gem5/configs/splash2/cluster.py

305 lines
10 KiB
Python
Raw Normal View History

# Copyright (c) 2006-2007 The Regents of The University of Michigan
# All rights reserved.
#
# 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.
#
# Authors: Ron Dreslinski
# Simple test script
#
# "m5 test.py"
import os
import optparse
import sys
import m5
from m5.objects import *
m5.util.addToPath('../common')
# --------------------
# Define Command Line Options
# ====================
parser = optparse.OptionParser()
parser.add_option("-d", "--detailed", action="store_true")
parser.add_option("-t", "--timing", action="store_true")
parser.add_option("-m", "--maxtick", type="int")
parser.add_option("-c", "--numclusters",
help="Number of clusters", type="int")
parser.add_option("-n", "--numcpus",
help="Number of cpus in total", type="int")
parser.add_option("-f", "--frequency",
default = "1GHz",
help="Frequency of each CPU")
parser.add_option("--l1size",
default = "32kB")
parser.add_option("--l1latency",
default = 1)
parser.add_option("--l2size",
default = "256kB")
parser.add_option("--l2latency",
default = 10)
parser.add_option("--rootdir",
help="ROot directory of Splash2",
default="/dist/splash2/codes/")
parser.add_option("-b", "--benchmark",
help="Splash 2 benchmark to run")
(options, args) = parser.parse_args()
if args:
print "Error: script doesn't take any positional arguments"
sys.exit(1)
# --------------------
# Define Splash2 Benchmarks
# ====================
class Cholesky(LiveProcess):
executable = options.rootdir + '/kernels/cholesky/CHOLESKY'
cmd = 'CHOLESKY -p' + str(options.numcpus) + ' '\
+ options.rootdir + '/kernels/cholesky/inputs/tk23.O'
class FFT(LiveProcess):
executable = options.rootdir + 'kernels/fft/FFT'
cmd = 'FFT -p' + str(options.numcpus) + ' -m18'
class LU_contig(LiveProcess):
executable = options.rootdir + 'kernels/lu/contiguous_blocks/LU'
cmd = 'LU -p' + str(options.numcpus)
class LU_noncontig(LiveProcess):
executable = options.rootdir + 'kernels/lu/non_contiguous_blocks/LU'
cmd = 'LU -p' + str(options.numcpus)
class Radix(LiveProcess):
executable = options.rootdir + 'kernels/radix/RADIX'
cmd = 'RADIX -n524288 -p' + str(options.numcpus)
class Barnes(LiveProcess):
executable = options.rootdir + 'apps/barnes/BARNES'
cmd = 'BARNES'
input = options.rootdir + 'apps/barnes/input.p' + str(options.numcpus)
class FMM(LiveProcess):
executable = options.rootdir + 'apps/fmm/FMM'
cmd = 'FMM'
input = options.rootdir + 'apps/fmm/inputs/input.2048.p' + str(options.numcpus)
class Ocean_contig(LiveProcess):
executable = options.rootdir + 'apps/ocean/contiguous_partitions/OCEAN'
cmd = 'OCEAN -p' + str(options.numcpus)
class Ocean_noncontig(LiveProcess):
executable = options.rootdir + 'apps/ocean/non_contiguous_partitions/OCEAN'
cmd = 'OCEAN -p' + str(options.numcpus)
class Raytrace(LiveProcess):
executable = options.rootdir + 'apps/raytrace/RAYTRACE'
cmd = 'RAYTRACE -p' + str(options.numcpus) + ' ' \
+ options.rootdir + 'apps/raytrace/inputs/teapot.env'
class Water_nsquared(LiveProcess):
executable = options.rootdir + 'apps/water-nsquared/WATER-NSQUARED'
cmd = 'WATER-NSQUARED'
input = options.rootdir + 'apps/water-nsquared/input.p' + str(options.numcpus)
class Water_spatial(LiveProcess):
executable = options.rootdir + 'apps/water-spatial/WATER-SPATIAL'
cmd = 'WATER-SPATIAL'
input = options.rootdir + 'apps/water-spatial/input.p' + str(options.numcpus)
# --------------------
# Base L1 Cache Definition
# ====================
class L1(BaseCache):
latency = options.l1latency
block_size = 64
mshrs = 12
tgts_per_mshr = 8
# ----------------------
# Base L2 Cache Definition
# ----------------------
class L2(BaseCache):
block_size = 64
latency = options.l2latency
mshrs = 92
tgts_per_mshr = 16
write_buffers = 8
# ----------------------
# Define the clusters with their cpus
# ----------------------
class Cluster:
pass
cpusPerCluster = options.numcpus/options.numclusters
busFrequency = Frequency(options.frequency)
busFrequency *= cpusPerCluster
all_cpus = []
all_l1s = []
all_l1buses = []
if options.timing:
clusters = [ Cluster() for i in xrange(options.numclusters)]
for j in xrange(options.numclusters):
clusters[j].id = j
for cluster in clusters:
Bus: Split the bus into a non-coherent and coherent bus This patch introduces a class hierarchy of buses, a non-coherent one, and a coherent one, splitting the existing bus functionality. By doing so it also enables further specialisation of the two types of buses. A non-coherent bus connects a number of non-snooping masters and slaves, and routes the request and response packets based on the address. The request packets issued by the master connected to a non-coherent bus could still snoop in caches attached to a coherent bus, as is the case with the I/O bus and memory bus in most system configurations. No snoops will, however, reach any master on the non-coherent bus itself. The non-coherent bus can be used as a template for modelling PCI, PCIe, and non-coherent AMBA and OCP buses, and is typically used for the I/O buses. A coherent bus connects a number of (potentially) snooping masters and slaves, and routes the request and response packets based on the address, and also forwards all requests to the snoopers and deals with the snoop responses. The coherent bus can be used as a template for modelling QPI, HyperTransport, ACE and coherent OCP buses, and is typically used for the L1-to-L2 buses and as the main system interconnect. The configuration scripts are updated to use a NoncoherentBus for all peripheral and I/O buses. A bit of minor tidying up has also been done. --HG-- rename : src/mem/bus.cc => src/mem/coherent_bus.cc rename : src/mem/bus.hh => src/mem/coherent_bus.hh rename : src/mem/bus.cc => src/mem/noncoherent_bus.cc rename : src/mem/bus.hh => src/mem/noncoherent_bus.hh
2012-05-31 19:30:04 +02:00
cluster.clusterbus = CoherentBus(clock=busFrequency)
all_l1buses += [cluster.clusterbus]
cluster.cpus = [TimingSimpleCPU(cpu_id = i + cluster.id,
clock=options.frequency)
for i in xrange(cpusPerCluster)]
all_cpus += cluster.cpus
cluster.l1 = L1(size=options.l1size, assoc = 4)
all_l1s += [cluster.l1]
elif options.detailed:
clusters = [ Cluster() for i in xrange(options.numclusters)]
for j in xrange(options.numclusters):
clusters[j].id = j
for cluster in clusters:
Bus: Split the bus into a non-coherent and coherent bus This patch introduces a class hierarchy of buses, a non-coherent one, and a coherent one, splitting the existing bus functionality. By doing so it also enables further specialisation of the two types of buses. A non-coherent bus connects a number of non-snooping masters and slaves, and routes the request and response packets based on the address. The request packets issued by the master connected to a non-coherent bus could still snoop in caches attached to a coherent bus, as is the case with the I/O bus and memory bus in most system configurations. No snoops will, however, reach any master on the non-coherent bus itself. The non-coherent bus can be used as a template for modelling PCI, PCIe, and non-coherent AMBA and OCP buses, and is typically used for the I/O buses. A coherent bus connects a number of (potentially) snooping masters and slaves, and routes the request and response packets based on the address, and also forwards all requests to the snoopers and deals with the snoop responses. The coherent bus can be used as a template for modelling QPI, HyperTransport, ACE and coherent OCP buses, and is typically used for the L1-to-L2 buses and as the main system interconnect. The configuration scripts are updated to use a NoncoherentBus for all peripheral and I/O buses. A bit of minor tidying up has also been done. --HG-- rename : src/mem/bus.cc => src/mem/coherent_bus.cc rename : src/mem/bus.hh => src/mem/coherent_bus.hh rename : src/mem/bus.cc => src/mem/noncoherent_bus.cc rename : src/mem/bus.hh => src/mem/noncoherent_bus.hh
2012-05-31 19:30:04 +02:00
cluster.clusterbus = CoherentBus(clock=busFrequency)
all_l1buses += [cluster.clusterbus]
cluster.cpus = [DerivO3CPU(cpu_id = i + cluster.id,
clock=options.frequency)
for i in xrange(cpusPerCluster)]
all_cpus += cluster.cpus
cluster.l1 = L1(size=options.l1size, assoc = 4)
all_l1s += [cluster.l1]
else:
clusters = [ Cluster() for i in xrange(options.numclusters)]
for j in xrange(options.numclusters):
clusters[j].id = j
for cluster in clusters:
Bus: Split the bus into a non-coherent and coherent bus This patch introduces a class hierarchy of buses, a non-coherent one, and a coherent one, splitting the existing bus functionality. By doing so it also enables further specialisation of the two types of buses. A non-coherent bus connects a number of non-snooping masters and slaves, and routes the request and response packets based on the address. The request packets issued by the master connected to a non-coherent bus could still snoop in caches attached to a coherent bus, as is the case with the I/O bus and memory bus in most system configurations. No snoops will, however, reach any master on the non-coherent bus itself. The non-coherent bus can be used as a template for modelling PCI, PCIe, and non-coherent AMBA and OCP buses, and is typically used for the I/O buses. A coherent bus connects a number of (potentially) snooping masters and slaves, and routes the request and response packets based on the address, and also forwards all requests to the snoopers and deals with the snoop responses. The coherent bus can be used as a template for modelling QPI, HyperTransport, ACE and coherent OCP buses, and is typically used for the L1-to-L2 buses and as the main system interconnect. The configuration scripts are updated to use a NoncoherentBus for all peripheral and I/O buses. A bit of minor tidying up has also been done. --HG-- rename : src/mem/bus.cc => src/mem/coherent_bus.cc rename : src/mem/bus.hh => src/mem/coherent_bus.hh rename : src/mem/bus.cc => src/mem/noncoherent_bus.cc rename : src/mem/bus.hh => src/mem/noncoherent_bus.hh
2012-05-31 19:30:04 +02:00
cluster.clusterbus = CoherentBus(clock=busFrequency)
all_l1buses += [cluster.clusterbus]
cluster.cpus = [AtomicSimpleCPU(cpu_id = i + cluster.id,
clock=options.frequency)
for i in xrange(cpusPerCluster)]
all_cpus += cluster.cpus
cluster.l1 = L1(size=options.l1size, assoc = 4)
all_l1s += [cluster.l1]
# ----------------------
# Create a system, and add system wide objects
# ----------------------
MEM: Enable multiple distributed generalized memories This patch removes the assumption on having on single instance of PhysicalMemory, and enables a distributed memory where the individual memories in the system are each responsible for a single contiguous address range. All memories inherit from an AbstractMemory that encompasses the basic behaviuor of a random access memory, and provides untimed access methods. What was previously called PhysicalMemory is now SimpleMemory, and a subclass of AbstractMemory. All future types of memory controllers should inherit from AbstractMemory. To enable e.g. the atomic CPU and RubyPort to access the now distributed memory, the system has a wrapper class, called PhysicalMemory that is aware of all the memories in the system and their associated address ranges. This class thus acts as an infinitely-fast bus and performs address decoding for these "shortcut" accesses. Each memory can specify that it should not be part of the global address map (used e.g. by the functional memories by some testers). Moreover, each memory can be configured to be reported to the OS configuration table, useful for populating ATAG structures, and any potential ACPI tables. Checkpointing support currently assumes that all memories have the same size and organisation when creating and resuming from the checkpoint. A future patch will enable a more flexible re-organisation. --HG-- rename : src/mem/PhysicalMemory.py => src/mem/AbstractMemory.py rename : src/mem/PhysicalMemory.py => src/mem/SimpleMemory.py rename : src/mem/physical.cc => src/mem/abstract_mem.cc rename : src/mem/physical.hh => src/mem/abstract_mem.hh rename : src/mem/physical.cc => src/mem/simple_mem.cc rename : src/mem/physical.hh => src/mem/simple_mem.hh
2012-04-06 19:46:31 +02:00
system = System(cpu = all_cpus, l1_ = all_l1s, l1bus_ = all_l1buses,
Bus: Split the bus into a non-coherent and coherent bus This patch introduces a class hierarchy of buses, a non-coherent one, and a coherent one, splitting the existing bus functionality. By doing so it also enables further specialisation of the two types of buses. A non-coherent bus connects a number of non-snooping masters and slaves, and routes the request and response packets based on the address. The request packets issued by the master connected to a non-coherent bus could still snoop in caches attached to a coherent bus, as is the case with the I/O bus and memory bus in most system configurations. No snoops will, however, reach any master on the non-coherent bus itself. The non-coherent bus can be used as a template for modelling PCI, PCIe, and non-coherent AMBA and OCP buses, and is typically used for the I/O buses. A coherent bus connects a number of (potentially) snooping masters and slaves, and routes the request and response packets based on the address, and also forwards all requests to the snoopers and deals with the snoop responses. The coherent bus can be used as a template for modelling QPI, HyperTransport, ACE and coherent OCP buses, and is typically used for the L1-to-L2 buses and as the main system interconnect. The configuration scripts are updated to use a NoncoherentBus for all peripheral and I/O buses. A bit of minor tidying up has also been done. --HG-- rename : src/mem/bus.cc => src/mem/coherent_bus.cc rename : src/mem/bus.hh => src/mem/coherent_bus.hh rename : src/mem/bus.cc => src/mem/noncoherent_bus.cc rename : src/mem/bus.hh => src/mem/noncoherent_bus.hh
