sanchayanmaity/gem5
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity
gem5/configs/ruby/MESI_Three_Level.py

242 lines
9.9 KiB
Python
Raw Normal View History

ruby: add a three level MESI protocol. The first two levels (L0, L1) are private to the core, the third level (L2)is possibly shared. The protocol supports clustered designs. For example, one can have two sets of two cores. Each core has an L0 and L1 cache. There are two L2 controllers where each set accesses only one of the L2 controllers.
2014-01-04 07:03:34 +01:00
# Copyright (c) 2006-2007 The Regents of The University of Michigan
# Copyright (c) 2009 Advanced Micro Devices, Inc.
# Copyright (c) 2013 Mark D. Hill and David A. Wood
# 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: Brad Beckmann
# Nilay Vaish
import math
import m5
from m5.objects import *
from m5.defines import buildEnv
from Ruby import create_topology
x86 isa: This patch attempts an implementation at mwait. Mwait works as follows: 1. A cpu monitors an address of interest (monitor instruction) 2. A cpu calls mwait - this loads the cache line into that cpu's cache. 3. The cpu goes to sleep. 4. When another processor requests write permission for the line, it is evicted from the sleeping cpu's cache. This eviction is forwarded to the sleeping cpu, which then wakes up. Committed by: Nilay Vaish <nilay@cs.wisc.edu>
2014-11-06 12:42:22 +01:00
from Ruby import send_evicts
ruby: add a three level MESI protocol. The first two levels (L0, L1) are private to the core, the third level (L2)is possibly shared. The protocol supports clustered designs. For example, one can have two sets of two cores. Each core has an L0 and L1 cache. There are two L2 controllers where each set accesses only one of the L2 controllers.
2014-01-04 07:03:34 +01:00
#
# Note: the L1 Cache latency is only used by the sequencer on fast path hits
#
class L0Cache(RubyCache):
latency = 1
class L1Cache(RubyCache):
latency = 5
#
# Note: the L2 Cache latency is not currently used
#
class L2Cache(RubyCache):
latency = 15
def define_options(parser):
parser.add_option("--num-clusters", type="int", default=1,
help="number of clusters in a design in which there are shared\
caches private to clusters")
return
ruby: single physical memory in fs mode Both ruby and the system used to maintain memory copies. With the changes carried for programmed io accesses, only one single memory is required for fs simulations. This patch sets the copy of memory that used to reside with the system to null, so that no space is allocated, but address checks can still be carried out. All the memory accesses now source and sink values to the memory maintained by ruby.
2014-11-06 12:41:44 +01:00
def create_system(options, full_system, system, dma_ports, ruby_system):
ruby: add a three level MESI protocol. The first two levels (L0, L1) are private to the core, the third level (L2)is possibly shared. The protocol supports clustered designs. For example, one can have two sets of two cores. Each core has an L0 and L1 cache. There are two L2 controllers where each set accesses only one of the L2 controllers.
2014-01-04 07:03:34 +01:00
if buildEnv['PROTOCOL'] != 'MESI_Three_Level':
fatal("This script requires the MESI_Three_Level protocol to be built.")
cpu_sequencers = []
