fn(
+        (rx, cur_byte, text),
+        async |(rx, cur_byte, text), mut sim| {
+            // ensure the receiver can properly handle slightly fast or slow baud rates
+            let baud_periods = [
+                BAUD_PERIOD_FAST,
+                BAUD_PERIOD_FAST,
+                BAUD_PERIOD_FAST,
+                BAUD_PERIOD_FAST,
+                BAUD_PERIOD,
+                BAUD_PERIOD,
+                BAUD_PERIOD,
+                BAUD_PERIOD,
+                BAUD_PERIOD_SLOW,
+                BAUD_PERIOD_SLOW,
+                BAUD_PERIOD_SLOW,
+                BAUD_PERIOD_SLOW,
+            ]
+            .iter()
+            .copied()
+            .cycle();
+            sim.write(rx, true).await;
+            sim.write(cur_byte, 0u8).await;
+            // allow time for reset and stuff
+            sim.advance_time(SimDuration::from_micros(100)).await;
+            for (byte, baud_period) in text.bytes().cycle().zip(baud_periods) {
+                sim.write(cur_byte, byte).await;
+                sim.write(rx, false).await; // start bit
+                sim.advance_time(baud_period).await;
+                for i in 0..u8::BITS {
+                    sim.write(rx, byte & (1 << i) != 0).await; // data bit
+                    sim.advance_time(baud_period).await;
+                }
+                sim.write(rx, true).await; // stop bit
+                sim.advance_time(baud_period).await;
+            }
+        },
+    );
+}
+
+#[hdl_module]
+fn receiver_no_queue_and_uart_clock_gen(text: Interned<str>) {
+    #[hdl]
+    let cd: ClockDomain = m.input();
+    #[hdl]
+    let rx: Bool = m.output();
+    #[hdl]
+    let output_byte: HdlOption<UInt<8>> = m.output();
+    #[hdl]
+    let cur_byte: UInt<8> = m.output();
+    #[hdl]
+    let reference_baud_rate_clk: Clock = m.output();
+
+    #[hdl]
+    let test_cases = instance(receiver_no_queue_test_cases(text));
+    connect(rx, test_cases.rx);
+    connect(cur_byte, test_cases.cur_byte);
+
+    #[hdl]
+    let reference_baud_rate_clk_gen = instance(reference_baud_rate_clk_gen());
+    connect(
+        reference_baud_rate_clk,
+        reference_baud_rate_clk_gen.reference_baud_rate_clk,
+    );
+
+    // only need one stage of synchronization in the simulator
+    #[hdl]
+    let rx_synchronized = reg_builder().clock_domain(cd).reset(true);
+
+    connect(rx_synchronized, rx);
+
+    #[hdl]
+    let receiver = instance(receiver_no_queue());
+    connect(receiver.cd, cd);
+    connect(receiver.rx_synchronized, rx_synchronized);
+    connect(output_byte, receiver.output_byte);
+    #[hdl]
+    let clock_gen = instance(uart_clock_gen(clock_input_properties(), BAUD_RATE));
+    connect(clock_gen.cd, cd);
+    connect(receiver.tick, clock_gen.tick);
+}
+
+#[test]
+#[hdl]
+fn test_receiver_no_queue_and_uart_clock_gen() {
+    let _n = SourceLocation::normalize_files_for_tests();
+    let text = "Testing 123, testing, testing.\n".intern();
+    let m = receiver_no_queue_and_uart_clock_gen(text);
+    let mut sim = Simulation::new(m);
+    let writer = RcWriter::default();
+    sim.add_trace_writer(VcdWriterDecls::new(writer.clone()));
+    struct DumpVcdOnDrop {
+        writer: Option<RcWriter>,
+    }
+    impl Drop for DumpVcdOnDrop {
+        fn drop(&mut self) {
+            if let Some(mut writer) = self.writer.take() {
+                let vcd = String::from_utf8(writer.take()).unwrap();
