Merge branch 'master' into nac3

This commit is contained in:
Sebastien Bourdeauducq 2021-11-12 19:46:37 +08:00
commit 75dad8090f
35 changed files with 303 additions and 206 deletions

View File

@ -2,6 +2,7 @@ import os
import subprocess
from migen import *
from migen.build.platforms.sinara import kasli
from misoc.interconnect.csr import *
from misoc.integration.builder import *
@ -57,11 +58,17 @@ def build_artiq_soc(soc, argdict):
builder = Builder(soc, **argdict)
builder.software_packages = []
builder.add_software_package("bootloader", os.path.join(firmware_dir, "bootloader"))
is_kasli_v1 = isinstance(soc.platform, kasli.Platform) and soc.platform.hw_rev in ("v1.0", "v1.1")
if isinstance(soc, AMPSoC):
builder.add_software_package("libm")
builder.add_software_package("libprintf")
builder.add_software_package("libunwind")
builder.add_software_package("ksupport", os.path.join(firmware_dir, "ksupport"))
kernel_cpu_type = "vexriscv" if is_kasli_v1 else "vexriscv-g"
builder.add_software_package("libm", cpu_type=kernel_cpu_type)
builder.add_software_package("libprintf", cpu_type=kernel_cpu_type)
builder.add_software_package("libunwind", cpu_type=kernel_cpu_type)
builder.add_software_package("ksupport", os.path.join(firmware_dir, "ksupport"), cpu_type=kernel_cpu_type)
# Generate unwinder for soft float target (ARTIQ runtime)
# If the kernel lacks FPU, then the runtime unwinder is already generated
if not is_kasli_v1:
builder.add_software_package("libunwind")
builder.add_software_package("runtime", os.path.join(firmware_dir, "runtime"))
else:
# Assume DRTIO satellite.

View File

@ -42,7 +42,7 @@ class Core:
ref_multiplier: KernelInvariant[int32]
coarse_ref_period: KernelInvariant[float]
def __init__(self, dmgr, host, ref_period, ref_multiplier=8, target="riscv"):
def __init__(self, dmgr, host, ref_period, ref_multiplier=8, target="rv32g"):
self.ref_period = ref_period
self.ref_multiplier = ref_multiplier
self.coarse_ref_period = ref_period*ref_multiplier

View File

@ -11,7 +11,7 @@ all:: bootloader.bin
$(RUSTOUT)/libbootloader.a:
$(cargo) --target-dir ./cargo \
--manifest-path $(BOOTLOADER_DIRECTORY)/Cargo.toml \
--target $(BOOTLOADER_DIRECTORY)/../riscv32ima-unknown-none-elf.json
--target $(BOOTLOADER_DIRECTORY)/../$(CARGO_TRIPLE).json
bootloader.elf: $(RUSTOUT)/libbootloader.a
$(link) -T $(BOOTLOADER_DIRECTORY)/bootloader.ld

View File

@ -22,11 +22,11 @@ all:: ksupport.elf
$(RUSTOUT)/libksupport.a:
$(cargo) --target-dir ./cargo \
--manifest-path $(KSUPPORT_DIRECTORY)/Cargo.toml \
--target $(KSUPPORT_DIRECTORY)/../riscv32ima-unknown-none-elf.json
--target $(KSUPPORT_DIRECTORY)/../$(CARGO_TRIPLE).json
ksupport.elf: $(RUSTOUT)/libksupport.a glue.o
$(link) -T $(KSUPPORT_DIRECTORY)/ksupport.ld \
-lunwind-elf -lprintf-float -lm
-lunwind-$(CPU)-elf -lprintf-float -lm
%.o: $(KSUPPORT_DIRECTORY)/%.c
$(compile)

View File

@ -23,6 +23,8 @@ mod imp {
pub const RTIO_I_STATUS_WAIT_STATUS: u8 = 4;
pub const RTIO_I_STATUS_DESTINATION_UNREACHABLE: u8 = 8;
const OFFSET_MULTIPLE: isize = (csr::CONFIG_DATA_WIDTH_BYTES / 4) as isize;
pub extern fn init() {
send(&RtioInitRequest);
}
@ -47,14 +49,14 @@ mod imp {
#[inline(always)]
pub unsafe fn rtio_o_data_write(offset: usize, data: u32) {
write_volatile(
csr::rtio::O_DATA_ADDR.offset((csr::rtio::O_DATA_SIZE - 1 - offset) as isize),
csr::rtio::O_DATA_ADDR.offset(OFFSET_MULTIPLE*(csr::rtio::O_DATA_SIZE - 1 - offset) as isize),
data);
}
#[inline(always)]
pub unsafe fn rtio_i_data_read(offset: usize) -> u32 {
read_volatile(
csr::rtio::I_DATA_ADDR.offset((csr::rtio::I_DATA_SIZE - 1 - offset) as isize))
csr::rtio::I_DATA_ADDR.offset(OFFSET_MULTIPLE*(csr::rtio::I_DATA_SIZE - 1 - offset) as isize))
}
#[inline(never)]

View File

@ -1,9 +1,9 @@
use core::ptr::{read_volatile, write_volatile};
use core::slice;
use board_misoc::{mem, cache};
use board_misoc::{mem, cache, csr::CONFIG_DATA_WIDTH_BYTES};
const SEND_MAILBOX: *mut usize = (mem::MAILBOX_BASE + 4) as *mut usize;
const RECV_MAILBOX: *mut usize = (mem::MAILBOX_BASE + 8) as *mut usize;
const SEND_MAILBOX: *mut usize = (mem::MAILBOX_BASE + CONFIG_DATA_WIDTH_BYTES as usize) as *mut usize;
const RECV_MAILBOX: *mut usize = (mem::MAILBOX_BASE + (CONFIG_DATA_WIDTH_BYTES * 2) as usize) as *mut usize;
const QUEUE_BEGIN: usize = 0x44000000;
const QUEUE_END: usize = 0x44ffff80;

