diff --git a/docs/diagrams/diagrams.odg b/docs/diagrams/diagrams.odg
index fdc1d82..5b42eb6 100644
Binary files a/docs/diagrams/diagrams.odg and b/docs/diagrams/diagrams.odg differ
diff --git a/docs/diagrams/rbuf.png b/docs/diagrams/rbuf.png
new file mode 100644
index 0000000..34d7777
Binary files /dev/null and b/docs/diagrams/rbuf.png differ
diff --git a/docs/diagrams/wbuf.png b/docs/diagrams/wbuf.png
new file mode 100644
index 0000000..d4e5e54
Binary files /dev/null and b/docs/diagrams/wbuf.png differ
diff --git a/docs/index.rst b/docs/index.rst
index 60b2da1..7713cdc 100644
--- a/docs/index.rst
+++ b/docs/index.rst
@@ -115,3 +115,10 @@ Links
     props/addrmap
     props/signal
     props/rhs_props
+
+.. toctree::
+    :hidden:
+    :caption: Extended Properties
+
+    udps/intro
+    udps/write_buffering
diff --git a/docs/props/field.rst b/docs/props/field.rst
index 7e3fbf3..7010f3d 100644
--- a/docs/props/field.rst
+++ b/docs/props/field.rst
@@ -60,7 +60,9 @@ swmod
 |OK|
 
 If true, infers an output signal ``hwif_out..swmod`` that is asserted as the
-field is being modified by software.
+field is being modified by software. This can be due to a software write
+operation, or a software read operation that has clear/set side-effects.
+
 
 .. wavedrom::
 
diff --git a/docs/udps/intro.rst b/docs/udps/intro.rst
new file mode 100644
index 0000000..c35e710
--- /dev/null
+++ b/docs/udps/intro.rst
@@ -0,0 +1,32 @@
+Introduction
+============
+
+Although the official SystemRDL spec defines numerous properties that allow you
+to define complex register map structures, sometimes they are not enough to
+accurately describe a necessary feature. Fortunately the SystemRDL spec allows
+the language to be extended using "User Defined Properties" (UDPs). The
+PeakRDL-regblock tool understands several UDPs that are described in this
+section.
+
+
+.. list-table:: Summary of UDPs
+    :header-rows: 1
+
+    *   - Name
+        - Component
+        - Type
+        - Description
+
+    *   - buffer_writes
+        - reg
+        - boolean
+        - If set, writes to the register are double-buffered.
+
+          See details here: :ref:`write_buffering`.
+
+    *   - wbuffer_trigger
+        - reg
+        - reference
+        - Defines the buffered write commit trigger.
+
+          See details here: :ref:`write_buffering`.
diff --git a/docs/udps/write_buffering.rst b/docs/udps/write_buffering.rst
new file mode 100644
index 0000000..69fd5a5
--- /dev/null
+++ b/docs/udps/write_buffering.rst
@@ -0,0 +1,173 @@
+.. _write_buffering:
+
+Write-buffered Registers
+========================
+
+In some situations, your hardware design may requre that fields be
+updated atomically (on the same clock cycle), but your cpuif is not wide enough
+to do so natively. Using some UDP extensions, this regblock generator implements
+a way for you to add write-buffering to specific registers, which allows the
+regblock to delay the effect of a software write operation until a defined
+trigger event.
+
+Some examples of when this is useful:
+    * You need to have software update a wide 64-bit register atomically, but
+      the CPU interface is only 32-bits.
+    * Software needs to be able to write multiple registers such that the
+      hardware is updated atomically.
+    * Software can pre-load one or more registers with their next value, and
+      trigger the update via an external hardware signal.
+
+If a register is write-buffered, a holding buffer stage is inserted between the
+decode logic and the field logic. This effectively defers any software write
+operations to that register until a trigger event occurs that releases it.
+Write buffering storage is unique to each register that enables it.
+If a register is not write buffered, this buffer stage is bypassed.
+
+.. figure:: ../diagrams/wbuf.png
+
+
+Properties
+----------
+The behavior of write-buffered registers is defined using the following two
+properties:
+
+``buffer_writes``
+    * Assigned value is a boolean.
+    * If true, enables double-buffering of writes to this register.
+    * Any software write operation to a buffered register is held back in a storage element
+      unique to the register.
+    * The software write operation is committed to the register once triggered to do so.
+    * Unless specified otherwise, the buffer trigger occurs when the highest
+      address of the buffered register is written.
+
+``wbuffer_trigger``
+    * Assigned value is a reference to a register, single-bit field, signal, or single-bit property.
+    * Controls when the double-buffer commits the software write operation to the register's fields.
+    * If reference is a single-bit value (signal, field, property reference),
+      then the assertion of that value triggers the buffer to be evicted.
+    * Signal references shall have either activehigh/activelow property set to define the polarity.
+    * If the reference is a reg, then buffer is evicted when the register's
+      highest address is written
+
+
+Other Rules
+^^^^^^^^^^^
+* It is an error to set ``buffer_writes`` if the register does not contain any
+  writable fields
+* If ``buffer_writes`` is false, then anything assigned to ``wbuffer_trigger``
+  is ignored.
+* The buffered register and the trigger reference shall both be within the same
+  internal device. ie: one cannot be in an external scope with respect to the
+  other.
+* Unless it is a register, the reference assigned to ``wbuffer_trigger`` shall
+  represent a single bit.
+* If a buffered register was not written, any trigger events are ignored.
+* It is valid for a buffered register to be partially written (either via write
+  strobes, or partial addressing).
+* The software write operation is not considered to take place until the buffer
+  is evicted by the trigger. This influences the behavior of properties like
+  ``swmod`` and ``swacc`` - they are not asserted until the register's fields
+  are actually written by the buffer.
+
+
+
+Examples
+--------
+Below are several examples of what you can do with registers that are
+write-buffered.
+
+Wide Atomic Register
+^^^^^^^^^^^^^^^^^^^^
+
+Without write-buffering, it is impossible to update the state of a 64-bit
+register using a 32-bit CPU interface in a single clock-cycle.
+In this example, it still requires two write-cycles to update the register, but
+the register's storage element is not updated until both sub-words are written.
+Upon writing the 2nd sub-word (the higher byte address), the write data for both
+write cycles are committed to the register's storage element together on the
+same clock cycle. The register is updated atomically.
+
+.. code-block:: systemrdl
+    :emphasize-lines: 4
+
+    reg {
+        regwidth = 64;
+        accesswidth = 32;
+        buffer_writes = true;
+        field {
+            sw=rw; hw=r;
+        } my_field[63:0] = 0;
+    };
+
+
+Atomic Group of Registers
+^^^^^^^^^^^^^^^^^^^^^^^^^
+
+Perhaps you have a group of registers that need their state to be updated
+atomically. Using the ``wbuffer_trigger`` property, you can define which register
+write operation triggers the group to be updated.
+
+
+.. code-block:: systemrdl
+    :emphasize-lines: 2, 18-20
+
+    reg my_buffered_reg {
+        buffer_writes = true;
+        field {
+            sw=rw; hw=r;
+        } my_field[31:0] = 0;
+    };
+
+    my_buffered_reg reg1;
+    my_buffered_reg reg2;
+    my_buffered_reg reg3;
+
+    reg {
+        field {
+            sw=rw; hw=r;
+        } my_field[31:0] = 0;
+    } reg4;
+
+    reg1->wbuffer_trigger = reg4;
+    reg2->wbuffer_trigger = reg4;
+    reg3->wbuffer_trigger = reg4;
+
+
+In this example software may pre-write information into reg1-reg3, but the
+register write operations do not take effect until software also writes to reg4.
+The write operation to reg4 triggers the buffered data to be committed to reg1-reg3.
+This is guaranteed to occur on the same clock-cycle.
+
+
+Externally Triggered Register Update
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+Some applications may require precise timing for when a register (or group of registers)
+update their value. Often software cannot offer such timing precision.
+
+In this example, the trigger event is bound to an external signal. When asserted,
+any pending write operation the buffered register will be committed.
+The hwif_out value presents the new register state on the clock cycle after the
+trigger is asserted.
+
+.. code-block:: systemrdl
+    :emphasize-lines: 2, 11-13
+
+    reg my_buffered_reg {
+        buffer_writes = true;
+        field {
+            sw=rw; hw=r;
+        } my_field[31:0] = 0;
+    };
+
+    my_buffered_reg reg1;
+    my_buffered_reg reg2;
+
+    signal {
+        activehigh;
+    } trigger_signal;
+    reg1->wbuffer_trigger = trigger_signal;
+    reg2->wbuffer_trigger = trigger_signal;
+
+After software writes to ``reg1`` & ``reg2``, the written data is held back in the write
+buffer until ``hwif_in..trigger_signal`` is asserted by the hardware.
diff --git a/hdl-src/regblock_udps.rdl b/hdl-src/regblock_udps.rdl
new file mode 100644
index 0000000..a85d9cd
--- /dev/null
+++ b/hdl-src/regblock_udps.rdl
@@ -0,0 +1,18 @@
+/*
+ * This file defines several property extensions that are understood by the
+ * PeakRDL-Regblock SystemVerilog code generator.
+ *
+ * Compile this file prior to your other SystemRDL sources.
+ *
+ * For more details, see: https://peakrdl-regblock.readthedocs.io/en/latest/udps/intro.html
+ */
+
+property buffer_writes {
+    component = reg;
+    type = boolean;
+};
+
+property wbuffer_trigger {
+    component = reg;
+    type = ref;
+};
diff --git a/setup.py b/setup.py
index eaec68a..d91cb73 100644
--- a/setup.py
+++ b/setup.py
@@ -24,7 +24,7 @@ setuptools.setup(
     include_package_data=True,
     python_requires='>=3.6',
     install_requires=[
-        "systemrdl-compiler>=1.25.0",
+        "systemrdl-compiler>=1.25.1",
         "Jinja2>=2.11",
     ],
     entry_points = {
@@ -42,6 +42,7 @@ setuptools.setup(
         "Programming Language :: Python :: 3.8",
         "Programming Language :: Python :: 3.9",
         "Programming Language :: Python :: 3.10",
+        "Programming Language :: Python :: 3.11",
         "Programming Language :: Python :: 3 :: Only",
         "Intended Audience :: Developers",
         "License :: OSI Approved :: GNU General Public License v3 (GPLv3)",
diff --git a/src/peakrdl_regblock/addr_decode.py b/src/peakrdl_regblock/addr_decode.py
index 79eb656..48361d3 100644
--- a/src/peakrdl_regblock/addr_decode.py
+++ b/src/peakrdl_regblock/addr_decode.py
@@ -120,6 +120,7 @@ class DecodeLogicGenerator(RDLForLoopGenerator):
             s = f"{self.addr_decode.get_access_strobe(node)} = cpuif_req_masked & (cpuif_addr == {self._get_address_str(node)});"
             self.add_content(s)
         else:
+            # Register is wide. Create a substrobe for each subword
             n_subwords = regwidth // accesswidth
             subword_stride = accesswidth // 8
             for i in range(n_subwords):
diff --git a/src/peakrdl_regblock/exporter.py b/src/peakrdl_regblock/exporter.py
index 76a07cc..78152e0 100644
--- a/src/peakrdl_regblock/exporter.py
+++ b/src/peakrdl_regblock/exporter.py
@@ -10,12 +10,13 @@ from .dereferencer import Dereferencer
 from .readback import Readback
 from .identifier_filter import kw_filter as kwf
 
