Subversion Repositories neorv32

<<<
:sectnums:
==== Processor-External Memory Interface (WISHBONE) (AXI4-Lite)

[cols="<3,<3,<4"]
[frame="topbot",grid="none"]
|=======================
| Hardware source file(s): | neorv32_wishbone.vhd |
| Software driver file(s): | none             | _implicitly used_
| Top entity port:         | `wb_tag_o`  | request tag output (3-bit)
|                          | `wb_adr_o`  | address output (32-bit)
|                          | `wb_dat_i`  | data input (32-bit)
|                          | `wb_dat_o`  | data output (32-bit)
|                          | `wb_we_o`   | write enable (1-bit)
|                          | `wb_sel_o`  | byte enable (4-bit)
|                          | `wb_stb_o`  | strobe (1-bit)
|                          | `wb_cyc_o`  | valid cycle (1-bit)
|                          | `wb_lock_o` | exclusive access request (1-bit)
|                          | `wb_ack_i`  | acknowledge (1-bit)
|                          | `wb_err_i`  | bus error (1-bit)
|                          | `fence_o`   | an executed `fence` instruction
|                          | `fencei_o`  | an executed `fence.i` instruction
| Configuration generics:  | _MEM_EXT_EN_         | enable external memory interface when _true_
|                          | _MEM_EXT_TIMEOUT_    | number of clock cycles after which an unacknowledged external bus access will auto-terminate (0 = disabled)
|                          | _MEM_EXT_PIPE_MODE_  | when _false_ (default): classic/standard Wishbone protocol; when _true_: pipelined Wishbone protocol
|                          | _MEM_EXT_BIG_ENDIAN_ | byte-order (Endianness) of external memory interface; true=BIG, false=little (default)
|                          | _MEM_EXT_ASYNC_RX_   | use registered RX path when _false_ (default); use async/direct RX path when _true_
| CPU interrupts:          | none |
|=======================


The external memory interface provides a Wishbone b4-compatible on-chip bus interface. The bus interface is
implemented when the _MEM_EXT_EN_ generic is _true_. This interface can be used to attach external memories,
custom hardware accelerators, additional IO devices or all other kinds of IP blocks.

The external interface is _not_ mapped to a _specific_ address space region. Instead, all CPU memory accesses that
do not target a processor-internal module are delegated to the external memory interface. In summary, a CPU load/store
access is delegated to the external bus interface if...

. it does not target the internal instruction memory IMEM (if implemented at all)
. **and** it does not target the internal data memory DMEM (if implemented at all)
. **and** it does not target the internal bootloader ROM or any of the IO devices - regardless if one or more of these components are
actually implemented or not.

[NOTE]
If the Execute In Place module (XIP) is implemented accesses map to this module are not forwarded to the
external memory interface. See section <<_execute_in_place_module_xip>> for more information.

[TIP]
See section <<_address_space>> for more information.


**Wishbone Bus Protocol**

The external memory interface either uses the **standard** ("classic") Wishbone transaction protocol (default) or
**pipelined** Wishbone transaction protocol. The transaction protocol is configured via the _MEM_EXT_PIPE_MODE_ generic:

When _MEM_EXT_PIPE_MODE_ is _false_, all bus control signals including _STB_ are active and remain stable until the
transfer is acknowledged/terminated. If _MEM_EXT_PIPE_MODE_ is _true_, all bus control except _STB_ are active
and remain until the transfer is acknowledged/terminated. In this case, _STB_ is asserted only during the very
first bus clock cycle.

.Exemplary Wishbone bus accesses using "classic" and "pipelined" protocol
[cols="^2,^2"]
[grid="none"]
|=======================
a| image::wishbone_classic_read.png[700,300]
a| image::wishbone_pipelined_write.png[700,300]
| **Classic** Wishbone read access | **Pipelined** Wishbone write access
|=======================


[TIP]
A detailed description of the implemented Wishbone bus protocol and the according interface signals
can be found in the data sheet "Wishbone B4 - WISHBONE System-on-Chip (SoC) Interconnection
Architecture for Portable IP Cores". A copy of this document can be found in the docs folder of this
project.


**Bus Access**

The NEORV32 Wishbone gateway does not support burst transfer yet, so there is always just one transfer in progress.
Hence, the Wishbone `STALL` signal is not implemented. An accessed Wishbone device does not have to respond immediately to a bus
request by sending an ACK. instead, there is a _time window_ where the device has to acknowledge the transfer. This time window
id configured by the _MEM_EXT_TIMEOUT_ top generic that defines the maximum time (in clock cycles) a bus access can be pending
before it is automatically terminated. If _MEM_EXT_TIMEOUT_ is set to zero, the timeout disabled an a bus access can take an
arbitrary number of cycles to complete.