View File

@ -1,19 +1,16 @@
extern crate build_misoc;
extern crate cc;
use std::env;
use std::path::Path;
fn main() {
build_misoc::cfg();
let triple = env::var("TARGET").unwrap();
let arch = triple.split("-").next().unwrap();
let vectors_path = Path::new(arch).join("vectors.S");
let vectors_path = "riscv32/vectors.S";
println!("cargo:rerun-if-changed={}", vectors_path.to_str().unwrap());
println!("cargo:rerun-if-changed={}", vectors_path);
cc::Build::new()
.flag("--target=riscv32-unknown-elf")
.file(vectors_path)
.file(Path::new(vectors_path))
.compile("vectors");
}

View File

@ -8,7 +8,7 @@ extern crate log;
extern crate smoltcp;
#[cfg(target_arch = "riscv32")]
#[path = "riscv32ima/mod.rs"]
#[path = "riscv32/mod.rs"]
mod arch;
#[cfg(target_arch = "riscv32")]

View File

@ -2,6 +2,7 @@
mod ddr {
use core::{ptr, fmt};
use csr::{dfii, ddrphy};
use csr::CONFIG_DATA_WIDTH_BYTES;
use sdram_phy::{self, spin_cycles};
use sdram_phy::{DFII_COMMAND_CS, DFII_COMMAND_WE, DFII_COMMAND_CAS, DFII_COMMAND_RAS,
DFII_COMMAND_WRDATA, DFII_COMMAND_RDDATA};
@ -14,6 +15,8 @@ mod ddr {
const DQS_SIGNAL_COUNT: usize = DFII_PIX_DATA_SIZE / 2;
const CSR_SEPARATION: isize = CONFIG_DATA_WIDTH_BYTES as isize / 4;
macro_rules! log {
($logger:expr, $( $arg:expr ),+) => (
if let &mut Some(ref mut f) = $logger {
@ -46,7 +49,7 @@ mod ddr {
for n in 0..DQS_SIGNAL_COUNT {
let dq_addr = dfii::PI0_RDDATA_ADDR
.offset((DQS_SIGNAL_COUNT - 1 - n) as isize);
.offset(CSR_SEPARATION * (DQS_SIGNAL_COUNT - 1 - n) as isize);
log!(logger, "Module {}:\n", DQS_SIGNAL_COUNT - 1 - n);
@ -100,7 +103,7 @@ mod ddr {
let mut failed = false;
for n in 0..DQS_SIGNAL_COUNT {
let dq_addr = dfii::PI0_RDDATA_ADDR
.offset((DQS_SIGNAL_COUNT - 1 - n) as isize);
.offset(CSR_SEPARATION * (DQS_SIGNAL_COUNT - 1 - n) as isize);
delay[n] = 0;
high_skew[n] = false;
@ -223,7 +226,7 @@ mod ddr {
// Write test pattern
for p in 0..DFII_NPHASES {
for offset in 0..DFII_PIX_DATA_SIZE {
let addr = DFII_PIX_WRDATA_ADDR[p].offset(offset as isize);
let addr = DFII_PIX_WRDATA_ADDR[p].offset(CSR_SEPARATION * offset as isize);
let data = prs[DFII_PIX_DATA_SIZE * p + offset];
ptr::write_volatile(addr, data as u32);
}
@ -258,7 +261,7 @@ mod ddr {
for p in 0..DFII_NPHASES {
for &offset in [n, n + DQS_SIGNAL_COUNT].iter() {
let addr = DFII_PIX_RDDATA_ADDR[p].offset(offset as isize);
let addr = DFII_PIX_RDDATA_ADDR[p].offset(CSR_SEPARATION * offset as isize);
let data = prs[DFII_PIX_DATA_SIZE * p + offset];
if ptr::read_volatile(addr) as u8 != data {
working = false;
@ -306,7 +309,7 @@ mod ddr {
// Write test pattern
for p in 0..DFII_NPHASES {
for offset in 0..DFII_PIX_DATA_SIZE {
let addr = DFII_PIX_WRDATA_ADDR[p].offset(offset as isize);
let addr = DFII_PIX_WRDATA_ADDR[p].offset(CSR_SEPARATION * offset as isize);
let data = prs[DFII_PIX_DATA_SIZE * p + offset];
ptr::write_volatile(addr, data as u32);
}
@ -349,7 +352,7 @@ mod ddr {
for p in 0..DFII_NPHASES {
for &offset in [n, n + DQS_SIGNAL_COUNT].iter() {
let addr = DFII_PIX_RDDATA_ADDR[p].offset(offset as isize);
let addr = DFII_PIX_RDDATA_ADDR[p].offset(CSR_SEPARATION * offset as isize);
let data = prs[DFII_PIX_DATA_SIZE * p + offset];
if ptr::read_volatile(addr) as u8 != data {
valid = false;

View File

@ -2230,6 +2230,12 @@ pub const R_OR1K_TLS_TPOFF: u8 = 32;
pub const R_OR1K_TLS_DTPOFF: u8 = 33;
pub const R_OR1K_TLS_DTPMOD: u8 = 34;
pub const R_OR1K_NUM: u8 = 35;
pub const EF_RISCV_RVC: u32 = 1;
pub const EF_RISCV_FLOAT_ABI: u32 = 6;
pub const EF_RISCV_FLOAT_ABI_SOFT: u32 = 0;
pub const EF_RISCV_FLOAT_ABI_SINGLE: u32 = 2;
pub const EF_RISCV_FLOAT_ABI_DOUBLE: u32 = 4;
pub const EF_RISCV_FLOAT_ABI_QUAD: u32 = 6;
pub const R_RISCV_NONE: u8 = 0;
pub const R_RISCV_32: u8 = 1;
pub const R_RISCV_64: u8 = 2;

View File

@ -219,7 +219,13 @@ impl<'a> Library<'a> {
#[cfg(not(target_arch = "riscv32"))]
const ARCH: u16 = EM_NONE;
if ehdr.e_ident != IDENT || ehdr.e_type != ET_DYN || ehdr.e_machine != ARCH {
#[cfg(all(target_feature = "f", target_feature = "d"))]
const FLAGS: u32 = EF_RISCV_FLOAT_ABI_DOUBLE;
#[cfg(not(all(target_feature = "f", target_feature = "d")))]
const FLAGS: u32 = EF_RISCV_FLOAT_ABI_SOFT;
if ehdr.e_ident != IDENT || ehdr.e_type != ET_DYN || ehdr.e_machine != ARCH || ehdr.e_flags != FLAGS {
return Err("not a shared library for current architecture")?
}

