1 |
60 |
zero_gravi |
:sectnums:
|
2 |
|
|
== NEORV32 Central Processing Unit (CPU)
|
3 |
|
|
|
4 |
|
|
image::riscv_logo.png[width=350,align=center]
|
5 |
|
|
|
6 |
|
|
**Key Features**
|
7 |
|
|
|
8 |
|
|
* 32-bit pipelined/multi-cycle in-order `rv32` RISC-V CPU
|
9 |
61 |
zero_gravi |
* Optional RISC-V extensions:
|
10 |
|
|
** `A` - atomic memory access operations
|
11 |
|
|
** `C` - 16-bit compressed instructions
|
12 |
|
|
** `I` - integer base ISA (always enabled)
|
13 |
|
|
** `E` - embedded CPU version (reduced register file size)
|
14 |
|
|
** `M` - integer multiplication and division hardware
|
15 |
|
|
** `U` - less-privileged _user_ mode
|
16 |
63 |
zero_gravi |
** `Zbb` - basic bit-manipulation operations
|
17 |
61 |
zero_gravi |
** `Zfinx` - single-precision floating-point unit
|
18 |
|
|
** `Zicsr` - control and status register access (privileged architecture)
|
19 |
|
|
** `Zifencei` - instruction stream synchronization
|
20 |
|
|
** `Zmmul` - integer multiplication hardware
|
21 |
|
|
** `PMP` - physical memory protection
|
22 |
|
|
** `HPM` - hardware performance monitors
|
23 |
|
|
** `DB` - debug mode
|
24 |
60 |
zero_gravi |
* Compatible to the RISC-V user specifications and a subset of the RISC-V privileged architecture specifications – passes the official RISC-V Architecture Tests (v2+)
|
25 |
|
|
* Official RISC-V open-source architecture ID
|
26 |
|
|
* Standard RISC-V interrupts (_external_, _timer_, _software_) plus 16 _fast_ interrupts and 1 non-maskable interrupt
|
27 |
|
|
* Supports most of the traps from the RISC-V specifications (including bus access exceptions) and traps on all unimplemented/illegal/malformed instructions
|
28 |
|
|
* Optional physical memory configuration (PMP), compatible to the RISC-V specifications
|
29 |
|
|
* Optional hardware performance monitors (HPM) for application benchmarking
|
30 |
|
|
* Separated interfaces for instruction fetch and data access (merged into single bus via a bus switch for
|
31 |
|
|
the NEORV32 processor)
|
32 |
|
|
* little-endian byte order
|
33 |
|
|
* Configurable hardware reset
|
34 |
64 |
zero_gravi |
* No hardware support of unaligned data/instruction accesses – they will trigger an exception.
|
35 |
60 |
zero_gravi |
|
36 |
|
|
[NOTE]
|
37 |
|
|
It is recommended to use the **NEORV32 Processor** as default top instance even if you only want to use the actual
|
38 |
|
|
CPU. Simply disable all the processor-internal modules via the generics and you will get a "CPU
|
39 |
|
|
wrapper" that provides a minimal CPU environment and an external bus interface (like AXI4). This
|
40 |
|
|
setup also allows to further use the default bootloader and software framework. From this base you
|
41 |
|
|
can start building your own SoC. Of course you can also use the CPU in it’s true stand-alone mode.
|
42 |
|
|
|
43 |
|
|
[NOTE]
|
44 |
|
|
This documentation assumes the reader is familiar with the official RISC-V "User" and "Privileged Architecture" specifications.
|
45 |
|
|
|
46 |
|
|
<<<
|
47 |
|
|
// ####################################################################################################################
|
48 |
|
|
:sectnums:
|
49 |
|
|
=== Architecture
|
50 |
|
|
|
51 |
|
|
The NEORV32 CPU was designed from scratch based only on the official ISA / privileged architecture
|
52 |
|
|
specifications. The following figure shows the simplified architecture of the CPU.
|
53 |
|
|
|
54 |
|
|
image::neorv32_cpu.png[align=center]
|
55 |
|
|
|
56 |
|
|
The CPU uses a pipelined architecture with basically two main stages. The first stage (IF – instruction fetch)
|
57 |
|
|
is responsible for fetching new instruction data from memory via the fetch engine. The instruction data is
|
58 |
|
|
stored to a FIFO – the instruction prefetch buffer. The issue engine takes this data and assembles 32-bit
|
59 |
|
|
instruction words for the next pipeline stage. Compressed instructions – if enabled – are also decompressed
|
60 |
|
|
in this stage. The second stage (EX – execution) is responsible for actually executing the fetched instructions
|
61 |
|
|
via the execute engine.
|
62 |
|
|
|
63 |
|
|
These two pipeline stages are based on a multi-cycle processing engine. So the processing of each stage for a
|
64 |
|
|
certain operations can take several cycles. Since the IF and EX stages are decoupled via the instruction
|
65 |
|
|
prefetch buffer, both stages can operate in parallel and with overlapping operations. Hence, the optimal CPI
|
66 |
|
|
(cycles per instructions) is 2, but it can be significantly higher: For instance when executing loads/stores
|
67 |
|
|
multi-cycle operations like divisions or when the instruction fetch engine has to reload the prefetch buffers
|
68 |
|
|
due to a taken branch.
|
69 |
|
|
|
70 |
|
|
Basically, the NEORV32 CPU is somewhere between a classical pipelined architecture, where each stage
|
71 |
|
|
requires exactly one processing cycle (if not stalled) and a classical multi-cycle architecture, which executes
|
72 |
|
|
every single instruction in a series of consecutive micro-operations. The combination of these two classical
|
73 |
|
|
design paradigms allows an increased instruction execution in contrast to a pure multi-cycle approach (due to
|
74 |
|
|
the pipelined approach) at a reduced hardware footprint (due to the multi-cycle approach).
|
75 |
|
|
|
76 |
|
|
The CPU provides independent interfaces for instruction fetch and data access. These two bus interfaces are
|
77 |
|
|
merged into a single processor-internal bus via a bus switch. Hence, memory locations including peripheral
|
78 |
|
|
devices are mapped to a single 32-bit address space making the architecture a modified Von-Neumann
|
79 |
|
|
Architecture.
|
80 |
|
|
|
81 |
|
|
|
82 |
|
|
// ####################################################################################################################
|
83 |
|
|
:sectnums:
|
84 |
|
|
=== RISC-V Compatibility
|
85 |
|
|
|
86 |
|
|
The NEORV32 CPU passes the rv32_m/I, rv32_m/M, rv32_m/C, rv32_m/privilege, and
|
87 |
|
|
rv32_m/Zifencei tests of the official RISC-V Architecture Tests (GitHub). The port files for the
|
88 |
62 |
zero_gravi |
NEORV32 processor are located in the repository's `sw/isa-test` folder.
|
89 |
|
|
|
90 |
|
|
[NOTE]
|
91 |
|
|
See section https://stnolting.github.io/neorv32/ug/#_risc_v_architecture_test_framework[User Guide: RISC-V Architecture Test Framework]
|
92 |
60 |
zero_gravi |
for information how to run the tests on the NEORV32.
|
93 |
|
|
|
94 |
|
|
.**RISC-V `rv32_m/C` Tests**
|
95 |
|
|
...................................
|
96 |
|
|
Check cadd-01 ... OK
|
97 |
|
|
Check caddi-01 ... OK
|
98 |
|
|
Check caddi16sp-01 ... OK
|
99 |
|
|
Check caddi4spn-01 ... OK
|
100 |
|
|
Check cand-01 ... OK
|
101 |
|
|
Check candi-01 ... OK
|
102 |
|
|
Check cbeqz-01 ... OK
|
103 |
|
|
Check cbnez-01 ... OK
|
104 |
|
|
Check cebreak-01 ... OK
|
105 |
|
|
Check cj-01 ... OK
|
106 |
|
|
Check cjal-01 ... OK
|
107 |
|
|
Check cjalr-01 ... OK
|
108 |
|
|
Check cjr-01 ... OK
|
109 |
|
|
Check cli-01 ... OK
|
110 |
|
|
Check clui-01 ... OK
|
111 |
|
|
Check clw-01 ... OK
|
112 |
|
|
Check clwsp-01 ... OK
|
113 |
|
|
Check cmv-01 ... OK
|
114 |
|
|
Check cnop-01 ... OK
|
115 |
|
|
Check cor-01 ... OK
|
116 |
|
|
Check cslli-01 ... OK
|
117 |
|
|
Check csrai-01 ... OK
|
118 |
|
|
Check csrli-01 ... OK
|
119 |
|
|
Check csub-01 ... OK
|
120 |
|
|
Check csw-01 ... OK
|
121 |
|
|
Check cswsp-01 ... OK
|
122 |
|
|
Check cxor-01 ... OK
|
123 |
|
|
--------------------------------
|
124 |
|
|
OK: 27/27 RISCV_TARGET=neorv32 RISCV_DEVICE=C XLEN=32
|
125 |
|
|
...................................
|
126 |
|
|
|
127 |
|
|
.**RISC-V `rv32_m/I` Tests**
|
128 |
|
|
...................................
|
129 |
|
|
Check add-01 ... OK
|
130 |
|
|
Check addi-01 ... OK
|
131 |
|
|
Check and-01 ... OK
|
132 |
|
|
Check andi-01 ... OK
|
133 |
|
|
Check auipc-01 ... OK
|
134 |
|
|
Check beq-01 ... OK
|
135 |
|
|
Check bge-01 ... OK
|
136 |
|
|
Check bgeu-01 ... OK
|
137 |
|
|
Check blt-01 ... OK
|
138 |
|
|
Check bltu-01 ... OK
|
139 |
|
|
Check bne-01 ... OK
|
140 |
|
|
Check fence-01 ... OK
|
141 |
|
|
Check jal-01 ... OK
|
142 |
|
|
Check jalr-01 ... OK
|
143 |
|
|
Check lb-align-01 ... OK
|
144 |
|
|
Check lbu-align-01 ... OK
|
145 |
|
|
Check lh-align-01 ... OK
|
146 |
|
|
Check lhu-align-01 ... OK
|
147 |
|
|
Check lui-01 ... OK
|
148 |
|
|
Check lw-align-01 ... OK
|
149 |
|
|
Check or-01 ... OK
|
150 |
|
|
Check ori-01 ... OK
|
151 |
|
|
Check sb-align-01 ... OK
|
152 |
|
|
Check sh-align-01 ... OK
|
153 |
|
|
Check sll-01 ... OK
|
154 |
|
|
Check slli-01 ... OK
|
155 |
|
|
Check slt-01 ... OK
|
156 |
|
|
Check slti-01 ... OK
|
157 |
|
|
Check sltiu-01 ... OK
|
158 |
|
|
Check sltu-01 ... OK
|
159 |
|
|
Check sra-01 ... OK
|
160 |
|
|
Check srai-01 ... OK
|
161 |
|
|
Check srl-01 ... OK
|
162 |
|
|
Check srli-01 ... OK
|
163 |
|
|
Check sub-01 ... OK
|
164 |
|
|
Check sw-align-01 ... OK
|
165 |
|
|
Check xor-01 ... OK
|
166 |
|
|
Check xori-01 ... OK
|
167 |
|
|
--------------------------------
|
168 |
|
|
OK: 38/38 RISCV_TARGET=neorv32 RISCV_DEVICE=I XLEN=32
|
169 |
|
|
...................................