-from .cpuif import CpuifBase
-from .cpuif.apb4 import APB4_Cpuif
-from .hwif import Hwif
 from .utils import get_always_ff_event
 from .scan_design import DesignScanner
 from .validate_design import DesignValidator
+from .cpuif import CpuifBase
+from .cpuif.apb4 import APB4_Cpuif
+from .hwif import Hwif
+from .write_buffering import WriteBuffering
 
 class RegblockExporter:
     def __init__(self, **kwargs: Any) -> None:
@@ -30,6 +31,7 @@ class RegblockExporter:
         self.address_decode = AddressDecode(self)
         self.field_logic = FieldLogic(self)
         self.readback = None # type: Readback
+        self.write_buffering = None # type: WriteBuffering
         self.dereferencer = Dereferencer(self)
         self.min_read_latency = 0
         self.min_write_latency = 0
@@ -143,6 +145,7 @@ class RegblockExporter:
             self,
             retime_read_fanin
         )
+        self.write_buffering = WriteBuffering(self)
 
         # Validate that there are no unsupported constructs
         validator = DesignValidator(self)
@@ -153,8 +156,11 @@ class RegblockExporter:
             "module_name": module_name,
             "user_out_of_hier_signals": scanner.out_of_hier_signals.values(),
             "has_writable_msb0_fields": scanner.has_writable_msb0_fields,
+            "has_buffered_write_regs": scanner.has_buffered_write_regs,
+            "has_buffered_read_regs": scanner.has_buffered_read_regs,
             "cpuif": self.cpuif,
             "hwif": self.hwif,
+            "write_buffering": self.write_buffering,
             "get_resetsignal": self.dereferencer.get_resetsignal,
             "address_decode": self.address_decode,
             "field_logic": self.field_logic,
diff --git a/src/peakrdl_regblock/field_logic/__init__.py b/src/peakrdl_regblock/field_logic/__init__.py
index 38da2a4..87afb80 100644
--- a/src/peakrdl_regblock/field_logic/__init__.py
+++ b/src/peakrdl_regblock/field_logic/__init__.py
@@ -1,7 +1,6 @@
 from typing import TYPE_CHECKING
 
-from systemrdl.rdltypes import PropertyReference, PrecedenceType, InterruptType
-from systemrdl.node import Node
+from systemrdl.rdltypes import PrecedenceType, InterruptType
 
 from .bases import AssignmentPrecedence, NextStateConditional
 from . import sw_onread
@@ -186,19 +185,35 @@ class FieldLogic:
         """
         w_modifiable = field.is_sw_writable
         r_modifiable = (field.get_property('onread') is not None)
-        strb = self.exp.dereferencer.get_access_strobe(field)
+        buffer_writes = field.parent.get_property('buffer_writes')
 
         if w_modifiable and not r_modifiable:
             # assert swmod only on sw write
-            return f"{strb} && decoded_req_is_wr"
+            if buffer_writes:
+                # Write strobe arrives from buffer layer instead
+                wstrb = self.exp.write_buffering.get_write_strobe(field)
+                return wstrb
+            else:
+                # Unbuffered. Use decoder strobe directly
+                astrb = self.exp.dereferencer.get_access_strobe(field)
+                return f"{astrb} && decoded_req_is_wr"
 
         if w_modifiable and r_modifiable:
-            # assert swmod on all sw actions
-            return strb
+            # assert swmod on both sw read and write
+            if buffer_writes:
+                astrb = self.exp.dereferencer.get_access_strobe(field)
+                wstrb = self.exp.write_buffering.get_write_strobe(field)
+                rstrb = f"{astrb} && !decoded_req_is_wr"
+                return f"{wstrb} || {rstrb}"
+            else:
+                # Unbuffered. Use decoder strobe directly
+                astrb = self.exp.dereferencer.get_access_strobe(field)
+                return astrb
 
         if not w_modifiable and r_modifiable:
             # assert swmod only on sw read
-            return f"{strb} && !decoded_req_is_wr"
+            astrb = self.exp.dereferencer.get_access_strobe(field)
+            return f"{astrb} && !decoded_req_is_wr"
 
