configs: Add an example elastic trace generation script (#415)

Current [TraceCPU
documentation](https://www.gem5.org/documentation/general_docs/cpu_models/TraceCPU)
still references the deprecated **se.py/fs.py** scripts for elastic
trace generation (script paths are also outdated).

With this PR we provide a simpler Arm based elastic trace generation
script that can
be used out of the box by a user or that can be extended as needed.
This commit is contained in:
Bobby R. Bruce
2023-10-09 14:11:33 -07:00
committed by GitHub
4 changed files with 357 additions and 114 deletions

View File

@@ -0,0 +1,58 @@
# Copyright (c) 2012, 2017-2018, 2023 Arm Limited
# All rights reserved.
#
# The license below extends only to copyright in the software and shall
# not be construed as granting a license to any other intellectual
# property including but not limited to intellectual property relating
# to a hardware implementation of the functionality of the software
# licensed hereunder. You may use the software subject to the license
# terms below provided that you ensure that this notice is replicated
# unmodified and in its entirety in all distributions of the software,
# modified or unmodified, in source code or in binary form.
#
# Redistribution and use in source and binary forms, with or without
# modification, are permitted provided that the following conditions are
# met: redistributions of source code must retain the above copyright
# notice, this list of conditions and the following disclaimer;
# redistributions in binary form must reproduce the above copyright
# notice, this list of conditions and the following disclaimer in the
# documentation and/or other materials provided with the distribution;
# neither the name of the copyright holders nor the names of its
# contributors may be used to endorse or promote products derived from
# this software without specific prior written permission.
#
# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
# "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
# LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
# A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
# OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
# SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
# LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
# DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
# THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
from m5.objects import *
from .O3_ARM_v7a import O3_ARM_v7a_3
# O3_ARM_v7a_3 adapted to generate elastic traces
class O3_ARM_v7a_3_Etrace(O3_ARM_v7a_3):
# Make the number of entries in the ROB, LQ and SQ very
# large so that there are no stalls due to resource
# limitation as such stalls will get captured in the trace
# as compute delay. For replay, ROB, LQ and SQ sizes are
# modelled in the Trace CPU.
numROBEntries = 512
LQEntries = 128
SQEntries = 128
def attach_probe_listener(self, inst_trace_file, data_trace_file):
# Attach the elastic trace probe listener. Set the protobuf trace
# file names. Set the dependency window size equal to the cpu it
# is attached to.
self.traceListener = m5.objects.ElasticTrace(
instFetchTraceFile=inst_trace_file,
dataDepTraceFile=data_trace_file,
depWindowSize=3 * self.numROBEntries,
)