|
170 |
|
|
|
171 |
|
|
.**RISC-V `rv32_m/M` Tests**
|
172 |
|
|
...................................
|
173 |
|
|
Check div-01 ... OK
|
174 |
|
|
Check divu-01 ... OK
|
175 |
|
|
Check mul-01 ... OK
|
176 |
|
|
Check mulh-01 ... OK
|
177 |
|
|
Check mulhsu-01 ... OK
|
178 |
|
|
Check mulhu-01 ... OK
|
179 |
|
|
Check rem-01 ... OK
|
180 |
|
|
Check remu-01 ... OK
|
181 |
|
|
--------------------------------
|
182 |
|
|
OK: 8/8 RISCV_TARGET=neorv32 RISCV_DEVICE=M XLEN=32
|
183 |
|
|
...................................
|
184 |
|
|
|
185 |
|
|
.**RISC-V `rv32_m/privilege` Tests**
|
186 |
|
|
...................................
|
187 |
|
|
Check ebreak ... OK
|
188 |
|
|
Check ecall ... OK
|
189 |
|
|
Check misalign-beq-01 ... OK
|
190 |
|
|
Check misalign-bge-01 ... OK
|
191 |
|
|
Check misalign-bgeu-01 ... OK
|
192 |
|
|
Check misalign-blt-01 ... OK
|
193 |
|
|
Check misalign-bltu-01 ... OK
|
194 |
|
|
Check misalign-bne-01 ... OK
|
195 |
|
|
Check misalign-jal-01 ... OK
|
196 |
|
|
Check misalign-lh-01 ... OK
|
197 |
|
|
Check misalign-lhu-01 ... OK
|
198 |
|
|
Check misalign-lw-01 ... OK
|
199 |
|
|
Check misalign-sh-01 ... OK
|
200 |
|
|
Check misalign-sw-01 ... OK
|
201 |
|
|
Check misalign1-jalr-01 ... OK
|
202 |
|
|
Check misalign2-jalr-01 ... OK
|
203 |
|
|
--------------------------------
|
204 |
|
|
OK: 16/16 RISCV_TARGET=neorv32 RISCV_DEVICE=privilege XLEN=32
|
205 |
|
|
...................................
|
206 |
|
|
|
207 |
|
|
.**RISC-V `rv32_m/Zifencei` Tests**
|
208 |
|
|
...................................
|
209 |
|
|
Check Fencei ... OK
|
210 |
|
|
--------------------------------
|
211 |
|
|
OK: 1/1 RISCV_TARGET=neorv32 RISCV_DEVICE=Zifencei XLEN=32
|
212 |
|
|
...................................
|
213 |
|
|
|
214 |
|
|
|
215 |
|
|
<<<
|
216 |
|
|
:sectnums:
|
217 |
|
|
==== RISC-V Incompatibility Issues and Limitations
|
218 |
|
|
|
219 |
64 |
zero_gravi |
This list shows the currently identified issues regarding full RISC-V-compatibility. More specific information
|
220 |
60 |
zero_gravi |
can be found in section <<_instruction_sets_and_extensions>>.
|
221 |
|
|
|
222 |
64 |
zero_gravi |
.Hardwired R/W CSRs
|
223 |
60 |
zero_gravi |
[IMPORTANT]
|
224 |
64 |
zero_gravi |
The `misa`, `mip` and `mtval` CSRs in the NEORV32 are _read-only_.
|
225 |
|
|
Any write access to it (in machine mode) to them are ignored and will _not_ cause any exceptions or side-effects.
|
226 |
60 |
zero_gravi |
|
227 |
64 |
zero_gravi |
.Physical memory protection
|
228 |
60 |
zero_gravi |
[IMPORTANT]
|
229 |
|
|
The physical memory protection (see section <<_machine_physical_memory_protection>>)
|
230 |
|
|
only supports the modes _OFF_ and _NAPOT_ yet and a minimal granularity of 8 bytes per region.
|
231 |
|
|
|
232 |
64 |
zero_gravi |
.Atomic memory operations
|
233 |
60 |
zero_gravi |
[IMPORTANT]
|
234 |
64 |
zero_gravi |
The `A` CPU extension only implements the `lr.w` and `sc.w` instructions yet.
|
235 |
|
|
However, these instructions are sufficient to emulate all further atomic memory operations.
|
236 |
60 |
zero_gravi |
|
237 |
64 |
zero_gravi |
.Instruction Misalignment
|
238 |
|
|
[NOTE]
|
239 |
|
|
This is not a real RISC-V incompatibility, but something that might not be clear when studying the RISC-V privileged
|
240 |
|
|
architecture specifications: for 32-bit only instructions (no `C` extension) the misaligned instruction exception
|
241 |
|
|
is raised if bit 1 of the access address is set (i.e. not on 32-bit boundary). If the `C` extension is implemented
|
242 |
|
|
there will be no misaligned instruction exceptions _at all_.
|
243 |
|
|
In both cases bit 0 of the program counter and all related registers is hardwired to zero.
|
244 |
60 |
zero_gravi |
|
245 |
64 |
zero_gravi |
|
246 |
60 |
zero_gravi |
<<<
|
247 |
|
|
// ####################################################################################################################
|
248 |
|
|
:sectnums:
|
249 |
|
|
=== CPU Top Entity - Signals
|
250 |
|
|
|
251 |
|
|
The following table shows all interface signals of the CPU top entity `rtl/core/neorv32_cpu.vhd`. The
|
252 |
|
|
type of all signals is _std_ulogic_ or _std_ulogic_vector_, respectively. The "Dir." column shows the signal
|
253 |
|
|
direction seen from the CPU.
|
254 |
|
|
|
255 |
|
|
.NEORV32 CPU top entity signals
|
256 |
|
|
[cols="<2,^1,^1,<6"]
|
257 |
|
|
[options="header", grid="rows"]
|
258 |
|
|
|=======================
|
259 |
|
|
| Signal | Width | Dir. | Function
|
260 |
|
|
4+^| **Global Signals**
|
261 |
|
|
| `clk_i` | 1 | in | global clock line, all registers triggering on rising edge
|
262 |
|
|
| `rstn_i` | 1 | in | global reset, low-active
|
263 |
|
|
| `sleep_o` | 1 | out | CPU is in sleep mode when set
|
264 |
|
|
4+^| **Instruction Bus Interface (<<_bus_interface>>)**
|
265 |
|
|
| `i_bus_addr_o` | 32 | out | destination address
|
266 |
|
|
| `i_bus_rdata_i` | 32 | in | read data
|
267 |
|
|
| `i_bus_wdata_o` | 32 | out | write data (always zero)
|
268 |
|
|
| `i_bus_ben_o` | 4 | out | byte enable
|
269 |
|
|
| `i_bus_we_o` | 1 | out | write transaction (always zero)
|
270 |
|
|
| `i_bus_re_o` | 1 | out | read transaction
|
271 |
|
|
| `i_bus_lock_o` | 1 | out | exclusive access request (always zero)
|
272 |
|
|
| `i_bus_ack_i` | 1 | in | bus transfer acknowledge from accessed peripheral
|
273 |
|
|
| `i_bus_err_i` | 1 | in | bus transfer terminate from accessed peripheral
|
274 |
|
|
| `i_bus_fence_o` | 1 | out | indicates an executed _fence.i_ instruction
|
275 |
|
|
| `i_bus_priv_o` | 2 | out | current CPU privilege level
|
276 |
|
|
4+^| **Data Bus Interface (<<_bus_interface>>)**
|
277 |
|
|
| `d_bus_addr_o` | 32 | out | destination address
|
278 |
|
|
| `d_bus_rdata_i` | 32 | in | read data
|
279 |
|
|
| `d_bus_wdata_o` | 32 | out | write data
|
280 |
|
|
| `d_bus_ben_o` | 4 | out | byte enable
|
281 |
|
|
| `d_bus_we_o` | 1 | out | write transaction
|
282 |
|
|
| `d_bus_re_o` | 1 | out | read transaction
|
283 |
|
|
| `d_bus_lock_o` | 1 | out | exclusive access request
|
284 |
|
|
| `d_bus_ack_i` | 1 | in | bus transfer acknowledge from accessed peripheral
|
285 |
|
|
| `d_bus_err_i` | 1 | in | bus transfer terminate from accessed peripheral
|
286 |
|
|
| `d_bus_fence_o` | 1 | out | indicates an executed _fence_ instruction
|
287 |
|
|
| `d_bus_priv_o` | 2 | out | current CPU privilege level
|
288 |
|
|
4+^| **System Time (see <<_timeh>> CSR)**
|
289 |
|
|
| `time_i` | 64 | in | system time input (from MTIME)
|
290 |
|
|
4+^| **Interrupts, RISC-V-compatible (<<_traps_exceptions_and_interrupts>>)**
|
291 |
|
|
| `msw_irq_i` | 1 | in | RISC-V machine software interrupt
|
292 |
|
|
| `mext_irq_i` | 1 | in | RISC-V machine external interrupt
|
293 |
|
|
| `mtime_irq_i` | 1 | in | RISC-V machine timer interrupt
|
294 |
|
|
4+^| **Fast Interrupts, NEORV32-specific (<<_traps_exceptions_and_interrupts>>)**
|
295 |
|
|
| `firq_i` | 16 | in | fast interrupt request signals
|
296 |
|
|
4+^| **Enter Debug Mode Request (<<_on_chip_debugger_ocd>>)**
|
297 |
|
|
| `db_halt_req_i` | 1 | in | request CPU to halt and enter debug mode
|
298 |
|
|
|=======================
|
299 |
|
|
|
300 |
|
|
<<<
|
301 |
|
|
// ####################################################################################################################
|
302 |
|
|
:sectnums:
|
303 |
|
|
=== CPU Top Entity - Generics
|
304 |
|
|
|
305 |
|
|
Most of the CPU configuration generics are a subset of the actual Processor configuration generics (see section <<_processor_top_entity_generics>>).
