cluster.py revision 4876:a18cedc19da5 
1# Copyright (c) 2006-2007 The Regents of The University of Michigan
2# All rights reserved.
3#
4# Redistribution and use in source and binary forms, with or without
5# modification, are permitted provided that the following conditions are
6# met: redistributions of source code must retain the above copyright
7# notice, this list of conditions and the following disclaimer;
8# redistributions in binary form must reproduce the above copyright
9# notice, this list of conditions and the following disclaimer in the
10# documentation and/or other materials provided with the distribution;
11# neither the name of the copyright holders nor the names of its
12# contributors may be used to endorse or promote products derived from
13# this software without specific prior written permission.
14#
15# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
16# "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
17# LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
18# A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
19# OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
20# SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
21# LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
22# DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
23# THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
24# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
25# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
26#
27# Authors: Ron Dreslinski
28
29# Simple test script
30#
31# "m5 test.py"
32
33import m5
34from m5.objects import *
35import os, optparse, sys
36m5.AddToPath('../common')
37
38# --------------------
39# Define Command Line Options
40# ====================
41
42parser = optparse.OptionParser()
43
44parser.add_option("-d", "--detailed", action="store_true")
45parser.add_option("-t", "--timing", action="store_true")
46parser.add_option("-m", "--maxtick", type="int")
47parser.add_option("-c", "--numclusters",
48                  help="Number of clusters", type="int")
49parser.add_option("-n", "--numcpus",
50                  help="Number of cpus in total", type="int")
51parser.add_option("-f", "--frequency",
52                  default = "1GHz",
53                  help="Frequency of each CPU")
54parser.add_option("--l1size",
55                  default = "32kB")
56parser.add_option("--l1latency",
57                  default = 1)
58parser.add_option("--l2size",
59                  default = "256kB")
60parser.add_option("--l2latency",
61                  default = 10)
62parser.add_option("--rootdir",
63                  help="ROot directory of Splash2",
64                  default="/dist/splash2/codes/")
65parser.add_option("-b", "--benchmark",
66                  help="Splash 2 benchmark to run")
67
68(options, args) = parser.parse_args()
69
70if args:
71    print "Error: script doesn't take any positional arguments"
72    sys.exit(1)
73
74# --------------------
75# Define Splash2 Benchmarks
76# ====================
77class Cholesky(LiveProcess):
78        executable = options.rootdir + '/kernels/cholesky/CHOLESKY'
79        cmd = 'CHOLESKY -p' + str(options.numcpus) + ' '\
80             + options.rootdir + '/kernels/cholesky/inputs/tk23.O'
81
82class FFT(LiveProcess):
83        executable = options.rootdir + 'kernels/fft/FFT'
84        cmd = 'FFT -p' + str(options.numcpus) + ' -m18'
85
86class LU_contig(LiveProcess):
87        executable = options.rootdir + 'kernels/lu/contiguous_blocks/LU'
88        cmd = 'LU -p' + str(options.numcpus)
89
90class LU_noncontig(LiveProcess):
91        executable = options.rootdir + 'kernels/lu/non_contiguous_blocks/LU'
92        cmd = 'LU -p' + str(options.numcpus)
93
94class Radix(LiveProcess):
95        executable = options.rootdir + 'kernels/radix/RADIX'
96        cmd = 'RADIX -n524288 -p' + str(options.numcpus)
97
98class Barnes(LiveProcess):
99        executable = options.rootdir + 'apps/barnes/BARNES'
100        cmd = 'BARNES'
101        input = options.rootdir + 'apps/barnes/input.p' + str(options.numcpus)
102
103class FMM(LiveProcess):
104        executable = options.rootdir + 'apps/fmm/FMM'
105        cmd = 'FMM'
106        input = options.rootdir + 'apps/fmm/inputs/input.2048.p' + str(options.numcpus)
107
108class Ocean_contig(LiveProcess):
109        executable = options.rootdir + 'apps/ocean/contiguous_partitions/OCEAN'
110        cmd = 'OCEAN -p' + str(options.numcpus)
111
112class Ocean_noncontig(LiveProcess):
113        executable = options.rootdir + 'apps/ocean/non_contiguous_partitions/OCEAN'
114        cmd = 'OCEAN -p' + str(options.numcpus)
115
116class Raytrace(LiveProcess):
117        executable = options.rootdir + 'apps/raytrace/RAYTRACE'
118        cmd = 'RAYTRACE -p' + str(options.numcpus) + ' ' \
119             + options.rootdir + 'apps/raytrace/inputs/teapot.env'
120
121class Water_nsquared(LiveProcess):
122        executable = options.rootdir + 'apps/water-nsquared/WATER-NSQUARED'
123        cmd = 'WATER-NSQUARED'
124        input = options.rootdir + 'apps/water-nsquared/input.p' + str(options.numcpus)
125
126class Water_spatial(LiveProcess):
127        executable = options.rootdir + 'apps/water-spatial/WATER-SPATIAL'
128        cmd = 'WATER-SPATIAL'
129        input = options.rootdir + 'apps/water-spatial/input.p' + str(options.numcpus)
130
131
132# --------------------
133# Base L1 Cache Definition
134# ====================
135
136class L1(BaseCache):
137    latency = options.l1latency
138    block_size = 64
139    mshrs = 12
140    tgts_per_mshr = 8