         # Not sw modifiable
         return "1'b0"
diff --git a/src/peakrdl_regblock/field_logic/generators.py b/src/peakrdl_regblock/field_logic/generators.py
index a84b70f..a163c10 100644
--- a/src/peakrdl_regblock/field_logic/generators.py
+++ b/src/peakrdl_regblock/field_logic/generators.py
@@ -85,7 +85,7 @@ class FieldLogicGenerator(RDLForLoopGenerator):
         super().__init__()
         self.field_logic = field_logic
         self.exp = field_logic.exp
-        self.field_storage_template = self.field_logic.exp.jj_env.get_template(
+        self.field_storage_template = self.exp.jj_env.get_template(
             "field_logic/templates/field_storage.sv"
         )
         self.intr_fields = [] # type: List[FieldNode]
diff --git a/src/peakrdl_regblock/field_logic/sw_onwrite.py b/src/peakrdl_regblock/field_logic/sw_onwrite.py
index df2ad3d..1a22ffa 100644
--- a/src/peakrdl_regblock/field_logic/sw_onwrite.py
+++ b/src/peakrdl_regblock/field_logic/sw_onwrite.py
@@ -15,60 +15,104 @@ class _OnWrite(NextStateConditional):
         return field.is_sw_writable and field.get_property('onwrite') == self.onwritetype
 
     def get_predicate(self, field: 'FieldNode') -> str:
-        strb = self.exp.dereferencer.get_access_strobe(field)
+        if field.parent.get_property('buffer_writes'):
+            # Is buffered write. Use alternate strobe
+            wstrb = self.exp.write_buffering.get_write_strobe(field)
 
-        if field.get_property('swwe') or field.get_property('swwel'):
-            # dereferencer will wrap swwel complement if necessary
-            qualifier = self.exp.dereferencer.get_field_propref_value(field, 'swwe')
-            return f"{strb} && decoded_req_is_wr && {qualifier}"
+            if field.get_property('swwe') or field.get_property('swwel'):
+                # dereferencer will wrap swwel complement if necessary
+                qualifier = self.exp.dereferencer.get_field_propref_value(field, 'swwe')
+                return f"{wstrb} && {qualifier}"
 
-        return f"{strb} && decoded_req_is_wr"
+            return wstrb
+        else:
+            # is regular register
+            strb = self.exp.dereferencer.get_access_strobe(field)
 
-    def _wbus_bitslice(self, field: 'FieldNode', subword_idx: int) -> str:
+            if field.get_property('swwe') or field.get_property('swwel'):
+                # dereferencer will wrap swwel complement if necessary
+                qualifier = self.exp.dereferencer.get_field_propref_value(field, 'swwe')
+                return f"{strb} && decoded_req_is_wr && {qualifier}"
+
+            return f"{strb} && decoded_req_is_wr"
+
+    def _wbus_bitslice(self, field: 'FieldNode', subword_idx: int = 0) -> str:
         # Get the source bitslice range from the internal cpuif's data bus
-        # For normal fields this ends up passing-through the field's low/high
-        # values unchanged.
-        # For fields within a wide register (accesswidth < regwidth), low/high
-        # may be shifted down and clamped depending on which sub-word is being accessed
-        accesswidth = field.parent.get_property('accesswidth')
+        if field.parent.get_property('buffer_writes'):
+            # register is buffered.
+            # write buffer is the full width of the register. no need to deal with subwords
+            high = field.high
+            low = field.low
+            if field.msb < field.lsb:
+                # slice is for an msb0 field.
+                # mirror it
+                regwidth = field.parent.get_property('regwidth')
+                low = regwidth - 1 - low
+                high = regwidth - 1 - high
+                low, high = high, low
+        else:
+            # Regular non-buffered register
+            # For normal fields this ends up passing-through the field's low/high
+            # values unchanged.
+            # For fields within a wide register (accesswidth < regwidth), low/high
+            # may be shifted down and clamped depending on which sub-word is being accessed
+            accesswidth = field.parent.get_property('accesswidth')
 
-        # Shift based on subword
-        high = field.high - (subword_idx * accesswidth)
-        low = field.low - (subword_idx * accesswidth)
+            # Shift based on subword
+            high = field.high - (subword_idx * accesswidth)
+            low = field.low - (subword_idx * accesswidth)
 
-        # clamp to accesswidth
-        high = max(min(high, accesswidth), 0)
-        low = max(min(low, accesswidth), 0)
+            # clamp to accesswidth
+            high = max(min(high, accesswidth), 0)
+            low = max(min(low, accesswidth), 0)
 
-        if field.msb < field.lsb:
-            # slice is for an msb0 field.
-            # mirror it
-            bus_width = self.exp.cpuif.data_width
-            low = bus_width - 1 - low
-            high = bus_width - 1 - high
-            low, high = high, low
+            if field.msb < field.lsb:
+                # slice is for an msb0 field.
+                # mirror it
+                bus_width = self.exp.cpuif.data_width
+                low = bus_width - 1 - low
+                high = bus_width - 1 - high
+                low, high = high, low
 
         return f"[{high}:{low}]"
 
     def _wr_data(self, field: 'FieldNode', subword_idx: int=0) -> str:
-        bslice = self._wbus_bitslice(field, subword_idx)
-
-        if field.msb < field.lsb:
-            # Field gets bitswapped since it is in [low:high] orientation
-            value = f"decoded_wr_data_bswap{bslice}"
+        if field.parent.get_property('buffer_writes'):
+            # Is buffered. Use value from write buffer
+            # No need to check msb0 ordering. Bus is pre-swapped, and bitslice
+            # accounts for it
+            bslice = self._wbus_bitslice(field)
+            wbuf_prefix = self.exp.write_buffering.get_wbuf_prefix(field)
+            return wbuf_prefix + ".data" + bslice
         else:
-            value = f"decoded_wr_data{bslice}"
-        return value
+            # Regular non-buffered register
+            bslice = self._wbus_bitslice(field, subword_idx)
+
+            if field.msb < field.lsb:
+                # Field gets bitswapped since it is in [low:high] orientation
+                value = "decoded_wr_data_bswap" + bslice
+            else:
+                value = "decoded_wr_data" + bslice
+            return value
 
     def _wr_biten(self, field: 'FieldNode', subword_idx: int=0) -> str:
-        bslice = self._wbus_bitslice(field, subword_idx)
-
-        if field.msb < field.lsb:
-            # Field gets bitswapped since it is in [low:high] orientation
-            value = f"decoded_wr_biten_bswap{bslice}"
+        if field.parent.get_property('buffer_writes'):
+            # Is buffered. Use value from write buffer
+            # No need to check msb0 ordering. Bus is pre-swapped, and bitslice
+            # accounts for it
+            bslice = self._wbus_bitslice(field)
+            wbuf_prefix = self.exp.write_buffering.get_wbuf_prefix(field)
+            return wbuf_prefix + ".biten" + bslice
         else:
-            value = f"decoded_wr_biten{bslice}"
-        return value
+            # Regular non-buffered register
+            bslice = self._wbus_bitslice(field, subword_idx)
+
+            if field.msb < field.lsb:
+                # Field gets bitswapped since it is in [low:high] orientation
+                value = "decoded_wr_biten_bswap" + bslice
+            else:
+                value = "decoded_wr_biten" + bslice
+            return value
 
     def get_assignments(self, field: 'FieldNode') -> List[str]:
         accesswidth = field.parent.get_property("accesswidth")
@@ -92,6 +136,7 @@ class _OnWrite(NextStateConditional):
         raise NotImplementedError
 
