gem5/configs/ruby/MOESI_hammer.py

249 lines
10 KiB
Python
Raw Normal View History

# Copyright (c) 2006-2007 The Regents of The University of Michigan
# Copyright (c) 2009 Advanced Micro Devices, Inc.
# 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
import math
import m5
from m5.objects import *
from m5.defines import buildEnv
from Ruby import create_topology
#
# Note: the L1 Cache latency is only used by the sequencer on fast path hits
#
class L1Cache(RubyCache):
latency = 2
#
# Note: the L2 Cache latency is not currently used
#
class L2Cache(RubyCache):
latency = 10
#
# Probe filter is a cache, latency is not used
#
class ProbeFilter(RubyCache):
latency = 1
def define_options(parser):
parser.add_option("--allow-atomic-migration", action="store_true",
help="allow migratory sharing for atomic only accessed blocks")
parser.add_option("--pf-on", action="store_true",
help="Hammer: enable Probe Filter")
parser.add_option("--dir-on", action="store_true",
help="Hammer: enable Full-bit Directory")
def create_system(options, system, dma_ports, ruby_system):
if buildEnv['PROTOCOL'] != 'MOESI_hammer':
panic("This script requires the MOESI_hammer 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.
#
l1_cntrl_nodes = []
dir_cntrl_nodes = []
dma_cntrl_nodes = []
#
# Must create the individual controllers before the network to ensure the
# controller constructors are called before the network constructor
#
2011-03-25 18:13:50 +01:00
block_size_bits = int(math.log(options.cacheline_size, 2))
network: convert links & switches to first class C++ SimObjects This patch converts links and switches from second class simobjects that were virtually ignored by the networks (both simple and Garnet) to first class simobjects that directly correspond to c++ ojbects manipulated by the topology and network classes. This is especially true for Garnet, where the links and switches directly correspond to specific C++ objects. By making this change, many aspects of the Topology class were simplified. --HG-- rename : src/mem/ruby/network/Network.cc => src/mem/ruby/network/BasicLink.cc rename : src/mem/ruby/network/Network.hh => src/mem/ruby/network/BasicLink.hh rename : src/mem/ruby/network/Network.cc => src/mem/ruby/network/garnet/fixed-pipeline/GarnetLink_d.cc rename : src/mem/ruby/network/Network.hh => src/mem/ruby/network/garnet/fixed-pipeline/GarnetLink_d.hh rename : src/mem/ruby/network/garnet/fixed-pipeline/GarnetNetwork_d.py => src/mem/ruby/network/garnet/fixed-pipeline/GarnetLink_d.py rename : src/mem/ruby/network/garnet/fixed-pipeline/GarnetNetwork_d.py => src/mem/ruby/network/garnet/fixed-pipeline/GarnetRouter_d.py rename : src/mem/ruby/network/Network.cc => src/mem/ruby/network/garnet/flexible-pipeline/GarnetLink.cc rename : src/mem/ruby/network/Network.hh => src/mem/ruby/network/garnet/flexible-pipeline/GarnetLink.hh rename : src/mem/ruby/network/garnet/fixed-pipeline/GarnetNetwork_d.py => src/mem/ruby/network/garnet/flexible-pipeline/GarnetLink.py rename : src/mem/ruby/network/garnet/fixed-pipeline/GarnetNetwork_d.py => src/mem/ruby/network/garnet/flexible-pipeline/GarnetRouter.py
2011-04-29 02:18:14 +02:00
for i in xrange(options.num_cpus):
#
# First create the Ruby objects associated with this cpu
#
l1i_cache = L1Cache(size = options.l1i_size,
2011-03-25 18:13:50 +01:00
assoc = options.l1i_assoc,
start_index_bit = block_size_bits,
is_icache = True)
l1d_cache = L1Cache(size = options.l1d_size,
2011-03-25 18:13:50 +01:00
assoc = options.l1d_assoc,
start_index_bit = block_size_bits)
l2_cache = L2Cache(size = options.l2_size,
2011-03-25 18:13:50 +01:00
assoc = options.l2_assoc,
start_index_bit = block_size_bits)
l1_cntrl = L1Cache_Controller(version = i,
L1Icache = l1i_cache,
L1Dcache = l1d_cache,
L2cache = l2_cache,
no_mig_atomic = not \
options.allow_atomic_migration,
send_evictions = (
options.cpu_type == "detailed"),
transitions_per_cycle = options.ports,
clk_domain=system.cpu[i].clk_domain,
ruby_system = ruby_system)
cpu_seq = RubySequencer(version = i,
icache = l1i_cache,
dcache = l1d_cache,
clk_domain=system.cpu[i].clk_domain,
ruby_system = ruby_system)
l1_cntrl.sequencer = cpu_seq
if options.recycle_latency:
l1_cntrl.recycle_latency = options.recycle_latency
exec("ruby_system.l1_cntrl%d = l1_cntrl" % i)
# Add controllers and sequencers to the appropriate lists
cpu_sequencers.append(cpu_seq)
l1_cntrl_nodes.append(l1_cntrl)
