config, arm: Add multi-core KVM support to bL config

Add support for KVM in the big.LITTLE(tm) example configuration. This
replaces the --atomic option with a --cpu-type option that can be used
to switch between atomic, kvm, and timing simulation.

When running in KVM mode, the simulation script automatically assigns
separate event queues (threads) to each of the simulated CPUs. All
simulated devices, including CPU child devices (e.g., interrupt
controllers and caches), are assigned to event queue 0.

Change-Id: Ic9a3f564db91f5a3d3cb754c5a02fdd5c17d5fdf
Signed-off-by: Andreas Sandberg <andreas.sandberg@arm.com>
Reviewed-by: Curtis Dunham <curtis.dunham@arm.com>
Reviewed-by: Sascha Bischoff <sascha.bischoff@arm.com>
Reviewed-by: Gabor Dozsa <gabor.dozsa@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/2561
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Reviewed-by: Weiping Liao <weipingliao@google.com>
This commit is contained in:
Andreas Sandberg 2017-03-17 11:10:55 +00:00
parent 3547af6e44
commit 9a13acaa36
2 changed files with 89 additions and 24 deletions

View file

@ -1,4 +1,4 @@
# Copyright (c) 2016 ARM Limited # Copyright (c) 2016-2017 ARM Limited
# All rights reserved. # All rights reserved.
# #
# The license below extends only to copyright in the software and shall # The license below extends only to copyright in the software and shall
@ -44,6 +44,8 @@ m5.util.addToPath('../../')
from common.Caches import * from common.Caches import *
from common import CpuConfig from common import CpuConfig
have_kvm = "kvm" in CpuConfig.cpu_names()
class L1I(L1_ICache): class L1I(L1_ICache):
tag_latency = 1 tag_latency = 1
data_latency = 1 data_latency = 1
@ -170,6 +172,14 @@ class AtomicCluster(CpuCluster):
def addL1(self): def addL1(self):
pass pass
class KvmCluster(CpuCluster):
def __init__(self, system, num_cpus, cpu_clock, cpu_voltage="1.0V"):
cpu_config = [ CpuConfig.get("kvm"), None, None, None, None ]
super(KvmCluster, self).__init__(system, num_cpus, cpu_clock,
cpu_voltage, *cpu_config)
def addL1(self):
pass
class SimpleSystem(LinuxArmSystem): class SimpleSystem(LinuxArmSystem):
cache_line_size = 64 cache_line_size = 64