|
306 |
|
|
and are not listed here. However, the CPU provides some _specific_ generics that are used to configure the CPU for the
|
307 |
|
|
NEORV32 processor setup. These generics are assigned by the processor setup only and are not available for user defined configuration.
|
308 |
|
|
The _specific_ generics are listed below.
|
309 |
|
|
|
310 |
|
|
[cols="4,4,2"]
|
311 |
|
|
[frame="all",grid="none"]
|
312 |
|
|
|======
|
313 |
|
|
| **CPU_BOOT_ADDR** | _std_ulogic_vector(31 downto 0)_ | 0x00000000
|
314 |
|
|
3+| This address defines the reset address at which the CPU starts fetching instructions after reset. In terms of the NEORV32 processor, this
|
315 |
|
|
generic is configured with the base address of the bootloader ROM (default) or with the base address of the processor-internal instruction
|
316 |
61 |
zero_gravi |
memory (IMEM) if the bootloader is disabled (_INT_BOOTLOADER_EN_ = _false_). See section <<_address_space>> for more information.
|
317 |
60 |
zero_gravi |
|======
|
318 |
|
|
|
319 |
|
|
[cols="4,4,2"]
|
320 |
|
|
[frame="all",grid="none"]
|
321 |
|
|
|======
|
322 |
|
|
| **CPU_DEBUG_ADDR** | _std_ulogic_vector(31 downto 0)_ | 0x00000000
|
323 |
|
|
3+| This address defines the entry address for the "execution based" on-chip debugger. By default, this generic is configured with the base address
|
324 |
|
|
of the debugger memory. See section <<_on_chip_debugger_ocd>> for more information.
|
325 |
|
|
|======
|
326 |
|
|
|
327 |
|
|
[cols="4,4,2"]
|
328 |
|
|
[frame="all",grid="none"]
|
329 |
|
|
|======
|
330 |
|
|
| **CPU_EXTENSION_RISCV_DEBUG** | _boolean_ | false
|
331 |
|
|
3+| Implement RISC-V-compatible "debug" CPU operation mode. See section <<_cpu_debug_mode>> for more information.
|
332 |
|
|
|======
|
333 |
|
|
|
334 |
|
|
|
335 |
|
|
<<<
|
336 |
|
|
// ####################################################################################################################
|
337 |
|
|
:sectnums:
|
338 |
|
|
=== Instruction Sets and Extensions
|
339 |
|
|
|
340 |
|
|
The NEORV32 is an RISC-V `rv32i` architecture that provides several optional RISC-V CPU and ISA
|
341 |
|
|
(instruction set architecture) extensions. For more information regarding the RISC-V ISA extensions please
|
342 |
|
|
see the The _RISC-V Instruction Set Manual – Volume I: Unprivileged ISA_ and _The RISC-V Instruction Set Manual
|
343 |
|
|
Volume II: Privileged Architecture_, which are available in the projects `docs/references` folder.
|
344 |
|
|
|
345 |
|
|
[TIP]
|
346 |
63 |
zero_gravi |
The CPU can discover available ISA extensions via the <<_misa>> CSR and the
|
347 |
64 |
zero_gravi |
`CPU` <<_system_configuration_information_memory_sysinfo, SYSINFO>> register
|
348 |
63 |
zero_gravi |
or by executing an instruction and checking for an _illegal instruction exception_.
|
349 |
60 |
zero_gravi |
|
350 |
63 |
zero_gravi |
[NOTE]
|
351 |
|
|
Executing an instruction from an extension that is not implemented or not enabled (for example via the according
|
352 |
|
|
top entity generic) will raise an _illegal instruction_ exception.
|
353 |
60 |
zero_gravi |
|
354 |
63 |
zero_gravi |
|
355 |
60 |
zero_gravi |
==== **`A`** - Atomic Memory Access
|
356 |
|
|
|
357 |
|
|
Atomic memory access instructions (for implementing semaphores and mutexes) are available when the
|
358 |
|
|
`CPU_EXTENSION_RISCV_A` configuration generic is _true_. In this case the following additional instructions
|
359 |
|
|
are available:
|
360 |
|
|
|
361 |
|
|
* `lr.w`: load-reservate
|
362 |
|
|
* `sc.w`: store-conditional
|
363 |
|
|
|
364 |
|
|
[NOTE]
|
365 |
|
|
Even though only `lr.w` and `sc.w` instructions are implemented yet, all further atomic operations
|
366 |
|
|
(load-modify-write instruction) can be emulated using these two instruction. Furthermore, the
|
367 |
|
|
instruction’s ordering flags (`aq` and `lr`) are ignored by the CPU hardware. Using any other (not yet
|
368 |
|
|
implemented) AMO (atomic memory operation) will trigger an illegal instruction exception.
|
369 |
|
|
|
370 |
|
|
[NOTE]
|
371 |
|
|
The atomic instructions have special requirements for memory system / bus interconnect. More
|
372 |
|
|
information can be found in sections <<_bus_interface>> and <<_processor_external_memory_interface_wishbone_axi4_lite>>, respectively.
|
373 |
|
|
|
374 |
|
|
|
375 |
|
|
==== **`C`** - Compressed Instructions
|
376 |
|
|
|
377 |
|
|
Compressed 16-bit instructions are available when the `CPU_EXTENSION_RISCV_C` configuration generic is
|
378 |
|
|
_true_. In this case the following instructions are available:
|
379 |
|
|
|
380 |
|
|
* `c.addi4spn`, `c.lw`, `c.sw`, `c.nop`, `c.addi`, `c.jal`, `c.li`, `c.addi16sp`, `c.lui`, `c.srli`, `c.srai` `c.andi`, `c.sub`,
|
381 |
|
|
`c.xor`, `c.or`, `c.and`, `c.j`, `c.beqz`, `c.bnez`, `c.slli`, `c.lwsp`, `c.jr`, `c.mv`, `c.ebreak`, `c.jalr`, `c.add`, `c.swsp`
|
382 |
|
|
|
383 |
|
|
[NOTE]
|
384 |
|
|
When the compressed instructions extension is enabled, branches to an _unaligned_ and _uncompressed_ address require
|
385 |
|
|
an additional instruction fetch to load the required second half-word of that instruction. The performance can be increased
|
386 |
|
|
again by forcing a 32-bit alignment of branch target addresses. By default, this is enforced via the GCC `-falign-functions=4`,
|
387 |
|
|
`-falign-labels=4`, `-falign-loops=4` and `-falign-jumps=4` compile flags (via the makefile).
|
388 |
|
|
|
389 |
|
|
|
390 |
|
|
==== **`E`** - Embedded CPU
|
391 |
|
|
|
392 |
|
|
The embedded CPU extensions reduces the size of the general purpose register file from 32 entries to 16 entries to reduce hardware
|
393 |
|
|
requirements. This extensions is enabled when the `CPU_EXTENSION_RISCV_E` configuration generic is _true_. Accesses to registers beyond
|
394 |
|
|
`x15` will raise and _illegal instruction exception_.
|
395 |
|
|
|
396 |
63 |
zero_gravi |
[IMPORTANT]
|
397 |
|
|
Due to the reduced register file size an alternate toolchain ABI (**`ilp32e`**) is required.
|
398 |
60 |
zero_gravi |
|
399 |
|
|
|
400 |
|
|
==== **`I`** - Base Integer ISA
|
401 |
|
|
The CPU always supports the complete `rv32i` base integer instruction set. This base set is always enabled
|
402 |
|
|
regardless of the setting of the remaining exceptions. The base instruction set includes the following
|
403 |
|
|
instructions:
|
404 |
|
|
|
405 |
|
|
* immediates: `lui`, `auipc`
|
406 |
|
|
* jumps: `jal`, `jalr`
|
407 |
|
|
* branches: `beq`, `bne`, `blt`, `bge`, `bltu`, `bgeu`
|
408 |
|
|
* memory: `lb`, `lh`, `lw`, `lbu`, `lhu`, `sb`, `sh`, `sw`
|
409 |
|
|
* alu: `addi`, `slti`, `sltiu`, `xori`, `ori`, `andi`, `slli`, `srli`, `srai`, `add`, `sub`, `sll`, `slt`, `sltu`, `xor`, `srl`, `sra`, `or`, `and`
|
410 |
|
|
* environment: `ecall`, `ebreak`, `fence`
|
411 |
|
|
|
412 |
|
|
[NOTE]
|
413 |
61 |
zero_gravi |
In order to keep the hardware footprint low, the CPU's shift unit uses a bit-serial serial approach. Hence, shift operations
|
414 |
|
|
take up to 32 cycles (plus overhead) depending on the actual shift amount. Alternatively, the shift operations can be processed
|
415 |
|
|
completely in parallels by a fast (but large) barrel shifter when the `FAST_SHIFT_EN` generic is _true_. In that case, shift operations
|
416 |
62 |
zero_gravi |
complete within 2 cycles (plus overhead) regardless of the actual shift amount.
|
417 |
60 |
zero_gravi |
|
418 |
|
|
[NOTE]
|
419 |
|
|
Internally, the `fence` instruction does not perform any operation inside the CPU. It only sets the
|
420 |
|
|
top’s `d_bus_fence_o` signal high for one cycle to inform the memory system a `fence` instruction has been
|
421 |
|
|
executed. Any flags within the `fence` instruction word are ignore by the hardware.
|
422 |
|
|
|
423 |
|
|
|
424 |
|
|
==== **`M`** - Integer Multiplication and Division
|
425 |
|
|
|
426 |
|
|
Hardware-accelerated integer multiplication and division instructions are available when the
|
427 |
|
|
`CPU_EXTENSION_RISCV_M` configuration generic is _true_. In this case the following instructions are
|
428 |
|
|
available:
|
429 |
|
|
|
430 |
61 |
zero_gravi |
* multiplication: `mul`, `mulh`, `mulhsu`, `mulhu`
|
431 |
|
|
* division: `div`, `divu`, `rem`, `remu`
|
432 |
60 |
zero_gravi |
|
433 |
|
|
[NOTE]
|
434 |
|
|
By default, multiplication and division operations are executed in a bit-serial approach.
|
435 |
|
|
Alternatively, the multiplier core can be implemented using DSP blocks if the `FAST_MUL_EN`
|
436 |
|
|
generic is _true_ allowing faster execution. Multiplications and divisions
|
437 |
|
|
always require a fixed amount of cycles to complete - regardless of the input operands.