View File

@ -0,0 +1,40 @@
{
"arch": "riscv32",
"code-model": "medium",
"cpu": "generic-rv32",
"crt-static-respected": true,
"data-layout": "e-m:e-p:32:32-i64:64-n32-S128",
"dynamic-linking": true,
"executables": true,
"features": "+m,+a,+f,+d",
"has-elf-tls": true,
"has-rpath": true,
"llvm-abiname": "ilp32d",
"llvm-target": "riscv32-unknown-linux",
"max-atomic-width": 32,
"position-independent-executables": true,
"pre-link-args": {
"gcc": [
"-Wl,--as-needed",
"-Wl,-z,noexecstack"
]
},
"relro-level": "full",
"target-family": "unix",
"target-pointer-width": "32",
"unsupported-abis": [
"cdecl",
"stdcall",
"fastcall",
"vectorcall",
"thiscall",
"aapcs",
"win64",
"sysv64",
"ptx-kernel",
"msp430-interrupt",
"x86-interrupt",
"amdgpu-kernel"
]
}

View File

@ -1,32 +1,31 @@
{
"arch": "riscv32",
"cpu": "generic-rv32",
"data-layout": "e-m:e-p:32:32-i64:64-n32-S128",
"eh-frame-header": false,
"emit-debug-gdb-scripts": false,
"executables": true,
"features": "+m,+a,-c",
"is-builtin": false,
"linker": "rust-lld",
"linker-flavor": "ld.lld",
"llvm-target": "riscv32",
"max-atomic-width": 32,
"panic-strategy": "unwind",
"relocation-model": "static",
"target-pointer-width": "32",
"unsupported-abis": [
"cdecl",
"stdcall",
"fastcall",
"vectorcall",
"thiscall",
"aapcs",
"win64",
"sysv64",
"ptx-kernel",
"msp430-interrupt",
"x86-interrupt",
"amdgpu-kernel"
]
}
"arch": "riscv32",
"cpu": "generic-rv32",
"data-layout": "e-m:e-p:32:32-i64:64-n32-S128",
"eh-frame-header": false,
"emit-debug-gdb-scripts": false,
"executables": true,
"features": "+m,+a,-c",
"is-builtin": false,
"linker": "rust-lld",
"linker-flavor": "ld.lld",
"llvm-target": "riscv32",
"max-atomic-width": 32,
"panic-strategy": "unwind",
"relocation-model": "static",
"target-pointer-width": "32",
"unsupported-abis": [
"cdecl",
"stdcall",
"fastcall",
"vectorcall",
"thiscall",
"aapcs",
"win64",
"sysv64",
"ptx-kernel",
"msp430-interrupt",
"x86-interrupt",
"amdgpu-kernel"
]
}

View File

@ -18,7 +18,7 @@ all:: runtime.bin runtime.fbi
$(RUSTOUT)/libruntime.a:
$(cargo) --target-dir ./cargo \
--manifest-path $(RUNTIME_DIRECTORY)/Cargo.toml \
--target $(RUNTIME_DIRECTORY)/../riscv32ima-unknown-none-elf.json
--target $(RUNTIME_DIRECTORY)/../$(CARGO_TRIPLE).json
runtime.elf: $(RUSTOUT)/libruntime.a ksupport_data.o
$(link) -T $(RUNTIME_DIRECTORY)/runtime.ld \

View File

@ -13,7 +13,7 @@ all:: satman.bin satman.fbi
$(RUSTOUT)/libsatman.a:
$(cargo) --target-dir ./cargo \
--manifest-path $(SATMAN_DIRECTORY)/Cargo.toml \
--target $(SATMAN_DIRECTORY)/../riscv32ima-unknown-none-elf.json
--target $(SATMAN_DIRECTORY)/../$(CARGO_TRIPLE).json
satman.elf: $(RUSTOUT)/libsatman.a
$(link) -T $(SATMAN_DIRECTORY)/satman.ld

View File

@ -11,14 +11,16 @@ from artiq.coredevice import jsondesc
def process_header(output, description):
if description["target"] not in ("kasli", "kasli_soc"):
if description["target"] == "kasli":
if description["hw_rev"] in ("v1.0", "v1.1"):
cpu_target = "rv32ima"
else:
cpu_target = "rv32g"
elif description["target"] == "kasli_soc":
cpu_target = "cortexa9"
else:
raise NotImplementedError
cpu_target = {
"kasli": "riscv",
"kasli_soc": "cortexa9"
}[description["target"]]
print(textwrap.dedent("""
# Autogenerated for the {variant} variant
core_addr = "{core_addr}"

View File

@ -1,4 +1,5 @@
from migen import *
from migen.build.platforms.sinara import kasli
from misoc.interconnect.csr import *
from misoc.interconnect import wishbone
from misoc.cores import vexriscv
@ -14,21 +15,22 @@ class KernelCPU(Module):
# # #
self._wb_slaves = WishboneSlaveManager(0x80000000)
# CPU core
self.clock_domains.cd_sys_kernel = ClockDomain()
self.comb += [
self.cd_sys_kernel.clk.eq(ClockSignal()),
self.cd_sys_kernel.rst.eq(self._reset.storage)
]
kasli_v1 = isinstance(platform, kasli.Platform) and platform.hw_rev in ("v1.0", "v1.1")
self.submodules.cpu = ClockDomainsRenamer("sys_kernel")(
vexriscv.VexRiscv(
platform,
exec_address))
vexriscv.VexRiscv(platform, exec_address,
variant="VexRiscv_IMA" if kasli_v1 else "VexRiscv_G"))
self.cpu_dw = len(self.cpu.dbus.dat_w)
self._wb_slaves = WishboneSlaveManager(0x80000000, dw=self.cpu_dw)
# DRAM access
self.wb_sdram = wishbone.Interface()
self.wb_sdram = wishbone.Interface(data_width=self.cpu_dw, adr_width=32-log2_int(self.cpu_dw//8))
self.add_wb_slave(main_mem_origin, 0x10000000, self.wb_sdram)
def get_csrs(self):
@ -37,7 +39,7 @@ class KernelCPU(Module):
def do_finalize(self):
self.submodules.wishbonecon = wishbone.InterconnectShared(
[self.cpu.ibus, self.cpu.dbus],
self._wb_slaves.get_interconnect_slaves(), register=True)
self._wb_slaves.get_interconnect_slaves(), register=True, dw=self.cpu_dw)
def add_wb_slave(self, origin, length, interface):
if self.finalized:

View File

@ -3,9 +3,9 @@ from misoc.interconnect import wishbone
class Mailbox(Module):
def __init__(self, size=1):
self.i1 = wishbone.Interface()
self.i2 = wishbone.Interface()
def __init__(self, size=1, adr_width=30):
self.i1 = wishbone.Interface(data_width=32, adr_width=adr_width)
self.i2 = wishbone.Interface(data_width=32, adr_width=adr_width)
# # #

View File

@ -1,3 +1,4 @@
from migen import *
from misoc.cores import timer
from misoc.interconnect import wishbone
@ -19,21 +20,24 @@ class AMPSoC:
self.csr_devices.append("kernel_cpu")
mailbox_size = 3
self.submodules.mailbox = Mailbox(mailbox_size)
self.add_wb_slave(self.mem_map["mailbox"], 4*mailbox_size,
self.csr_separation = self.kernel_cpu.cpu_dw//8
self.submodules.mailbox = Mailbox(mailbox_size, adr_width=32-log2_int(self.csr_separation))
self.add_wb_slave(self.mem_map["mailbox"], self.csr_separation*mailbox_size,
self.mailbox.i1)
self.kernel_cpu.add_wb_slave(self.mem_map["mailbox"], 4*mailbox_size,
self.kernel_cpu.add_wb_slave(self.mem_map["mailbox"], self.csr_separation*mailbox_size,
self.mailbox.i2)
self.add_memory_region("mailbox",
self.mem_map["mailbox"] | 0x80000000,
4*mailbox_size)
self.csr_separation*mailbox_size)
def register_kernel_cpu_csrdevice(self, name, csrs=None):
if csrs is None:
csrs = getattr(self, name).get_csrs()
bank = wishbone.CSRBank(csrs)
csr_bus = wishbone.Interface(data_width=32, adr_width=32-log2_int(self.csr_separation))
bank = wishbone.CSRBank(csrs, bus=csr_bus)
self.submodules += bank
self.kernel_cpu.add_wb_slave(self.mem_map[name], 4*2**bank.decode_bits, bank.bus)
self.kernel_cpu.add_wb_slave(self.mem_map[name], self.csr_separation*2**bank.decode_bits, bank.bus)
self.add_csr_region(name,
self.mem_map[name] | 0x80000000, 32,
csrs)

View File

@ -212,14 +212,15 @@ class Receiver(Module, AutoCSR):
# TODO: FullMemoryWE should be applied by migen.build
@FullMemoryWE()
class DRTIOAuxController(Module):
def __init__(self, link_layer):
self.bus = wishbone.Interface()
def __init__(self, link_layer, dw=32):
wsb = log2_int(dw//8)
self.bus = wishbone.Interface(data_width=dw, adr_width=32-wsb)
self.submodules.transmitter = Transmitter(link_layer, len(self.bus.dat_w))
self.submodules.receiver = Receiver(link_layer, len(self.bus.dat_w))
tx_sdram_if = wishbone.SRAM(self.transmitter.mem, read_only=False)
rx_sdram_if = wishbone.SRAM(self.receiver.mem, read_only=True)
wsb = log2_int(len(self.bus.dat_w)//8)
tx_sdram_if = wishbone.SRAM(self.transmitter.mem, read_only=False, data_width=dw)
rx_sdram_if = wishbone.SRAM(self.receiver.mem, read_only=True, data_width=dw)
decoder = wishbone.Decoder(self.bus,
[(lambda a: a[log2_int(max_packet)-wsb] == 0, tx_sdram_if.bus),
(lambda a: a[log2_int(max_packet)-wsb] == 1, rx_sdram_if.bus)],

View File

@ -3,6 +3,7 @@ from migen.genlib.record import Record, layout_len
from misoc.interconnect.csr import *
from misoc.interconnect import stream
from artiq.gateware.rtio import dma
from artiq.gateware.rtio.cri import commands as cri_commands
from artiq.coredevice.comm_analyzer import MessageType, ExceptionType
@ -42,20 +43,6 @@ assert layout_len(exception_layout) == message_len
assert layout_len(stopped_layout) == message_len
def convert_signal(signal):
assert len(signal) % 8 == 0
nbytes = len(signal)//8
assert nbytes % 4 == 0
nwords = nbytes//4
signal_words = []
for i in range(nwords):
signal_bytes = []
for j in range(4):
signal_bytes.append(signal[8*(j+i*4):8*((j+i*4)+1)])
signal_words.extend(reversed(signal_bytes))
return Cat(*signal_words)
class MessageEncoder(Module, AutoCSR):
def __init__(self, tsc, cri, enable):
self.source = stream.Endpoint([("data", message_len)])
@ -150,7 +137,7 @@ class MessageEncoder(Module, AutoCSR):
class DMAWriter(Module, AutoCSR):
def __init__(self, membus):
def __init__(self, membus, cpu_dw):
aw = len(membus.adr)
dw = len(membus.dat_w)
messages_per_dw = dw//message_len
@ -175,7 +162,7 @@ class DMAWriter(Module, AutoCSR):
membus.stb.eq(self.sink.stb),
self.sink.ack.eq(membus.ack),
membus.we.eq(1),
membus.dat_w.eq(convert_signal(self.sink.data))
membus.dat_w.eq(dma.convert_signal(self.sink.data, cpu_dw//8))
]
if messages_per_dw > 1:
for i in range(dw//8):
@ -207,7 +194,7 @@ class DMAWriter(Module, AutoCSR):
class Analyzer(Module, AutoCSR):
def __init__(self, tsc, cri, membus, fifo_depth=128):
def __init__(self, tsc, cri, membus, fifo_depth=128, cpu_dw=32):
# shutdown procedure: set enable to 0, wait until busy=0
self.enable = CSRStorage()
self.busy = CSRStatus()
@ -219,7 +206,7 @@ class Analyzer(Module, AutoCSR):
self.submodules.converter = stream.Converter(
message_len, len(membus.dat_w), reverse=True,
report_valid_token_count=True)
self.submodules.dma = DMAWriter(membus)
self.submodules.dma = DMAWriter(membus, cpu_dw)
enable_r = Signal()
self.sync += [