View File

@@ -338,56 +338,15 @@ class FastmodelCluster(CpuCluster):
pass
class BaseSimpleSystem(ArmSystem):
cache_line_size = 64
def __init__(self, mem_size, platform, **kwargs):
super(BaseSimpleSystem, self).__init__(**kwargs)
self.voltage_domain = VoltageDomain(voltage="1.0V")
self.clk_domain = SrcClockDomain(
clock="1GHz", voltage_domain=Parent.voltage_domain
)
if platform is None:
self.realview = VExpress_GEM5_V1()
else:
self.realview = platform
if hasattr(self.realview.gic, "cpu_addr"):
self.gic_cpu_addr = self.realview.gic.cpu_addr
self.terminal = Terminal()
self.vncserver = VncServer()
self.iobus = IOXBar()
# Device DMA -> MEM
self.mem_ranges = self.getMemRanges(int(Addr(mem_size)))
class ClusterSystem:
"""
Base class providing cpu clusters generation/handling methods to
SE/FS systems
"""
def __init__(self, **kwargs):
self._clusters = []
def getMemRanges(self, mem_size):
"""
Define system memory ranges. This depends on the physical
memory map provided by the realview platform and by the memory
size provided by the user (mem_size argument).
The method is iterating over all platform ranges until they cover
the entire user's memory requirements.
"""
mem_ranges = []
for mem_range in self.realview._mem_regions:
size_in_range = min(mem_size, mem_range.size())
mem_ranges.append(
AddrRange(start=mem_range.start, size=size_in_range)
)
mem_size -= size_in_range
if mem_size == 0:
return mem_ranges
raise ValueError("memory size too big for platform capabilities")
def numCpuClusters(self):
return len(self._clusters)
@@ -423,6 +382,80 @@ class BaseSimpleSystem(ArmSystem):
cluster.connectMemSide(cluster_mem_bus)
class SimpleSeSystem(System, ClusterSystem):
"""
Example system class for syscall emulation mode
"""
# Use a fixed cache line size of 64 bytes
cache_line_size = 64
def __init__(self, **kwargs):
System.__init__(self, **kwargs)
ClusterSystem.__init__(self, **kwargs)
# Create a voltage and clock domain for system components
self.voltage_domain = VoltageDomain(voltage="3.3V")
self.clk_domain = SrcClockDomain(
clock="1GHz", voltage_domain=self.voltage_domain
)
# Create the off-chip memory bus.
self.membus = SystemXBar()
def connect(self):
self.system_port = self.membus.cpu_side_ports
class BaseSimpleSystem(ArmSystem, ClusterSystem):
cache_line_size = 64
def __init__(self, mem_size, platform, **kwargs):
ArmSystem.__init__(self, **kwargs)
ClusterSystem.__init__(self, **kwargs)
self.voltage_domain = VoltageDomain(voltage="1.0V")
self.clk_domain = SrcClockDomain(
clock="1GHz", voltage_domain=Parent.voltage_domain
)
if platform is None:
self.realview = VExpress_GEM5_V1()
else:
self.realview = platform
if hasattr(self.realview.gic, "cpu_addr"):
self.gic_cpu_addr = self.realview.gic.cpu_addr
self.terminal = Terminal()
self.vncserver = VncServer()
self.iobus = IOXBar()
# Device DMA -> MEM
self.mem_ranges = self.getMemRanges(int(Addr(mem_size)))
def getMemRanges(self, mem_size):
"""
Define system memory ranges. This depends on the physical
memory map provided by the realview platform and by the memory
size provided by the user (mem_size argument).
The method is iterating over all platform ranges until they cover
the entire user's memory requirements.
"""
mem_ranges = []
for mem_range in self.realview._mem_regions:
size_in_range = min(mem_size, mem_range.size())
mem_ranges.append(
AddrRange(start=mem_range.start, size=size_in_range)
)
mem_size -= size_in_range
if mem_size == 0:
return mem_ranges
raise ValueError("memory size too big for platform capabilities")
class SimpleSystem(BaseSimpleSystem):
"""
Meant to be used with the classic memory model