|
438 |
|
|
|
439 |
|
|
|
440 |
61 |
zero_gravi |
==== **`Zmmul`** - Integer Multiplication
|
441 |
|
|
|
442 |
|
|
This is a _sub-extension_ of the `M` ISA extension. It implements the multiplication-only operations
|
443 |
|
|
of the `M` extensions and is intended for small scale applications, that require hardware-based
|
444 |
|
|
integer multiplications but not hardware-based divisions, which will be computed entirely in software.
|
445 |
|
|
This extension requires only ~50% of the hardware utilization of the `M` extension.
|
446 |
|
|
|
447 |
|
|
* multiplication: `mul`, `mulh`, `mulhsu`, `mulhu`
|
448 |
|
|
|
449 |
63 |
zero_gravi |
If `Zmmul` is enabled, executing any division instruction from the `M` ISA extension (`div`, `divu`, `rem`, `remu`)
|
450 |
|
|
will raise an _illegal instruction exception_.
|
451 |
61 |
zero_gravi |
|
452 |
63 |
zero_gravi |
Note that `M` and `Zmmul` extensions _cannot_ be enabled at the same time.
|
453 |
61 |
zero_gravi |
|
454 |
|
|
[TIP]
|
455 |
|
|
If your RISC-V GCC toolchain does not (yet) support the `_Zmmul` ISA extensions, it can be "emulated"
|
456 |
|
|
using a `rv32im` machine architecture and setting the `-mno-div` compiler flag
|
457 |
|
|
(example `$ make MARCH=-march=rv32im USER_FLAGS+=-mno-div clean_all exe`).
|
458 |
|
|
|
459 |
|
|
|
460 |
60 |
zero_gravi |
==== **`U`** - Less-Privileged User Mode
|
461 |
|
|
|
462 |
63 |
zero_gravi |
Adds the less-privileged _user mode_ if the `CPU_EXTENSION_RISCV_U` configuration generic is _true_. For
|
463 |
60 |
zero_gravi |
instance, use-level code cannot access machine-mode CSRs. Furthermore, access to the address space (like
|
464 |
|
|
peripheral/IO devices) can be limited via the physical memory protection (_PMP_) unit for code running in user mode.
|
465 |
|
|
|
466 |
|
|
|
467 |
|
|
==== **`X`** - NEORV32-Specific (Custom) Extensions
|
468 |
|
|
|
469 |
|
|
The NEORV32-specific extensions are always enabled and are indicated by the set `X` bit in the `misa` CSR.
|
470 |
|
|
|
471 |
63 |
zero_gravi |
The most important points of the NEORV32-specific extensions are:
|
472 |
|
|
* The CPU provides 16 _fast interrupt_ interrupts (`FIRQ)`, which are controlled via custom bits in the `mie`
|
473 |
64 |
zero_gravi |
and `mip` CSR. This extension is mapped to _reserved_ CSR bits, that are available for custom use (according to the
|
474 |
60 |
zero_gravi |
RISC-V specs). Also, custom trap codes for `mcause` are implemented.
|
475 |
63 |
zero_gravi |
* All undefined/unimplemented/malformed/illegal instructions do raise an illegal instruction exception (see <<_full_virtualization>>).
|
476 |
60 |
zero_gravi |
|
477 |
|
|
|
478 |
63 |
zero_gravi |
==== **`Zfinx`** Single-Precision Floating-Point Operations
|
479 |
60 |
zero_gravi |
|
480 |
63 |
zero_gravi |
[WARNING]
|
481 |
|
|
The NEORV32 `Zfinx` extension is specification-compliant and operational but still _experimental_.
|
482 |
60 |
zero_gravi |
|
483 |
|
|
The `Zfinx` floating-point extension is an alternative of the `F` floating-point instruction that also uses the
|
484 |
|
|
integer register file `x` to store and operate on floating-point data (hence, `F-in-x`). Since not dedicated floating-point `f`
|
485 |
|
|
register file exists, the `Zfinx` extension requires less hardware resources and features faster context changes.
|
486 |
|
|
This also implies that there are NO dedicated `f` register file related load/store or move instructions. The
|
487 |
|
|
official RISC-V specifications can be found here: https://github.com/riscv/riscv-zfinx
|
488 |
|
|
|
489 |
|
|
The NEORV32 floating-point unit used by the `Zfinx` extension is compatible to the _IEEE-754_ specifications.
|
490 |
|
|
|
491 |
|
|
The `Zfinx` extensions only supports single-precision (`.s` suffix) yet (so it is a direct alternative to the `F`
|
492 |
|
|
extension). The `Zfinx` extension is implemented when the `CPU_EXTENSION_RISCV_Zfinx` configuration
|
493 |
|
|
generic is _true_. In this case the following instructions and CSRs are available:
|
494 |
|
|
|
495 |
|
|
* conversion: `fcvt.s.w`, `fcvt.s.wu`, `fcvt.w.s`, `fcvt.wu.s`
|
496 |
|
|
* comparison: `fmin.s`, `fmax.s`, `feq.s`, `flt.s`, `fle.s`
|
497 |
|
|
* computational: `fadd.s`, `fsub.s`, `fmul.s`
|
498 |
|
|
* sign-injection: `fsgnj.s`, `fsgnjn.s`, `fsgnjx.s`
|
499 |
|
|
* number classification: `fclass.s`
|
500 |
|
|
|
501 |
|
|
* additional CSRs: `fcsr`, `frm`, `fflags`
|
502 |
|
|
|
503 |
|
|
[WARNING]
|
504 |
|
|
Fused multiply-add instructions `f[n]m[add/sub].s` are not supported!
|
505 |
|
|
Division `fdiv.s` and square root `fsqrt.s` instructions are not supported yet!
|
506 |
|
|
|
507 |
|
|
[WARNING]
|
508 |
|
|
Subnormal numbers (also "de-normalized" numbers) are not supported by the NEORV32 FPU.
|
509 |
|
|
Subnormal numbers (exponent = 0) are _flushed to zero_ (setting them to +/- 0) before entering the
|
510 |
|
|
FPU's processing core. If a computational instruction (like `fmul.s`) generates a subnormal result, the
|
511 |
|
|
result is also flushed to zero during normalization.
|
512 |
|
|
|
513 |
|
|
[WARNING]
|
514 |
|
|
The `Zfinx` extension is not yet officially ratified, but is expected to stay unchanged. There is no
|
515 |
|
|
software support for the `Zfinx` extension in the upstream GCC RISC-V port yet. However, an
|
516 |
|
|
intrinsic library is provided to utilize the provided `Zfinx` floating-point extension from C-language
|
517 |
|
|
code (see `sw/example/floating_point_test`).
|
518 |
|
|
|
519 |
63 |
zero_gravi |
|
520 |
|
|
==== **`Zbb`** Basic Bit-Manipulation Operations
|
521 |
|
|
|
522 |
|
|
[WARNING]
|
523 |
|
|
The NEORV32 `Zbb` extension is specification-compliant and operational but still _experimental_.
|
524 |
|
|
|
525 |
|
|
The `Zbb` extension implements the _basic_ sub-set of the RISC-V bit-manipulation extensions `B`.
|
526 |
|
|
The official RISC-V specifications can be found here: https://github.com/riscv/riscv-bitmanip
|
527 |
|
|
|
528 |
|
|
The `Zbb` extension is implemented when the `CPU_EXTENSION_RISCV_Zbb` configuration
|
529 |
|
|
generic is _true_. In this case the following instructions are available:
|
530 |
|
|
|
531 |
|
|
* `andn`, `orn`, `xnor`
|
532 |
|
|
* `clz`, `ctz`, `cpop`
|
533 |
|
|
* `max`, `maxu`, `min`, `minu`
|
534 |
|
|
* `sext.b`, `sext.h`, `zext.h`
|
535 |
|
|
* `rol`, `ror`, `rori`
|
536 |
|
|
* `orc.b`, `rev8`
|
537 |
|
|
|
538 |
|
|
[TIP]
|
539 |
|
|
By default, the bit-manipulation unit uses an _iterative_ approach to compute shift-related operations
|
540 |
|
|
like `clz` and `rol`. To increase performance (at the cost of additional hardware resources) the
|
541 |
|
|
<<_fast_shift_en>> generic can be enabled to implement full-parallel logic (like barrel shifters) for all
|
542 |
|
|
shift-related `Zbb` instructions.
|
543 |
|
|
|
544 |
62 |
zero_gravi |
[IMPORTANT]
|
545 |
63 |
zero_gravi |
The `Zbb` extension is frozen but not officially ratified yet. There is no
|
546 |
|
|
software support for this extension in the upstream GCC RISC-V port yet. However, an
|
547 |
|
|
intrinsic library is provided to utilize the provided `Zbb` extension from C-language
|
548 |
|
|
code (see `sw/example/bitmanip_test`).
|
549 |
60 |
zero_gravi |
|
550 |
62 |
zero_gravi |
|
551 |
60 |
zero_gravi |
==== **`Zicsr`** Control and Status Register Access / Privileged Architecture
|
552 |
|
|
|
553 |
|
|
The CSR access instructions as well as the exception and interrupt system (= the privileged architecture) is implemented when the
|
554 |
|
|
`CPU_EXTENSION_RISCV_Zicsr` configuration generic is _true_. In this case the following instructions are
|
555 |
|
|
available:
|
556 |
|
|
|
557 |
|
|
* CSR access: `csrrw`, `csrrs`, `csrrc`, `csrrwi`, `csrrsi`, `csrrci`
|
558 |
|
|
* environment: `mret`, `wfi`
|
559 |
|
|
|
560 |
|
|
[WARNING]
|
561 |
|
|
If the `Zicsr` extension is disabled the CPU does not provide any kind of interrupt or exception
|
562 |
|
|
support at all. In order to provide the full spectrum of functions and to allow a secure executions
|
563 |
|
|
environment, the `Zicsr` extension should always be enabled.
|
564 |
|
|
|
565 |
|
|
[NOTE]
|
566 |
|
|
The "wait for interrupt instruction" `wfi` works like a sleep command. When executed, the CPU is
|
567 |
|
|
halted until a valid interrupt request occurs. To wake up again, the according interrupt source has to
|
568 |
|
|
be enabled via the `mie` CSR and the global interrupt enable flag in `mstatus` has to be set.
|
569 |
|
|
|
570 |
62 |
zero_gravi |
[IMPORTANT]
|
571 |
|
|
The `wfi` instruction will raise an illegal instruction exception when executed outside of machine-mode
|
572 |
|
|
and <<_mstatus>> bit `TW` (timeout wait) is set.
