This document describes integrating Kelvin as an AXI/TileLink peripheral in a bigger system.
We provide a scalar-only Kelvin configuration that can integrate with an AXI based system. The SystemVerilog can be generated with:
bazel build //hdl/chisel/src/kelvin:core_mini_axi_cc_library_emit_verilog
The interfaces to Kelvin are defined as follows:
Signal Bundle | Description |
---|---|
clk | The clock of the AXI Bus/Kelvin core. |
reset | The active-low reset signal for the AXI Bus/Kelvin core. |
s_axi | An AXI4 slave interface that can be used to write TCMs or touch Kelvin CSRs. |
m_axi | An AXI4 master interface used by Kelvin to read/write to memories/CSRs. |
irqn | Active-low interrupt to the Kelvin core. Can be triggered by peripherals or other host processor. |
wfi | Active-high signal from the Kelvin core, indicating that the core is waiting for an interrupt. While this is active, Kelvin is clock-gated. |
debug | Debug interface to monitor Kelvin instructions execution. This interface is typically only used for simulation. |
s_log | Debug interface to handle SLOG instruction. This interface is typically only used for simulation. |
halted | Output interface informing if the Core is running or not. Can be ignored. |
fault | Output interface to determine if the Core hit a fault. These signals should be connected to a system control CPU interrupt-line or status register for notification when Kelvin faults or is halted. |
AR / AW channel
Signal | Behaviour |
---|---|
addr | Address Kelvin wishes to read/write |
prot | Always 2 (unprivileged, insecure, data) |
id | Always 0 |
len | (Count of beats in the burst) - 1 |
size | Bytes-per-beat (1, 2, or 4) |
burst | Always 1 (INCR) |
lock | Always 0 (normal access) |
cache | Always 0 (Device non-bufferable) |
qos | Always 0 |
region | Always 0 |
R channel
Signal | Behaviour |
---|---|
data | Response data from the slave |
id | Ignored, but should be 0 as Kelvin only emits txns with an id of 0 |
resp | Response code |
last | Whether the beat is the last in the burst |
W channel
Signal | Behaviour |
---|---|
data | Data Kelvin wishes to write |
last | Whether the beat is the last in the burst |
strb | Which bytes in the data are valid |
B channel
Signal | Behaviour |
---|---|
id | Ignored, but should be 0 as Kelvin only emits txns with an id of 0 (an RTL assertion exists for this) |
resp | Response code |
Note: the USER signal is not supported on any of the channels.
AR / AW channel
Signal | Behaviour |
---|---|
addr | Address the master wishes to read / write to |
prot | Ignored |
id | Transaction ID, should be reflected in the response beats |
len | (Count of beats in the burst) - 1 |
size | Bytes-per-beat (1,2,4,8,16) |
burst | 0, 1, or 2 (FIXED, INCR, WRAP) |
lock | Ignored |
cache | Ignored |
qos | Ignored |
region | Ignored |
R channel
Signal | Behaviour |
---|---|
data | Response data from Kelvin |
id | Transaction ID, should match with the id field from AR |
resp | Response code (0/OKAY or 2/SLVERR) |
last | Whether the beat is the last in the burst |
W channel
Signal | Behaviour |
---|---|
data | Data the master wishes to write to Kelvin |
last | Whether the beat is the last in the burst |
strb | Which bytes in data are valid |
B channel
Signal | Behaviour |
---|---|
id | Transaction ID, should match with the id field from AW |
resp | Response code (0/OKAY or 2/SLVERR) |
Note: the USER signal is not supported on any of the channels.
Signal | Behaviour |
---|---|
en | 4-bit value, indicating which fetch lanes are active |
addr | 32-bit values, containing the PC for each fetch lane |
inst | 32-bit values, containing the instruction for each fetch lane |
cycles | cycle counter |
dbus | Information about internal LSU transactions |
-> valid | Whether the transaction is valid |
-> bits | addr: The 32-bit address for the transaction |
write: If the transaction is a write | |
wdata: 128-bit write data for the transaction | |
dispatch | Information about instructions which are dispatched for execution |
-> fire | If an instruction was dispatched in the slot, this cycle |
-> addr | The 32-bit address of the instruction |
-> inst | The 32-bit value of the instruction |
regfile | Information about writes to the integer register file |
-> writeAddr | Register addresses to which a future write is expected |
->-> valid | If an instruction was dispatched in this lane, which will write the regfile |
->-> bits | The 5-bit register address to which the write is expected |
-> writeData | For each port in the register file, information about writes |
->-> valid | If a write occurred on this port, this cycle |
->-> bits_addr | The 5-bit register address to which the write occurred |
->-> bits_data | The 32-bit value which was written to the register |
float | Information about write to the floating point register file |
-> writeAddr | Register addresses to which a future write is expected |
->-> valid | If an instruction was dispatched to floating point on this cycle |
->-> bits | The address of the register to which a write is expected |
-> writeData | For each port in the register file, information about writes |
->-> valid | If a write occured on this port, this cycle |
->-> bits_addr | The 5-bit register address to which the writh occurred |
->-> bits_data | The 32-bit value which was written to the register |
Memory accesses to Kelvin are defined as follows:
Region | Range | Size | Alignment | Description |
---|---|---|---|---|
ITCM | 0x0000 - 0x1FFF | 8kB | 4 bytes | ITCM storage for code executed by Kelvin. |
DTCM | 0x10000 - 0x17FFF | 32kB | 1 byte | DTCM storage for data used by Kelvin. |
CSR | 0x30000 - TBD | TBD | 4 bytes | CSR interface used to query/control Kelvin. |
Kelvin uses a synchronous reset strategy -- to ensure proper reset behavior, ensure that the clock runs for a cycle with reset active, before enabling either the internal clock gate (via CSR) or gating externally.
A note first -- in these examples, Kelvin is located in the overall system memory map at 0x70000000.
volatile uint8_t* kelvin_itcm = (uint8_t*)0x00000000L; for (int i = 0; i < kelvin_binary_len; ++i) { kelvin_itcm[i] = kelvin_binary[i]; }
If something like a DMA engine is present in your system, that is probably a better option for initializing the ITCM.
volatile uint32_t* kelvin_pc_csr = (uint32_t*)0x00030004L; *kelvin_pc_csr = start_addr;
volatile uint32_t* kelvin_reset_csr = (uint32_t*)0x00030000L; *kelvin_reset_csr = 1;
After this, ensure you wait a cycle to allow Kelvin‘s reset to occur. If you want to configure something like an interrupt that is connected to Kelvin’s fault or halted outputs, this is a good time.
volatile uint32_t* kelvin_reset_csr = (uint32_t*)0x00030000L; *kelvin_reset_csr = 0;
At this point, Kelvin will begin executing at the PC programmed in step 2.
io_halted
The status of Kelvin's execution can be checked by reading the status CSR:volatile uint32_t* kelvin_status_csr = (uint32_t*)0x00030008L; uint32_t status = *kelvin_status_csr; bool halted = status & 1; bool fault = status & 2;