cluster.py revision 9036:6385cf85bf12 
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 os
34import optparse
35import sys
36
37import m5
38from m5.objects import *
39
40m5.util.addToPath('../common')
41
42# --------------------
43# Define Command Line Options
44# ====================
45
46parser = optparse.OptionParser()
47
48parser.add_option("-d", "--detailed", action="store_true")
49parser.add_option("-t", "--timing", action="store_true")
50parser.add_option("-m", "--maxtick", type="int")
51parser.add_option("-c", "--numclusters",
52                  help="Number of clusters", type="int")
53parser.add_option("-n", "--numcpus",
54                  help="Number of cpus in total", type="int")
55parser.add_option("-f", "--frequency",
56                  default = "1GHz",
57                  help="Frequency of each CPU")
58parser.add_option("--l1size",
59                  default = "32kB")
60parser.add_option("--l1latency",
61                  default = 1)
62parser.add_option("--l2size",
63                  default = "256kB")
64parser.add_option("--l2latency",
65                  default = 10)
66parser.add_option("--rootdir",
67                  help="ROot directory of Splash2",
68                  default="/dist/splash2/codes/")
69parser.add_option("-b", "--benchmark",
70                  help="Splash 2 benchmark to run")
71
72(options, args) = parser.parse_args()
73
74if args:
75    print "Error: script doesn't take any positional arguments"
76    sys.exit(1)
77
78# --------------------
79# Define Splash2 Benchmarks
80# ====================
81class Cholesky(LiveProcess):
82        executable = options.rootdir + '/kernels/cholesky/CHOLESKY'
83        cmd = 'CHOLESKY -p' + str(options.numcpus) + ' '\
84             + options.rootdir + '/kernels/cholesky/inputs/tk23.O'
85
86class FFT(LiveProcess):
87        executable = options.rootdir + 'kernels/fft/FFT'
88        cmd = 'FFT -p' + str(options.numcpus) + ' -m18'
89
90class LU_contig(LiveProcess):
91        executable = options.rootdir + 'kernels/lu/contiguous_blocks/LU'
92        cmd = 'LU -p' + str(options.numcpus)
93
94class LU_noncontig(LiveProcess):
95        executable = options.rootdir + 'kernels/lu/non_contiguous_blocks/LU'
96        cmd = 'LU -p' + str(options.numcpus)
97
98class Radix(LiveProcess):
99        executable = options.rootdir + 'kernels/radix/RADIX'
100        cmd = 'RADIX -n524288 -p' + str(options.numcpus)
101
102class Barnes(LiveProcess):
103        executable = options.rootdir + 'apps/barnes/BARNES'
104        cmd = 'BARNES'
105        input = options.rootdir + 'apps/barnes/input.p' + str(options.numcpus)
106
107class FMM(LiveProcess):
108        executable = options.rootdir + 'apps/fmm/FMM'
109        cmd = 'FMM'
110        input = options.rootdir + 'apps/fmm/inputs/input.2048.p' + str(options.numcpus)
111
112class Ocean_contig(LiveProcess):
113        executable = options.rootdir + 'apps/ocean/contiguous_partitions/OCEAN'
114        cmd = 'OCEAN -p' + str(options.numcpus)
115
116class Ocean_noncontig(LiveProcess):
117        executable = options.rootdir + 'apps/ocean/non_contiguous_partitions/OCEAN'
118        cmd = 'OCEAN -p' + str(options.numcpus)
119
120class Raytrace(LiveProcess):
121        executable = options.rootdir + 'apps/raytrace/RAYTRACE'
122        cmd = 'RAYTRACE -p' + str(options.numcpus) + ' ' \
123             + options.rootdir + 'apps/raytrace/inputs/teapot.env'
124
125class Water_nsquared(LiveProcess):
126        executable = options.rootdir + 'apps/water-nsquared/WATER-NSQUARED'
127        cmd = 'WATER-NSQUARED'
128        input = options.rootdir + 'apps/water-nsquared/input.p' + str(options.numcpus)
129
130class Water_spatial(LiveProcess):
131        executable = options.rootdir + 'apps/water-spatial/WATER-SPATIAL'
132        cmd = 'WATER-SPATIAL'
133        input = options.rootdir + 'apps/water-spatial/input.p' + str(options.numcpus)
134
135
136# --------------------
137# Base L1 Cache Definition
138# ====================
139
140class L1(BaseCache):
141    latency = options.l1latency
142    block_size = 64
143    mshrs = 12
144    tgts_per_mshr = 8
145
146# ----------------------
147# Base L2 Cache Definition
148# ----------------------
149
150class L2(BaseCache):
151    block_size = 64
152    latency = options.l2latency
153    mshrs = 92
154    tgts_per_mshr = 16
155    write_buffers = 8
156
157# ----------------------
158# Define the clusters with their cpus
159# ----------------------
160class Cluster:
161    pass
162
163cpusPerCluster = options.numcpus/options.numclusters
164
165busFrequency = Frequency(options.frequency)
166busFrequency *= cpusPerCluster
167
168all_cpus = []
169all_l1s = []
170all_l1buses = []
171if options.timing:
172    clusters = [ Cluster() for i in xrange(options.numclusters)]
173    for j in xrange(options.numclusters):