|
573 |
60 |
zero_gravi |
|
574 |
62 |
zero_gravi |
|
575 |
60 |
zero_gravi |
==== **`Zifencei`** Instruction Stream Synchronization
|
576 |
|
|
|
577 |
|
|
The `Zifencei` CPU extension is implemented if the `CPU_EXTENSION_RISCV_Zifencei` configuration
|
578 |
|
|
generic is _true_. It allows manual synchronization of the instruction stream via the following instruction:
|
579 |
|
|
|
580 |
|
|
* `fence.i`
|
581 |
|
|
|
582 |
|
|
[NOTE]
|
583 |
|
|
The `fence.i` instruction resets the CPU's internal instruction fetch engine and flushes the prefetch buffer.
|
584 |
64 |
zero_gravi |
This allows a clean re-fetch of modified instructions from memory. Also, the top's `i_bus_fencei_o` signal is set
|
585 |
|
|
high for one cycle to inform the memory system (like the i-cache to perform a flush/reload.
|
586 |
|
|
Any additional flags within the `fence.i` instruction word are ignore by the hardware.
|
587 |
60 |
zero_gravi |
|
588 |
|
|
|
589 |
|
|
==== **`PMP`** Physical Memory Protection
|
590 |
|
|
|
591 |
|
|
The NEORV32 physical memory protection (PMP) is compatible to the PMP specified by the RISC-V specs.
|
592 |
|
|
The CPU PMP only supports _NAPOT_ mode yet and a minimal region size (granularity) of 8 bytes. Larger minimal sizes can be configured
|
593 |
|
|
via the top `PMP_MIN_GRANULARITY` generic to reduce hardware requirements. The physical memory protection system is implemented when the
|
594 |
|
|
`PMP_NUM_REGIONS` configuration generic is >0. In this case the following additional CSRs are available:
|
595 |
|
|
|
596 |
|
|
* `pmpcfg*` (0..15, depending on configuration): PMP configuration registers
|
597 |
|
|
* `pmpaddr*` (0..63, depending on configuration): PMP address registers
|
598 |
|
|
|
599 |
|
|
See section <<_machine_physical_memory_protection>> for more information regarding the PMP CSRs.
|
600 |
|
|
|
601 |
|
|
**Configuration**
|
602 |
|
|
|
603 |
|
|
The actual number of regions and the minimal region granularity are defined via the top entity
|
604 |
|
|
`PMP_MIN_GRANULARITY` and `PMP_NUM_REGIONS` generics. `PMP_MIN_GRANULARITY` defines the minimal available
|
605 |
|
|
granularity of each region in bytes. `PMP_NUM_REGIONS` defines the total number of implemented regions and thus, the
|
606 |
|
|
number of available `pmpcfg*` and `pmpaddr*` CSRs.
|
607 |
|
|
|
608 |
|
|
When implementing more PMP regions that a _certain critical limit_ *an additional register stage
|
609 |
|
|
is automatically inserted* into the CPU's memory interfaces to reduce critical path length. Unfortunately, this will also
|
610 |
|
|
increase the latency of instruction fetches and data access by +1 cycle.
|
611 |
|
|
|
612 |
|
|
The critical limit can be adapted for custom use by a constant from the main VHDL package file
|
613 |
|
|
(`rtl/core/neorv32_package.vhd`). The default value is 8:
|
614 |
|
|
|
615 |
|
|
[source,vhdl]
|
616 |
|
|
----
|
617 |
|
|
-- "critical" number of PMP regions --
|
618 |
|
|
constant pmp_num_regions_critical_c : natural := 8;
|
619 |
|
|
----
|
620 |
|
|
|
621 |
|
|
**Operation**
|
622 |
|
|
|
623 |
|
|
Any memory access address (from the CPU's instruction fetch or data access interface) is tested if it is accessing any
|
624 |
|
|
of the specified (configured via `pmpaddr*` and enabled via `pmpcfg*`) PMP regions. If an
|
625 |
|
|
address accesses one of these regions, the configured access rights (attributes in `pmpcfg*`) are checked:
|
626 |
|
|
|
627 |
|
|
* a write access (store) will fail if no write attribute is set
|
628 |
|
|
* a read access (load) will fail if no read attribute is set
|
629 |
|
|
* an instruction fetch access will fail if no execute attribute is set
|
630 |
|
|
|
631 |
|
|
If an access to a protected region does not have the according access rights (attributes) it will raise the according
|
632 |
|
|
_instruction/load/store access fault exception_.
|
633 |
|
|
|
634 |
|
|
By default, all PMP checks are enforced for user-level programs only. If you wish to enforce the physical
|
635 |
|
|
memory protection also for machine-level programs you need to active the _locked bit_ in the according
|
636 |
|
|
`pmpcfg*` configuration.
|
637 |
|
|
|
638 |
|
|
[IMPORTANT]
|
639 |
|
|
After updating the address configuration registers `pmpaddr*` the system requires up to 33 cycles for
|
640 |
|
|
internal (iterative) computations before the configuration becomes valid.
|
641 |
|
|
|
642 |
|
|
[NOTE]
|
643 |
|
|
For more information regarding RISC-V physical memory protection see the official _The RISC-V
|
644 |
|
|
Instruction Set Manual – Volume II: Privileged Architecture_ specifications.
|
645 |
|
|
|
646 |
|
|
|
647 |
|
|
==== **`HPM`** Hardware Performance Monitors
|
648 |
|
|
|
649 |
|
|
In additions to the mandatory cycles (`[m]cycle[h]`) and instruction (`[m]instret[h]`) counters the NEORV32 CPU provides
|
650 |
|
|
up to 29 hardware performance monitors (HPM 3..31), which can be used to benchmark applications. Each HPM consists of an
|
651 |
|
|
N-bit wide counter (split in a high-word 32-bit CSR and a low-word 32-bit CSR), where N is defined via the top's
|
652 |
|
|
`HPM_CNT_WIDTH` generic (0..64-bit), and a corresponding event configuration CSR. The event configuration
|
653 |
|
|
CSR defines the architectural events that lead to an increment of the associated HPM counter.
|
654 |
|
|
|
655 |
|
|
The cycle, time and instructions-retired counters (`[m]cycle[h]`, `time[h]`, `[m]instret[h]`) are
|
656 |
62 |
zero_gravi |
mandatory performance monitors on every RISC-V platform and have fixed increment events. For example,
|
657 |
60 |
zero_gravi |
the instructions-retired counter increments with each executed instructions. The actual hardware performance
|
658 |
|
|
monitors are optional and can be configured to increment on arbitrary hardware events. The number of
|
659 |
|
|
available HPM is configured via the top's `HPM_NUM_CNTS` generic at synthesis time. Assigning a zero will exclude
|
660 |
|
|
all HPM logic from the design.
|
661 |
|
|
|
662 |
|
|
Depending on the configuration, the following additional CSR are available:
|
663 |
|
|
|
664 |
62 |
zero_gravi |
* counters: `mhpmcounter*[h]` (3..31, depending on configuration)
|
665 |
60 |
zero_gravi |
* event configuration: `mhpmevent*` (3..31, depending on configuration)
|
666 |
|
|
|
667 |
62 |
zero_gravi |
[IMPORTANT]
|
668 |
|
|
The HPM counter CSR can only be accessed in machine-mode. Hence, the according `mcounteren` CSR bits
|
669 |
|
|
are always zero and read-only.
|
670 |
|
|
|
671 |
60 |
zero_gravi |
Auto-increment of the HPMs can be individually deactivated via the `mcountinhibit` CSR.
|
672 |
|
|
|
673 |
62 |
zero_gravi |
If `HPM_NUM_CNTS` is lower than the maximum value (=29) the remaining HPM CSRs are not implemented and the
|
674 |
|
|
according `mcountinhibit` CSR bits are hardwired to zero.
|
675 |
|
|
However, accessing their associated CSRs will not raise an illegal instruction exception (if in machine mode).
|
676 |
|
|
The according CSRs are read-only and will always return 0.
|
677 |
60 |
zero_gravi |
|
678 |
|
|
[NOTE]
|
679 |
|
|
For a list of all allocated HPM-related CSRs and all provided event configurations see section <<_hardware_performance_monitors_hpm>>.
|
680 |
|
|
|
681 |
|
|
|
682 |
|
|
<<<
|
683 |
|
|
// ####################################################################################################################
|
684 |
|
|
:sectnums:
|
685 |
|
|
=== Instruction Timing
|
686 |
|
|
|
687 |
|
|
The instruction timing listed in the table below shows the required clock cycles for executing a certain
|
688 |
|
|
instruction. These instruction cycles assume a bus access without additional wait states and a filled
|
689 |
|
|
pipeline.
|
690 |
|
|
|
691 |
|
|
Average CPI (cycles per instructions) values for "real applications" like for executing the CoreMark benchmark for different CPU
|
692 |
|
|
configurations are presented in <<_cpu_performance>>.