+                println!("####### VCD:\n{vcd}\n#######");
+            }
+        }
+    }
+    let mut writer = DumpVcdOnDrop {
+        writer: Some(writer),
+    };
+    sim.write_clock(sim.io().cd.clk, false);
+    sim.write_reset(sim.io().cd.rst, true);
+    let mut received_text = String::new();
+    for _cycle in 0..10000 {
+        sim.advance_time(CLOCK_HALF_PERIOD);
+        sim.write_clock(sim.io().cd.clk, true);
+        sim.advance_time(CLOCK_HALF_PERIOD);
+        sim.write_clock(sim.io().cd.clk, false);
+        sim.write_reset(sim.io().cd.rst, false);
+        #[hdl(sim)]
+        if let HdlSome(byte) = sim.read(sim.io().output_byte) {
+            let byte = byte.as_int();
+            let expected_byte = sim.read(sim.io().cur_byte).as_int();
+            assert_eq!(
+                byte, expected_byte,
+                "byte={:#x} {:?}, expected_byte={:#x} {:?}",
+                byte, byte as char, expected_byte, expected_byte as char,
+            );
+            assert!(byte.is_ascii());
+            received_text.push(byte as char);
+        }
+    }
+    println!("{received_text:?}");
+    assert!(received_text.len() >= text.len());
+    let mut expected_text = text.repeat(received_text.len().div_ceil(text.len()));
+    expected_text.truncate(received_text.len());
+    assert_eq!(received_text, expected_text);
+    let vcd = String::from_utf8(writer.writer.take().unwrap().take()).unwrap();
+    println!("####### VCD:\n{vcd}\n#######");
+    if vcd != include_str!("expected/receiver_no_queue_and_uart_clock_gen.vcd") {
+        panic!();
+    }
+}
+
+#[hdl_module(extern)]
+fn receiver_test_cases() {
+    #[hdl]
+    let clk: Clock = m.input();
+    #[hdl]
+    let rx: Bool = m.output();
+    #[hdl]
+    let cur_byte: UInt<8> = m.output();
+    #[hdl]
+    let ready: Bool = m.output();
+    #[hdl]
+    let text: SimOnly<String> = m.input();
+    m.register_clock_for_past(clk);
+    m.extern_module_simulation_fn(
+        (clk, rx, cur_byte, ready, text),
+        async |(clk, rx, cur_byte, ready, text), mut sim| {
+            sim.write(rx, true).await;
+            sim.write(ready, false).await;
+            loop {
+                sim.write(cur_byte, 0u8).await;
+                sim.wait_for_clock_edge(clk).await;
+                sim.write(ready, true).await;
+                let text = loop {
+                    sim.wait_for_clock_edge(clk).await;
+                    let text = sim.read_past(text, clk).await;
+                    if text.is_empty() {
+                        continue;
+                    }
+                    break text;
+                };
+                sim.write(ready, false).await;
+                sim.wait_for_clock_edge(clk).await;
+                for byte in text.bytes() {
+                    sim.write(cur_byte, byte).await;
+                    sim.write(rx, false).await; // start bit
+                    sim.advance_time(BAUD_PERIOD).await;
+                    for i in 0..u8::BITS {
+                        sim.write(rx, byte & (1 << i) != 0).await; // data bit
+                        sim.advance_time(BAUD_PERIOD).await;
+                    }
+                    sim.write(rx, true).await; // stop bit
+                    sim.advance_time(BAUD_PERIOD).await;
+                }
+            }
+        },
+    );
+}
+
+#[hdl_module]
+fn receiver_and_uart_clock_gen(queue_capacity: NonZeroUsize) {
