configs/splash2/cluster.py

   1 # Copyright (c) 2006 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
  33 import m5
  34 from m5.objects import *
  35 import os, optparse, sys
  36 m5.AddToPath('../common')
  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("-p", "--protocol",
  55                   default="moesi",
  56                   help="The coherence protocol to use for the L1'a (i.e. MOESI, MOSI)")
  57 parser.add_option("--l1size",
  58                   default = "32kB")
  59 parser.add_option("--l1latency",
  60                   default = 1)
  61 parser.add_option("--l2size",
  62                   default = "256kB")
  63 parser.add_option("--l2latency",
  64                   default = 10)
  65 parser.add_option("--rootdir",
  66                   help="ROot directory of Splash2",
  67                   default="/dist/splash2/codes/")
  68 parser.add_option("-b", "--benchmark",
  69                   help="Splash 2 benchmark to run")
  70
  71 (options, args) = parser.parse_args()
  72
  73 if args:
  74     print "Error: script doesn't take any positional arguments"
  75     sys.exit(1)
  76
  77 # --------------------
  78 # Define Splash2 Benchmarks
  79 # ====================
  80 class Cholesky(LiveProcess):
  81         executable = options.rootdir + '/kernels/cholesky/CHOLESKY'
  82         cmd = 'CHOLESKY -p' + str(options.numcpus) + ' '\
  83              + options.rootdir + '/kernels/cholesky/inputs/tk23.O'
  84
  85 class FFT(LiveProcess):
  86         executable = options.rootdir + 'kernels/fft/FFT'
  87         cmd = 'FFT -p' + str(options.numcpus) + ' -m18'
  88
  89 class LU_contig(LiveProcess):
  90         executable = options.rootdir + 'kernels/lu/contiguous_blocks/LU'
  91         cmd = 'LU -p' + str(options.numcpus)
  92
  93 class LU_noncontig(LiveProcess):
  94         executable = options.rootdir + 'kernels/lu/non_contiguous_blocks/LU'
  95         cmd = 'LU -p' + str(options.numcpus)
  96
  97 class Radix(LiveProcess):
  98         executable = options.rootdir + 'kernels/radix/RADIX'
  99         cmd = 'RADIX -n524288 -p' + str(options.numcpus)
 100
 101 class Barnes(LiveProcess):
 102         executable = options.rootdir + 'apps/barnes/BARNES'
 103         cmd = 'BARNES'
 104         input = options.rootdir + 'apps/barnes/input.p' + str(options.numcpus)
 105
 106 class FMM(LiveProcess):
 107         executable = options.rootdir + 'apps/fmm/FMM'
 108         cmd = 'FMM'
 109         input = options.rootdir + 'apps/fmm/inputs/input.2048.p' + str(options.numcpus)
 110
 111 class Ocean_contig(LiveProcess):
 112         executable = options.rootdir + 'apps/ocean/contiguous_partitions/OCEAN'
 113         cmd = 'OCEAN -p' + str(options.numcpus)
 114
 115 class Ocean_noncontig(LiveProcess):
 116         executable = options.rootdir + 'apps/ocean/non_contiguous_partitions/OCEAN'
 117         cmd = 'OCEAN -p' + str(options.numcpus)
 118
 119 class Raytrace(LiveProcess):
 120         executable = options.rootdir + 'apps/raytrace/RAYTRACE'
 121         cmd = 'RAYTRACE -p' + str(options.numcpus) + ' ' \
 122              + options.rootdir + 'apps/raytrace/inputs/teapot.env'
 123
 124 class Water_nsquared(LiveProcess):
 125         executable = options.rootdir + 'apps/water-nsquared/WATER-NSQUARED'
 126         cmd = 'WATER-NSQUARED'
 127         input = options.rootdir + 'apps/water-nsquared/input.p' + str(options.numcpus)
 128
 129 class Water_spatial(LiveProcess):
 130         executable = options.rootdir + 'apps/water-spatial/WATER-SPATIAL'
 131         cmd = 'WATER-SPATIAL'
 132         input = options.rootdir + 'apps/water-spatial/input.p' + str(options.numcpus)
 133
 134
 135 # --------------------
 136 # Base L1 Cache Definition
 137 # ====================
 138
 139 class L1(BaseCache):
 140     latency = options.l1latency
 141     block_size = 64
 142     mshrs = 12
 143     tgts_per_mshr = 8
 144     protocol = CoherenceProtocol(protocol=options.protocol)
 145