View file

@ -44,6 +44,7 @@ import argparse
import os import os
import sys import sys
import m5 import m5
import m5.util
from m5.objects import * from m5.objects import *
m5.util.addToPath("../../") m5.util.addToPath("../../")
@ -52,6 +53,7 @@ from common import SysPaths
from common import CpuConfig from common import CpuConfig
import devices import devices
from devices import AtomicCluster, KvmCluster
default_dtb = 'armv8_gem5_v1_big_little_2_2.dtb' default_dtb = 'armv8_gem5_v1_big_little_2_2.dtb'
@ -61,6 +63,21 @@ default_rcs = 'bootscript.rcS'
default_mem_size= "2GB" default_mem_size= "2GB"
def _to_ticks(value):
"""Helper function to convert a latency from string format to Ticks"""
return m5.ticks.fromSeconds(m5.util.convert.anyToLatency(value))
def _using_pdes(root):
"""Determine if the simulator is using multiple parallel event queues"""
for obj in root.descendants():
if not m5.proxy.isproxy(obj.eventq_index) and \
obj.eventq_index != root.eventq_index:
return True
return False
class BigCluster(devices.CpuCluster): class BigCluster(devices.CpuCluster):
def __init__(self, system, num_cpus, cpu_clock, def __init__(self, system, num_cpus, cpu_clock,
@ -107,6 +124,15 @@ def createSystem(caches, kernel, bootscript, disks=[]):
return sys return sys
cpu_types = {
"atomic" : (AtomicCluster, AtomicCluster),
"timing" : (BigCluster, LittleCluster),
}
# Only add the KVM CPU if it has been compiled into gem5
if devices.have_kvm:
cpu_types["kvm"] = (KvmCluster, KvmCluster)
def addOptions(parser): def addOptions(parser):
parser.add_argument("--restore-from", type=str, default=None, parser.add_argument("--restore-from", type=str, default=None,
@ -119,8 +145,9 @@ def addOptions(parser):
help="Disks to instantiate") help="Disks to instantiate")
parser.add_argument("--bootscript", type=str, default=default_rcs, parser.add_argument("--bootscript", type=str, default=default_rcs,
help="Linux bootscript") help="Linux bootscript")
parser.add_argument("--atomic", action="store_true", default=False, parser.add_argument("--cpu-type", type=str, choices=cpu_types.keys(),
help="Use atomic CPUs") default="timing",
help="CPU simulation mode. Default: %(default)s")
parser.add_argument("--kernel-init", type=str, default="/sbin/init", parser.add_argument("--kernel-init", type=str, default="/sbin/init",
help="Override init") help="Override init")
parser.add_argument("--big-cpus", type=int, default=1, parser.add_argument("--big-cpus", type=int, default=1,
@ -135,9 +162,11 @@ def addOptions(parser):
help="Big CPU clock frequency") help="Big CPU clock frequency")
parser.add_argument("--little-cpu-clock", type=str, default="1GHz", parser.add_argument("--little-cpu-clock", type=str, default="1GHz",
help="Little CPU clock frequency") help="Little CPU clock frequency")
parser.add_argument("--sim-quantum", type=str, default="1ms",
help="Simulation quantum for parallel simulation. " \
"Default: %(default)s")
return parser return parser
def build(options): def build(options):
m5.ticks.fixGlobalFrequency() m5.ticks.fixGlobalFrequency()
@ -165,35 +194,31 @@ def build(options):
root.system = system root.system = system
system.boot_osflags = " ".join(kernel_cmd) system.boot_osflags = " ".join(kernel_cmd)
AtomicCluster = devices.AtomicCluster
if options.big_cpus + options.little_cpus == 0: if options.big_cpus + options.little_cpus == 0:
m5.util.panic("Empty CPU clusters") m5.util.panic("Empty CPU clusters")
big_model, little_model = cpu_types[options.cpu_type]
all_cpus = []
# big cluster # big cluster
if options.big_cpus > 0: if options.big_cpus > 0:
if options.atomic: system.bigCluster = big_model(system, options.big_cpus,
system.bigCluster = AtomicCluster(system, options.big_cpus, options.big_cpu_clock)
options.big_cpu_clock) system.mem_mode = system.bigCluster.memoryMode()
else: all_cpus += system.bigCluster.cpus
system.bigCluster = BigCluster(system, options.big_cpus,
options.big_cpu_clock)
mem_mode = system.bigCluster.memoryMode()
# little cluster # little cluster
if options.little_cpus > 0: if options.little_cpus > 0:
if options.atomic: system.littleCluster = little_model(system, options.little_cpus,
system.littleCluster = AtomicCluster(system, options.little_cpus, options.little_cpu_clock)
options.little_cpu_clock) system.mem_mode = system.littleCluster.memoryMode()
all_cpus += system.littleCluster.cpus
else: # Figure out the memory mode
system.littleCluster = LittleCluster(system, options.little_cpus, if options.big_cpus > 0 and options.little_cpus > 0 and \
options.little_cpu_clock) system.littleCluster.memoryMode() != system.littleCluster.memoryMode():
mem_mode = system.littleCluster.memoryMode() m5.util.panic("Memory mode missmatch among CPU clusters")
if options.big_cpus > 0 and options.little_cpus > 0:
if system.bigCluster.memoryMode() != system.littleCluster.memoryMode():
m5.util.panic("Memory mode missmatch among CPU clusters")
system.mem_mode = mem_mode
# create caches # create caches
system.addCaches(options.caches, options.last_cache_level) system.addCaches(options.caches, options.last_cache_level)
@ -203,13 +228,43 @@ def build(options):
if options.little_cpus > 0 and system.littleCluster.requireCaches(): if options.little_cpus > 0 and system.littleCluster.requireCaches():
m5.util.panic("Little CPU model requires caches") m5.util.panic("Little CPU model requires caches")
# Create a KVM VM and do KVM-specific configuration
if issubclass(big_model, KvmCluster):
_build_kvm(system, all_cpus)
# Linux device tree # Linux device tree
system.dtb_filename = SysPaths.binary(options.dtb) system.dtb_filename = SysPaths.binary(options.dtb)
return root return root
def _build_kvm(system, cpus):
system.kvm_vm = KvmVM()
# Assign KVM CPUs to their own event queues / threads. This
# has to be done after creating caches and other child objects
# since these mustn't inherit the CPU event queue.
if len(cpus) > 1:
device_eq = 0
first_cpu_eq = 1
for idx, cpu in enumerate(cpus):
# Child objects usually inherit the parent's event
# queue. Override that and use the same event queue for
# all devices.
for obj in cpu.descendants():
obj.eventq_index = device_eq
cpu.eventq_index = first_cpu_eq + idx
def instantiate(options, checkpoint_dir=None): def instantiate(options, checkpoint_dir=None):
# Setup the simulation quantum if we are running in PDES-mode
# (e.g., when using KVM)
root = Root.getInstance()
if root and _using_pdes(root):
m5.util.inform("Running in PDES mode with a %s simulation quantum.",
options.sim_quantum)
root.sim_quantum = _to_ticks(options.sim_quantum)
# Get and load from the chkpt or simpoint checkpoint # Get and load from the chkpt or simpoint checkpoint
if options.restore_from: if options.restore_from:
if checkpoint_dir and not os.path.isabs(options.restore_from): if checkpoint_dir and not os.path.isabs(options.restore_from):