+    #[hdl]
+    let cd: ClockDomain = m.input();
+    #[hdl]
+    let rx: Bool = m.output();
+    #[hdl]
+    let reference_baud_rate_clk: Clock = m.output();
+    #[hdl]
+    let cur_byte: UInt<8> = m.output();
+    #[hdl]
+    let ready: Bool = m.output();
+    #[hdl]
+    let text: SimOnly<String> = m.input();
+    #[hdl]
+    let output_byte: ReadyValid<UInt<8>> = m.output();
+    #[hdl]
+    let queue_status: ReceiverQueueStatus = m.output();
+
+    #[hdl]
+    let test_cases = instance(receiver_test_cases());
+    connect(test_cases.clk, cd.clk);
+    connect(rx, test_cases.rx);
+    connect(cur_byte, test_cases.cur_byte);
+    connect(ready, test_cases.ready);
+    connect(test_cases.text, text);
+
+    #[hdl]
+    let reference_baud_rate_clk_gen = instance(reference_baud_rate_clk_gen());
+    connect(
+        reference_baud_rate_clk,
+        reference_baud_rate_clk_gen.reference_baud_rate_clk,
+    );
+
+    // only need one stage of synchronization in the simulator
+    #[hdl]
+    let rx_synchronized = reg_builder().clock_domain(cd).reset(true);
+
+    connect(rx_synchronized, rx);
+
+    #[hdl]
+    let receiver = instance(receiver(queue_capacity));
+    connect(receiver.cd, cd);
+    connect(receiver.rx_synchronized, rx_synchronized);
+    connect(output_byte, receiver.output_byte);
+    connect(queue_status, receiver.queue_status);
+    #[hdl]
+    let clock_gen = instance(uart_clock_gen(clock_input_properties(), BAUD_RATE));
+    connect(clock_gen.cd, cd);
+    connect(receiver.tick, clock_gen.tick);
+}
+
+#[test]
+#[hdl]
+fn test_receiver_and_uart_clock_gen() {
+    let _n = SourceLocation::normalize_files_for_tests();
+    const QUEUE_SIZE: NonZeroUsize = NonZeroUsize::new(4).expect("not zero");
+    let m = receiver_and_uart_clock_gen(QUEUE_SIZE);
+    let mut sim = Simulation::new(m);
+    let writer = RcWriter::default();
+    sim.add_trace_writer(VcdWriterDecls::new(writer.clone()));
+    struct DumpVcdOnDrop {
+        writer: Option<RcWriter>,
+    }
+    impl Drop for DumpVcdOnDrop {
+        fn drop(&mut self) {
+            if let Some(mut writer) = self.writer.take() {
+                let vcd = String::from_utf8(writer.take()).unwrap();
+                println!("####### VCD:\n{vcd}\n#######");
+            }
+        }
+    }
+    let mut writer = DumpVcdOnDrop {
+        writer: Some(writer),
+    };
+    sim.write_clock(sim.io().cd.clk, false);
+    sim.write_reset(sim.io().cd.rst, true);
+    let empty_text = SimOnlyValue::new(String::new());
+    sim.write(sim.io().text, &empty_text);
+    sim.write_bool(sim.io().output_byte.ready, false);
+    sim.advance_time(CLOCK_HALF_PERIOD);
+    sim.write_clock(sim.io().cd.clk, true);
+    sim.advance_time(CLOCK_HALF_PERIOD);
+    sim.write_clock(sim.io().cd.clk, false);
+    sim.write_reset(sim.io().cd.rst, false);
+    const TEST_TEXT: &str = "Test\n";
+    const { assert!(TEST_TEXT.len() > QUEUE_SIZE.get()) }
+    for bytes_before_dequeue in 0..=TEST_TEXT.len() {
+        println!("\nbytes_before_dequeue={bytes_before_dequeue}");
+        for _cycle in 0..10 {
+            if sim.read_bool(sim.io().ready) {
+                break;
+            }
+            sim.advance_time(CLOCK_HALF_PERIOD);
+            sim.write_clock(sim.io().cd.clk, true);
+            sim.advance_time(CLOCK_HALF_PERIOD);
+            sim.write_clock(sim.io().cd.clk, false);
+        }