View File

@ -11,28 +11,20 @@ def _reverse_bytes(s, g):
return Cat(reversed(list(s[i*g:(i+1)*g] for i in range(len(s)//g))))
def reverse_bytes(s):
n = (len(s) + 7)//8
return Cat(*[s[i*8:min((i + 1)*8, len(s))]
for i in reversed(range(n))])
def convert_signal(signal):
def convert_signal(signal, granularity):
assert len(signal) % 8 == 0
nbytes = len(signal)//8
assert nbytes % 4 == 0
nwords = nbytes//4
assert nbytes % granularity == 0
nwords = nbytes//granularity
signal_words = []
for i in range(nwords):
signal_bytes = []
for j in range(4):
signal_bytes.append(signal[8*(j+i*4):8*((j+i*4)+1)])
signal_words.extend(reversed(signal_bytes))
return Cat(*signal_words)
signal_words.append(_reverse_bytes(
signal[i*granularity*8:(i+1)*granularity*8], 8))
return Cat(signal_words)
class WishboneReader(Module):
def __init__(self, bus):
def __init__(self, bus, cpu_dw):
self.bus = bus
aw = len(bus.adr)
@ -57,18 +49,18 @@ class WishboneReader(Module):
If(self.source.ack, data_reg_loaded.eq(0)),
If(bus.ack,
data_reg_loaded.eq(1),
self.source.data.eq(convert_signal(bus.dat_r)),
self.source.data.eq(convert_signal(bus.dat_r, cpu_dw//8)),
self.source.eop.eq(self.sink.eop)
)
]
class DMAReader(Module, AutoCSR):
def __init__(self, membus, enable):
def __init__(self, membus, enable, cpu_dw):
aw = len(membus.adr)
data_alignment = log2_int(len(membus.dat_w)//8)
self.submodules.wb_reader = WishboneReader(membus)
self.submodules.wb_reader = WishboneReader(membus, cpu_dw)
self.source = self.wb_reader.source
# All numbers in bytes
@ -344,11 +336,11 @@ class CRIMaster(Module, AutoCSR):
class DMA(Module):
def __init__(self, membus):
def __init__(self, membus, cpu_dw):
self.enable = CSR()
flow_enable = Signal()
self.submodules.dma = DMAReader(membus, flow_enable)
self.submodules.dma = DMAReader(membus, flow_enable, cpu_dw)
self.submodules.slicer = RecordSlicer(len(membus.dat_w))
self.submodules.time_offset = TimeOffset()
self.submodules.cri_master = CRIMaster()

View File

@ -104,9 +104,15 @@ class StandaloneBase(MiniSoC, AMPSoC):
}
mem_map.update(MiniSoC.mem_map)
def __init__(self, gateware_identifier_str=None, **kwargs):
def __init__(self, gateware_identifier_str=None, hw_rev="v2.0", **kwargs):
if hw_rev in ("v1.0", "v1.1"):
cpu_bus_width = 32
else:
cpu_bus_width = 64
MiniSoC.__init__(self,
cpu_type="vexriscv",
hw_rev=hw_rev,
cpu_bus_width=cpu_bus_width,
sdram_controller_type="minicon",
l2_size=128*1024,
integrated_sram_size=8192,
@ -138,7 +144,7 @@ class StandaloneBase(MiniSoC, AMPSoC):
self.csr_devices.append("rtio_core")
self.submodules.rtio = rtio.KernelInitiator(self.rtio_tsc)
self.submodules.rtio_dma = ClockDomainsRenamer("sys_kernel")(
rtio.DMA(self.get_native_sdram_if()))
rtio.DMA(self.get_native_sdram_if(), self.cpu_dw))
self.register_kernel_cpu_csrdevice("rtio")
self.register_kernel_cpu_csrdevice("rtio_dma")
self.submodules.cri_con = rtio.CRIInterconnectShared(
@ -156,7 +162,7 @@ class StandaloneBase(MiniSoC, AMPSoC):
self.rtio_crg.cd_rtio.clk)
self.submodules.rtio_analyzer = rtio.Analyzer(self.rtio_tsc, self.rtio_core.cri,
self.get_native_sdram_if())
self.get_native_sdram_if(), cpu_dw=self.cpu_dw)
self.csr_devices.append("rtio_analyzer")
@ -255,9 +261,15 @@ class MasterBase(MiniSoC, AMPSoC):
}
mem_map.update(MiniSoC.mem_map)
def __init__(self, rtio_clk_freq=125e6, enable_sata=False, gateware_identifier_str=None, **kwargs):
def __init__(self, rtio_clk_freq=125e6, enable_sata=False, gateware_identifier_str=None, hw_rev="v2.0", **kwargs):
if hw_rev in ("v1.0", "v1.1"):
cpu_bus_width = 32
else:
cpu_bus_width = 64
MiniSoC.__init__(self,
cpu_type="vexriscv",
hw_rev=hw_rev,
cpu_bus_width=cpu_bus_width,
sdram_controller_type="minicon",
l2_size=128*1024,
integrated_sram_size=8192,
@ -333,7 +345,7 @@ class MasterBase(MiniSoC, AMPSoC):
self.drtio_cri.append(core.cri)
self.csr_devices.append(core_name)
coreaux = cdr(DRTIOAuxController(core.link_layer))
coreaux = cdr(DRTIOAuxController(core.link_layer, self.cpu_dw))
setattr(self.submodules, coreaux_name, coreaux)
self.csr_devices.append(coreaux_name)
@ -374,7 +386,7 @@ class MasterBase(MiniSoC, AMPSoC):
self.submodules.rtio = rtio.KernelInitiator(self.rtio_tsc)
self.submodules.rtio_dma = ClockDomainsRenamer("sys_kernel")(
rtio.DMA(self.get_native_sdram_if()))
rtio.DMA(self.get_native_sdram_if(), self.cpu_dw))
self.register_kernel_cpu_csrdevice("rtio")
self.register_kernel_cpu_csrdevice("rtio_dma")
self.submodules.cri_con = rtio.CRIInterconnectShared(
@ -386,7 +398,7 @@ class MasterBase(MiniSoC, AMPSoC):
self.csr_devices.append("routing_table")
self.submodules.rtio_analyzer = rtio.Analyzer(self.rtio_tsc, self.cri_con.switch.slave,
self.get_native_sdram_if())
self.get_native_sdram_if(), cpu_dw=self.cpu_dw)
self.csr_devices.append("rtio_analyzer")
# Never running out of stupid features, GTs on A7 make you pack
@ -431,9 +443,15 @@ class SatelliteBase(BaseSoC):
}
mem_map.update(BaseSoC.mem_map)
def __init__(self, rtio_clk_freq=125e6, enable_sata=False, *, with_wrpll=False, gateware_identifier_str=None, **kwargs):
def __init__(self, rtio_clk_freq=125e6, enable_sata=False, *, with_wrpll=False, gateware_identifier_str=None, hw_rev="v2.0", **kwargs):
if hw_rev in ("v1.0", "v1.1"):
cpu_bus_width = 32
else:
cpu_bus_width = 64
BaseSoC.__init__(self,
cpu_type="vexriscv",
hw_rev=hw_rev,
cpu_bus_width=cpu_bus_width,
sdram_controller_type="minicon",
l2_size=128*1024,
**kwargs)
@ -521,7 +539,7 @@ class SatelliteBase(BaseSoC):
self.drtio_cri.append(core.cri)
self.csr_devices.append(corerep_name)
coreaux = cdr(DRTIOAuxController(core.link_layer))
coreaux = cdr(DRTIOAuxController(core.link_layer, self.cpu_dw))
setattr(self.submodules, coreaux_name, coreaux)
self.csr_devices.append(coreaux_name)

