There are several parts to this PR to work towards #1349 . (1) Make RubySystem::getBlockSizeBytes non-static by providing ways to access the block size or passing the block size explicitly to classes. The main changes are: - DataBlocks must be explicitly allocated. A default ctor still exists to avoid needing to heavily modify SLICC. The size can be set using a realloc function, operator=, or copy ctor. This is handled completely transparently meaning no protocol or config changes are required. - WriteMask now requires block size to be set. This is also handled transparently by modifying the SLICC parser to identify WriteMask types and call setBlockSize(). - AbstractCacheEntry and TBE classes now require block size to be set. This is handled transparently by modifying the SLICC parser to identify these classes and call initBlockSize() which calls setBlockSize() for any DataBlock or WriteMask. - All AbstractControllers now have a pointer to RubySystem. This is assigned in SLICC generated code and requires no changes to protocol or configs. - The Ruby Message class now requires block size in all constructors. This is added to the argument list automatically by the SLICC parser. (2) Relax dependence on common functions in src/mem/ruby/common/Address.hh so that RubySystem::getBlockSizeBits is no longer static. Many classes already have a way to get block size from the previous commit, so they simply multiple by 8 to get the number of bits. For handling SLICC and reducing the number of changes, define makeCacheLine, getOffset, etc. in RubyPort and AbstractController. The only protocol changes required are to change any "RubySystem::foo()" calls with "m_ruby_system->foo()". For classes which do not have a way to get access to block size but still used makeLineAddress, getOffset, etc., the block size must be passed to that class. This requires some changes to the SimObject interface for two commonly used classes: DirectoryMemory and RubyPrefecther, resulting in user-facing API changes User-facing API changes: - DirectoryMemory and RubyPrefetcher now require the cache line size as a non-optional argument. - RubySequencer SimObjects now require RubySystem as a non-optional argument. - TesterThread in the GPU ruby tester now requires the cache line size as a non-optional argument. (3) Removes static member variables in RubySystem which control randomization, cooldown, and warmup. These are mostly used by the Ruby Network. The network classes are modified to take these former static variables as parameters which are passed to the corresponding method (e.g., enqueue, delayHead, etc.) rather than needing a RubySystem object at all. Change-Id: Ia63c2ad5cf0bf9d1cbdffba5d3a679bb4d3b1220 (4) There are two major SLICC generated static methods: getNumControllers() on each cache controller which returns the number of controllers created by the configs at run time and the functions which access this method, which are MachineType_base_count and MachineType_base_number. These need to be removed to create multiple RubySystem objects otherwise NetDest, version value, and other objects are incorrect. To remove the static requirement, MachineType_base_count and MachineType_base_number are moved to RubySystem. Any class which needs to call these methods must now have a pointer to a RubySystem. To enable that, several changes are made: - RubyRequest and Message now require a RubySystem pointer in the constructor. The pointer is passed to fields in the Message class which require a RubySystem pointer (e.g., NetDest). SLICC is modified to do this automatically. - SLICC structures may now optionally take an "implicit constructor" which can be used to call a non-default constructor for locally defined variables (e.g., temporary variables within SLICC actions). A statement such as "NetDest bcast_dest;" in SLICC will implicitly append a call to the NetDest constructor taking RubySystem, for example. - RubySystem gets passed to Ruby network objects (Network, Topology).
256 lines
10 KiB
Python
256 lines
10 KiB
Python
# Copyright (c) 2017 Jason Power
|
|
# 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.
|
|
|
|
""" This file creates a set of Ruby caches, the Ruby network, and a simple
|
|
point-to-point topology.
|
|
See Part 3 in the Learning gem5 book:
|
|
http://gem5.org/documentation/learning_gem5/part3/MSIintro
|
|
|
|
IMPORTANT: If you modify this file, it's likely that the Learning gem5 book
|
|
also needs to be updated. For now, email Jason <jason@lowepower.com>
|
|
|
|
"""
|
|
|
|
import math
|
|
|
|
from m5.defines import buildEnv
|
|
from m5.objects import *
|
|
from m5.util import (
|
|
fatal,
|
|
panic,
|
|
)
|
|
|
|
|
|
class MyCacheSystem(RubySystem):
|
|
def __init__(self):
|
|
if buildEnv["PROTOCOL"] != "MSI":
|
|
fatal("This system assumes MSI from learning gem5!")