+        assert!(sim.read_bool(sim.io().ready));
+        let text = SimOnlyValue::new(TEST_TEXT[..bytes_before_dequeue].to_string());
+        dbg!(&text);
+        sim.write(sim.io().text, &text);
+        sim.advance_time(CLOCK_HALF_PERIOD);
+        sim.write_clock(sim.io().cd.clk, true);
+        sim.advance_time(CLOCK_HALF_PERIOD);
+        sim.write_clock(sim.io().cd.clk, false);
+        sim.write(sim.io().text, &empty_text);
+        for _cycle in 0..2000 {
+            if sim.read_bool(sim.io().ready) {
+                break;
+            }
+            sim.advance_time(CLOCK_HALF_PERIOD);
+            sim.write_clock(sim.io().cd.clk, true);
+            sim.advance_time(CLOCK_HALF_PERIOD);
+            sim.write_clock(sim.io().cd.clk, false);
+        }
+        assert!(sim.read_bool(sim.io().ready));
+        for _cycle in 0..10 {
+            sim.advance_time(CLOCK_HALF_PERIOD);
+            sim.write_clock(sim.io().cd.clk, true);
+            sim.advance_time(CLOCK_HALF_PERIOD);
+            sim.write_clock(sim.io().cd.clk, false);
+        }
+        dbg!(bytes_before_dequeue);
+        #[hdl(sim)]
+        match dbg!(sim.read(sim.io().queue_status)) {
+            ReceiverQueueStatus::QueueEmpty => {
+                assert_eq!(bytes_before_dequeue, 0);
+            }
+            ReceiverQueueStatus::QueueNotEmpty => {
+                assert!(bytes_before_dequeue > 0 && bytes_before_dequeue < QUEUE_SIZE.get() - 1);
+            }
+            ReceiverQueueStatus::QueueAlmostFull => {
+                assert_eq!(bytes_before_dequeue, QUEUE_SIZE.get() - 1);
+            }
+            ReceiverQueueStatus::QueueFull => {
+                assert_eq!(bytes_before_dequeue, QUEUE_SIZE.get());
+            }
+            ReceiverQueueStatus::QueueOverflowed => {
+                assert!(bytes_before_dequeue > QUEUE_SIZE.get());
+            }
+            ReceiverQueueStatus::Unknown => unreachable!(),
+        }
+        let expected_text = &TEST_TEXT[..bytes_before_dequeue.min(QUEUE_SIZE.get())];
+        dbg!(expected_text);
+        let mut received_text = String::new();
+        for expected_byte in expected_text.bytes() {
+            sim.write_bool(sim.io().output_byte.ready, true);
+            #[hdl(sim)]
+            if let HdlSome(byte) = sim.read(sim.io().output_byte.data) {
+                let byte = byte.as_int();
+                assert_eq!(
+                    byte, expected_byte,
+                    "byte={:#x} {:?}, expected_byte={:#x} {:?}",
+                    byte, byte as char, expected_byte, expected_byte as char,
+                );
+                assert!(byte.is_ascii());
+                received_text.push(byte as char);
+            } else {
+                break;
+            }
+            sim.advance_time(CLOCK_HALF_PERIOD);
+            sim.write_clock(sim.io().cd.clk, true);
+            sim.advance_time(CLOCK_HALF_PERIOD);
+            sim.write_clock(sim.io().cd.clk, false);
+        }
+        #[hdl(sim)]
+        if let HdlSome(_) = sim.read(sim.io().output_byte.data) {
+            panic!("queue should be empty now");
+        }
+        sim.write_bool(sim.io().output_byte.ready, false);
+    }
+    let vcd = String::from_utf8(writer.writer.take().unwrap().take()).unwrap();
+    println!("####### VCD:\n{vcd}\n#######");
+    if vcd != include_str!("expected/receiver_and_uart_clock_gen.vcd") {
+        panic!();