View File

@ -132,6 +132,7 @@ class _StandaloneBase(MiniSoC, AMPSoC):
def __init__(self, gateware_identifier_str=None, **kwargs):
MiniSoC.__init__(self,
cpu_type="vexriscv",
cpu_bus_width=64,
sdram_controller_type="minicon",
l2_size=128*1024,
integrated_sram_size=8192,
@ -176,7 +177,7 @@ class _StandaloneBase(MiniSoC, AMPSoC):
self.csr_devices.append("rtio_core")
self.submodules.rtio = rtio.KernelInitiator(self.rtio_tsc)
self.submodules.rtio_dma = ClockDomainsRenamer("sys_kernel")(
rtio.DMA(self.get_native_sdram_if()))
rtio.DMA(self.get_native_sdram_if(), self.cpu_dw))
self.register_kernel_cpu_csrdevice("rtio")
self.register_kernel_cpu_csrdevice("rtio_dma")
self.submodules.cri_con = rtio.CRIInterconnectShared(
@ -192,7 +193,7 @@ class _StandaloneBase(MiniSoC, AMPSoC):
self.rtio_crg.cd_rtio.clk)
self.submodules.rtio_analyzer = rtio.Analyzer(self.rtio_tsc, self.rtio_core.cri,
self.get_native_sdram_if())
self.get_native_sdram_if(), cpu_dw=self.cpu_dw)
self.csr_devices.append("rtio_analyzer")
@ -209,6 +210,7 @@ class _MasterBase(MiniSoC, AMPSoC):
def __init__(self, gateware_identifier_str=None, **kwargs):
MiniSoC.__init__(self,
cpu_type="vexriscv",
cpu_bus_width=64,
sdram_controller_type="minicon",
l2_size=128*1024,
integrated_sram_size=8192,
@ -263,7 +265,7 @@ class _MasterBase(MiniSoC, AMPSoC):
self.drtio_cri.append(core.cri)
self.csr_devices.append(core_name)
coreaux = cdr(DRTIOAuxController(core.link_layer))
coreaux = cdr(DRTIOAuxController(core.link_layer, self.cpu_dw))
setattr(self.submodules, coreaux_name, coreaux)
self.csr_devices.append(coreaux_name)
@ -321,7 +323,7 @@ class _MasterBase(MiniSoC, AMPSoC):
self.submodules.rtio = rtio.KernelInitiator(self.rtio_tsc)
self.submodules.rtio_dma = ClockDomainsRenamer("sys_kernel")(
rtio.DMA(self.get_native_sdram_if()))
rtio.DMA(self.get_native_sdram_if(), self.cpu_dw))
self.register_kernel_cpu_csrdevice("rtio")
self.register_kernel_cpu_csrdevice("rtio_dma")
self.submodules.cri_con = rtio.CRIInterconnectShared(
@ -342,6 +344,7 @@ class _SatelliteBase(BaseSoC):
def __init__(self, gateware_identifier_str=None, sma_as_sat=False, **kwargs):
BaseSoC.__init__(self,
cpu_type="vexriscv",
cpu_bus_width=64,
sdram_controller_type="minicon",
l2_size=128*1024,
integrated_sram_size=8192,
@ -404,7 +407,7 @@ class _SatelliteBase(BaseSoC):
self.drtio_cri.append(core.cri)
self.csr_devices.append(corerep_name)
coreaux = cdr(DRTIOAuxController(core.link_layer))
coreaux = cdr(DRTIOAuxController(core.link_layer, self.cpu_dw))
setattr(self.submodules, coreaux_name, coreaux)
self.csr_devices.append(coreaux_name)

View File

@ -43,6 +43,7 @@ class Master(MiniSoC, AMPSoC):
def __init__(self, gateware_identifier_str=None, **kwargs):
MiniSoC.__init__(self,
cpu_type="vexriscv",
cpu_bus_width=64,
sdram_controller_type="minicon",
l2_size=128*1024,
integrated_sram_size=8192,
@ -96,7 +97,7 @@ class Master(MiniSoC, AMPSoC):
drtio_cri.append(core.cri)
self.csr_devices.append(core_name)
coreaux = cdr(DRTIOAuxController(core.link_layer))
coreaux = cdr(DRTIOAuxController(core.link_layer, self.cpu_dw))
setattr(self.submodules, coreaux_name, coreaux)
self.csr_devices.append(coreaux_name)
@ -149,7 +150,7 @@ class Master(MiniSoC, AMPSoC):
self.submodules.rtio = rtio.KernelInitiator(self.rtio_tsc)
self.submodules.rtio_dma = ClockDomainsRenamer("sys_kernel")(
rtio.DMA(self.get_native_sdram_if()))
rtio.DMA(self.get_native_sdram_if(), self.cpu_dw))
self.register_kernel_cpu_csrdevice("rtio")
self.register_kernel_cpu_csrdevice("rtio_dma")
self.submodules.cri_con = rtio.CRIInterconnectShared(