 
+#-------------------------------------------------------------------------------
 class WriteOneSet(_OnWrite):
     comment = "SW write 1 set"
     onwritetype = OnWriteType.woset
diff --git a/src/peakrdl_regblock/module_tmpl.sv b/src/peakrdl_regblock/module_tmpl.sv
index c9bd456..3c0207b 100644
--- a/src/peakrdl_regblock/module_tmpl.sv
+++ b/src/peakrdl_regblock/module_tmpl.sv
@@ -108,6 +108,15 @@ module {{module_name}} (
     assign cpuif_wr_ack = decoded_req & decoded_req_is_wr;
     assign cpuif_wr_err = '0;
 
+{%- if has_buffered_write_regs %}
+    //--------------------------------------------------------------------------
+    // Write double-buffers
+    //--------------------------------------------------------------------------
+    {{write_buffering.get_storage_struct()|indent}}
+
+    {{write_buffering.get_implementation()|indent}}
+{%- endif %}
+
     //--------------------------------------------------------------------------
     // Field logic
     //--------------------------------------------------------------------------
@@ -124,7 +133,6 @@ module {{module_name}} (
     logic readback_done;
     logic [{{cpuif.data_width-1}}:0] readback_data;
     {{readback.get_implementation()|indent}}
-
 {% if retime_read_response %}
     always_ff {{get_always_ff_event(cpuif.reset)}} begin
         if({{get_resetsignal(cpuif.reset)}}) begin
@@ -141,6 +149,5 @@ module {{module_name}} (
     assign cpuif_rd_ack = readback_done;
     assign cpuif_rd_data = readback_data;
     assign cpuif_rd_err = readback_err;
-{% endif %}
-
+{%- endif %}
 endmodule
diff --git a/src/peakrdl_regblock/readback/templates/readback.sv b/src/peakrdl_regblock/readback/templates/readback.sv
index fc706c8..29ca77d 100644
--- a/src/peakrdl_regblock/readback/templates/readback.sv
+++ b/src/peakrdl_regblock/readback/templates/readback.sv
@@ -3,7 +3,9 @@
 logic [{{cpuif.data_width-1}}:0] readback_array[{{array_size}}];
 {{array_assignments}}
 
-{% if do_fanin_stage %}
+
+{%- if do_fanin_stage %}
+
 // fanin stage
 logic [{{cpuif.data_width-1}}:0] readback_array_c[{{fanin_array_size}}];
 for(genvar g=0; g<{{fanin_loop_iter}}; g++) begin
@@ -48,6 +50,7 @@ always_comb begin
 end
 
 {%- else %}
+
 // Reduce the array
 always_comb begin
     automatic logic [{{cpuif.data_width-1}}:0] readback_data_var;
diff --git a/src/peakrdl_regblock/scan_design.py b/src/peakrdl_regblock/scan_design.py
index f2960bf..946334a 100644
--- a/src/peakrdl_regblock/scan_design.py
+++ b/src/peakrdl_regblock/scan_design.py
@@ -26,6 +26,8 @@ class DesignScanner(RDLListener):
         self.out_of_hier_signals = OrderedDict() # type: OrderedDict[str, SignalNode]
 
         self.has_writable_msb0_fields = False
+        self.has_buffered_write_regs = False
+        self.has_buffered_read_regs = False
 
     def _get_out_of_hier_field_reset(self) -> None:
         current_node = self.exp.top_node.parent
@@ -69,19 +71,7 @@ class DesignScanner(RDLListener):
         if node.external and (node != self.exp.top_node):
             # Do not inspect external components. None of my business
             return WalkerAction.SkipDescendants
-        return None
 
-    def enter_Reg(self, node: 'RegNode') -> None:
-        # The CPUIF's bus width is sized according to the largest accesswidth in the design
-        accesswidth = node.get_property('accesswidth')
-        self.cpuif_data_width = max(self.cpuif_data_width, accesswidth)
-
-    def enter_Signal(self, node: 'SignalNode') -> None:
-        if node.get_property('field_reset'):
-            path = node.get_path()
-            self.in_hier_signal_paths.add(path)
-
-    def enter_Field(self, node: 'FieldNode') -> None:
         # Collect any signals that are referenced by a property
         for prop_name in node.list_properties():
             value = node.get_property(prop_name)
@@ -93,5 +83,21 @@ class DesignScanner(RDLListener):
                 else:
                     self.in_hier_signal_paths.add(path)
 
+        return None
+
+    def enter_Reg(self, node: 'RegNode') -> None:
+        # The CPUIF's bus width is sized according to the largest accesswidth in the design
+        accesswidth = node.get_property('accesswidth')
+        self.cpuif_data_width = max(self.cpuif_data_width, accesswidth)
+
+        self.has_buffered_write_regs = self.has_buffered_write_regs or node.get_property('buffer_writes')
+        self.has_buffered_read_regs = self.has_buffered_read_regs or node.get_property('buffer_reads')
+
+    def enter_Signal(self, node: 'SignalNode') -> None:
+        if node.get_property('field_reset'):
+            path = node.get_path()
+            self.in_hier_signal_paths.add(path)
+
+    def enter_Field(self, node: 'FieldNode') -> None:
         if node.is_sw_writable and (node.msb < node.lsb):
             self.has_writable_msb0_fields = True
diff --git a/src/peakrdl_regblock/udps.py b/src/peakrdl_regblock/udps.py
index e06de80..dc754a5 100644
--- a/src/peakrdl_regblock/udps.py
+++ b/src/peakrdl_regblock/udps.py
@@ -3,7 +3,8 @@ from typing import Any
 from systemrdl.udp import UDPDefinition
 from systemrdl.component import Reg
 from systemrdl.rdltypes.references import RefType, PropertyReference
-from systemrdl.node import Node, RegNode, VectorNode
+from systemrdl.rdltypes import NoValue
+from systemrdl.node import Node, RegNode, VectorNode, SignalNode
 
 
 class xBufferTrigger(UDPDefinition):
@@ -13,7 +14,12 @@ class xBufferTrigger(UDPDefinition):
     def validate(self, node: Node, value: Any) -> None:
         # TODO: Reference shall not cross an internal/external boundary
 