+    }
+}
+
+#[hdl_module]
+fn simple_uart_loopback(
+    config: PhantomConst<MemoryInterfaceConfig>,
+    clock_input_properties: PhantomConst<peripherals::ClockInputProperties>,
+    baud_rate: f64,
+    receiver_queue_size: NonZeroUsize,
+) {
+    #[hdl]
+    let cd: ClockDomain = m.input();
+    #[hdl]
+    let memory_interface: MemoryInterface<PhantomConst<MemoryInterfaceConfig>> =
+        m.input(MemoryInterface[config]);
+    #[hdl]
+    let tx: Bool = m.output();
+    #[hdl]
+    let inner = instance(simple_uart(
+        config,
+        clock_input_properties,
+        baud_rate,
+        receiver_queue_size,
+    ));
+    connect(inner.cd, cd);
+    connect(inner.memory_interface, memory_interface);
+    connect(tx, inner.uart.tx);
+    connect(inner.uart.rx, inner.uart.tx);
+}
+
+const SIMPLE_UART_MEMORY_INTERFACE_CONFIG: MemoryInterfaceConfig =
+    simple_uart_memory_interface_config(FETCH_BLOCK_ID_WIDTH, Wrapping(0));
+
+struct MemoryOperationRunner<Sim: BorrowMut<Simulation<T>>, T: BundleType> {
+    sim: Sim,
+    clk: Expr<Clock>,
+    memory_interface: Expr<MemoryInterface<PhantomConst<MemoryInterfaceConfig>>>,
+    next_op_id: u64,
+    _phantom: PhantomData<T>,
+}
+
+const BUS_WIDTH_IN_BYTES: usize = SIMPLE_UART_MEMORY_INTERFACE_CONFIG.bus_width_in_bytes();
+
+impl<Sim: BorrowMut<Simulation<T>>, T: BundleType> MemoryOperationRunner<Sim, T> {
+    #[hdl]
+    fn run_memory_operation(
+        &mut self,
+        kind: impl ToSimValue<Type = MemoryOperationKind>,
+        addr: u64,
+        write_data: [u8; BUS_WIDTH_IN_BYTES],
+        rw_mask: [bool; BUS_WIDTH_IN_BYTES],
+        timeout_cycles: usize,
+    ) -> Result<[u8; BUS_WIDTH_IN_BYTES], SimValue<MemoryOperationErrorKind>> {
+        let kind = kind.into_sim_value();
+        assert_eq!(addr % BUS_WIDTH_IN_BYTES as u64, 0);
+        let config = self.memory_interface.ty().config;
+        let op_id = self.next_op_id.cast_to(UInt[FETCH_BLOCK_ID_WIDTH]);
+        self.next_op_id += 1;
+        println!(
+            "started: MemoryOperationStart {{\n    \
+            kind: {kind:?},\n    \
+            addr: {addr:#x},\n    \
+            write_data: {write_data:?},\n    \
+            rw_mask: {rw_mask:?},\n    \
+            op_id: {op_id},\n\
+            }}"
+        );
+        let start_value = #[hdl(sim)]
+        MemoryOperationStart::<_> {
+            kind,
+            addr,
+            write_data: &write_data[..],
+            rw_mask: &rw_mask[..],
+            op_id: op_id,
+            config,
+        };
+        #[hdl]
+        let MemoryInterface::<_> {
+            start,
+            finish,
+            next_op_ids,
+            config: _,
+        } = self.memory_interface;
+        let sim = self.sim.borrow_mut();
+        let check_io = |sim: &mut Simulation<T>,
+                        expected_start_ready: bool,
+                        expected_finish_data_is_some: bool,
+                        expected_next_op_ids: &[_]| {
+            assert_eq!(sim.read_bool(start.ready), expected_start_ready);
+            let finish_data = sim.read(finish.data);
+            let finish_data_is_some = #[hdl(sim)]
+            if let HdlSome(_) = &finish_data {
+                true
+            } else {
+                false
+            };
+            assert_eq!(
+                finish_data_is_some, expected_finish_data_is_some,
+                "finish.data: {finish_data:?}"