|
693 |
|
|
|
694 |
|
|
.Clock cycles per instruction
|
695 |
|
|
[cols="<2,^1,^4,<3"]
|
696 |
|
|
[options="header", grid="rows"]
|
697 |
|
|
|=======================
|
698 |
|
|
| Class | ISA | Instruction(s) | Execution cycles
|
699 |
|
|
| ALU | `I/E` | `addi` `slti` `sltiu` `xori` `ori` `andi` `add` `sub` `slt` `sltu` `xor` `or` `and` `lui` `auipc` | 2
|
700 |
|
|
| ALU | `C` | `c.addi4spn` `c.nop` `c.addi` `c.li` `c.addi16sp` `c.lui` `c.andi` `c.sub` `c.xor` `c.or` `c.and` `c.add` `c.mv` | 2
|
701 |
|
|
| ALU | `I/E` | `slli` `srli` `srai` `sll` `srl` `sra` | 3 + SAfootnote:[Shift amount.]/4 + SA%4; FAST_SHIFTfootnote:[Barrel shift when `FAST_SHIFT_EN` is enabled.]: 4; TINY_SHIFTfootnote:[Serial shift when `TINY_SHIFT_EN` is enabled.]: 2..32
|
702 |
61 |
zero_gravi |
| ALU | `C` | `c.srli` `c.srai` `c.slli` | 3 + SAfootnote:[Shift amount (0..31).]; FAST_SHIFTfootnote:[Barrel shifter when `FAST_SHIFT_EN` is enabled.]:
|
703 |
60 |
zero_gravi |
| Branches | `I/E` | `beq` `bne` `blt` `bge` `bltu` `bgeu` | Taken: 5 + MLfootnote:[Memory latency.]; Not taken: 3
|
704 |
|
|
| Branches | `C` | `c.beqz` `c.bnez` | Taken: 5 + MLfootnote:[Memory latency.]; Not taken: 3
|
705 |
|
|
| Jumps / Calls | `I/E` | `jal` `jalr` | 4 + ML
|
706 |
|
|
| Jumps / Calls | `C` | `c.jal` `c.j` `c.jr` `c.jalr` | 4 + ML
|
707 |
|
|
| Memory access | `I/E` | `lb` `lh` `lw` `lbu` `lhu` `sb` `sh` `sw` | 4 + ML
|
708 |
|
|
| Memory access | `C` | `c.lw` `c.sw` `c.lwsp` `c.swsp` | 4 + ML
|
709 |
|
|
| Memory access | `A` | `lr.w` `sc.w` | 4 + ML
|
710 |
|
|
| Multiplication | `M` | `mul` `mulh` `mulhsu` `mulhu` | 2+31+3; FAST_MULfootnote:[DSP-based multiplication; enabled via `FAST_MUL_EN`.]: 5
|
711 |
|
|
| Division | `M` | `div` `divu` `rem` `remu` | 22+32+4
|
712 |
|
|
| Bit-manipulation - arithmetic/logic | `B(Zbb)` | `sext.b` `sext.h` `min` `minu` `max` `maxu` `andn` `orn` `xnor` `zext`(pack) `rev8`(grevi) `orc.b`(gorci) | 3
|
713 |
|
|
| Bit-manipulation - shifts | `B(Zbb)` | `clz` `ctz` | 3 + 0..32
|
714 |
|
|
| Bit-manipulation - shifts | `B(Zbb)` | `cpop` | 3 + 32
|
715 |
|
|
| Bit-manipulation - shifts | `B(Zbb)` | `rol` `ror` `rori` | 3 + SA
|
716 |
|
|
| Bit-manipulation - single-bit | `B(Zbs)` | `sbset[i]` `sbclr[i]` `sbinv[i]` `sbext[i]` | 3
|
717 |
|
|
| Bit-manipulation - shifted-add | `B(Zba)` | `sh1add` `sh2add` `sh3add` | 3
|
718 |
|
|
| CSR access | `Zicsr` | `csrrw` `csrrs` `csrrc` `csrrwi` `csrrsi` `csrrci` | 4
|
719 |
|
|
| System | `I/E`+`Zicsr` | `ecall` `ebreak` | 4
|
720 |
|
|
| System | `I/E` | `fence` | 3
|
721 |
|
|
| System | `C`+`Zicsr` | `c.break` | 4
|
722 |
|
|
| System | `Zicsr` | `mret` `wfi` | 5
|
723 |
|
|
| System | `Zifencei` | `fence.i` | 5
|
724 |
|
|
| Floating-point - artihmetic | `Zfinx` | `fadd.s` | 110
|
725 |
|
|
| Floating-point - artihmetic | `Zfinx` | `fsub.s` | 112
|
726 |
|
|
| Floating-point - artihmetic | `Zfinx` | `fmul.s` | 22
|
727 |
|
|
| Floating-point - compare | `Zfinx` | `fmin.s` `fmax.s` `feq.s` `flt.s` `fle.s` | 13
|
728 |
|
|
| Floating-point - misc | `Zfinx` | `fsgnj.s` `fsgnjn.s` `fsgnjx.s` `fclass.s` | 12
|
729 |
|
|
| Floating-point - conversion | `Zfinx` | `fcvt.w.s` `fcvt.wu.s` | 47
|
730 |
|
|
| Floating-point - conversion | `Zfinx` | `fcvt.s.w` `fcvt.s.wu` | 48
|
731 |
63 |
zero_gravi |
| Basic bit-manip - logic | `Zbb` | `andn` `orn` `xnor` | 3
|
732 |
|
|
| Basic bit-manip - shift | `Zbb` | `clz` `ctz` `cpop` `rol` `ror` `rori` | 4+SA, FAST_SHIFT: 4
|
733 |
|
|
| Basic bit-manip - arith | `Zbb` | `max` `maxu` `min` `minu` | 3
|
734 |
|
|
| Basic bit-manip - misc | `Zbb` | `sext.b` `sext.h` `zext.h` `orc.b` `rev8` | 3
|
735 |
60 |
zero_gravi |
|=======================
|
736 |
|
|
|
737 |
|
|
[NOTE]
|
738 |
|
|
The presented values of the *floating-point execution cycles* are average values – obtained from
|
739 |
|
|
4096 instruction executions using pseudo-random input values. The execution time for emulating the
|
740 |
|
|
instructions (using pure-software libraries) is ~17..140 times higher.
|
741 |
|
|
|
742 |
|
|
|
743 |
|
|
|
744 |
|
|
// ####################################################################################################################
|
745 |
|
|
include::cpu_csr.adoc[]
|
746 |
|
|
|
747 |
|
|
|
748 |
|
|
|
749 |
|
|
<<<
|
750 |
|
|
// ####################################################################################################################
|
751 |
|
|
:sectnums:
|
752 |
62 |
zero_gravi |
==== Full Virtualization
|
753 |
60 |
zero_gravi |
|
754 |
62 |
zero_gravi |
Just like the RISC-V ISA the NEORV32 aims to support _ maximum virtualization_ capabilities
|
755 |
|
|
on CPU _and_ SoC level. The CPU supports **all** traps specified by the official RISC-V specifications.footnote:[If the `Zicsr` CPU
|
756 |
|
|
extension is enabled (implementing the full set of the privileged architecture).]
|
757 |
|
|
Thus, the CPU provides defined hardware fall-backs for any expected and unexpected situation (e.g. executing an
|
758 |
|
|
malformed instruction word or accessing a not-allocated address). For any kind of trap the core is always in a
|
759 |
|
|
defined and fully synchronized state throughout the whole architecture (i.e. there are no out-of-order operations that
|
760 |
|
|
have to be made undone). This allows predictable execution behavior - and thus, defined operations to resolve the cause
|
761 |
|
|
of the trap - at any time improving overall _execution safety_.
|
762 |
60 |
zero_gravi |
|
763 |
62 |
zero_gravi |
**NEORV32-Specific Virtualization Features**
|
764 |
60 |
zero_gravi |
|
765 |
62 |
zero_gravi |
* Due to the acknowledged memory accesses the CPU is _always_ sync with the memory system
|
766 |
|
|
(i.e. there is no speculative execution / no out-of-order states).
|
767 |
|
|
* The CPU supports _all_ RISC-V bus exceptions including access exceptions that are triggered if an
|
768 |
|
|
accessed address does not respond or encounters an internal error during access.
|
769 |
|
|
* The CPU raises an illegal instruction trap for _all_ unimplemented/malformed/illegal instructions.
|
770 |
|
|
* To be continued...
|
771 |
60 |
zero_gravi |
|
772 |
|
|
|
773 |
|
|
<<<
|
774 |
|
|
// ####################################################################################################################
|
775 |
|
|
:sectnums:
|
776 |
|
|
==== Traps, Exceptions and Interrupts
|
777 |
|
|
|
778 |
61 |
zero_gravi |
In this document the following nomenclature regarding traps is used:
|
779 |
60 |
zero_gravi |
|
780 |
64 |
zero_gravi |
* _interrupts_ = asynchronous exceptions
|
781 |
60 |
zero_gravi |
* _exceptions_ = synchronous exceptions
|
782 |
|
|
* _traps_ = exceptions + interrupts (synchronous or asynchronous exceptions)
|
783 |
|
|
|
784 |
61 |
zero_gravi |
Whenever an exception or interrupt is triggered, the CPU transfers control to the address stored in `mtvec`
|
785 |
|
|
CSR. The cause of the according interrupt or exception can be determined via the content of `mcause`
|
786 |
|
|
CSR. The address that reflects the current program counter when a trap was taken is stored to `mepc` CSR.
|
787 |
|
|
Additional information regarding the cause of the trap can be retrieved from `mtval` CSR.
|
788 |
60 |
zero_gravi |
|
789 |
61 |
zero_gravi |
The traps are prioritized. If several _exceptions_ occur at once only the one with highest priority is triggered
|
790 |
|
|
while all remaining exceptions are ignored. If several _interrupts_ trigger at once, the one with highest priority
|
791 |
64 |
zero_gravi |
is serviced first while the remaining ones stay _pending_. After completing the interrupt handler the interrupt with
|
792 |
61 |
zero_gravi |
the second highest priority will get serviced and so on until no further interrupt are pending.
|
793 |
60 |
zero_gravi |
|
794 |
64 |
zero_gravi |
.RISC-V interrupts
|
795 |
61 |
zero_gravi |
[IMPORTANT]
|
796 |
64 |
zero_gravi |
All RISC-V defined machine level interrupts request signals are high-active. A request has to stay at high-level until
|
797 |
|
|
it is acknowledged by the CPU (for example by writing to a specific memory-mapped register).
|
798 |
60 |
zero_gravi |
|
799 |
61 |
zero_gravi |
.Instruction Atomicity
|
800 |
|
|
[NOTE]
|
801 |
|
|
All instructions execute as atomic operations – interrupts can only trigger between two instructions.
|
802 |
64 |
zero_gravi |
So if there is a permanent interrupt request, exactly one instruction from the interrupt program will be executed before
|
803 |
|
|
a new interrupt handler can start.
|
804 |
60 |
zero_gravi |
|
805 |
|
|
|
806 |
61 |
zero_gravi |
:sectnums:
|
807 |
|
|
==== Memory Access Exceptions**
|
808 |
60 |
zero_gravi |
|
809 |
61 |
zero_gravi |
If a load operation causes any exception, the instruction's destination register is
|
810 |
|
|
_not written_ at all. Load exceptions caused by a misalignment or a physical memory protection fault do not
|
811 |
|
|
trigger a bus read-operation at all. Exceptions caused by a store address misalignment or a store physical
|
812 |
64 |
zero_gravi |
memory protection fault do not trigger a bus write-operation at all.
|
813 |
60 |
zero_gravi |
|
814 |
|
|
|
815 |
61 |
zero_gravi |
:sectnums:
|
816 |
|
|
==== Custom Fast Interrupt Request Lines
|
817 |
60 |
zero_gravi |
|
818 |
61 |
zero_gravi |
As a custom extension, the NEORV32 CPU features 16 fast interrupt request (FIRQ) lines via the `firq_i` CPU top
|
819 |
60 |
zero_gravi |
entity signals. These interrupts have custom configuration and status flags in the `mie` and `mip` CSRs and also
|
820 |
64 |
zero_gravi |
provide custom trap codes in `mcause`. These FIRQs are reserved for processor-internal usage only.
|
821 |
60 |
zero_gravi |
|
822 |
64 |
zero_gravi |
[NOTE]
|
823 |
|
|
The fast interrupt request lines trigger on a **rising-edge**.