View File

@ -57,6 +57,7 @@ class SatelliteBase(MiniSoC):
def __init__(self, rtio_clk_freq=125e6, identifier_suffix="", gateware_identifier_str=None, with_sfp=False, *, with_wrpll, **kwargs):
MiniSoC.__init__(self,
cpu_type="vexriscv",
cpu_bus_width=64,
sdram_controller_type="minicon",
l2_size=128*1024,
integrated_sram_size=8192,
@ -117,7 +118,7 @@ class SatelliteBase(MiniSoC):
self.drtio_cri.append(core.cri)
self.csr_devices.append(corerep_name)
coreaux = cdr(DRTIOAuxController(core.link_layer))
coreaux = cdr(DRTIOAuxController(core.link_layer, self.cpu_dw))
setattr(self.submodules, coreaux_name, coreaux)
self.csr_devices.append(coreaux_name)

View File

@ -37,6 +37,7 @@ class _SatelliteBase(BaseSoC):
def __init__(self, rtio_clk_freq, *, with_wrpll, gateware_identifier_str, **kwargs):
BaseSoC.__init__(self,
cpu_type="vexriscv",
cpu_bus_width=64,
**kwargs)
add_identifier(self, gateware_identifier_str=gateware_identifier_str)
self.rtio_clk_freq = rtio_clk_freq
@ -79,7 +80,7 @@ class _SatelliteBase(BaseSoC):
self.submodules.drtiosat = core
self.csr_devices.append("drtiosat")
coreaux = cdr(DRTIOAuxController(core.link_layer))
coreaux = cdr(DRTIOAuxController(core.link_layer, self.cpu_dw))
self.submodules.drtioaux0 = coreaux
self.csr_devices.append("drtioaux0")

View File

@ -33,63 +33,67 @@ class Loopback(Module):
class TB(Module):
def __init__(self, nwords):
def __init__(self, nwords, dw):
self.submodules.link_layer = Loopback(nwords)
self.submodules.aux_controller = ClockDomainsRenamer(
{"rtio": "sys", "rtio_rx": "sys"})(DRTIOAuxController(self.link_layer))
{"rtio": "sys", "rtio_rx": "sys"})(DRTIOAuxController(self.link_layer, dw))
class TestAuxController(unittest.TestCase):
def test_aux_controller(self):
dut = TB(4)
dut = {
32: TB(4, 32),
64: TB(4, 64)
}
def link_init():
def link_init(dw):
for i in range(8):
yield
yield dut.link_layer.ready.eq(1)
yield dut[dw].link_layer.ready.eq(1)
def send_packet(packet):
def send_packet(packet, dw):
for i, d in enumerate(packet):
yield from dut.aux_controller.bus.write(i, d)
yield from dut.aux_controller.transmitter.aux_tx_length.write(len(packet)*4)
yield from dut.aux_controller.transmitter.aux_tx.write(1)
yield from dut[dw].aux_controller.bus.write(i, d)
yield from dut[dw].aux_controller.transmitter.aux_tx_length.write(len(packet)*dw//8)
yield from dut[dw].aux_controller.transmitter.aux_tx.write(1)
yield
while (yield from dut.aux_controller.transmitter.aux_tx.read()):
while (yield from dut[dw].aux_controller.transmitter.aux_tx.read()):
yield
def receive_packet():
while not (yield from dut.aux_controller.receiver.aux_rx_present.read()):
def receive_packet(dw):
while not (yield from dut[dw].aux_controller.receiver.aux_rx_present.read()):
yield
length = yield from dut.aux_controller.receiver.aux_rx_length.read()
length = yield from dut[dw].aux_controller.receiver.aux_rx_length.read()
r = []
for i in range(length//4):
r.append((yield from dut.aux_controller.bus.read(256+i)))
yield from dut.aux_controller.receiver.aux_rx_present.write(1)
for i in range(length//(dw//8)):
r.append((yield from dut[dw].aux_controller.bus.read(256+i)))
yield from dut[dw].aux_controller.receiver.aux_rx_present.write(1)
return r
prng = random.Random(0)
def send_and_check_packet():
data = [prng.randrange(2**32-1) for _ in range(prng.randrange(1, 16))]
yield from send_packet(data)
received = yield from receive_packet()
def send_and_check_packet(dw):
data = [prng.randrange(2**dw-1) for _ in range(prng.randrange(1, 16))]
yield from send_packet(data, dw)
received = yield from receive_packet(dw)
self.assertEqual(data, received)
def sim():
yield from link_init()
def sim(dw):
yield from link_init(dw)
for i in range(8):
yield from send_and_check_packet()
yield from send_and_check_packet(dw)
@passive
def rt_traffic():
def rt_traffic(dw):
while True:
while prng.randrange(4):
yield
yield dut.link_layer.tx_rt_frame.eq(1)
yield dut[dw].link_layer.tx_rt_frame.eq(1)
yield
while prng.randrange(4):
yield
yield dut.link_layer.tx_rt_frame.eq(0)
yield dut[dw].link_layer.tx_rt_frame.eq(0)
yield
run_simulation(dut, [sim(), rt_traffic()])
run_simulation(dut[32], [sim(32), rt_traffic(32)])
run_simulation(dut[64], [sim(64), rt_traffic(64)])