+            );
+
+            #[hdl(sim)]
+            if let HdlSome(next_op_ids) = sim.read(next_op_ids) {
+                let next_op_ids: Vec<_> = ArrayVec::elements_sim_ref(&next_op_ids)
+                    .iter()
+                    .map(|v| {
+                        v.cast_to_static::<UInt<{ FETCH_BLOCK_ID_WIDTH }>>()
+                            .as_int()
+                    })
+                    .collect();
+                assert_eq!(next_op_ids, expected_next_op_ids);
+            } else {
+                panic!("next_op_ids should be HdlSome");
+            }
+        };
+        check_io(sim, true, false, &[]);
+        sim.write(
+            start.data,
+            #[hdl(sim)]
+            (start.ty().data).HdlSome(&start_value),
+        );
+        sim.write(finish.ready, true);
+        sim.advance_time(CLOCK_HALF_PERIOD);
+        sim.write_clock(self.clk, true);
+        sim.advance_time(CLOCK_HALF_PERIOD);
+        sim.write_clock(self.clk, false);
+        sim.write(
+            start.data,
+            #[hdl(sim)]
+            (start.ty().data).HdlNone(),
+        );
+        check_io(
+            sim,
+            false,
+            false,
+            &[start_value
+                .op_id
+                .cast_to_static::<UInt<{ FETCH_BLOCK_ID_WIDTH }>>()
+                .as_int()],
+        );
+        for _ in 0..timeout_cycles {
+            sim.advance_time(CLOCK_HALF_PERIOD);
+            sim.write_clock(self.clk, true);
+            sim.advance_time(CLOCK_HALF_PERIOD);
+            sim.write_clock(self.clk, false);
+            #[hdl(sim)]
+            if let HdlSome(finish_data) = sim.read(finish.data) {
+                #[hdl(sim)]
+                let MemoryOperationFinish::<_> {
+                    kind,
+                    read_data,
+                    config: _,
+                } = finish_data;
+                assert_eq!(read_data.len(), BUS_WIDTH_IN_BYTES);
+                let read_data = std::array::from_fn(|i| read_data[i].as_int());
+                println!(
+                    "finished: MemoryOperationFinish {{\n    \
+                    kind: {kind:?},\n    \
+                    read_data: {read_data:?},\n\
+                    }}"
+                );
+                check_io(
+                    sim,
+                    false,
+                    true,
+                    &[start_value
+                        .op_id
+                        .cast_to_static::<UInt<{ FETCH_BLOCK_ID_WIDTH }>>()
+                        .as_int()],
+                );
+                sim.advance_time(CLOCK_HALF_PERIOD);
+                sim.write_clock(self.clk, true);
+                sim.advance_time(CLOCK_HALF_PERIOD);
+                sim.write_clock(self.clk, false);
+                sim.write(finish.ready, false);
+                check_io(sim, true, false, &[]);
+                return #[hdl(sim)]
+                match kind {
+                    MemoryOperationFinishKind::Success(kind) => {
+                        assert_eq!(
+                            SimValue::bits(&kind),
+                            SimValue::bits(&start_value.kind),
+                            "finish_data.kind={kind:?}",
+                        );
+                        Ok(read_data)
+                    }
+                    MemoryOperationFinishKind::Error(error) => Err(error),