|
824 |
60 |
zero_gravi |
|
825 |
|
|
|
826 |
|
|
<<<
|
827 |
|
|
// ####################################################################################################################
|
828 |
|
|
:sectnums!:
|
829 |
|
|
===== NEORV32 Trap Listing
|
830 |
|
|
|
831 |
|
|
.NEORV32 trap listing
|
832 |
|
|
[cols="3,6,5,14,11,4,4"]
|
833 |
|
|
[options="header",grid="rows"]
|
834 |
|
|
|=======================
|
835 |
64 |
zero_gravi |
| Prio. | `mcause` | [RISC-V] | ID [C] | Cause | `mepc` | `mtval`
|
836 |
|
|
| 1 | `0x00000000` | 0.0 | _TRAP_CODE_I_MISALIGNED_ | instruction address misaligned | _B-ADR_ | _PC_
|
837 |
|
|
| 2 | `0x00000001` | 0.1 | _TRAP_CODE_I_ACCESS_ | instruction access fault | _B-ADR_ | _PC_
|
838 |
|
|
| 3 | `0x00000002` | 0.2 | _TRAP_CODE_I_ILLEGAL_ | illegal instruction | _PC_ | _Inst_
|
839 |
|
|
| 4 | `0x0000000B` | 0.11 | _TRAP_CODE_MENV_CALL_ | environment call from M-mode (`ecall` in machine-mode) | _PC_ | _PC_
|
840 |
|
|
| 5 | `0x00000008` | 0.8 | _TRAP_CODE_UENV_CALL_ | environment call from U-mode (`ecall` in user-mode) | _PC_ | _PC_
|
841 |
|
|
| 6 | `0x00000003` | 0.3 | _TRAP_CODE_BREAKPOINT_ | breakpoint (EBREAK) | _PC_ | _PC_
|
842 |
|
|
| 7 | `0x00000006` | 0.6 | _TRAP_CODE_S_MISALIGNED_ | store address misaligned | _B-ADR_ | _B-ADR_
|
843 |
|
|
| 8 | `0x00000004` | 0.4 | _TRAP_CODE_L_MISALIGNED_ | load address misaligned | _B-ADR_ | _B-ADR_
|
844 |
|
|
| 9 | `0x00000007` | 0.7 | _TRAP_CODE_S_ACCESS_ | store access fault | _B-ADR_ | _B-ADR_
|
845 |
|
|
| 10 | `0x00000005` | 0.5 | _TRAP_CODE_L_ACCESS_ | load access fault | _B-ADR_ | _B-ADR_
|
846 |
|
|
| 11 | `0x80000010` | 1.16 | _TRAP_CODE_FIRQ_0_ | fast interrupt request channel 0 | _I-PC_ | _0_
|
847 |
|
|
| 12 | `0x80000011` | 1.17 | _TRAP_CODE_FIRQ_1_ | fast interrupt request channel 1 | _I-PC_ | _0_
|
848 |
|
|
| 13 | `0x80000012` | 1.18 | _TRAP_CODE_FIRQ_2_ | fast interrupt request channel 2 | _I-PC_ | _0_
|
849 |
|
|
| 14 | `0x80000013` | 1.19 | _TRAP_CODE_FIRQ_3_ | fast interrupt request channel 3 | _I-PC_ | _0_
|
850 |
|
|
| 15 | `0x80000014` | 1.20 | _TRAP_CODE_FIRQ_4_ | fast interrupt request channel 4 | _I-PC_ | _0_
|
851 |
|
|
| 16 | `0x80000015` | 1.21 | _TRAP_CODE_FIRQ_5_ | fast interrupt request channel 5 | _I-PC_ | _0_
|
852 |
|
|
| 17 | `0x80000016` | 1.22 | _TRAP_CODE_FIRQ_6_ | fast interrupt request channel 6 | _I-PC_ | _0_
|
853 |
|
|
| 18 | `0x80000017` | 1.23 | _TRAP_CODE_FIRQ_7_ | fast interrupt request channel 7 | _I-PC_ | _0_
|
854 |
|
|
| 19 | `0x80000018` | 1.24 | _TRAP_CODE_FIRQ_8_ | fast interrupt request channel 8 | _I-PC_ | _0_
|
855 |
|
|
| 20 | `0x80000019` | 1.25 | _TRAP_CODE_FIRQ_9_ | fast interrupt request channel 9 | _I-PC_ | _0_
|
856 |
|
|
| 21 | `0x8000001a` | 1.26 | _TRAP_CODE_FIRQ_10_ | fast interrupt request channel 10 | _I-PC_ | _0_
|
857 |
|
|
| 22 | `0x8000001b` | 1.27 | _TRAP_CODE_FIRQ_11_ | fast interrupt request channel 11 | _I-PC_ | _0_
|
858 |
|
|
| 23 | `0x8000001c` | 1.28 | _TRAP_CODE_FIRQ_12_ | fast interrupt request channel 12 | _I-PC_ | _0_
|
859 |
|
|
| 24 | `0x8000001d` | 1.29 | _TRAP_CODE_FIRQ_13_ | fast interrupt request channel 13 | _I-PC_ | _0_
|
860 |
|
|
| 25 | `0x8000001e` | 1.30 | _TRAP_CODE_FIRQ_14_ | fast interrupt request channel 14 | _I-PC_ | _0_
|
861 |
|
|
| 26 | `0x8000001f` | 1.31 | _TRAP_CODE_FIRQ_15_ | fast interrupt request channel 15 | _I-PC_ | _0_
|
862 |
|
|
| 27 | `0x8000000B` | 1.11 | _TRAP_CODE_MEI_ | machine external interrupt | _I-PC_ | _0_
|
863 |
|
|
| 28 | `0x80000003` | 1.3 | _TRAP_CODE_MSI_ | machine software interrupt | _I-PC_ | _0_
|
864 |
|
|
| 29 | `0x80000007` | 1.7 | _TRAP_CODE_MTI_ | machine timer interrupt | _I-PC_ | _0_
|
865 |
60 |
zero_gravi |
|=======================
|
866 |
|
|
|
867 |
|
|
**Notes**
|
868 |
|
|
|
869 |
|
|
The "Prio." column shows the priority of each trap. The highest priority is 1. The "`mcause`" column shows the
|
870 |
|
|
cause ID of the according trap that is written to `mcause` CSR. The "[RISC-V]" columns show the interrupt/exception code value from the
|
871 |
|
|
official RISC-V privileged architecture manual. The "[C]" names are defined by the NEORV32 core library (`sw/lib/include/neorv32.h`) and can
|
872 |
|
|
be used in plain C code. The "`mepc`" and "`mtval`" columns show the value written to
|
873 |
|
|
`mepc` and `mtval` CSRs when a trap is triggered:
|
874 |
|
|
|
875 |
|
|
* _I-PC_ - address of interrupted instruction (instruction has not been execute/completed yet)
|
876 |
|
|
* _B-ADR_- bad memory access address that cause the trap
|
877 |
|
|
* _PC_ - address of instruction that caused the trap
|
878 |
|
|
* _0_ - zero
|
879 |
|
|
* _Inst_ - the faulting instruction itself
|
880 |
|
|
|
881 |
|
|
|
882 |
|
|
|
883 |
|
|
<<<
|
884 |
|
|
// ####################################################################################################################
|
885 |
|
|
:sectnums:
|
886 |
|
|
==== Bus Interface
|
887 |
|
|
|
888 |
|
|
The CPU provides two independent bus interfaces: One for fetching instructions (`i_bus_*`) and one for
|
889 |
|
|
accessing data (`d_bus_*`) via load and store operations. Both interfaces use the same interface protocol.
|
890 |
|
|
|
891 |
|
|
:sectnums:
|
892 |
|
|
===== Address Space
|
893 |
|
|
|
894 |
|
|
The CPU is a 32-bit architecture with separated instruction and data interfaces making it a Harvard
|
895 |
|
|
Architecture. Each of this interfaces can access an address space of up to 2^32^ bytes (4GB). The memory
|
896 |
|
|
system is based on 32-bit words with a minimal granularity of 1 byte. Please note, that the NEORV32 CPU
|
897 |
|
|
does not support unaligned memory accesses _in hardware_ – however, a software-based handling can be
|
898 |
|
|
implemented as any unaligned memory access will trigger an according exception.
|
899 |
|
|
|
900 |
|
|
:sectnums:
|
901 |
|
|
===== Interface Signals
|
902 |
|
|
|
903 |
|
|
The following table shows the signals of the data and instruction interfaces seen from the CPU
|
904 |
|
|
(`*_o` signals are driven by the CPU / outputs, `*_i` signals are read by the CPU / inputs).
|
905 |
|
|
|
906 |
|
|
.CPU bus interface
|
907 |
|
|
[cols="<2,^1,<7"]
|
908 |
|
|
[options="header",grid="rows"]
|
909 |
|
|
|=======================
|
910 |
|
|
| Signal | Size | Function
|
911 |
|
|
| `bus_addr_o` | 32 | access address
|
912 |
|
|
| `bus_rdata_i` | 32 | data input for read operations
|
913 |
|
|
| `bus_wdata_o` | 32 | data output for write operations
|
914 |
|
|
| `bus_ben_o` | 4 | byte enable signal for write operations
|
915 |
|
|
| `bus_we_o` | 1 | bus write access
|
916 |
|
|
| `bus_re_o` | 1 | bus read access
|
917 |
|
|
| `bus_lock_o` | 1 | exclusive access request
|
918 |
|
|
| `bus_ack_i` | 1 | accessed peripheral indicates a successful completion of the bus transaction
|
919 |
|
|
| `bus_err_i` | 1 | accessed peripheral indicates an error during the bus transaction
|
920 |
|
|
| `bus_fence_o` | 1 | this signal is set for one cycle when the CPU executes a data/instruction fence operation
|
921 |
|
|
| `bus_priv_o` | 2 | current CPU privilege level
|
922 |
|
|
|=======================
|
923 |
|
|
|
924 |
|
|
[NOTE]
|
925 |
|
|
Currently, there a no pipelined or overlapping operations implemented within the same bus interface.
|
926 |
|
|
So only a single transfer request can be "on the fly".
|
927 |
|
|
|
928 |
|
|
:sectnums:
|
929 |
|
|
===== Protocol
|
930 |
|
|
|
931 |
|
|
A bus request is triggered either by the `bus_re_o` signal (for reading data) or by the `bus_we_o` signal (for
|
932 |
|
|
writing data). These signals are active for exactly one cycle and initiate either a read or a write transaction. The transaction is
|
933 |
|
|
completed when the accessed peripheral either sets the `bus_ack_i` signal (-> successful completion) or the
|
934 |
|
|
`bus_err_i` signal is set (-> failed completion). All these control signals are only active (= high) for one
|
935 |
|
|
single cycle. An error indicated via the `bus_err_i` signal during a transfer will trigger the according instruction bus
|
936 |
|
|
access fault or load/store bus access fault exception.