#
# The ruby network creation expects the list of nodes in the system to be
# consistent with the NetDest list. Therefore the l1 controller nodes must be
# listed before the directory nodes and directory nodes before dma nodes, etc.
#
l0_cntrl_nodes = []
l1_cntrl_nodes = []
l2_cntrl_nodes = []
dir_cntrl_nodes = []
dma_cntrl_nodes = []
assert (options.num_cpus % options.num_clusters == 0)
num_cpus_per_cluster = options.num_cpus / options.num_clusters
assert (options.num_l2caches % options.num_clusters == 0)
num_l2caches_per_cluster = options.num_l2caches / options.num_clusters
l2_bits = int(math.log(num_l2caches_per_cluster, 2))
block_size_bits = int(math.log(options.cacheline_size, 2))
l2_index_start = block_size_bits + l2_bits
#
# Must create the individual controllers before the network to ensure the
# controller constructors are called before the network constructor
#
for i in xrange(options.num_clusters):
for j in xrange(num_cpus_per_cluster):
#
# First create the Ruby objects associated with this cpu
#
l0i_cache = L0Cache(size = '4096B', assoc = 1, is_icache = True,
start_index_bit = block_size_bits, replacement_policy="LRU")
l0d_cache = L0Cache(size = '4096B', assoc = 1, is_icache = False,
start_index_bit = block_size_bits, replacement_policy="LRU")
l0_cntrl = L0Cache_Controller(version = i*num_cpus_per_cluster + j,
Icache = l0i_cache, Dcache = l0d_cache,
x86 isa: This patch attempts an implementation at mwait. Mwait works as follows: 1. A cpu monitors an address of interest (monitor instruction) 2. A cpu calls mwait - this loads the cache line into that cpu's cache. 3. The cpu goes to sleep. 4. When another processor requests write permission for the line, it is evicted from the sleeping cpu's cache. This eviction is forwarded to the sleeping cpu, which then wakes up. Committed by: Nilay Vaish <nilay@cs.wisc.edu>
2014-11-06 12:42:22 +01:00
send_evictions = send_evicts(options),
ruby: Fixes clock domains in configuration files This patch fixes scripts related to ruby by adding the ruby clock domain. Now the L1 controllers and the Sequencer shares the cpu clock domain, while the rest of the components use the ruby clock domain. Before this patch, running simulations with the cpu clock set at 2GHz or 1GHz will output the same time results and could distort power measurements. Committed by: Nilay Vaish <nilay@cs.wisc.edu>
2014-09-01 23:55:30 +02:00
clk_domain=system.cpu[i].clk_domain,
ruby: add a three level MESI protocol. The first two levels (L0, L1) are private to the core, the third level (L2)is possibly shared. The protocol supports clustered designs. For example, one can have two sets of two cores. Each core has an L0 and L1 cache. There are two L2 controllers where each set accesses only one of the L2 controllers.
2014-01-04 07:03:34 +01:00
ruby_system = ruby_system)
cpu_seq = RubySequencer(version = i, icache = l0i_cache,
ruby: Fixes clock domains in configuration files This patch fixes scripts related to ruby by adding the ruby clock domain. Now the L1 controllers and the Sequencer shares the cpu clock domain, while the rest of the components use the ruby clock domain. Before this patch, running simulations with the cpu clock set at 2GHz or 1GHz will output the same time results and could distort power measurements. Committed by: Nilay Vaish <nilay@cs.wisc.edu>
2014-09-01 23:55:30 +02:00
clk_domain=system.cpu[i].clk_domain,
ruby: add a three level MESI protocol. The first two levels (L0, L1) are private to the core, the third level (L2)is possibly shared. The protocol supports clustered designs. For example, one can have two sets of two cores. Each core has an L0 and L1 cache. There are two L2 controllers where each set accesses only one of the L2 controllers.
2014-01-04 07:03:34 +01:00
dcache = l0d_cache, ruby_system = ruby_system)
l0_cntrl.sequencer = cpu_seq
l1_cache = L1Cache(size = options.l1d_size, assoc = options.l1d_assoc,
start_index_bit = block_size_bits, is_icache = False)
l1_cntrl = L1Cache_Controller(version = i*num_cpus_per_cluster+j,
cache = l1_cache, l2_select_num_bits = l2_bits,
cluster_id = i, ruby_system = ruby_system)
exec("ruby_system.l0_cntrl%d = l0_cntrl" % (
i*num_cpus_per_cluster+j))
exec("ruby_system.l1_cntrl%d = l1_cntrl" % (
i*num_cpus_per_cluster+j))
#
# Add controllers and sequencers to the appropriate lists
#
cpu_sequencers.append(cpu_seq)
l0_cntrl_nodes.append(l0_cntrl)
l1_cntrl_nodes.append(l1_cntrl)
ruby: message buffers: significant changes This patch is the final patch in a series of patches. The aim of the series is to make ruby more configurable than it was. More specifically, the connections between controllers are not at all possible (unless one is ready to make significant changes to the coherence protocol). Moreover the buffers themselves are magically connected to the network inside the slicc code. These connections are not part of the configuration file. This patch makes changes so that these connections will now be made in the python configuration files associated with the protocols. This requires each state machine to expose the message buffers it uses for input and output. So, the patch makes these buffers configurable members of the machines. The patch drops the slicc code that usd to connect these buffers to the network. Now these buffers are exposed to the python configuration system as Master and Slave ports. In the configuration files, any master port can be connected any slave port. The file pyobject.cc has been modified to take care of allocating the actual message buffer. This is inline with how other port connections work.