When _MEM_EXT_TIMEOUT_ is greater than zero, the Wishbone gateway starts an internal countdown whenever the CPU
accesses a memory address via the external memory interface. If the accessed memory / device does not acknowledge (via `wb_ack_i`)
or terminate (via `wb_err_i`) the transfer within _MEM_EXT_TIMEOUT_ clock cycles, the bus access is automatically canceled
setting `wb_cyc_o` low again and a CPU load/store/instruction fetch bus access fault exception is raised.

[IMPORTANT]
Setting _MEM_EXT_TIMEOUT_ to zero will permanently stall the CPU if the targeted Wishbone device never responds. Hence,
_MEM_EXT_TIMEOUT_ should be always set to a value greater than zero. +
 +
This feature can be used as **safety guard** if the external memory system does not check for "address space holes". That means
that accessing addresses, which do not belong to a certain memory or device, do not permanently stall the processor due to an
unacknowledged/unterminated bus access. If the external memory system can guarantee to access **any** bus access
(even it targets an unimplemented address) the timeout feature should be disabled (_MEM_EXT_TIMEOUT_ = 0).


**Wishbone Tag**

The 3-bit wishbone `wb_tag_o` signal provides additional information regarding the access type. This signal
is compatible to the AXI4 _AxPROT_ signal.

* `wb_tag_o(0)` 1: privileged access (CPU is in machine mode); 0: unprivileged access
* `wb_tag_o(1)` always zero (indicating "secure access")
* `wb_tag_o(2)` 1: instruction fetch access, 0: data access


**Exclusive / Atomic Bus Access**

If the atomic memory access CPU extension (via _CPU_EXTENSION_RISCV_A_) is enabled, the CPU can
request an atomic/exclusive bus access via the external memory interface.

The load-reservate instruction (`lr.w`) will set the `wb_lock_o` signal telling the bus interconnect to establish a
reservation for the current accessed address (start of an exclusive access). This signal will stay asserted until
another memory access instruction is executed (for example a `sc.w`).

The memory system has to make sure that no other entity can access the reservated address until `wb_lock_o`
is released again. If this attempt fails, the memory system has to assert `wb_err_i` in order to indicate that the
reservation was broken.

[TIP]
See section <<_bus_interface>> for the CPU bus interface protocol.


**Endianness**

The NEORV32 CPU and the Processor setup are *little-endian* architectures. To allow direct connection
to a big-endian memory system the external bus interface provides an _Endianness configuration_. The
Endianness (of the external memory interface) can be configured via the _MEM_EXT_BIG_ENDIAN_ generic.
By default, the external memory interface uses little-endian byte-order (like the rest of the processor / CPU).

Application software can check the Endianness configuration of the external bus interface via the
SYSINFO module (see section <<_system_configuration_information_memory_sysinfo>> for more information).


**Gateway Latency**

By default, the Wishbone gateway introduces two additional latency cycles: processor-outgoing ("TX") and
processor-incoming ("RX") signals are fully registered. Thus, any access from the CPU to a processor-external devices
via Wishbone requires 2 additional clock cycles (at least; depending on device's latency).

If the attached Wishbone network / peripheral already provides output registers or if the Wishbone network is not relevant
for timing closure, the default buffering of incoming ("RX") data within the gateway can be disabled by implementing an
"asynchronous" RX path. The configuration is done via the _MEM_EXT_ASYNC_RX_ generic.


**AXI4-Lite Connectivity**

The AXI4-Lite wrapper (`rtl/system_integration/neorv32_SystemTop_axi4lite.vhd`) provides a Wishbone-to-
AXI4-Lite bridge, compatible with Xilinx Vivado (IP packager and block design editor). All entity signals of
this wrapper are of type _std_logic_ or _std_logic_vector_, respectively.

The AXI Interface has been verified using Xilinx Vivado IP Packager and Block Designer. The AXI
interface port signals are automatically detected when packaging the core.

.Example AXI SoC using Xilinx Vivado
image::neorv32_axi_soc.png[]

[WARNING]
Using the auto-termination timeout feature (_MEM_EXT_TIMEOUT_ greater than zero) is **not AXI4 compliant** as
the AXI protocol does not support canceling of bus transactions. Therefore, the NEORV32 top wrapper with AXI4-Lite interface
(`rtl/system_integration/neorv32_SystemTop_axi4lite`) configures _MEM_EXT_TIMEOUT_ = 0 by default.