|
937 |
|
|
|
938 |
|
|
[NOTE]
|
939 |
|
|
The transfer can be completed directly in the same cycle as it was initiated (via the `bus_re_o` or `bus_we_o`
|
940 |
|
|
signal) if the peripheral sets `bus_ack_i` or `bus_err_i` high for one cycle. However, in order to shorten the critical path such "asynchronous"
|
941 |
|
|
completion should be avoided. The default processor-internal module provide exactly **one cycle delay** between initiation and completion of transfers.
|
942 |
|
|
|
943 |
|
|
.Bus Keeper: Processor-internal memories and memory-mapped devices with variable / high latency
|
944 |
|
|
[IMPORTANT]
|
945 |
|
|
Processor-internal peripherals or memories do not have to respond within one cycle after the transfer initiation (= latency > 1 cycle).
|
946 |
|
|
However, the bus transaction has to be completed (= acknowledged) within a certain **response time window**. This time window is defined
|
947 |
|
|
by the global `max_proc_int_response_time_c` constant (default = 15 cycles) from the processor's VHDL package file (`rtl/neorv32_package.vhd`).
|
948 |
|
|
It defines the maximum number of cycles after which an _unacknowledged_ processor-internal bus transfer will timeout and raise a **bus fault exception**.
|
949 |
|
|
The _BUSKEEPER_ hardware module (`rtl/core/neorv32_bus_keeper.vhd`) keeps track of all _internal_ bus transactions. If any bus operations times out
|
950 |
|
|
(for example when accessing "address space holes") this unit will issue a bus error to the CPU that will raise the according instruction fetch or data access bus exception.
|
951 |
|
|
Note that **the bus keeper does not track external accesses via the external memory bus interface**. However, the external memory bus interface also provides
|
952 |
|
|
an _optional_ bus timeout (see section <<_processor_external_memory_interface_wishbone_axi4_lite>>).
|
953 |
|
|
|
954 |
|
|
**Exemplary Bus Accesses**
|
955 |
|
|
|
956 |
|
|
.Example bus accesses: see read/write access description below
|
957 |
|
|
[cols="^2,^2"]
|
958 |
|
|
[grid="none"]
|
959 |
|
|
|=======================
|
960 |
|
|
a| image::cpu_interface_read_long.png[read,300,150]
|
961 |
|
|
a| image::cpu_interface_write_long.png[write,300,150]
|
962 |
|
|
| Read access | Write access
|
963 |
|
|
|=======================
|
964 |
|
|
|
965 |
|
|
**Write Access**
|
966 |
|
|
|
967 |
|
|
For a write access, the accessed address (`bus_addr_o`), the data to be written (`bus_wdata_o`) and the byte
|
968 |
|
|
enable signals (`bus_ben_o`) are set when bus_we_o goes high. These three signals are kept stable until the
|
969 |
|
|
transaction is completed. In the example the accessed peripheral cannot answer directly in the next
|
970 |
|
|
cycle after issuing. Here, the transaction is successful and the peripheral sets the `bus_ack_i` signal several
|
971 |
|
|
cycles after issuing.
|
972 |
|
|
|
973 |
|
|
**Read Access**
|
974 |
|
|
|
975 |
|
|
For a read access, the accessed address (`bus_addr_o`) is set when `bus_re_o` goes high. The address is kept
|
976 |
|
|
stable until the transaction is completed. In the example the accessed peripheral cannot answer
|
977 |
|
|
directly in the next cycle after issuing. The peripheral hast to apply the read data right in the same cycle as
|
978 |
|
|
the bus transaction is completed (here, the transaction is successful and the peripheral sets the `bus_ack_i`
|
979 |
|
|
signal).
|
980 |
|
|
|
981 |
|
|
**Access Boundaries**
|
982 |
|
|
|
983 |
|
|
The instruction interface will always access memory on word (= 32-bit) boundaries even if fetching
|
984 |
|
|
compressed (16-bit) instructions. The data interface can access memory on byte (= 8-bit), half-word (= 16-
|
985 |
|
|
bit) and word (= 32-bit) boundaries.
|
986 |
|
|
|
987 |
|
|
**Exclusive (Atomic) Access**
|
988 |
|
|
|
989 |
|
|
The CPU can access memory in an exclusive manner by generating a load-reservate and store-conditional
|
990 |
|
|
combination. Normally, these combinations should target the same memory address.
|
991 |
|
|
|
992 |
|
|
The CPU starts an exclusive access to memory via the _load-reservate instruction_ (`lr.w`). This instruction
|
993 |
|
|
will set the CPU-internal _exclusive access lock_, which directly drives the `d_bus_lock_o`. It is the task of
|
994 |
|
|
the memory system to manage this exclusive access reservation by storing the according access address and
|
995 |
|
|
the source of the access itself (for example via the CPU ID in a multi-core system).
|
996 |
|
|
|
997 |
|
|
When the CPU executes a _store-conditional instruction_ (`sc.w`) the _CPU-internal exclusive access lock_ is
|
998 |
|
|
evaluated to check if the exclusive access was successful. If the lock is still OK, the instruction will write-back
|
999 |
|
|
zero and will allow the according store operation to the memory system. If the lock is broken, the
|
1000 |
|
|
instruction will write-back non-zero and will not generate an actual memory store operation.
|
1001 |
|
|
|
1002 |
|
|
The CPU-internal exclusive access lock is broken if at least one of the situations appear.
|
1003 |
|
|
|
1004 |
|
|
* when executing any other memory-access operation than `lr.w`
|
1005 |
|
|
* when any trap (sync. or async.) is triggered (for example to force a context switch)
|
1006 |
|
|
* when the memory system signals a bus error (via the `bus_err_i` signal)
|
1007 |
|
|
|
1008 |
|
|
[TIP]
|
1009 |
|
|
For more information regarding the SoC-level behavior and requirements of atomic operations see
|
1010 |
|
|
section <<_processor_external_memory_interface_wishbone_axi4_lite>>.
|
1011 |
|
|
|
1012 |
|
|
**Memory Barriers**
|
1013 |
|
|
|
1014 |
|
|
Whenever the CPU executes a fence instruction, the according interface signal is set high for one cycle
|
1015 |
|
|
(`d_bus_fence_o` for a _fence_ instruction; `i_bus_fence_o` for a _fencei_ instruction). It is the task of the
|
1016 |
|
|
memory system to perform the necessary operations (like a cache flush and refill).
|
1017 |
|
|
|
1018 |
|
|
|
1019 |
|
|
|
1020 |
|
|
<<<
|
1021 |
|
|
// ####################################################################################################################
|
1022 |
|
|
:sectnums:
|
1023 |
|
|
==== CPU Hardware Reset
|
1024 |
|
|
|
1025 |
|
|
In order to reduce routing constraints (and by this the actual hardware requirements), most uncritical
|
1026 |
|
|
registers of the NEORV32 CPU as well as most register of the whole NEORV32 Processor do not use **a
|
1027 |
|
|
dedicated hardware reset**. "Uncritical registers" in this context means that the initial value of these registers
|
1028 |
|
|
after power-up is not relevant for a defined CPU boot process.
|
1029 |
|
|
|
1030 |
|
|
**Rational**
|
1031 |
|
|
|
1032 |
|
|
A good example to illustrate the concept of uncritical registers is a pipelined processing engine. Each stage
|
1033 |
|
|
of the engine features an N-bit _data register_ and a 1-bit _status register_. The status register is set when the
|
1034 |
|
|
data in the according data register is valid. At the end of the pipeline the status register might trigger a writeback
|
1035 |
|
|
of the processing result to some kind of memory. The initial status of the data registers after power-up is
|
1036 |
|
|
irrelevant as long as the status registers are all reset to a defined value that indicates there is no valid data in
|
1037 |
|
|
the pipeline’s data register. Therefore, the pipeline data register do no require a dedicated reset as they do not
|
1038 |
|
|
control the actual operation (in contrast to the status register). This makes the pipeline data registers from
|
1039 |
|
|
this example "uncritical registers".
|
1040 |
|
|
|
1041 |
|
|
**NEORV32 CPU Reset**
|
1042 |
|
|
|
1043 |
|
|
In terms of the NEORV32 CPU, there are several pipeline registers, state machine registers and even status
|
1044 |
|
|
and control registers (CSRs) that do not require a defined initial state to ensure a correct boot process. The
|
1045 |
|
|
pipeline register will get initialized by the CPU’s internal state machines, which are initialized from the main
|
1046 |
|
|
control engine that actually features a defined reset. The initialization of most of the CPU's core CSRs (like
|
1047 |
|
|
interrupt control) is done by the software (to be more specific, this is done by the `crt0.S` start-up code).
|
1048 |
|
|
|
1049 |
|
|
During the very early boot process (where `crt0.S` is running) there is no chance for undefined behavior due to
|
1050 |
|
|
the lack of dedicated hardware resets of certain CSRs. For example the machine interrupt-enable CSR (`mie`)
|
1051 |
|
|
does not provide a dedicated reset. The value after reset of this register is uncritical as interrupts cannot fire
|
1052 |
|
|
because the global interrupt enabled flag in the status register (`mstatsus(mie)`) provides a dedicated
|
1053 |
|
|
hardware reset setting it to low (globally disabling interrupts).
|
1054 |
|
|
|
1055 |
|
|
**Reset Configuration**
|
1056 |
|
|
|
1057 |
|
|
Most CPU-internal register do feature an asynchronous reset in the VHDL code, but the "don't care" value
|
1058 |
|
|
(VHDL `'-'`) is used for initialization of the uncritical register, effectively generating a flip-flop without a
|
1059 |
|
|
reset. However, certain applications or situations (like advanced gate-level / timing simulations) might
|
1060 |
|
|
require a more deterministic reset state. For this case, a defined reset level (reset-to-low) of all registers can
|
1061 |
|
|
be enabled via a constant in the main VHDL package file (`rtl/core/neorv32_package.vhd`):
|
1062 |
|
|
|
1063 |
|
|
[source,vhdl]
|
1064 |
|
|
----
|
1065 |
|
|
-- "critical" number of PMP regions --
|
1066 |
|
|
constant dedicated_reset_c : boolean := false; -- use dedicated hardware reset value
|
1067 |
|
|
for UNCRITICAL registers (FALSE=reset value is irrelevant (might simplify HW),
|
1068 |
|
|
default; TRUE=defined LOW reset value)
|
1069 |
|
|
----
|