View File

@ -31,24 +31,25 @@ def encode_record(channel, timestamp, address, data):
return encode_n(len(r)+1, 1, 1) + r
def pack(x, size):
def pack(x, size, dw):
r = []
for i in range((len(x)+size-1)//size):
n = 0
for j in range(i*size, (i+1)*size):
n <<= 8
for j in range(i*size//(dw//8), (i+1)*size//(dw//8)):
n <<= dw
try:
n |= x[j]
encoded = int.from_bytes(x[j*(dw//8): (j+1)*(dw//8)], "little")
n |= encoded
except IndexError:
pass
r.append(n)
return r
def encode_sequence(writes, ws):
def encode_sequence(writes, ws, dw):
sequence = [b for write in writes for b in encode_record(*write)]
sequence.append(0)
return pack(sequence, ws)
return pack(sequence, ws, dw)
def do_dma(dut, address):
@ -84,9 +85,9 @@ prng = random.Random(0)
class TB(Module):
def __init__(self, ws):
sequence1 = encode_sequence(test_writes1, ws)
sequence2 = encode_sequence(test_writes2, ws)
def __init__(self, ws, dw):
sequence1 = encode_sequence(test_writes1, ws, dw)
sequence2 = encode_sequence(test_writes2, ws, dw)
offset = 512//ws
assert len(sequence1) < offset
sequence = (
@ -97,7 +98,7 @@ class TB(Module):
bus = wishbone.Interface(ws*8)
self.submodules.memory = wishbone.SRAM(
1024, init=sequence, bus=bus)
self.submodules.dut = dma.DMA(bus)
self.submodules.dut = dma.DMA(bus, dw)
test_writes_full_stack = [
@ -109,7 +110,7 @@ test_writes_full_stack = [
class FullStackTB(Module):
def __init__(self, ws):
def __init__(self, ws, dw):
self.ttl0 = Signal()
self.ttl1 = Signal()
@ -121,12 +122,12 @@ class FullStackTB(Module):
rtio.Channel.from_phy(self.phy1)
]
sequence = encode_sequence(test_writes_full_stack, ws)
sequence = encode_sequence(test_writes_full_stack, ws, dw)
bus = wishbone.Interface(ws*8)
bus = wishbone.Interface(ws*8, 32-log2_int(dw//8))
self.submodules.memory = wishbone.SRAM(
256, init=sequence, bus=bus)
self.submodules.dut = dma.DMA(bus)
self.submodules.dut = dma.DMA(bus, dw)
self.submodules.tsc = rtio.TSC("async")
self.submodules.rtio = rtio.Core(self.tsc, rtio_channels)
self.comb += self.dut.cri.connect(self.rtio.cri)
@ -134,16 +135,22 @@ class FullStackTB(Module):
class TestDMA(unittest.TestCase):
def test_dma_noerror(self):
tb = TB(64)
tb = {
32: TB(64, 32),
64: TB(64, 64)
}
def do_writes():
yield from do_dma(tb.dut, 0)
yield from do_dma(tb.dut, 512)
def do_writes(dw):
yield from do_dma(tb[dw].dut, 0)
yield from do_dma(tb[dw].dut, 512)
received = []
received = {
32: [],
64: []
}
@passive
def rtio_sim():
dut_cri = tb.dut.cri
def rtio_sim(dw):
dut_cri = tb[dw].dut.cri
while True:
cmd = yield dut_cri.cmd
if cmd == cri.commands["nop"]:
@ -153,7 +160,7 @@ class TestDMA(unittest.TestCase):
timestamp = yield dut_cri.o_timestamp
address = yield dut_cri.o_address
data = yield dut_cri.o_data
received.append((channel, timestamp, address, data))
received[dw].append((channel, timestamp, address, data))
yield dut_cri.o_status.eq(1)
for i in range(prng.randrange(10)):
@ -163,32 +170,46 @@ class TestDMA(unittest.TestCase):
self.fail("unexpected RTIO command")
yield
run_simulation(tb, [do_writes(), rtio_sim()])
self.assertEqual(received, test_writes1 + test_writes2)
run_simulation(tb[32], [do_writes(32), rtio_sim(32)])
self.assertEqual(received[32], test_writes1 + test_writes2)
run_simulation(tb[64], [do_writes(64), rtio_sim(64)])
self.assertEqual(received[64], test_writes1 + test_writes2)
def test_full_stack(self):
tb = FullStackTB(64)
tb = {
32: FullStackTB(64, 32),
64: FullStackTB(64, 64)
}
ttl_changes = []
ttl_changes = {
32: [],
64: []
}
@passive
def monitor():
def monitor(dw):
old_ttl_states = [0, 0]
for time in itertools.count():
ttl_states = [
(yield tb.ttl0),
(yield tb.ttl1)
(yield tb[dw].ttl0),
(yield tb[dw].ttl1)
]
for i, (old, new) in enumerate(zip(old_ttl_states, ttl_states)):
if new != old:
ttl_changes.append((time, i))
ttl_changes[dw].append((time, i))
old_ttl_states = ttl_states
yield
run_simulation(tb, {"sys": [
do_dma(tb.dut, 0), monitor(),
run_simulation(tb[32], {"sys": [
do_dma(tb[32].dut, 0), monitor(32),
(None for _ in range(70)),
]}, {"sys": 8, "rsys": 8, "rtio": 8, "rio": 8, "rio_phy": 8})
run_simulation(tb[64], {"sys": [
do_dma(tb[64].dut, 0), monitor(64),
(None for _ in range(70)),
]}, {"sys": 8, "rsys": 8, "rtio": 8, "rio": 8, "rio_phy": 8})
correct_changes = [(timestamp + 11, channel)
for channel, timestamp, _, _ in test_writes_full_stack]
self.assertEqual(ttl_changes, correct_changes)
self.assertEqual(ttl_changes[32], correct_changes)
self.assertEqual(ttl_changes[64], correct_changes)

View File

@ -44,11 +44,11 @@
"src-migen": {
"flake": false,
"locked": {
"lastModified": 1634182166,
"narHash": "sha256-Iw2d8fCgwuuIevkugSqd8Iplj6N+2nR1pn+or5E38Fk=",
"lastModified": 1636715924,
"narHash": "sha256-V3ThFSo2d7OC4SHE0lCkKGQKeFXmvxtwZRWe5NMU3nM=",
"owner": "m-labs",
"repo": "migen",
"rev": "7507a2bb16dd2cac63535175ce67fb30dfdae1c0",
"rev": "9a0be7a4210ff96043412539eb5388659b81831d",
"type": "github"
},
"original": {