Cloverleaf is a hydrodynamics benchmark with three workloads. Almost all the time is spent in the second workload. The overall profile suggests a runable process on every core.
Topdown metrics show a very memory bound application.with little time in retiring instructions.
AMD metrics show floating point code without many branches. There is a reasonably high L2 miss rate.
elapsed 5873.670
on_cpu 0.972 # 15.56 / 16 cores
utime 91203.508
stime 189.167
nvcsw 3950849 # 83.79%
nivcsw 764581 # 16.21%
inblock 8 # 0.00/sec
onblock 47808 # 8.14/sec
cpu-clock 91496573378200 # 91496.573 seconds
task-clock 91501724220020 # 91501.724 seconds
page faults 11117501 # 121.500/sec
context switches 4744569 # 51.852/sec
cpu migrations 100280 # 1.096/sec
major page faults 55 # 0.001/sec
minor page faults 11117446 # 121.500/sec
alignment faults 0 # 0.000/sec
emulation faults 0 # 0.000/sec
branches 5634504972305 # 52.699 branches per 1000 inst
branch misses 34518004796 # 0.61% branch miss
conditional 4723812723676 # 44.181 conditional branches per 1000 inst
indirect 450208732 # 0.004 indirect branches per 1000 inst
cpu-cycles 416297076847393 # 4.42 GHz
instructions 106923594819988 # 0.26 IPC
slots 832408287294990 #
retiring 36729774536367 # 4.4% ( 4.7%)
-- ucode 16033010471 # 0.0%
-- fastpath 36713741525896 # 4.4%
frontend 19057802930313 # 2.3% ( 2.4%)
-- latency 13207251742008 # 1.6%
-- bandwidth 5850551188305 # 0.7%
backend 733264248976694 # 88.1% (92.9%)
-- cpu 125417997459295 # 15.1%
-- memory 607846251517399 # 73.0%
speculation 665948773067 # 0.1% ( 0.1%)
-- branch mispredict 591782708845 # 0.1%
-- pipeline restart 74166064222 # 0.0%
smt-contention 42689832224123 # 5.1% ( 0.0%)
cpu-cycles 416281856017187 # 4.42 GHz
instructions 106924788382785 # 0.26 IPC
instructions 35640174860193 # 85.328 l2 access per 1000 inst
l2 hit from l1 1723378795510 # 34.64% l2 miss
l2 miss from l1 239934204030 #
l2 hit from l2 pf 504278478852 #
l3 hit from l2 pf 37875169844 #
l3 miss from l2 pf 775570101725 #
instructions 35622630123443 # 271.330 float per 1000 inst
float 512 60 # 0.000 AVX-512 per 1000 inst
float 256 672 # 0.000 AVX-256 per 1000 inst
float 128 9665485999618 # 271.330 AVX-128 per 1000 inst
float MMX 0 # 0.000 MMX per 1000 inst
float scalar 1 # 0.000 scalar per 1000 inst
Intel metrics similarly show missing L2 cache.
elapsed 9195.794
on_cpu 0.893 # 14.28 / 16 cores
utime 131158.728
stime 180.440
nvcsw 4459732 # 79.97%
nivcsw 1117243 # 20.03%
inblock 7232 # 0.79/sec
onblock 53072 # 5.77/sec
cpu-clock 131303755757174 # 131303.756 seconds
task-clock 131309362105649 # 131309.362 seconds
page faults 11289647 # 85.977/sec
context switches 5622700 # 42.820/sec
cpu migrations 690226 # 5.256/sec
major page faults 162 # 0.001/sec
minor page faults 11289485 # 85.976/sec
alignment faults 0 # 0.000/sec
emulation faults 0 # 0.000/sec
branches 6119594162100 # 52.521 branches per 1000 inst
branch misses 32286755507 # 0.53% branch miss
conditional 6119594182260 # 52.521 conditional branches per 1000 inst
indirect 2090955801758 # 17.945 indirect branches per 1000 inst
slots 474390606585344 #
retiring 71789897080078 # 15.1% (15.1%)
-- ucode 13179968617235 # 2.8%
-- fastpath 58609928462843 # 12.4%
frontend 29418725753857 # 6.2% ( 6.2%)
-- latency 18241180851442 # 3.8%
-- bandwidth 11177544902415 # 2.4%
backend 378719773656819 # 79.8% (79.8%)
-- cpu 62995505319387 # 13.3%
-- memory 315724268337432 # 66.6%
speculation 4567039699865 # 1.0% ( 1.0%)
-- branch mispredict 3065558847064 # 0.6%
-- pipeline restart 1501480852801 # 0.3%
smt-contention 0 # 0.0% ( 0.0%)
cpu-cycles 345428388454584 # 2.51 GHz
instructions 122694459382065 # 0.36 IPC
l2 access 3480548081213 # 62.611 l2 access per 1000 inst
l2 miss 1791504627857 # 51.47% l2 miss
Process overview crashed part way through the second workload so we don’t have a full account.
elapsed 9195.794
on_cpu 0.893 # 14.28 / 16 cores
utime 131158.728
stime 180.440
nvcsw 4459732 # 79.97%
nivcsw 1117243 # 20.03%
inblock 7232 # 0.79/sec
onblock 53072 # 5.77/sec
cpu-clock 131303755757174 # 131303.756 seconds
task-clock 131309362105649 # 131309.362 seconds
page faults 11289647 # 85.977/sec
context switches 5622700 # 42.820/sec
cpu migrations 690226 # 5.256/sec
major page faults 162 # 0.001/sec
minor page faults 11289485 # 85.976/sec
alignment faults 0 # 0.000/sec
emulation faults 0 # 0.000/sec
branches 6119594162100 # 52.521 branches per 1000 inst
branch misses 32286755507 # 0.53% branch miss
conditional 6119594182260 # 52.521 conditional branches per 1000 inst
indirect 2090955801758 # 17.945 indirect branches per 1000 inst
slots 474390606585344 #
retiring 71789897080078 # 15.1% (15.1%)
-- ucode 13179968617235 # 2.8%
-- fastpath 58609928462843 # 12.4%
frontend 29418725753857 # 6.2% ( 6.2%)
-- latency 18241180851442 # 3.8%
-- bandwidth 11177544902415 # 2.4%
backend 378719773656819 # 79.8% (79.8%)
-- cpu 62995505319387 # 13.3%
-- memory 315724268337432 # 66.6%
speculation 4567039699865 # 1.0% ( 1.0%)
-- branch mispredict 3065558847064 # 0.6%
-- pipeline restart 1501480852801 # 0.3%
smt-contention 0 # 0.0% ( 0.0%)
cpu-cycles 345428388454584 # 2.51 GHz
instructions 122694459382065 # 0.36 IPC
l2 access 3480548081213 # 62.611 l2 access per 1000 inst
l2 miss 1791504627857 # 51.47% l2 miss