174        clusters[j].id = j
175    for cluster in clusters:
176        cluster.clusterbus = CoherentBus(clock=busFrequency)
177        all_l1buses += [cluster.clusterbus]
178        cluster.cpus = [TimingSimpleCPU(cpu_id = i + cluster.id,
179                                        clock=options.frequency)
180                        for i in xrange(cpusPerCluster)]
181        all_cpus += cluster.cpus
182        cluster.l1 = L1(size=options.l1size, assoc = 4)
183        all_l1s += [cluster.l1]
184elif options.detailed:
185    clusters = [ Cluster() for i in xrange(options.numclusters)]
186    for j in xrange(options.numclusters):
187        clusters[j].id = j
188    for cluster in clusters:
189        cluster.clusterbus = CoherentBus(clock=busFrequency)
190        all_l1buses += [cluster.clusterbus]
191        cluster.cpus = [DerivO3CPU(cpu_id = i + cluster.id,
192                                   clock=options.frequency)
193                        for i in xrange(cpusPerCluster)]
194        all_cpus += cluster.cpus
195        cluster.l1 = L1(size=options.l1size, assoc = 4)
196        all_l1s += [cluster.l1]
197else:
198    clusters = [ Cluster() for i in xrange(options.numclusters)]
199    for j in xrange(options.numclusters):
200        clusters[j].id = j
201    for cluster in clusters:
202        cluster.clusterbus = CoherentBus(clock=busFrequency)
203        all_l1buses += [cluster.clusterbus]
204        cluster.cpus = [AtomicSimpleCPU(cpu_id = i + cluster.id,
205                                        clock=options.frequency)
206                        for i in xrange(cpusPerCluster)]
207        all_cpus += cluster.cpus
208        cluster.l1 = L1(size=options.l1size, assoc = 4)
209        all_l1s += [cluster.l1]
210
211# ----------------------
212# Create a system, and add system wide objects
213# ----------------------
214system = System(cpu = all_cpus, l1_ = all_l1s, l1bus_ = all_l1buses,
215                physmem = SimpleMemory(),
216                membus = CoherentBus(clock = busFrequency))
217
218system.toL2bus = CoherentBus(clock = busFrequency)
219system.l2 = L2(size = options.l2size, assoc = 8)
220
221# ----------------------
222# Connect the L2 cache and memory together
223# ----------------------
224
225system.physmem.port = system.membus.master
226system.l2.cpu_side = system.toL2bus.slave
227system.l2.mem_side = system.membus.master
228
229# ----------------------
230# Connect the L2 cache and clusters together
231# ----------------------
232for cluster in clusters:
233    cluster.l1.cpu_side = cluster.clusterbus.master
234    cluster.l1.mem_side = system.toL2bus.slave
235    for cpu in cluster.cpus:
236        cpu.icache_port = cluster.clusterbus.slave
237        cpu.dcache_port = cluster.clusterbus.slave
238
239# ----------------------
240# Define the root
241# ----------------------
242
243root = Root(full_system = False, system = system)
244
245# --------------------
246# Pick the correct Splash2 Benchmarks
247# ====================
248if options.benchmark == 'Cholesky':
249    root.workload = Cholesky()
250elif options.benchmark == 'FFT':
251    root.workload = FFT()
252elif options.benchmark == 'LUContig':
253    root.workload = LU_contig()
254elif options.benchmark == 'LUNoncontig':
255    root.workload = LU_noncontig()
256elif options.benchmark == 'Radix':
257    root.workload = Radix()
258elif options.benchmark == 'Barnes':
259    root.workload = Barnes()
260elif options.benchmark == 'FMM':
261    root.workload = FMM()
262elif options.benchmark == 'OceanContig':
263    root.workload = Ocean_contig()
264elif options.benchmark == 'OceanNoncontig':
265    root.workload = Ocean_noncontig()
266elif options.benchmark == 'Raytrace':
267    root.workload = Raytrace()
268elif options.benchmark == 'WaterNSquared':
269    root.workload = Water_nsquared()
270elif options.benchmark == 'WaterSpatial':
271    root.workload = Water_spatial()
272else:
273    m5.util.panic("""
274The --benchmark environment variable was set to something improper.
275Use Cholesky, FFT, LUContig, LUNoncontig, Radix, Barnes, FMM, OceanContig,
276OceanNoncontig, Raytrace, WaterNSquared, or WaterSpatial
277""")
278
279# --------------------
280# Assign the workload to the cpus
281# ====================
282
283for cluster in clusters:
284    for cpu in cluster.cpus:
285        cpu.workload = root.workload
286
287# ----------------------
288# Run the simulation
289# ----------------------
290
291if options.timing or options.detailed:
292    root.system.mem_mode = 'timing'
293
294# instantiate configuration
295m5.instantiate()
296
297# simulate until program terminates
298if options.maxtick:
299    exit_event = m5.simulate(options.maxtick)
300else:
301    exit_event = m5.simulate(m5.MaxTick)
302
303print 'Exiting @ tick', m5.curTick(), 'because', exit_event.getCause()
304
305