+                };
+            }
+            check_io(
+                sim,
+                false,
+                false,
+                &[start_value
+                    .op_id
+                    .cast_to_static::<UInt<{ FETCH_BLOCK_ID_WIDTH }>>()
+                    .as_int()],
+            );
+        }
+        panic!("memory operation took too long");
+    }
+    #[hdl]
+    fn rw_bytes(
+        &mut self,
+        start_addr: u64,
+        is_read: bool,
+        bytes: &mut [u8],
+        timeout_cycles: usize,
+    ) -> Result<(), SimValue<MemoryOperationErrorKind>> {
+        let offset_in_chunk = start_addr as usize % BUS_WIDTH_IN_BYTES;
+        let masked_addr = start_addr - offset_in_chunk as u64;
+        assert!(
+            offset_in_chunk
+                .checked_add(bytes.len())
+                .is_some_and(|v| v <= BUS_WIDTH_IN_BYTES)
+        );
+        let bytes_range = offset_in_chunk..(offset_in_chunk + bytes.len());
+        let mut write_data = [0u8; BUS_WIDTH_IN_BYTES];
+        let mut rw_mask = [false; BUS_WIDTH_IN_BYTES];
+        rw_mask[bytes_range.clone()].fill(true);
+        if !is_read {
+            write_data[bytes_range.clone()].copy_from_slice(bytes);
+        }
+        let read_bytes = self.run_memory_operation(
+            if is_read {
+                #[hdl(sim)]
+                MemoryOperationKind.Read()
+            } else {
+                #[hdl(sim)]
+                MemoryOperationKind.Write()
+            },
+            masked_addr,
+            write_data,
+            rw_mask,
+            timeout_cycles,
+        )?;
+        if is_read {
+            bytes.copy_from_slice(&read_bytes[bytes_range]);
+        }
+        Ok(())
+    }
+    #[hdl]
+    fn read_bytes<const N: usize>(
+        &mut self,
+        start_addr: u64,
+        timeout_cycles: usize,
+    ) -> Result<[u8; N], SimValue<MemoryOperationErrorKind>> {
+        let mut retval = [0; _];
+        self.rw_bytes(start_addr, true, &mut retval, timeout_cycles)?;
+        Ok(retval)
+    }
+    #[hdl]
+    fn write_bytes<const N: usize>(
+        &mut self,
+        start_addr: u64,
+        mut bytes: [u8; N],
+        timeout_cycles: usize,
+    ) -> Result<(), SimValue<MemoryOperationErrorKind>> {
+        self.rw_bytes(start_addr, false, &mut bytes, timeout_cycles)
+    }
+}
+
+#[test]
+#[hdl]
+fn test_simple_uart() {
+    let _n = SourceLocation::normalize_files_for_tests();
+    const RECEIVER_QUEUE_SIZE: NonZeroUsize = NonZeroUsize::new(16).expect("not zero");
+    let m = simple_uart_loopback(
+        PhantomConst::new_sized(SIMPLE_UART_MEMORY_INTERFACE_CONFIG),
+        clock_input_properties(),
+        BAUD_RATE,
+        RECEIVER_QUEUE_SIZE,
+    );
+    let mut sim = Simulation::new(m);
+    let writer = RcWriter::default();
+    sim.add_trace_writer(VcdWriterDecls::new(writer.clone()));
+    struct DumpVcdOnDrop {
+        writer: Option<RcWriter>,
+    }
+    impl Drop for DumpVcdOnDrop {
+        fn drop(&mut self) {
+            if let Some(mut writer) = self.writer.take() {
+                let vcd = String::from_utf8(writer.take()).unwrap();
+                println!("####### VCD:\n{vcd}\n#######");
+            }
+        }
+    }
+    let mut writer = DumpVcdOnDrop {
+        writer: Some(writer),
+    };
+    sim.write(
+        sim.io().memory_interface.start.data,
+        sim.io().ty().memory_interface.start.data.HdlNone(),