View File

@@ -0,0 +1,191 @@
# Copyright (c) 2016-2017, 2022-2023 Arm Limited
# All rights reserved.
#
# The license below extends only to copyright in the software and shall
# not be construed as granting a license to any other intellectual
# property including but not limited to intellectual property relating
# to a hardware implementation of the functionality of the software
# licensed hereunder. You may use the software subject to the license
# terms below provided that you ensure that this notice is replicated
# unmodified and in its entirety in all distributions of the software,
# modified or unmodified, in source code or in binary form.
#
# Redistribution and use in source and binary forms, with or without
# modification, are permitted provided that the following conditions are
# met: redistributions of source code must retain the above copyright
# notice, this list of conditions and the following disclaimer;
# redistributions in binary form must reproduce the above copyright
# notice, this list of conditions and the following disclaimer in the
# documentation and/or other materials provided with the distribution;
# neither the name of the copyright holders nor the names of its
# contributors may be used to endorse or promote products derived from
# this software without specific prior written permission.
#
# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
# "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
# LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
# A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
# OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
# SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
# LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
# DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
# THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
import os
import m5
from m5.util import addToPath
from m5.objects import *
import argparse
import shlex
m5.util.addToPath("../..")
from common import ObjectList
import devices
def get_processes(cmd):
"""Interprets commands to run and returns a list of processes"""
cwd = os.getcwd()
multiprocesses = []
for idx, c in enumerate(cmd):
argv = shlex.split(c)
process = Process(pid=100 + idx, cwd=cwd, cmd=argv, executable=argv[0])
process.gid = os.getgid()
print("info: %d. command and arguments: %s" % (idx + 1, process.cmd))
multiprocesses.append(process)
return multiprocesses
def create(args):
"""Create and configure the system object."""
system = devices.SimpleSeSystem(
mem_mode="timing",
)
# Add CPUs to the system. A cluster of CPUs typically have
# private L1 caches and a shared L2 cache.
system.cpu_cluster = devices.ArmCpuCluster(
system,
args.num_cores,
args.cpu_freq,
"1.2V",
ObjectList.cpu_list.get("O3_ARM_v7a_3_Etrace"),
devices.L1I,
devices.L1D,
devices.L2,
)
# Attach the elastic trace probe listener to every CPU in the cluster
for cpu in system.cpu_cluster:
cpu.attach_probe_listener(args.inst_trace_file, args.data_trace_file)
# As elastic trace generation is enabled, make sure the memory system is
# minimal so that compute delays do not include memory access latencies.
# Configure the compulsory L1 caches for the O3CPU, do not configure
# any more caches.
system.addCaches(True, last_cache_level=1)
# For elastic trace, over-riding Simple Memory latency to 1ns."
system.memory = SimpleMemory(
range=AddrRange(start=0, size=args.mem_size),
latency="1ns",
port=system.membus.mem_side_ports,
)
# Parse the command line and get a list of Processes instances
# that we can pass to gem5.
processes = get_processes(args.commands_to_run)
if len(processes) != args.num_cores:
print(
"Error: Cannot map %d command(s) onto %d CPU(s)"
% (len(processes), args.num_cores)
)
sys.exit(1)
system.workload = SEWorkload.init_compatible(processes[0].executable)
# Assign one workload to each CPU
for cpu, workload in zip(system.cpu_cluster.cpus, processes):
cpu.workload = workload
return system
def main():
parser = argparse.ArgumentParser(epilog=__doc__)
parser.add_argument(
"commands_to_run",
metavar="command(s)",
nargs="+",
help="Command(s) to run",
)
parser.add_argument(
"--inst-trace-file",
action="store",
type=str,
help="""Instruction fetch trace file input to
Elastic Trace probe in a capture simulation and
Trace CPU in a replay simulation""",
default="fetchtrace.proto.gz",
)
parser.add_argument(
"--data-trace-file",
action="store",
type=str,
help="""Data dependency trace file input to
Elastic Trace probe in a capture simulation and
Trace CPU in a replay simulation""",
default="deptrace.proto.gz",
)
parser.add_argument("--cpu-freq", type=str, default="4GHz")
parser.add_argument(
"--num-cores", type=int, default=1, help="Number of CPU cores"
)
parser.add_argument(
"--mem-size",
action="store",
type=str,
default="2GB",
help="Specify the physical memory size",
)
args = parser.parse_args()
# Create a single root node for gem5's object hierarchy. There can
# only exist one root node in the simulator at any given
# time. Tell gem5 that we want to use syscall emulation mode
# instead of full system mode.
root = Root(full_system=False)
# Populate the root node with a system. A system corresponds to a
# single node with shared memory.
root.system = create(args)
# Instantiate the C++ object hierarchy. After this point,
# SimObjects can't be instantiated anymore.
m5.instantiate()
# Start the simulator. This gives control to the C++ world and
# starts the simulator. The returned event tells the simulation
# script why the simulator exited.
event = m5.simulate()
# Print the reason for the simulation exit. Some exit codes are
# requests for service (e.g., checkpoints) from the simulation
# script. We'll just ignore them here and exit.
print(f"{event.getCause()} ({event.getCode()}) @ {m5.curTick()}")
if __name__ == "__m5_main__":
main()

