configs/splash2/cluster.py

   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 # Simple test script
  28 #
  29 # "m5 test.py"
  30
  31 import os
  32 import optparse
  33 import sys
  34
  35 import m5
  36 from m5.objects import *
  37
  38 # --------------------
  39 # Define Command Line Options
  40 # ====================
  41
  42 parser = optparse.OptionParser()
  43
  44 parser.add_option("-d", "--detailed", action="store_true")
  45 parser.add_option("-t", "--timing", action="store_true")
  46 parser.add_option("-m", "--maxtick", type="int")
  47 parser.add_option("-c", "--numclusters",
  48                   help="Number of clusters", type="int")
  49 parser.add_option("-n", "--numcpus",
  50                   help="Number of cpus in total", type="int")
  51 parser.add_option("-f", "--frequency",
  52                   default = "1GHz",
  53                   help="Frequency of each CPU")
  54 parser.add_option("--l1size",
  55                   default = "32kB")
  56 parser.add_option("--l1latency",
  57                   default = 1)
  58 parser.add_option("--l2size",
  59                   default = "256kB")
  60 parser.add_option("--l2latency",
  61                   default = 10)
  62 parser.add_option("--rootdir",
  63                   help="ROot directory of Splash2",
  64                   default="/dist/splash2/codes/")
  65 parser.add_option("-b", "--benchmark",
  66                   help="Splash 2 benchmark to run")
  67
  68 (options, args) = parser.parse_args()
  69
  70 if args:
  71     print("Error: script doesn't take any positional arguments")
  72     sys.exit(1)
  73
  74 # --------------------
  75 # Define Splash2 Benchmarks
  76 # ====================
  77 class Cholesky(Process):
  78         executable = options.rootdir + '/kernels/cholesky/CHOLESKY'
  79         cmd = 'CHOLESKY -p' + str(options.numcpus) + ' '\
  80              + options.rootdir + '/kernels/cholesky/inputs/tk23.O'
  81
  82 class FFT(Process):
  83         executable = options.rootdir + 'kernels/fft/FFT'
  84         cmd = 'FFT -p' + str(options.numcpus) + ' -m18'
  85
  86 class LU_contig(Process):
  87         executable = options.rootdir + 'kernels/lu/contiguous_blocks/LU'
  88         cmd = 'LU -p' + str(options.numcpus)
  89
  90 class LU_noncontig(Process):
  91         executable = options.rootdir + 'kernels/lu/non_contiguous_blocks/LU'
  92         cmd = 'LU -p' + str(options.numcpus)
  93
  94 class Radix(Process):
  95         executable = options.rootdir + 'kernels/radix/RADIX'
  96         cmd = 'RADIX -n524288 -p' + str(options.numcpus)
  97
  98 class Barnes(Process):
  99         executable = options.rootdir + 'apps/barnes/BARNES'
 100         cmd = 'BARNES'
 101         input = options.rootdir + 'apps/barnes/input.p' + str(options.numcpus)
 102
 103 class FMM(Process):
 104         executable = options.rootdir + 'apps/fmm/FMM'
 105         cmd = 'FMM'
 106         input = options.rootdir + 'apps/fmm/inputs/input.2048.p' + str(options.numcpus)
 107
 108 class Ocean_contig(Process):
 109         executable = options.rootdir + 'apps/ocean/contiguous_partitions/OCEAN'
 110         cmd = 'OCEAN -p' + str(options.numcpus)
 111
 112 class Ocean_noncontig(Process):
 113         executable = options.rootdir + 'apps/ocean/non_contiguous_partitions/OCEAN'
 114         cmd = 'OCEAN -p' + str(options.numcpus)
 115
 116 class Raytrace(Process):
 117         executable = options.rootdir + 'apps/raytrace/RAYTRACE'
 118         cmd = 'RAYTRACE -p' + str(options.numcpus) + ' ' \
 119              + options.rootdir + 'apps/raytrace/inputs/teapot.env'
 120
 121 class Water_nsquared(Process):
 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
 126 class Water_spatial(Process):
 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
 136 class L1(Cache):
 137     latency = options.l1latency
 138     mshrs = 12
 139     tgts_per_mshr = 8
 140
 141 # ----------------------
 142 # Base L2 Cache Definition
 143 # ----------------------
 144
 145 class L2(Cache):
 146     latency = options.l2latency
 147     mshrs = 92
 148     tgts_per_mshr = 16
 149     write_buffers = 8
 150
 151 # ----------------------
 152 # Define the clusters with their cpus
 153 # ----------------------
 154 class Cluster:
 155     pass
 156
 157 cpusPerCluster = options.numcpus/options.numclusters
 158
 159 busFrequency = Frequency(options.frequency)
 160 busFrequency *= cpusPerCluster
 161
 162 all_cpus = []
 163 all_l1s = []
 164 all_l1buses = []
 165 if options.timing:
 166     clusters = [ Cluster() for i in range(options.numclusters)]
 167     for j in range(options.numclusters):
 168         clusters[j].id = j
 169     for cluster in clusters:
 170         cluster.clusterbus = L2XBar(clock=busFrequency)
 171         all_l1buses += [cluster.clusterbus]
 172         cluster.cpus = [TimingSimpleCPU(cpu_id = i + cluster.id,