+    );
+    sim.write(sim.io().memory_interface.finish.ready, false);
+    sim.write_clock(sim.io().cd.clk, false);
+    sim.write_reset(sim.io().cd.rst, true);
+    sim.advance_time(CLOCK_HALF_PERIOD);
+    sim.write_clock(sim.io().cd.clk, true);
+    sim.advance_time(CLOCK_HALF_PERIOD);
+    sim.write_clock(sim.io().cd.clk, false);
+    sim.write_reset(sim.io().cd.rst, false);
+    let mut mem_op_runner = MemoryOperationRunner {
+        clk: sim.io().cd.clk,
+        memory_interface: sim.io().memory_interface,
+        next_op_id: 0,
+        _phantom: PhantomData,
+        sim,
+    };
+
+    let empty_status = ReceiverQueueStatus::sim_as_u8(
+        #[hdl(sim)]
+        ReceiverQueueStatus.QueueEmpty(),
+    );
+
+    assert_eq!(
+        mem_op_runner
+            .read_bytes(SIMPLE_UART_STATUS_OFFSET, 1)
+            .unwrap(),
+        [empty_status],
+    );
+
+    mem_op_runner
+        .write_bytes(SIMPLE_UART_STATUS_OFFSET, [0], 1)
+        .unwrap_err();
+
+    assert_eq!(
+        mem_op_runner
+            .read_bytes(SIMPLE_UART_RECEIVE_OFFSET, 1)
+            .unwrap(),
+        [0],
+    );
+
+    const { assert!(SIMPLE_UART_RECEIVE_OFFSET + 1 == SIMPLE_UART_STATUS_OFFSET) };
+
+    assert_eq!(
+        mem_op_runner
+            .read_bytes(SIMPLE_UART_RECEIVE_OFFSET, 1)
+            .unwrap(),
+        [0, empty_status],
+    );
+
+    for i in 0..2 * BUS_WIDTH_IN_BYTES as u64 {
+        mem_op_runner
+            .read_bytes::<1>(SIMPLE_UART_USED_SIZE.get() + i, 1)
+            .unwrap_err();
+        mem_op_runner
+            .write_bytes(SIMPLE_UART_USED_SIZE.get() + i, [0], 1)
+            .unwrap_err();
+    }
+
+    const TEST_TEXT: &str = "Test test test test\n";
+
+    const DELAY: usize = 2;
+
+    const {
+        assert!(
+            TEST_TEXT.len() > RECEIVER_QUEUE_SIZE.get() + DELAY,
+            "TEST_TEXT must be long enough to make the queue overflow when accounting for DELAY"
+        )
+    };
+
+    for (i, b) in TEST_TEXT.bytes().enumerate() {
+        let current_count = i + 1;
+        mem_op_runner
+            .write_bytes(SIMPLE_UART_TRANSMIT_OFFSET, [b], 300)
+            .unwrap();
+        let expected_status = ReceiverQueueStatus::for_queue_len_as_u8(
+            current_count.saturating_sub(DELAY),
+            RECEIVER_QUEUE_SIZE,
+        );
+        assert_eq!(
+            mem_op_runner
+                .read_bytes(SIMPLE_UART_STATUS_OFFSET, 1)
+                .unwrap(),
+            [expected_status],
+        );
+    }
+
+    for (i, mut expected_byte) in TEST_TEXT.bytes().enumerate() {
+        let read = mem_op_runner
+            .read_bytes(SIMPLE_UART_RECEIVE_OFFSET, 1)
+            .unwrap();
+        let current_count = if i == 0 {
+            RECEIVER_QUEUE_SIZE.get() + 1 // initial read starts with overflow
+        } else {
+            RECEIVER_QUEUE_SIZE.get().saturating_sub(i)
+        };
+        if current_count == 0 {
+            expected_byte = 0;
+        }
+        let expected_status =
+            ReceiverQueueStatus::for_queue_len_as_u8(current_count, RECEIVER_QUEUE_SIZE);
+        assert_eq!(read, [expected_byte, expected_status]);
+    }
+
+    let vcd = String::from_utf8(writer.writer.take().unwrap().take()).unwrap();
+    println!("####### VCD:\n{vcd}\n#######");
+    if vcd != include_str!("expected/simple_uart.vcd") {
+        panic!();
+    }
+}