View File

@@ -64,72 +64,6 @@ cpu_types = {
}
class SimpleSeSystem(System):
"""
Example system class for syscall emulation mode
"""
# Use a fixed cache line size of 64 bytes
cache_line_size = 64
def __init__(self, args, **kwargs):
super(SimpleSeSystem, self).__init__(**kwargs)
# Setup book keeping to be able to use CpuClusters from the
# devices module.
self._clusters = []
self._num_cpus = 0
# Create a voltage and clock domain for system components
self.voltage_domain = VoltageDomain(voltage="3.3V")
self.clk_domain = SrcClockDomain(
clock="1GHz", voltage_domain=self.voltage_domain
)
# Create the off-chip memory bus.
self.membus = SystemXBar()
# Wire up the system port that gem5 uses to load the kernel
# and to perform debug accesses.
self.system_port = self.membus.cpu_side_ports
# Add CPUs to the system. A cluster of CPUs typically have
# private L1 caches and a shared L2 cache.
self.cpu_cluster = devices.ArmCpuCluster(
self,
args.num_cores,
args.cpu_freq,
"1.2V",
*cpu_types[args.cpu],
tarmac_gen=args.tarmac_gen,
tarmac_dest=args.tarmac_dest,
)
# Create a cache hierarchy (unless we are simulating a
# functional CPU in atomic memory mode) for the CPU cluster
# and connect it to the shared memory bus.
if self.cpu_cluster.memory_mode() == "timing":
self.cpu_cluster.addL1()
self.cpu_cluster.addL2(self.cpu_cluster.clk_domain)
self.cpu_cluster.connectMemSide(self.membus)
# Tell gem5 about the memory mode used by the CPUs we are
# simulating.
self.mem_mode = self.cpu_cluster.memory_mode()
def numCpuClusters(self):
return len(self._clusters)
def addCpuCluster(self, cpu_cluster):
assert cpu_cluster not in self._clusters
assert len(cpu_cluster) > 0
self._clusters.append(cpu_cluster)
self._num_cpus += len(cpu_cluster)
def numCpus(self):
return self._num_cpus
def get_processes(cmd):
"""Interprets commands to run and returns a list of processes"""
@@ -150,7 +84,31 @@ def get_processes(cmd):
def create(args):
"""Create and configure the system object."""
system = SimpleSeSystem(args)
cpu_class = cpu_types[args.cpu][0]
mem_mode = cpu_class.memory_mode()
# Only simulate caches when using a timing CPU (e.g., the HPI model)
want_caches = True if mem_mode == "timing" else False
system = devices.SimpleSeSystem(
mem_mode=mem_mode,
)
# Add CPUs to the system. A cluster of CPUs typically have
# private L1 caches and a shared L2 cache.
system.cpu_cluster = devices.ArmCpuCluster(
system,
args.num_cores,
args.cpu_freq,
"1.2V",
*cpu_types[args.cpu],
tarmac_gen=args.tarmac_gen,
tarmac_dest=args.tarmac_dest,
)
# Create a cache hierarchy for the cluster. We are assuming that
# clusters have core-private L1 caches and an L2 that's shared
# within the cluster.
system.addCaches(want_caches, last_cache_level=2)
# Tell components about the expected physical memory ranges. This
# is, for example, used by the MemConfig helper to determine where
@@ -160,6 +118,9 @@ def create(args):
# Configure the off-chip memory system.
MemConfig.config_mem(args, system)
# Wire up the system's memory system
system.connect()
# Parse the command line and get a list of Processes instances
# that we can pass to gem5.
processes = get_processes(args.commands_to_run)