 173                                         clock=options.frequency)
 174                         for i in range(cpusPerCluster)]
 175         all_cpus += cluster.cpus
 176         cluster.l1 = L1(size=options.l1size, assoc = 4)
 177         all_l1s += [cluster.l1]
 178 elif options.detailed:
 179     clusters = [ Cluster() for i in range(options.numclusters)]
 180     for j in range(options.numclusters):
 181         clusters[j].id = j
 182     for cluster in clusters:
 183         cluster.clusterbus = L2XBar(clock=busFrequency)
 184         all_l1buses += [cluster.clusterbus]
 185         cluster.cpus = [DerivO3CPU(cpu_id = i + cluster.id,
 186                                    clock=options.frequency)
 187                         for i in range(cpusPerCluster)]
 188         all_cpus += cluster.cpus
 189         cluster.l1 = L1(size=options.l1size, assoc = 4)
 190         all_l1s += [cluster.l1]
 191 else:
 192     clusters = [ Cluster() for i in range(options.numclusters)]
 193     for j in range(options.numclusters):
 194         clusters[j].id = j
 195     for cluster in clusters:
 196         cluster.clusterbus = L2XBar(clock=busFrequency)
 197         all_l1buses += [cluster.clusterbus]
 198         cluster.cpus = [AtomicSimpleCPU(cpu_id = i + cluster.id,
 199                                         clock=options.frequency)
 200                         for i in range(cpusPerCluster)]
 201         all_cpus += cluster.cpus
 202         cluster.l1 = L1(size=options.l1size, assoc = 4)
 203         all_l1s += [cluster.l1]
 204
 205 # ----------------------
 206 # Create a system, and add system wide objects
 207 # ----------------------
 208 system = System(cpu = all_cpus, l1_ = all_l1s, l1bus_ = all_l1buses,
 209                 physmem = SimpleMemory(),
 210                 membus = SystemXBar(clock = busFrequency))
 211 system.clock = '1GHz'
 212
 213 system.toL2bus = L2XBar(clock = busFrequency)
 214 system.l2 = L2(size = options.l2size, assoc = 8)
 215
 216 # ----------------------
 217 # Connect the L2 cache and memory together
 218 # ----------------------
 219
 220 system.physmem.port = system.membus.master
 221 system.l2.cpu_side = system.toL2bus.slave
 222 system.l2.mem_side = system.membus.master
 223
 224 # ----------------------
 225 # Connect the L2 cache and clusters together
 226 # ----------------------
 227 for cluster in clusters:
 228     cluster.l1.cpu_side = cluster.clusterbus.master
 229     cluster.l1.mem_side = system.toL2bus.slave
 230     for cpu in cluster.cpus:
 231         cpu.icache_port = cluster.clusterbus.slave
 232         cpu.dcache_port = cluster.clusterbus.slave
 233
 234 # ----------------------
 235 # Define the root
 236 # ----------------------
 237
 238 root = Root(full_system = False, system = system)
 239
 240 # --------------------
 241 # Pick the correct Splash2 Benchmarks
 242 # ====================
 243 if options.benchmark == 'Cholesky':
 244     root.workload = Cholesky()
 245 elif options.benchmark == 'FFT':
 246     root.workload = FFT()
 247 elif options.benchmark == 'LUContig':
 248     root.workload = LU_contig()
 249 elif options.benchmark == 'LUNoncontig':
 250     root.workload = LU_noncontig()
 251 elif options.benchmark == 'Radix':
 252     root.workload = Radix()
 253 elif options.benchmark == 'Barnes':
 254     root.workload = Barnes()
 255 elif options.benchmark == 'FMM':
 256     root.workload = FMM()
 257 elif options.benchmark == 'OceanContig':
 258     root.workload = Ocean_contig()
 259 elif options.benchmark == 'OceanNoncontig':
 260     root.workload = Ocean_noncontig()
 261 elif options.benchmark == 'Raytrace':
 262     root.workload = Raytrace()
 263 elif options.benchmark == 'WaterNSquared':
 264     root.workload = Water_nsquared()
 265 elif options.benchmark == 'WaterSpatial':
 266     root.workload = Water_spatial()
 267 else:
 268     m5.util.panic("""
 269 The --benchmark environment variable was set to something improper.
 270 Use Cholesky, FFT, LUContig, LUNoncontig, Radix, Barnes, FMM, OceanContig,
 271 OceanNoncontig, Raytrace, WaterNSquared, or WaterSpatial
 272 """)
 273
 274 # --------------------
 275 # Assign the workload to the cpus
 276 # ====================
 277
 278 for cluster in clusters:
 279     for cpu in cluster.cpus:
 280         cpu.workload = root.workload
 281
 282 system.workload = SEWorkload.init_compatible(root.workload.executable)
 283
 284 # ----------------------
 285 # Run the simulation
 286 # ----------------------
 287
 288 if options.timing or options.detailed:
 289     root.system.mem_mode = 'timing'
 290
 291 # instantiate configuration
 292 m5.instantiate()
 293
 294 # simulate until program terminates
 295 if options.maxtick:
 296     exit_event = m5.simulate(options.maxtick)
 297 else:
 298     exit_event = m5.simulate(m5.MaxTick)
 299
 300 print('Exiting @ tick', m5.curTick(), 'because', exit_event.getCause())
 301