|
|
|
|
super().__init__()
|
|
|
|
def setup(self, system, cpus, mem_ctrls):
|
|
"""Set up the Ruby cache subsystem. Note: This can't be done in the
|
|
constructor because many of these items require a pointer to the
|
|
ruby system (self). This causes infinite recursion in initialize()
|
|
if we do this in the __init__.
|
|
"""
|
|
# Ruby's global network.
|
|
self.network = MyNetwork(self)
|
|
|
|
# MSI uses 3 virtual networks. One for requests (lowest priority), one
|
|
# for responses (highest priority), and one for "forwards" or
|
|
# cache-to-cache requests. See *.sm files for details.
|
|
self.number_of_virtual_networks = 3
|
|
self.network.number_of_virtual_networks = 3
|
|
|
|
# There is a single global list of all of the controllers to make it
|
|
# easier to connect everything to the global network. This can be
|
|
# customized depending on the topology/network requirements.
|
|
# Create one controller for each L1 cache (and the cache mem obj.)
|
|
# Create a single directory controller (Really the memory cntrl)
|
|
self.controllers = [L1Cache(system, self, cpu) for cpu in cpus] + [
|
|
DirController(self, system.mem_ranges, mem_ctrls)
|
|
]
|
|
|
|
# Create one sequencer per CPU. In many systems this is more
|
|
# complicated since you have to create sequencers for DMA controllers
|
|
# and other controllers, too.
|
|
self.sequencers = [
|
|
RubySequencer(
|
|
version=i,
|
|
# I/D cache is combined and grab from ctrl
|
|
dcache=self.controllers[i].cacheMemory,
|
|
clk_domain=self.controllers[i].clk_domain,
|
|
ruby_system=self,
|
|
)
|
|
for i in range(len(cpus))
|
|
]
|
|
|
|
# We know that we put the controllers in an order such that the first
|
|
# N of them are the L1 caches which need a sequencer pointer
|
|
for i, c in enumerate(self.controllers[0 : len(self.sequencers)]):
|
|
c.sequencer = self.sequencers[i]
|
|
|
|
self.num_of_sequencers = len(self.sequencers)
|
|
|
|
# Create the network and connect the controllers.
|
|
# NOTE: This is quite different if using Garnet!
|
|
self.network.connectControllers(self.controllers)
|
|
self.network.setup_buffers()
|
|
|
|
# Set up a proxy port for the system_port. Used for load binaries and
|
|
# other functional-only things.
|
|
self.sys_port_proxy = RubyPortProxy()
|
|
system.system_port = self.sys_port_proxy.in_ports
|
|
|
|
# Connect the cpu's cache, interrupt, and TLB ports to Ruby
|
|
for i, cpu in enumerate(cpus):
|
|
self.sequencers[i].connectCpuPorts(cpu)
|
|
|
|
|
|
class L1Cache(L1Cache_Controller):
|
|
_version = 0
|
|
|
|
@classmethod
|
|
def versionCount(cls):
|
|
cls._version += 1 # Use count for this particular type
|
|
return cls._version - 1
|
|
|
|
def __init__(self, system, ruby_system, cpu):
|
|
"""CPUs are needed to grab the clock domain and system is needed for
|
|
the cache block size.
|
|
"""
|
|
super().__init__()
|
|
|
|
self.version = self.versionCount()
|
|
# This is the cache memory object that stores the cache data and tags
|
|
self.cacheMemory = RubyCache(
|
|
size="16KiB",
|
|
assoc=8,
|
|
start_index_bit=self.getBlockSizeBits(system),
|
|
)
|
|
self.clk_domain = cpu.clk_domain
|
|
self.send_evictions = self.sendEvicts(cpu)
|
|
self.ruby_system = ruby_system
|
|
self.connectQueues(ruby_system)
|
|
|
|
def getBlockSizeBits(self, system):
|
|
bits = int(math.log(system.cache_line_size, 2))
|
|
if 2**bits != system.cache_line_size.value:
|
|
panic("Cache line size not a power of 2!")
|
|
return bits
|
|
|
|
def sendEvicts(self, cpu):
|
|
"""True if the CPU model or ISA requires sending evictions from caches
|
|
to the CPU. Two scenarios warrant forwarding evictions to the CPU:
|
|
1. The O3 model must keep the LSQ coherent with the caches
|
|
2. The x86 mwait instruction is built on top of coherence
|
|
3. The local exclusive monitor in ARM systems
|
|
|
|
As this is an X86 simulation we return True.
|
|
"""
|
|
return True
|
|
|
|
def connectQueues(self, ruby_system):
|
|
"""Connect all of the queues for this controller."""