-        if isinstance(value, VectorNode):
+        if value is NoValue:
+            self.msg.error(
+                "Double-buffer trigger property is missing a value assignment",
+                self.get_src_ref(node)
+            )
+        elif isinstance(value, VectorNode):
             # Trigger can reference a vector, but only if it is a single-bit
             if value.width != 1:
                 self.msg.error(
@@ -24,6 +30,14 @@ class xBufferTrigger(UDPDefinition):
                     ),
                     self.get_src_ref(node)
                 )
+
+            if isinstance(value, SignalNode):
+                if not value.get_property('activehigh') and not value.get_property('activelow'):
+                    self.msg.error(
+                        "Trigger was asigned a signal, but it does not specify whether it is activehigh/activelow",
+                        self.get_src_ref(node)
+                    )
+
         elif isinstance(value, PropertyReference) and value.width is not None:
             # Trigger can reference a property, but only if it is a single-bit
             if value.width != 1:
@@ -54,7 +68,6 @@ class BufferWrites(UDPDefinition):
     name = "buffer_writes"
     valid_components = {Reg}
     valid_type = bool
-    default_assignment = True
 
     def validate(self, node: 'Node', value: Any) -> None:
         assert isinstance(node, RegNode)
@@ -64,7 +77,6 @@ class BufferWrites(UDPDefinition):
                     "'buffer_writes' is set to true, but this register does not contain any writable fields.",
                     self.get_src_ref(node)
                 )
-        # TODO: Should I limit the use of other properties on double-buffered registers?
 
     def get_unassigned_default(self, node: 'Node') -> Any:
         return False
@@ -84,7 +96,6 @@ class BufferReads(UDPDefinition):
     name = "buffer_reads"
     valid_components = {Reg}
     valid_type = bool
-    default_assignment = True
 
     def validate(self, node: 'Node', value: Any) -> None:
         assert isinstance(node, RegNode)
@@ -95,8 +106,6 @@ class BufferReads(UDPDefinition):
                     self.get_src_ref(node)
                 )
 
-        # TODO: Should I limit the use of other properties on double-buffered registers?
-
     def get_unassigned_default(self, node: 'Node') -> Any:
         return False
 
diff --git a/src/peakrdl_regblock/validate_design.py b/src/peakrdl_regblock/validate_design.py
index 377d585..94ac16c 100644
--- a/src/peakrdl_regblock/validate_design.py
+++ b/src/peakrdl_regblock/validate_design.py
@@ -77,23 +77,30 @@ class DesignValidator(RDLListener):
 