 146 # ----------------------
 147 # Base L2 Cache Definition
 148 # ----------------------
 149
 150 class 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 # ----------------------
 160 class Cluster:
 161     pass
 162
 163 cpusPerCluster = options.numcpus/options.numclusters
 164
 165 busFrequency = Frequency(options.frequency)
 166 busFrequency *= cpusPerCluster
 167
 168 all_cpus = []
 169 all_l1s = []
 170 all_l1buses = []
 171 if 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 = Bus(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]
 184 elif 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 = Bus(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]
 197 else:
 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 = Bus(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 # ----------------------
 214 system = System(cpu = all_cpus, l1_ = all_l1s, l1bus_ = all_l1buses, physmem = PhysicalMemory(),
 215                 membus = Bus(clock = busFrequency))
 216
 217 system.toL2bus = Bus(clock = busFrequency)
 218 system.l2 = L2(size = options.l2size, assoc = 8)
 219
 220 # ----------------------
 221 # Connect the L2 cache and memory together
 222 # ----------------------
 223
 224 system.physmem.port = system.membus.port
 225 system.l2.cpu_side = system.toL2bus.port
 226 system.l2.mem_side = system.membus.port
 227
 228 # ----------------------
 229 # Connect the L2 cache and clusters together
 230 # ----------------------
 231 for cluster in clusters:
 232     cluster.l1.cpu_side = cluster.clusterbus.port
 233     cluster.l1.mem_side = system.toL2bus.port
 234     for cpu in cluster.cpus:
 235         cpu.icache_port = cluster.clusterbus.port
 236         cpu.dcache_port = cluster.clusterbus.port
 237         cpu.mem = cluster.l1
 238
 239 # ----------------------
 240 # Define the root
 241 # ----------------------
 242
 243 root = Root(system = system)
 244
 245 # --------------------
 246 # Pick the correct Splash2 Benchmarks
 247 # ====================
 248 if options.benchmark == 'Cholesky':
 249     root.workload = Cholesky()
 250 elif options.benchmark == 'FFT':
 251     root.workload = FFT()
 252 elif options.benchmark == 'LUContig':
 253     root.workload = LU_contig()
 254 elif options.benchmark == 'LUNoncontig':
 255     root.workload = LU_noncontig()
 256 elif options.benchmark == 'Radix':
 257     root.workload = Radix()
 258 elif options.benchmark == 'Barnes':
 259     root.workload = Barnes()
 260 elif options.benchmark == 'FMM':
 261     root.workload = FMM()
 262 elif options.benchmark == 'OceanContig':
 263     root.workload = Ocean_contig()
 264 elif options.benchmark == 'OceanNoncontig':
 265     root.workload = Ocean_noncontig()
 266 elif options.benchmark == 'Raytrace':
 267     root.workload = Raytrace()
 268 elif options.benchmark == 'WaterNSquared':
 269     root.workload = Water_nsquared()
 270 elif options.benchmark == 'WaterSpatial':
 271     root.workload = Water_spatial()
 272 else:
 273     panic("The --benchmark environment variable was set to something" \
 274           +" improper.\nUse Cholesky, FFT, LUContig, LUNoncontig, Radix" \
 275           +", Barnes, FMM, OceanContig,\nOceanNoncontig, Raytrace," \
 276           +" WaterNSquared, or WaterSpatial\n")
 277
 278 # --------------------
 279 # Assign the workload to the cpus
 280 # ====================
 281
 282 for cluster in clusters:
 283     for cpu in cluster.cpus:
 284         cpu.workload = root.workload
 285
 286 # ----------------------
 287 # Run the simulation
 288 # ----------------------
 289
 290 if options.timing or options.detailed:
 291     root.system.mem_mode = 'timing'
 292
 293 # instantiate configuration
 294 m5.instantiate(root)
 295
 296 # simulate until program terminates
 297 if options.maxtick:
 298     exit_event = m5.simulate(options.maxtick)
 299 else:
 300     exit_event = m5.simulate(m5.MaxTick)
 301
 302 print 'Exiting @ tick', m5.curTick(), 'because', exit_event.getCause()
 303