|
|
# mandatoryQueue is a special variable. It is used by the sequencer to
|
|
# send RubyRequests from the CPU (or other processor). It isn't
|
|
# explicitly connected to anything.
|
|
self.mandatoryQueue = MessageBuffer()
|
|
|
|
# All message buffers must be created and connected to the
|
|
# general Ruby network. In this case, "in_port/out_port" don't
|
|
# mean the same thing as normal gem5 ports. If a MessageBuffer
|
|
# is a "to" buffer (i.e., out) then you use the "out_port",
|
|
# otherwise, the in_port.
|
|
self.requestToDir = MessageBuffer(ordered=True)
|
|
self.requestToDir.out_port = ruby_system.network.in_port
|
|
self.responseToDirOrSibling = MessageBuffer(ordered=True)
|
|
self.responseToDirOrSibling.out_port = ruby_system.network.in_port
|
|
self.forwardFromDir = MessageBuffer(ordered=True)
|
|
self.forwardFromDir.in_port = ruby_system.network.out_port
|
|
self.responseFromDirOrSibling = MessageBuffer(ordered=True)
|
|
self.responseFromDirOrSibling.in_port = ruby_system.network.out_port
|
|
|
|
|
|
class DirController(Directory_Controller):
|
|
_version = 0
|
|
|
|
@classmethod
|
|
def versionCount(cls):
|
|
cls._version += 1 # Use count for this particular type
|
|
return cls._version - 1
|
|
|
|
def __init__(self, ruby_system, ranges, mem_ctrls):
|
|
"""ranges are the memory ranges assigned to this controller."""
|
|
if len(mem_ctrls) > 1:
|
|
panic("This cache system can only be connected to one mem ctrl")
|
|
super().__init__()
|
|
self.version = self.versionCount()
|
|
self.addr_ranges = ranges
|
|
self.ruby_system = ruby_system
|
|
self.directory = RubyDirectoryMemory(
|
|
block_size=ruby_system.block_size_bytes
|
|
)
|
|
# Connect this directory to the memory side.
|
|
self.memory = mem_ctrls[0].port
|
|
self.connectQueues(ruby_system)
|
|
|
|
def connectQueues(self, ruby_system):
|
|
self.requestFromCache = MessageBuffer(ordered=True)
|
|
self.requestFromCache.in_port = ruby_system.network.out_port
|
|
self.responseFromCache = MessageBuffer(ordered=True)
|
|
self.responseFromCache.in_port = ruby_system.network.out_port
|
|
|
|
self.responseToCache = MessageBuffer(ordered=True)
|
|
self.responseToCache.out_port = ruby_system.network.in_port
|
|
self.forwardToCache = MessageBuffer(ordered=True)
|
|
self.forwardToCache.out_port = ruby_system.network.in_port
|
|
|
|
# These are other special message buffers. They are used to send
|
|
# requests to memory and responses from memory back to the controller.
|
|
# Any messages sent or received on the memory port (see self.memory
|
|
# above) will be directed through these message buffers.
|
|
self.requestToMemory = MessageBuffer()
|
|
self.responseFromMemory = MessageBuffer()
|
|
|
|
|
|
class MyNetwork(SimpleNetwork):
|
|
"""A simple point-to-point network. This doesn't not use garnet."""
|
|
|
|
def __init__(self, ruby_system):
|
|
super().__init__()
|
|
self.netifs = []
|
|
self.ruby_system = ruby_system
|
|
|
|
def connectControllers(self, controllers):
|
|
"""Connect all of the controllers to routers and connec the routers
|
|
together in a point-to-point network.
|
|
"""
|
|
# Create one router/switch per controller in the system
|
|
self.routers = [Switch(router_id=i) for i in range(len(controllers))]
|
|
|
|
# Make a link from each controller to the router. The link goes
|
|
# externally to the network.
|
|
self.ext_links = [
|
|
SimpleExtLink(link_id=i, ext_node=c, int_node=self.routers[i])
|
|
for i, c in enumerate(controllers)
|
|
]
|
|
|
|
# Make an "internal" link (internal to the network) between every pair
|
|
# of routers.
|
|
link_count = 0
|
|
int_links = []
|
|
for ri in self.routers:
|
|
for rj in self.routers:
|
|
if ri == rj:
|
|
continue # Don't connect a router to itself!
|
|
link_count += 1
|
|
int_links.append(
|
|
SimpleIntLink(link_id=link_count, src_node=ri, dst_node=rj)
|
|
)
|
|
self.int_links = int_links
|