2014-09-01 23:55:47 +02:00
# Connect the L0 and L1 controllers
l0_cntrl.bufferToL1 = l1_cntrl.bufferFromL0
l0_cntrl.bufferFromL1 = l1_cntrl.bufferToL0
# Connect the L1 controllers and the network
l1_cntrl.requestToL2 = ruby_system.network.slave
l1_cntrl.responseToL2 = ruby_system.network.slave
l1_cntrl.unblockToL2 = ruby_system.network.slave
l1_cntrl.requestFromL2 = ruby_system.network.master
l1_cntrl.responseFromL2 = ruby_system.network.master
ruby: add a three level MESI protocol. The first two levels (L0, L1) are private to the core, the third level (L2)is possibly shared. The protocol supports clustered designs. For example, one can have two sets of two cores. Each core has an L0 and L1 cache. There are two L2 controllers where each set accesses only one of the L2 controllers.
2014-01-04 07:03:34 +01:00
for j in xrange(num_l2caches_per_cluster):
l2_cache = L2Cache(size = options.l2_size,
assoc = options.l2_assoc,
start_index_bit = l2_index_start)
l2_cntrl = L2Cache_Controller(
version = i * num_l2caches_per_cluster + j,
L2cache = l2_cache, cluster_id = i,
transitions_per_cycle=options.ports,
ruby_system = ruby_system)
exec("ruby_system.l2_cntrl%d = l2_cntrl" % (
i * num_l2caches_per_cluster + j))
l2_cntrl_nodes.append(l2_cntrl)
ruby: message buffers: significant changes This patch is the final patch in a series of patches. The aim of the series is to make ruby more configurable than it was. More specifically, the connections between controllers are not at all possible (unless one is ready to make significant changes to the coherence protocol). Moreover the buffers themselves are magically connected to the network inside the slicc code. These connections are not part of the configuration file. This patch makes changes so that these connections will now be made in the python configuration files associated with the protocols. This requires each state machine to expose the message buffers it uses for input and output. So, the patch makes these buffers configurable members of the machines. The patch drops the slicc code that usd to connect these buffers to the network. Now these buffers are exposed to the python configuration system as Master and Slave ports. In the configuration files, any master port can be connected any slave port. The file pyobject.cc has been modified to take care of allocating the actual message buffer. This is inline with how other port connections work.
2014-09-01 23:55:47 +02:00
# Connect the L2 controllers and the network
l2_cntrl.DirRequestFromL2Cache = ruby_system.network.slave
l2_cntrl.L1RequestFromL2Cache = ruby_system.network.slave
l2_cntrl.responseFromL2Cache = ruby_system.network.slave
l2_cntrl.unblockToL2Cache = ruby_system.network.master
l2_cntrl.L1RequestToL2Cache = ruby_system.network.master
l2_cntrl.responseToL2Cache = ruby_system.network.master
ruby: add a three level MESI protocol. The first two levels (L0, L1) are private to the core, the third level (L2)is possibly shared. The protocol supports clustered designs. For example, one can have two sets of two cores. Each core has an L0 and L1 cache. There are two L2 controllers where each set accesses only one of the L2 controllers.
2014-01-04 07:03:34 +01:00
phys_mem_size = sum(map(lambda r: r.size(), system.mem_ranges))
assert(phys_mem_size % options.num_dirs == 0)
mem_module_size = phys_mem_size / options.num_dirs
# Run each of the ruby memory controllers at a ratio of the frequency of
# the ruby system
# clk_divider value is a fix to pass regression.
ruby_system.memctrl_clk_domain = DerivedClockDomain(
clk_domain=ruby_system.clk_domain,
clk_divider=3)
for i in xrange(options.num_dirs):
#
# Create the Ruby objects associated with the directory controller
#
dir_size = MemorySize('0B')
dir_size.value = mem_module_size
dir_cntrl = Directory_Controller(version = i,
ruby: interface with classic memory controller This patch is the final in the series. The whole series and this patch in particular were written with the aim of interfacing ruby's directory controller with the memory controller in the classic memory system. This is being done since ruby's memory controller has not being kept up to date with the changes going on in DRAMs. Classic's memory controller is more up to date and supports multiple different types of DRAM. This also brings classic and ruby ever more close. The patch also changes ruby's memory controller to expose the same interface.