 
     def enter_Field(self, node: 'FieldNode') -> None:
-        # 10.6.1-f: Any field that is software-writable or clear on read shall
-        # not span multiple software accessible sub-words (e.g., a 64-bit
-        # register with a 32-bit access width may not have a writable field with
-        # bits in both the upper and lower half of the register).
-        #
-        # Interpreting this further - this rule applies any time a field is
-        # software-modifiable by any means, including rclr, rset, ruser
-        # TODO: suppress this check for registers that have the appropriate
-        # buffer_writes/buffer_reads UDP set
         parent_accesswidth = node.parent.get_property('accesswidth')
         parent_regwidth = node.parent.get_property('regwidth')
-        if ((parent_accesswidth < parent_regwidth)
-                and (node.lsb // parent_accesswidth) != (node.msb // parent_accesswidth)
-                and (node.is_sw_writable or node.get_property('onread') is not None)):
-            # Field spans across sub-words
-            self.msg.error(
-                f"Software-modifiable field '{node.inst_name}' shall not span "
-                "multiple software-accessible subwords.",
-                node.inst.inst_src_ref
-            )
+        if (
+            (parent_accesswidth < parent_regwidth)
+            and (node.lsb // parent_accesswidth) != (node.msb // parent_accesswidth)
+        ):
+            # field spans multiple sub-words
+            if node.is_sw_writable and not node.parent.get_property('buffer_writes'):
+                # ... and is writable without the protection of double-buffering
+                # Enforce 10.6.1-f
+                self.msg.error(
+                    f"Software-writable field '{node.inst_name}' shall not span"
+                    " multiple software-accessible subwords. Consider enabling"
+                    " write double-buffering.",
+                    node.inst.inst_src_ref
+                )
+
+            if node.get_property('onread') is not None and not node.parent.get_property('buffer_reads'):
+                # ... is modified by an onread action without the atomicity of read buffering
+                # Enforce 10.6.1-f
+                self.msg.error(
+                    f"The field '{node.inst_name}' spans multiple software-accessible"
+                    " subwords and is modified on-read, making it impossible to"
+                    " access its value correctly. Consider enabling read"
+                    " double-buffering.",
+                    node.inst.inst_src_ref
+                )
diff --git a/src/peakrdl_regblock/write_buffering/__init__.py b/src/peakrdl_regblock/write_buffering/__init__.py
new file mode 100644
index 0000000..c71ab48
--- /dev/null
+++ b/src/peakrdl_regblock/write_buffering/__init__.py
@@ -0,0 +1,80 @@
+from typing import TYPE_CHECKING, Union
+
+from systemrdl.node import AddrmapNode, RegNode, FieldNode, SignalNode
+
+from .storage_generator import WBufStorageStructGenerator
+from .implementation_generator import WBufLogicGenerator
+from ..utils import get_indexed_path
+
+if TYPE_CHECKING:
+    from ..exporter import RegblockExporter
+
+
+class WriteBuffering:
+    def __init__(self, exp:'RegblockExporter'):
+        self.exp = exp
+
+    @property
+    def top_node(self) -> 'AddrmapNode':
+        return self.exp.top_node
+
+
+    def get_storage_struct(self) -> str:
+        struct_gen = WBufStorageStructGenerator(self)
+        s = struct_gen.get_struct(self.top_node, "wbuf_storage_t")
+        assert s is not None
+        return s + "\nwbuf_storage_t wbuf_storage;"
+
+
+    def get_implementation(self) -> str:
+        gen = WBufLogicGenerator(self)
+        s = gen.get_content(self.top_node)
+        assert s is not None
+        return s
+
+    def get_wbuf_prefix(self, node: Union[RegNode, FieldNode]) -> str:
+        if isinstance(node, FieldNode):
+            node = node.parent
+        wbuf_prefix = "wbuf_storage." + get_indexed_path(self.top_node, node)
+        return wbuf_prefix
+
+    def get_write_strobe(self, node: Union[RegNode, FieldNode]) -> str:
+        prefix = self.get_wbuf_prefix(node)
+        return f"{prefix}.pending && {self.get_trigger(node)}"
+
+    def get_raw_trigger(self, node: 'RegNode') -> str:
+        trigger = node.get_property('wbuffer_trigger')
+
+        if isinstance(trigger, RegNode):
+            # Trigger is a register.
+            # trigger when uppermost address of the register is written
+            regwidth = node.get_property('regwidth')
+            accesswidth = node.get_property('accesswidth')
+            strb_prefix = self.exp.dereferencer.get_access_strobe(trigger, reduce_substrobes=False)
+
+            if accesswidth < regwidth:
+                n_subwords = regwidth // accesswidth
+                return f"{strb_prefix}[{n_subwords-1}] && decoded_req_is_wr"
+            else:
+                return f"{strb_prefix} && decoded_req_is_wr"
+        elif isinstance(trigger, SignalNode):
+            s = self.exp.dereferencer.get_value(trigger)
+            if trigger.get_property('activehigh'):
+                return s
+            else:
+                return f"~{s}"
+        else:
+            # Trigger is a field or propref bit
+            return self.exp.dereferencer.get_value(trigger)
+
+    def get_trigger(self, node: Union[RegNode, FieldNode]) -> str:
+        if isinstance(node, FieldNode):
+            node = node.parent
+        trigger = node.get_property('wbuffer_trigger')
+
+        if isinstance(trigger, RegNode) and trigger == node:
+            # register is its own trigger
+            # use the delayed trigger signal
+            return self.get_wbuf_prefix(node) + ".trigger_q"
+        else:
+            return self.get_raw_trigger(node)
diff --git a/src/peakrdl_regblock/write_buffering/implementation_generator.py b/src/peakrdl_regblock/write_buffering/implementation_generator.py
new file mode 100644
index 0000000..808b9d3
--- /dev/null
+++ b/src/peakrdl_regblock/write_buffering/implementation_generator.py
@@ -0,0 +1,60 @@
+from typing import TYPE_CHECKING
+from collections import namedtuple
+
+from systemrdl.component import Reg
+from systemrdl.node import RegNode
+
+from ..forloop_generator import RDLForLoopGenerator
+from ..utils import get_always_ff_event
+
+if TYPE_CHECKING:
+    from . import WriteBuffering
+
+class WBufLogicGenerator(RDLForLoopGenerator):
+    i_type = "genvar"
+    def __init__(self, wbuf: 'WriteBuffering') -> None:
+        super().__init__()
+        self.wbuf = wbuf
+        self.exp = wbuf.exp
+        self.template = self.exp.jj_env.get_template(
+            "write_buffering/template.sv"
+        )
+
+    def enter_Reg(self, node: 'RegNode') -> None:
+        super().enter_Reg(node)
+        assert isinstance(node.inst, Reg)
+
+        if not node.get_property('buffer_writes'):
+            return
+
+        regwidth = node.get_property('regwidth')
+        accesswidth = node.get_property('accesswidth')
+        strb_prefix = self.exp.dereferencer.get_access_strobe(node, reduce_substrobes=False)
+        Segment = namedtuple("Segment", ["strobe", "bslice"])
+        segments = []
+        if accesswidth < regwidth:
+            n_subwords = regwidth // accesswidth
+            for i in range(n_subwords):
+                strobe = strb_prefix + f"[{i}]"
+                if node.inst.is_msb0_order:
+                    bslice = f"[{regwidth - (accesswidth * i) - 1}: {regwidth - (accesswidth * (i+1))}]"
+                else:
+                    bslice = f"[{(accesswidth * (i + 1)) - 1}:{accesswidth * i}]"
+                segments.append(Segment(strobe, bslice))
+        else:
+            segments.append(Segment(strb_prefix, ""))
+
+        trigger = node.get_property('wbuffer_trigger')
+        is_own_trigger = (isinstance(trigger, RegNode) and trigger == node)
+
+        context = {
+            'wbuf': self.wbuf,
+            'wbuf_prefix': self.wbuf.get_wbuf_prefix(node),
+            'segments': segments,
+            'node': node,
+            'cpuif': self.exp.cpuif,
+            'get_resetsignal': self.exp.dereferencer.get_resetsignal,
+            'get_always_ff_event': lambda resetsignal : get_always_ff_event(self.exp.dereferencer, resetsignal),
+            'is_own_trigger': is_own_trigger,
+        }
+        self.add_content(self.template.render(context))
diff --git a/src/peakrdl_regblock/write_buffering/storage_generator.py b/src/peakrdl_regblock/write_buffering/storage_generator.py
new file mode 100644
index 0000000..6b48472
--- /dev/null
+++ b/src/peakrdl_regblock/write_buffering/storage_generator.py
@@ -0,0 +1,32 @@
+from typing import TYPE_CHECKING
+
+from systemrdl.node import FieldNode, RegNode
+
+from ..struct_generator import RDLStructGenerator
+
+if TYPE_CHECKING:
+    from . import WriteBuffering
+
+class WBufStorageStructGenerator(RDLStructGenerator):
+    def __init__(self, wbuf: 'WriteBuffering') -> None:
+        super().__init__()
+        self.wbuf = wbuf
+
+    def enter_Field(self, node: FieldNode) -> None:
+        # suppress parent class's field behavior
+        pass
+
+    def enter_Reg(self, node: RegNode) -> None:
+        super().enter_Reg(node)
+
+        if not node.get_property('buffer_writes'):
+            return
+
+        regwidth = node.get_property('regwidth')
+        self.add_member("data", regwidth)
+        self.add_member("biten", regwidth)
+        self.add_member("pending")
+
+        trigger = node.get_property('wbuffer_trigger')
+        if isinstance(trigger, RegNode) and trigger == node:
+            self.add_member("trigger_q")
diff --git a/src/peakrdl_regblock/write_buffering/template.sv b/src/peakrdl_regblock/write_buffering/template.sv
new file mode 100644
index 0000000..b2d01cd
--- /dev/null
+++ b/src/peakrdl_regblock/write_buffering/template.sv
@@ -0,0 +1,31 @@
+always_ff {{get_always_ff_event(cpuif.reset)}} begin
+    if({{get_resetsignal(cpuif.reset)}}) begin
+        {{wbuf_prefix}}.pending <= '0;
+        {{wbuf_prefix}}.data <= '0;
+        {{wbuf_prefix}}.biten <= '0;
+        {%- if is_own_trigger %}
+        {{wbuf_prefix}}.trigger_q <= '0;
+        {%- endif %}
+    end else begin
+        {%- for segment in segments %}
+        if({{segment.strobe}} && decoded_req_is_wr) begin
+            {{wbuf_prefix}}.pending <= '1;
+            {%- if node.inst.is_msb0_order %}
+            {{wbuf_prefix}}.data{{segment.bslice}} <= decoded_wr_data_bswap;
+            {{wbuf_prefix}}.biten{{segment.bslice}} <= decoded_wr_biten_bswap;
+            {%- else %}
+            {{wbuf_prefix}}.data{{segment.bslice}} <= decoded_wr_data;
+            {{wbuf_prefix}}.biten{{segment.bslice}} <= decoded_wr_biten;
+            {%- endif %}
+        end
+        {%- endfor %}
+        {% if is_own_trigger %}
+        {{wbuf_prefix}}.trigger_q <= {{wbuf.get_raw_trigger(node)}};
+        {%- endif %}
+        if({{wbuf.get_trigger(node)}}) begin
+            {{wbuf_prefix}}.pending <= '0;
+            {{wbuf_prefix}}.data <= '0;
+            {{wbuf_prefix}}.biten <= '0;
+        end
+    end
+end
diff --git a/tests/.coveragerc b/tests/.coveragerc
index f664bca..05e3012 100644
--- a/tests/.coveragerc
+++ b/tests/.coveragerc
@@ -2,6 +2,10 @@
 branch = True
 #relative_files = True
 