2012-05-31 19:30:04 +02:00
physmem = SimpleMemory(),
membus = CoherentBus(clock = busFrequency))
Bus: Split the bus into a non-coherent and coherent bus This patch introduces a class hierarchy of buses, a non-coherent one, and a coherent one, splitting the existing bus functionality. By doing so it also enables further specialisation of the two types of buses. A non-coherent bus connects a number of non-snooping masters and slaves, and routes the request and response packets based on the address. The request packets issued by the master connected to a non-coherent bus could still snoop in caches attached to a coherent bus, as is the case with the I/O bus and memory bus in most system configurations. No snoops will, however, reach any master on the non-coherent bus itself. The non-coherent bus can be used as a template for modelling PCI, PCIe, and non-coherent AMBA and OCP buses, and is typically used for the I/O buses. A coherent bus connects a number of (potentially) snooping masters and slaves, and routes the request and response packets based on the address, and also forwards all requests to the snoopers and deals with the snoop responses. The coherent bus can be used as a template for modelling QPI, HyperTransport, ACE and coherent OCP buses, and is typically used for the L1-to-L2 buses and as the main system interconnect. The configuration scripts are updated to use a NoncoherentBus for all peripheral and I/O buses. A bit of minor tidying up has also been done. --HG-- rename : src/mem/bus.cc => src/mem/coherent_bus.cc rename : src/mem/bus.hh => src/mem/coherent_bus.hh rename : src/mem/bus.cc => src/mem/noncoherent_bus.cc rename : src/mem/bus.hh => src/mem/noncoherent_bus.hh
2012-05-31 19:30:04 +02:00
system.toL2bus = CoherentBus(clock = busFrequency)
system.l2 = L2(size = options.l2size, assoc = 8)
# ----------------------
# Connect the L2 cache and memory together
# ----------------------
system.physmem.port = system.membus.master
system.l2.cpu_side = system.toL2bus.slave
system.l2.mem_side = system.membus.master
# ----------------------
# Connect the L2 cache and clusters together
# ----------------------
for cluster in clusters:
cluster.l1.cpu_side = cluster.clusterbus.master
cluster.l1.mem_side = system.toL2bus.slave
for cpu in cluster.cpus:
cpu.icache_port = cluster.clusterbus.slave
cpu.dcache_port = cluster.clusterbus.slave
# ----------------------
# Define the root
# ----------------------
root = Root(full_system = False, system = system)
# --------------------
# Pick the correct Splash2 Benchmarks
# ====================
if options.benchmark == 'Cholesky':
root.workload = Cholesky()
elif options.benchmark == 'FFT':
root.workload = FFT()
elif options.benchmark == 'LUContig':
root.workload = LU_contig()
elif options.benchmark == 'LUNoncontig':
root.workload = LU_noncontig()
elif options.benchmark == 'Radix':
root.workload = Radix()
elif options.benchmark == 'Barnes':
root.workload = Barnes()
elif options.benchmark == 'FMM':
root.workload = FMM()
elif options.benchmark == 'OceanContig':
root.workload = Ocean_contig()
elif options.benchmark == 'OceanNoncontig':
root.workload = Ocean_noncontig()
elif options.benchmark == 'Raytrace':
root.workload = Raytrace()
elif options.benchmark == 'WaterNSquared':
root.workload = Water_nsquared()
elif options.benchmark == 'WaterSpatial':
root.workload = Water_spatial()
else:
m5.util.panic("""
The --benchmark environment variable was set to something improper.
Use Cholesky, FFT, LUContig, LUNoncontig, Radix, Barnes, FMM, OceanContig,
OceanNoncontig, Raytrace, WaterNSquared, or WaterSpatial
""")
# --------------------
# Assign the workload to the cpus
# ====================
for cluster in clusters:
for cpu in cluster.cpus:
cpu.workload = root.workload
# ----------------------
# Run the simulation
# ----------------------
if options.timing or options.detailed:
root.system.mem_mode = 'timing'
# instantiate configuration
m5.instantiate()
# simulate until program terminates
if options.maxtick:
exit_event = m5.simulate(options.maxtick)
else:
exit_event = m5.simulate(m5.MaxTick)
print 'Exiting @ tick', m5.curTick(), 'because', exit_event.getCause()