141
142# ----------------------
143# Base L2 Cache Definition
144# ----------------------
145
146class L2(BaseCache):
147    block_size = 64
148    latency = options.l2latency
149    mshrs = 92
150    tgts_per_mshr = 16
151    write_buffers = 8
152
153# ----------------------
154# Define the clusters with their cpus
155# ----------------------
156class Cluster:
157    pass
158
159cpusPerCluster = options.numcpus/options.numclusters
160
161busFrequency = Frequency(options.frequency)
162busFrequency *= cpusPerCluster
163
164all_cpus = []
165all_l1s = []
166all_l1buses = []
167if options.timing:
168    clusters = [ Cluster() for i in xrange(options.numclusters)]
169    for j in xrange(options.numclusters):
170        clusters[j].id = j
171    for cluster in clusters:
172        cluster.clusterbus = Bus(clock=busFrequency)
173        all_l1buses += [cluster.clusterbus]
174        cluster.cpus = [TimingSimpleCPU(cpu_id = i + cluster.id,
175                                        clock=options.frequency)
176                        for i in xrange(cpusPerCluster)]
177        all_cpus += cluster.cpus
178        cluster.l1 = L1(size=options.l1size, assoc = 4)
179        all_l1s += [cluster.l1]
180elif options.detailed:
181    clusters = [ Cluster() for i in xrange(options.numclusters)]
182    for j in xrange(options.numclusters):
183        clusters[j].id = j
184    for cluster in clusters:
185        cluster.clusterbus = Bus(clock=busFrequency)
186        all_l1buses += [cluster.clusterbus]
187        cluster.cpus = [DerivO3CPU(cpu_id = i + cluster.id,
188                                   clock=options.frequency)
189                        for i in xrange(cpusPerCluster)]
190        all_cpus += cluster.cpus
191        cluster.l1 = L1(size=options.l1size, assoc = 4)
192        all_l1s += [cluster.l1]
193else:
194    clusters = [ Cluster() for i in xrange(options.numclusters)]
195    for j in xrange(options.numclusters):
196        clusters[j].id = j
197    for cluster in clusters:
198        cluster.clusterbus = Bus(clock=busFrequency)
199        all_l1buses += [cluster.clusterbus]
200        cluster.cpus = [AtomicSimpleCPU(cpu_id = i + cluster.id,
201                                        clock=options.frequency)
202                        for i in xrange(cpusPerCluster)]
203        all_cpus += cluster.cpus
204        cluster.l1 = L1(size=options.l1size, assoc = 4)
205        all_l1s += [cluster.l1]
206
207# ----------------------
208# Create a system, and add system wide objects
209# ----------------------
210system = System(cpu = all_cpus, l1_ = all_l1s, l1bus_ = all_l1buses, physmem = PhysicalMemory(),
211                membus = Bus(clock = busFrequency))
212
213system.toL2bus = Bus(clock = busFrequency)
214system.l2 = L2(size = options.l2size, assoc = 8)
215
216# ----------------------
217# Connect the L2 cache and memory together
218# ----------------------
219
220system.physmem.port = system.membus.port
221system.l2.cpu_side = system.toL2bus.port
222system.l2.mem_side = system.membus.port
223
224# ----------------------
225# Connect the L2 cache and clusters together
226# ----------------------
227for cluster in clusters:
228    cluster.l1.cpu_side = cluster.clusterbus.port
229    cluster.l1.mem_side = system.toL2bus.port
230    for cpu in cluster.cpus:
231        cpu.icache_port = cluster.clusterbus.port
232        cpu.dcache_port = cluster.clusterbus.port
233        cpu.mem = cluster.l1
234
235# ----------------------
236# Define the root
237# ----------------------
238
239root = Root(system = system)
240
241# --------------------
242# Pick the correct Splash2 Benchmarks
243# ====================
244if options.benchmark == 'Cholesky':
245    root.workload = Cholesky()
246elif options.benchmark == 'FFT':
247    root.workload = FFT()
248elif options.benchmark == 'LUContig':
249    root.workload = LU_contig()
250elif options.benchmark == 'LUNoncontig':
251    root.workload = LU_noncontig()
252elif options.benchmark == 'Radix':
253    root.workload = Radix()
254elif options.benchmark == 'Barnes':
255    root.workload = Barnes()
256elif options.benchmark == 'FMM':
257    root.workload = FMM()
258elif options.benchmark == 'OceanContig':
259    root.workload = Ocean_contig()
260elif options.benchmark == 'OceanNoncontig':
261    root.workload = Ocean_noncontig()
262elif options.benchmark == 'Raytrace':
263    root.workload = Raytrace()
264elif options.benchmark == 'WaterNSquared':
265    root.workload = Water_nsquared()
266elif options.benchmark == 'WaterSpatial':
267    root.workload = Water_spatial()
268else:
269    panic("The --benchmark environment variable was set to something" \
270          +" improper.\nUse Cholesky, FFT, LUContig, LUNoncontig, Radix" \
271          +", Barnes, FMM, OceanContig,\nOceanNoncontig, Raytrace," \
272          +" WaterNSquared, or WaterSpatial\n")
273
274# --------------------
275# Assign the workload to the cpus
276# ====================
277
278for cluster in clusters:
279    for cpu in cluster.cpus:
280        cpu.workload = root.workload
281
282# ----------------------
283# Run the simulation
284# ----------------------
285
286if options.timing or options.detailed:
287    root.system.mem_mode = 'timing'
288
289# instantiate configuration
290m5.instantiate(root)
291
292# simulate until program terminates
293if options.maxtick:
294    exit_event = m5.simulate(options.maxtick)
295else:
296    exit_event = m5.simulate(m5.MaxTick)
297
298print 'Exiting @ tick', m5.curTick(), 'because', exit_event.getCause()
299
300