+omit =
+    # to be covered elsewhere
+    */__peakrdl__.py
+
 [paths]
 source =
     ../src/peakrdl_regblock/
diff --git a/tests/lib/base_testcase.py b/tests/lib/base_testcase.py
index 97582e8..7bb6489 100644
--- a/tests/lib/base_testcase.py
+++ b/tests/lib/base_testcase.py
@@ -86,8 +86,13 @@ class BaseTestCase(unittest.TestCase):
             rdl_file = glob.glob(os.path.join(this_dir, "*.rdl"))[0]
 
         rdlc = RDLCompiler()
+
+        # Load the UDPs
         for udp in ALL_UDPS:
             rdlc.register_udp(udp)
+        # ... including the definition
+        udp_file = os.path.join(this_dir, "../../hdl-src/regblock_udps.rdl")
+        rdlc.compile_file(udp_file)
 
         rdlc.compile_file(rdl_file)
         root = rdlc.elaborate(cls.rdl_elab_target, "regblock", cls.rdl_elab_params)
diff --git a/tests/test_write_buffer/__init__.py b/tests/test_write_buffer/__init__.py
new file mode 100644
index 0000000..e69de29
diff --git a/tests/test_write_buffer/regblock.rdl b/tests/test_write_buffer/regblock.rdl
new file mode 100644
index 0000000..ca267bc
--- /dev/null
+++ b/tests/test_write_buffer/regblock.rdl
@@ -0,0 +1,106 @@
+addrmap top {
+    default regwidth = 16;
+    default accesswidth = 16;
+    default sw=rw;
+    default hw=r;
+    //--------------------------------------------------------------------------
+    // Wide registers
+    //--------------------------------------------------------------------------
+    reg {
+        regwidth = 64;
+        buffer_writes = true;
+        field {} f1[63:0] = 0;
+    } reg1;
+
+    reg {
+        regwidth = 64;
+        buffer_writes = true;
+        field {} f1[0:63] = 0;
+    } reg1_msb0;
+
+    reg {
+        regwidth = 32;
+        buffer_writes = true;
+        field {} f1[19:8] = 0;
+        field {} f2[23:20] = 0;
+    } reg2;
+
+    reg {
+        regwidth = 32;
+        buffer_writes = true;
+        field {} f1[8:19] = 0;
+        field {} f2[20:23] = 0;
+    } reg2_msb0;
+
+    //--------------------------------------------------------------------------
+    // Alternate Triggers
+    //--------------------------------------------------------------------------
+    reg myreg {
+        buffer_writes;
+        field {} f1[15:0] = 0;
+    };
+
+    // Trigger via another register
+    myreg g1_r1;
+    myreg g1_r2;
+    g1_r1->buffer_writes = false;
+    g1_r2->wbuffer_trigger = g1_r1;
+
+    // triger from signal
+    signal {
+        activehigh;
+    } trigger_sig;
+    signal {
+        activelow;
+    } trigger_sig_n;
+    myreg g2_r1;
+    myreg g2_r2;
+    g2_r1->wbuffer_trigger = trigger_sig;
+    g2_r2->wbuffer_trigger = trigger_sig_n;
+
+    // trigger from field
+    myreg g3_r1;
+    reg {
+        field {
+            sw=w; hw=r; singlepulse;
+        } trig = 0;
+    } g3_trig;
+    g3_r1->wbuffer_trigger = g3_trig.trig;
+
+    // trigger from propref
+    myreg g4_r1;
+    reg {
+        field {
+            hw=na;
+        } trig_vec[3:0] = 0;
+    } g4_trig;
+    g4_r1->wbuffer_trigger = g4_trig.trig_vec->anded;
+
+    //--------------------------------------------------------------------------
+    // swmod behavior
+    //--------------------------------------------------------------------------
+    myreg g5_r1;
+    g5_r1->wbuffer_trigger = trigger_sig;
+    reg {
+        field{
+            sw=rw;
+            hw=na;
+            counter;
+        } c[3:0] = 0;
+    } g5_modcount;
+    g5_modcount.c->incr = g5_r1.f1->swmod;
+
+    myreg g6_r1;
+    g6_r1.f1->rclr;
+    g6_r1->wbuffer_trigger = trigger_sig;
+    reg {
+        field{
+            sw=rw;
+            hw=na;
+            counter;
+        } c[3:0] = 0;
+    } g6_modcount;
+    g6_modcount.c->incr = g6_r1.f1->swmod;
+
+
+};
diff --git a/tests/test_write_buffer/tb_template.sv b/tests/test_write_buffer/tb_template.sv
new file mode 100644
index 0000000..c202570
--- /dev/null
+++ b/tests/test_write_buffer/tb_template.sv
@@ -0,0 +1,220 @@
+{% extends "lib/tb_base.sv" %}
+
+{% block seq %}
+    {% sv_line_anchor %}
+    cb.hwif_in.trigger_sig_n <= '1;
+    ##1;
+    cb.rst <= '0;
+    ##1;
+
+    //--------------------------------------------------------------------------
+    // Wide registers
+    //--------------------------------------------------------------------------
+
+    // reg1
+    cpuif.assert_read('h0, 'h0);
+    cpuif.assert_read('h2, 'h0);
+    cpuif.assert_read('h4, 'h0);
+    cpuif.assert_read('h6, 'h0);
+    assert(cb.hwif_out.reg1.f1.value == 0);
+    cpuif.write('h0, 'h1234);
+    cpuif.assert_read('h0, 'h0);
+    cpuif.assert_read('h2, 'h0);
+    cpuif.assert_read('h4, 'h0);
+    cpuif.assert_read('h6, 'h0);
+    assert(cb.hwif_out.reg1.f1.value == 0);
+    cpuif.write('h2, 'h5678);
+    cpuif.assert_read('h0, 'h0);
+    cpuif.assert_read('h2, 'h0);
+    cpuif.assert_read('h4, 'h0);
+    cpuif.assert_read('h6, 'h0);
+    assert(cb.hwif_out.reg1.f1.value == 0);
+    cpuif.write('h4, 'h9ABC);
+    cpuif.assert_read('h0, 'h0);
+    cpuif.assert_read('h2, 'h0);
+    cpuif.assert_read('h4, 'h0);
+    cpuif.assert_read('h6, 'h0);
+    assert(cb.hwif_out.reg1.f1.value == 0);
+    cpuif.write('h6, 'hDEF1);
+    @cb; @cb;
+    assert(cb.hwif_out.reg1.f1.value == 64'hDEF19ABC56781234);
+    cpuif.assert_read('h0, 'h1234);
+    cpuif.assert_read('h2, 'h5678);
+    cpuif.assert_read('h4, 'h9ABC);
+    cpuif.assert_read('h6, 'hDEF1);
+
+    // reg1_msb0
+    cpuif.assert_read('h8, 'h0);
+    cpuif.assert_read('hA, 'h0);
+    cpuif.assert_read('hC, 'h0);
+    cpuif.assert_read('hE, 'h0);
+    assert(cb.hwif_out.reg1_msb0.f1.value == 0);
+    cpuif.write('h8, 'h1234);
+    cpuif.assert_read('h8, 'h0);
+    cpuif.assert_read('hA, 'h0);
+    cpuif.assert_read('hC, 'h0);
+    cpuif.assert_read('hE, 'h0);
+    assert(cb.hwif_out.reg1_msb0.f1.value == 0);