# Connect the L1 controller and the network
# Connect the buffers from the controller to network
l1_cntrl.requestFromCache = ruby_system.network.slave
l1_cntrl.responseFromCache = ruby_system.network.slave
l1_cntrl.unblockFromCache = ruby_system.network.slave
# Connect the buffers from the network to the controller
l1_cntrl.forwardToCache = ruby_system.network.master
l1_cntrl.responseToCache = ruby_system.network.master
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
#
# determine size and index bits for probe filter
# By default, the probe filter size is configured to be twice the
# size of the L2 cache.
#
pf_size = MemorySize(options.l2_size)
pf_size.value = pf_size.value * 2
dir_bits = int(math.log(options.num_dirs, 2))
pf_bits = int(math.log(pf_size.value, 2))
if options.numa_high_bit:
if options.pf_on or options.dir_on:
# if numa high bit explicitly set, make sure it does not overlap
# with the probe filter index
assert(options.numa_high_bit - dir_bits > pf_bits)
# set the probe filter start bit to just above the block offset
pf_start_bit = block_size_bits
else:
if dir_bits > 0:
pf_start_bit = dir_bits + block_size_bits - 1
else:
pf_start_bit = block_size_bits
sim: Add the notion of clock domains to all ClockedObjects This patch adds the notion of source- and derived-clock domains to the ClockedObjects. As such, all clock information is moved to the clock domain, and the ClockedObjects are grouped into domains. The clock domains are either source domains, with a specific clock period, or derived domains that have a parent domain and a divider (potentially chained). For piece of logic that runs at a derived clock (a ratio of the clock its parent is running at) the necessary derived clock domain is created from its corresponding parent clock domain. For now, the derived clock domain only supports a divider, thus ensuring a lower speed compared to its parent. Multiplier functionality implies a PLL logic that has not been modelled yet (create a separate clock instead). The clock domains should be used as a mechanism to provide a controllable clock source that affects clock for every clocked object lying beneath it. The clock of the domain can (in a future patch) be controlled by a handler responsible for dynamic frequency scaling of the respective clock domains. All the config scripts have been retro-fitted with clock domains. For the System a default SrcClockDomain is created. For CPUs that run at a different speed than the system, there is a seperate clock domain created. This domain incorporates the CPU and the associated caches. As before, Ruby runs under its own clock domain. The clock period of all domains are pre-computed, such that no virtual functions or multiplications are needed when calling clockPeriod. Instead, the clock period is pre-computed when any changes occur. For this to be possible, each clock domain tracks its children.
2013-06-27 11:49:49 +02:00
# 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
#
sim: Add the notion of clock domains to all ClockedObjects This patch adds the notion of source- and derived-clock domains to the ClockedObjects. As such, all clock information is moved to the clock domain, and the ClockedObjects are grouped into domains. The clock domains are either source domains, with a specific clock period, or derived domains that have a parent domain and a divider (potentially chained). For piece of logic that runs at a derived clock (a ratio of the clock its parent is running at) the necessary derived clock domain is created from its corresponding parent clock domain. For now, the derived clock domain only supports a divider, thus ensuring a lower speed compared to its parent. Multiplier functionality implies a PLL logic that has not been modelled yet (create a separate clock instead). The clock domains should be used as a mechanism to provide a controllable clock source that affects clock for every clocked object lying beneath it. The clock of the domain can (in a future patch) be controlled by a handler responsible for dynamic frequency scaling of the respective clock domains. All the config scripts have been retro-fitted with clock domains. For the System a default SrcClockDomain is created. For CPUs that run at a different speed than the system, there is a seperate clock domain created. This domain incorporates the CPU and the associated caches. As before, Ruby runs under its own clock domain. The clock period of all domains are pre-computed, such that no virtual functions or multiplications are needed when calling clockPeriod. Instead, the clock period is pre-computed when any changes occur. For this to be possible, each clock domain tracks its children.
2013-06-27 11:49:49 +02:00
mem_cntrl = RubyMemoryControl(
clk_domain = ruby_system.memctrl_clk_domain,
version = i,
ruby_system = ruby_system)
dir_size = MemorySize('0B')
dir_size.value = mem_module_size
pf = ProbeFilter(size = pf_size, assoc = 4,
start_index_bit = pf_start_bit)
dir_cntrl = Directory_Controller(version = i,
directory = \
RubyDirectoryMemory( \
version = i,
size = dir_size,
use_map = options.use_map,
map_levels = \
2011-02-07 07:14:19 +01:00
options.map_levels,
numa_high_bit = \
options.numa_high_bit),
probeFilter = pf,
memBuffer = mem_cntrl,
probe_filter_enabled = options.pf_on,
full_bit_dir_enabled = options.dir_on,
transitions_per_cycle = options.ports,
ruby_system = ruby_system)
if options.recycle_latency:
dir_cntrl.recycle_latency = options.recycle_latency
exec("ruby_system.dir_cntrl%d = dir_cntrl" % i)
dir_cntrl_nodes.append(dir_cntrl)
# Connect the directory controller to the network
dir_cntrl.forwardFromDir = ruby_system.network.slave
dir_cntrl.responseFromDir = ruby_system.network.slave
dir_cntrl.dmaResponseFromDir = ruby_system.network.slave
dir_cntrl.unblockToDir = ruby_system.network.master
dir_cntrl.responseToDir = ruby_system.network.master
dir_cntrl.requestToDir = ruby_system.network.master
dir_cntrl.dmaRequestToDir = ruby_system.network.master
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)
if options.recycle_latency:
dma_cntrl.recycle_latency = options.recycle_latency
# Connect the dma controller to the network
dma_cntrl.responseFromDir = ruby_system.network.slave
dma_cntrl.requestToDir = ruby_system.network.master
all_cntrls = l1_cntrl_nodes + dir_cntrl_nodes + dma_cntrl_nodes
topology = create_topology(all_cntrls, options)
return (cpu_sequencers, dir_cntrl_nodes, topology)