2014-11-06 12:42:21 +01:00
directory = RubyDirectoryMemory(
version = i, size = dir_size),
ruby: add a three level MESI protocol. The first two levels (L0, L1) are private to the core, the third level (L2)is possibly shared. The protocol supports clustered designs. For example, one can have two sets of two cores. Each core has an L0 and L1 cache. There are two L2 controllers where each set accesses only one of the L2 controllers.
2014-01-04 07:03:34 +01:00
transitions_per_cycle = options.ports,
ruby_system = ruby_system)
exec("ruby_system.dir_cntrl%d = dir_cntrl" % i)
dir_cntrl_nodes.append(dir_cntrl)
ruby: message buffers: significant changes This patch is the final patch in a series of patches. The aim of the series is to make ruby more configurable than it was. More specifically, the connections between controllers are not at all possible (unless one is ready to make significant changes to the coherence protocol). Moreover the buffers themselves are magically connected to the network inside the slicc code. These connections are not part of the configuration file. This patch makes changes so that these connections will now be made in the python configuration files associated with the protocols. This requires each state machine to expose the message buffers it uses for input and output. So, the patch makes these buffers configurable members of the machines. The patch drops the slicc code that usd to connect these buffers to the network. Now these buffers are exposed to the python configuration system as Master and Slave ports. In the configuration files, any master port can be connected any slave port. The file pyobject.cc has been modified to take care of allocating the actual message buffer. This is inline with how other port connections work.
2014-09-01 23:55:47 +02:00
# Connect the directory controllers and the network
dir_cntrl.requestToDir = ruby_system.network.master
dir_cntrl.responseToDir = ruby_system.network.master
dir_cntrl.responseFromDir = ruby_system.network.slave
ruby: add a three level MESI protocol. The first two levels (L0, L1) are private to the core, the third level (L2)is possibly shared. The protocol supports clustered designs. For example, one can have two sets of two cores. Each core has an L0 and L1 cache. There are two L2 controllers where each set accesses only one of the L2 controllers.
2014-01-04 07:03:34 +01:00
for i, dma_port in enumerate(dma_ports):
#
# Create the Ruby objects associated with the dma controller
#
dma_seq = DMASequencer(version = i,
ruby_system = ruby_system)
dma_cntrl = DMA_Controller(version = i,
dma_sequencer = dma_seq,
transitions_per_cycle = options.ports,
ruby_system = ruby_system)
exec("ruby_system.dma_cntrl%d = dma_cntrl" % i)
exec("ruby_system.dma_cntrl%d.dma_sequencer.slave = dma_port" % i)
dma_cntrl_nodes.append(dma_cntrl)
all_cntrls = l0_cntrl_nodes + \
l1_cntrl_nodes + \
l2_cntrl_nodes + \
dir_cntrl_nodes + \
dma_cntrl_nodes
ruby: single physical memory in fs mode Both ruby and the system used to maintain memory copies. With the changes carried for programmed io accesses, only one single memory is required for fs simulations. This patch sets the copy of memory that used to reside with the system to null, so that no space is allocated, but address checks can still be carried out. All the memory accesses now source and sink values to the memory maintained by ruby.
2014-11-06 12:41:44 +01:00
# Create the io controller and the sequencer
if full_system:
io_seq = DMASequencer(version=len(dma_ports), ruby_system=ruby_system)
ruby_system._io_port = io_seq
io_controller = DMA_Controller(version = len(dma_ports),
dma_sequencer = io_seq,
ruby_system = ruby_system)
ruby_system.io_controller = io_controller
# Connect the dma controller to the network
io_controller.responseFromDir = ruby_system.network.master
io_controller.requestToDir = ruby_system.network.slave
all_cntrls = all_cntrls + [io_controller]
ruby: add a three level MESI protocol. The first two levels (L0, L1) are private to the core, the third level (L2)is possibly shared. The protocol supports clustered designs. For example, one can have two sets of two cores. Each core has an L0 and L1 cache. There are two L2 controllers where each set accesses only one of the L2 controllers.
2014-01-04 07:03:34 +01:00
topology = create_topology(all_cntrls, options)
return (cpu_sequencers, dir_cntrl_nodes, topology)
Reference in a new issue Copy permalink