+    cpuif.write('hA, 'h5678);
+    cpuif.assert_read('h8, 'h0);
+    cpuif.assert_read('hA, 'h0);
+    cpuif.assert_read('hC, 'h0);
+    cpuif.assert_read('hE, 'h0);
+    assert(cb.hwif_out.reg1_msb0.f1.value == 0);
+    cpuif.write('hC, 'h9ABC);
+    cpuif.assert_read('h8, 'h0);
+    cpuif.assert_read('hA, 'h0);
+    cpuif.assert_read('hC, 'h0);
+    cpuif.assert_read('hE, 'h0);
+    assert(cb.hwif_out.reg1_msb0.f1.value == 0);
+    cpuif.write('hE, 'hDEF1);
+    @cb; @cb;
+    assert({<<{cb.hwif_out.reg1_msb0.f1.value}} == 64'hDEF19ABC56781234);
+    cpuif.assert_read('h8, 'h1234);
+    cpuif.assert_read('hA, 'h5678);
+    cpuif.assert_read('hC, 'h9ABC);
+    cpuif.assert_read('hE, 'hDEF1);
+
+    // reg2
+    cpuif.assert_read('h10, 'h0);
+    cpuif.assert_read('h12, 'h0);
+    assert(cb.hwif_out.reg2.f1.value == 0);
+    assert(cb.hwif_out.reg2.f2.value == 0);
+    cpuif.write('h10, 'h34AA);
+    cpuif.assert_read('h10, 'h0);
+    cpuif.assert_read('h12, 'h0);
+    assert(cb.hwif_out.reg2.f1.value == 0);
+    assert(cb.hwif_out.reg2.f2.value == 0);
+    cpuif.write('h12, 'hAA12);
+    @cb; @cb;
+    assert(cb.hwif_out.reg2.f1.value == 12'h234);
+    assert(cb.hwif_out.reg2.f2.value == 4'h1);
+    cpuif.assert_read('h10, 'h3400);
+    cpuif.assert_read('h12, 'h0012);
+
+    // reg2_msb0
+    cpuif.assert_read('h14, 'h0);
+    cpuif.assert_read('h16, 'h0);
+    assert(cb.hwif_out.reg2_msb0.f1.value == 0);
+    assert(cb.hwif_out.reg2_msb0.f2.value == 0);
+    cpuif.write('h14, 'h34AA);
+    cpuif.assert_read('h14, 'h0);
+    cpuif.assert_read('h16, 'h0);
+    assert(cb.hwif_out.reg2_msb0.f1.value == 0);
+    assert(cb.hwif_out.reg2_msb0.f2.value == 0);
+    cpuif.write('h16, 'hAA12);
+    @cb; @cb;
+    assert({<<{cb.hwif_out.reg2_msb0.f1.value}} == 12'h234);
+    assert({<<{cb.hwif_out.reg2_msb0.f2.value}} == 4'h1);
+    cpuif.assert_read('h14, 'h3400);
+    cpuif.assert_read('h16, 'h0012);
+
+    //--------------------------------------------------------------------------
+    // Alternate Triggers
+    //--------------------------------------------------------------------------
+
+    // g1
+    cpuif.assert_read('h18, 'h0);
+    cpuif.assert_read('h1A, 'h0);
+    assert(cb.hwif_out.g1_r1.f1.value == 0);
+    assert(cb.hwif_out.g1_r2.f1.value == 0);
+    cpuif.write('h1A, 'h1234);
+    cpuif.assert_read('h18, 'h0);
+    cpuif.assert_read('h1A, 'h0);
+    assert(cb.hwif_out.g1_r1.f1.value == 0);
+    assert(cb.hwif_out.g1_r2.f1.value == 0);
+    cpuif.write('h18, 'hABCD);
+    @cb;
+    assert(cb.hwif_out.g1_r1.f1.value == 'hABCD);
+    assert(cb.hwif_out.g1_r2.f1.value == 'h1234);
+
+    // g2
+    cpuif.assert_read('h1C, 'h0);
+    cpuif.assert_read('h1E, 'h0);
+    assert(cb.hwif_out.g2_r1.f1.value == 0);
+    assert(cb.hwif_out.g2_r2.f1.value == 0);
+    cpuif.write('h1C, 'h5678);
+    cpuif.write('h1E, 'h9876);
+    cpuif.assert_read('h1C, 'h0);
+    cpuif.assert_read('h1E, 'h0);
+    assert(cb.hwif_out.g2_r1.f1.value == 0);
+    assert(cb.hwif_out.g2_r2.f1.value == 0);
+    cb.hwif_in.trigger_sig <= '1;
+    cb.hwif_in.trigger_sig_n <= '0;
+    @cb;
+    cb.hwif_in.trigger_sig <= '0;
+    cb.hwif_in.trigger_sig_n <= '1;
+    @cb;
+    assert(cb.hwif_out.g2_r1.f1.value == 'h5678);
+    assert(cb.hwif_out.g2_r2.f1.value == 'h9876);
+
+    // g3
+    cpuif.assert_read('h20, 'h0);
+    assert(cb.hwif_out.g3_r1.f1.value == 0);
+    cpuif.write('h20, 'hFEDC);
+    @cb; @cb;
+    assert(cb.hwif_out.g3_r1.f1.value == 0);
+    cpuif.assert_read('h20, 'h0);
+    cpuif.write('h22, 'h0000);
+    @cb; @cb;
+    assert(cb.hwif_out.g3_r1.f1.value == 0);
+    cpuif.assert_read('h20, 'h0);
+    cpuif.write('h22, 'h0001);
+    @cb; @cb;
+    assert(cb.hwif_out.g3_r1.f1.value == 'hFEDC);
+    cpuif.assert_read('h20, 'hFEDC);
+
+    // g4
+    cpuif.assert_read('h24, 'h0);
+    assert(cb.hwif_out.g4_r1.f1.value == 0);
+    cpuif.write('h24, 'hCAFE);
+    @cb; @cb;
+    assert(cb.hwif_out.g4_r1.f1.value == 0);
+    cpuif.assert_read('h24, 'h0);
+    cpuif.write('h26, 'h0000);
+    @cb; @cb;
+    assert(cb.hwif_out.g4_r1.f1.value == 0);
+    cpuif.assert_read('h24, 'h0);
+    cpuif.write('h26, 'h000E);
+    @cb; @cb;
+    assert(cb.hwif_out.g4_r1.f1.value == 0);
+    cpuif.assert_read('h24, 'h0);
+    cpuif.write('h26, 'h000F);
+    @cb; @cb;
+    assert(cb.hwif_out.g4_r1.f1.value == 'hCAFE);
+    cpuif.assert_read('h24, 'hCAFE);
+
+    //--------------------------------------------------------------------------
+    // swmod behavior
+    //--------------------------------------------------------------------------
+    // g5
+    cpuif.assert_read('h28, 'h0);
+    cpuif.assert_read('h2A, 'h0);
+    cpuif.write('h28, 'h1234);
+    cpuif.write('h28, 'h5678);
+    cpuif.assert_read('h28, 'h0);
+    cpuif.assert_read('h2A, 'h0);
+    cb.hwif_in.trigger_sig <= '1;
+    @cb;
+    cb.hwif_in.trigger_sig <= '0;
+    cpuif.assert_read('h28, 'h5678);
+    cpuif.assert_read('h2A, 'h1);
+
+    // g6
+    cpuif.assert_read('h2E, 'h0);
+    cpuif.assert_read('h2C, 'h0);
+    cpuif.assert_read('h2E, 'h1);
+    cpuif.write('h2C, 'h5678);
+    cpuif.write('h2C, 'h1234);
+    cpuif.assert_read('h2E, 'h1);
+    cpuif.assert_read('h2C, 'h0);
+    cpuif.assert_read('h2E, 'h2);
+    cb.hwif_in.trigger_sig <= '1;
+    @cb;
+    cb.hwif_in.trigger_sig <= '0;
+    cpuif.assert_read('h2E, 'h3);
+    cpuif.assert_read('h2C, 'h1234);
+    cpuif.assert_read('h2E, 'h4);
+
+
+{% endblock %}
diff --git a/tests/test_write_buffer/testcase.py b/tests/test_write_buffer/testcase.py
new file mode 100644
index 0000000..835b5ef
--- /dev/null
+++ b/tests/test_write_buffer/testcase.py
@@ -0,0 +1,5 @@
+from ..lib.sim_testcase import SimTestCase
+
+class Test(SimTestCase):
+    def test_dut(self):
+        self.run_test()