| Wiki Markup |
|---|
{viewpdf:name= |
...
Graphite_Day-Intro_and_Arch.pdf |
...
} The main page |
...
is [http://wiki.github.com/mit-carbon/Graphite/|http://wiki.github.com/mit-carbon/Graphite/ |
...
gettingstarted] The mailing list is: graphite-sim@googlegroups.com |
...
Here are my steps do generate fully functional Ubuntu Debian w/ Graphite. |
...
# Install new clean VM from ([http://www.debian.org/CD/netinst/]) |
...
# Follow instalation steps from Graphite site |
...
## upload PIN code (a tool for the dynamic instrumentation of programs from Intel) code. use version specified by Graphite-team even if it is not the latest one. Untar code and place it somewhere to stay permanently. I put it at \~/balewski/pin/ |
...
## Install the libraries needed for compiling (g++, make, etc…) using apt- |
...
get ## Install git-core. Although instruction says it is not essential, I could not compile the code on 32-bit architecture w/o it. |
...
## Upload Graphite tarball from github, unpack, place in permanent directory |
...
## Edit graphite/Makefile.config and set: |
...
### PIN_HOME to point to your PIN-code |
...
### change target architecture if you run it on 32-bit machine |
...
## type make and wait \~5 minutes |
...
# To test the code works execute: |
...
make cannon_app_test SIZE=4 |
...
You should see the output like this |
...
{code |
...
} balewski@debian5:~/graphite-2$ make cannon_app_test SIZE=4 tests/unit/Makefile:17: warning: overriding commands for target `clean' tests/apps/Makefile:16: warning: ignoring old commands for target `clean' tests/benchmarks/Makefile:9: warning: overriding commands for target `clean' tests/unit/Makefile:17: warning: ignoring old commands for target `clean' Makefile:23: warning: overriding commands for target `clean' tests/benchmarks/Makefile:9: warning: ignoring old commands for target `clean' make -C /home/balewski/graphite-2/tests/apps//cannon make[1]: Entering directory `/home/balewski/graphite-2/tests/apps/cannon' /home/balewski/graphite-2/tests/apps/cannon/../../../common/Makefile.common:82: cannon.d: No such file or directory cc -MM -MG -I/home/balewski/graphite-2/tests/apps/cannon/../../../common/user -I/home/balewski/graphite-2/tests/apps/cannon/../../../common/misc -c -Wall -O2 -m32 -DTARGET_IA32 -std=c99 cannon.c | sed -e 's,^\([^:]*\)\.o[ ]*:,./\1.o ./\1.d:,' >cannon.d make[1]: Leaving directory `/home/balewski/graphite-2/tests/apps/cannon' make[1]: Entering directory `/home/balewski/graphite-2/tests/apps/cannon' cc -I/home/balewski/graphite-2/tests/apps/cannon/../../../common/user -I/home/balewski/graphite-2/tests/apps/cannon/../../../common/misc -c -Wall -O2 -m32 -DTARGET_IA32 -std=c99 -c -o cannon.o cannon.c cannon.c: In function ‘do_cannon’: cannon.c:79: warning: ‘matSize’ may be used uninitialized in this function make -C /home/balewski/graphite-2/tests/apps/cannon/../../../common make[2]: Entering directory `/home/balewski/graphite-2/common' make[2]: Nothing to be done for `all'. make[2]: Leaving directory `/home/balewski/graphite-2/common' make -C /home/balewski/graphite-2/tests/apps/cannon/../../../pin make[2]: Entering directory `/home/balewski/graphite-2/pin' make[2]: Nothing to be done for `all'. make[2]: Leaving directory `/home/balewski/graphite-2/pin' if [ ! -e cannon ] || [ cannon.o -nt cannon ] || [ /home/balewski/graphite-2/tests/apps/cannon/../../../lib/libcarbon_sim.a -nt cannon ] || [ /home/balewski/graphite-2/tests/apps/cannon/../../../lib/pin_sim.so -nt cannon ]; \ then g++ cannon.o -o cannon -static -u CarbonStartSim -u CarbonStopSim -upthread_create -upthread_join -L/home/balewski/graphite-2/tests/apps/cannon/../../../lib -los-services -L /home/balewski/graphite-2/tests/apps/cannon/../../../os-services-25032-gcc.4.0.0-linux-ia32_intel64/ia32 -m32 -L/home/balewski/graphite-2/tests/apps/cannon/../../../contrib/orion -L/home/balewski/graphite-2/tests/apps/cannon/../../../lib -pthread -lcarbon_sim -los-services -lboost_filesystem-mt -lboost_system-mt -pthread -lorion; \ fi /home/balewski/graphite-2/tests/apps/cannon/../../../lib/libcarbon_sim.a(socktransport.o): In function `SockTransport::Socket::connect(char const*, int)': socktransport.cc:(.text+0x1a0c): warning: Using 'gethostbyname' in statically linked applications requires at runtime the shared libraries from the glibc version used for linking cd /home/balewski/graphite-2/tests/apps/cannon/../../.. ; /home/balewski/graphite-2/tests/apps/cannon/../../../tools/spawn.py 1 /home/balewski/graphite-2/tests/apps/cannon/../../../carbon_sim.cfg /home/balewski/pin/ia32/bin/pinbin -mt -t /home/balewski/graphite-2/tests/apps/cannon/../../../lib/pin_sim -c /home/balewski/graphite-2/tests/apps/cannon/../../../carbon_sim.cfg --general/total_cores=5 --general/num_processes=1 --general/enable_shared_mem=true -- /home/balewski/graphite-2/tests/apps/cannon/cannon -t 4 -m 4 [spawn.py] 'GRAPHITE_ROOT' undefined. Setting 'GRAPHITE_ROOT' to '/home/balewski/graphite-2 [spawn.py] Starting process: 0 : export CARBON_PROCESS_INDEX=0; export LD_LIBRARY_PATH="/afs/csail/group/carbon/tools/boost_1_38_0/stage/lib"; /home/balewski/pin/ia32/bin/pinbin -mt -t /home/balewski/graphite-2/tests/apps/cannon/../../../lib/pin_sim -c /home/balewski/graphite-2/tests/apps/cannon/../../../carbon_sim.cfg --general/total_cores=5 --general/num_processes=1 --general/enable_shared_mem=true -- /home/balewski/graphite-2/tests/apps/cannon/cannon -t 4 -m 4 [[Graphite]] --> [ Core IDs' with memory controllers = (0 1 2 3 4 5 6 ) ] Starting iteration 0... Allocating and Initializing matrix a Allocating and Initializing matrix b Allocating and Initializing matrix c Initializing thread structures Thread 0 starting to retrieve initial data Thread 0 finished retrieving initial data, starting computation Thread 1 starting to retrieve initial data Thread 2 starting to retrieve initial data Thread 3 starting to retrieve initial data Thread 1 finished retrieving initial data, starting computation Thread 2 finished retrieving initial data, starting computation Thread 3 finished retrieving initial data, starting computation Done sending, waiting for worker threads to complete 24.000000 Exiting... [spawn.py] Exited with return code: 0 {code} |
...
Also the log file should show some performance benchmarks |
...
{code |
...
}cat output_files/sim.out Simulation timers: start time 3623124 stop time 5712947 shutdown time 6039413 | Core 0 | Core 1 | Core 2 | Core 3 | Core 4 | TS 0 | MCP | Core Performance Model Summary | | | | | | | | Instructions | 15981 | 6188 | 6186 | 6156 | 6175 | 2754 | 0 | Completion Time | 123142 | 105450 | 104754 | 103939 | 101949 | 14449 | 0 | Average Frequency | 1 | 1 | 1 | 1 | 1 | 0 | 0 | Recv Instructions | 4 | 0 | 0 | 0 | 0 | 0 | 0 | Recv Instruction Costs | 14336 | 0 | 0 | 0 | 0 | 0 | 0 | Sync Instructions | 3 | 9 | 8 | 7 | 8 | 0 | 0 | Sync Instruction Costs | 54655 | 45850 | 42055 | 42596 | 42078 | 0 | 0 | Branch predictor stats | | | | | | | | num correct | 1886 | 694 | 695 | 691 | 698 | 267 | 0 | num incorrect | 384 | 195 | 195 | 196 | 191 | 39 | 0 | type | one-bit (1024) | one-bit (1024) | one-bit (1024) | one-bit (1024) | one-bit (1024) | one-bit (1024) | one-bit (1024) | Network summary | | | | | | | | Network model 0 | | | | | | | | num packets received | 8 | 17 | 17 | 17 | 17 | 0 | 131 | num bytes received | 384 | 764 | 764 | 764 | 764 | 0 | 9856 | average latency (in clock cycles) | 9 | 8.47059 | 8.47059 | 9.52941 | 9.52941 | nan | 14.8244 | average latency (in ns) | 9 | 8.47059 | 8.47059 | 9.52941 | 9.52941 | nan | 14.8244 | Static Power | 0.0176561 | 0.0176561 | 0.0176561 | 0.0176561 | 0.0176561 | 0.0176561 | 0.0176561 | Dynamic Energy | 1.89929e-06 | 1.86787e-06 | 9.37239e-07 | 1.2594e-06 | 5.22373e-07 | 0 | 0 | Activity Counters | | | | | | | | Switch Allocator Traversals | 153 | 122 | 64 | 85 | 39 | 0 | 0 | Crossbar Traversals | 1167 | 1148 | 576 | 774 | 321 | 0 | 0 | Link Traversals | 1167 | 1148 | 576 | 774 | 321 | 0 | 0 | Network model 1 | | | | | | | | num packets received | 0 | 0 | 0 | 0 | 0 | 0 | 0 | num bytes received | 0 | 0 | 0 | 0 | 0 | 0 | 0 | average latency (in clock cycles) | nan | nan | nan | nan | nan | nan | nan | average latency (in ns) | nan | nan | nan | nan | nan | nan | nan | Static Power | 0.0176561 | 0.0176561 | 0.0176561 | 0.0176561 | 0.0176561 | 0.0176561 | 0.0176561 | Dynamic Energy | 0 | 0 | 0 | 0 | 0 | 0 | 0 | Activity Counters | | | | | | | | Switch Allocator Traversals | 0 | 0 | 0 | 0 | 0 | 0 | 0 | Crossbar Traversals | 0 | 0 | 0 | 0 | 0 | 0 | 0 | Link Traversals | 0 | 0 | 0 | 0 | 0 | 0 | 0 | Network model 2 | | | | | | | | num packets received | 345 | 368 | 827 | 384 | 332 | 188 | 138 | num bytes received | 19257 | 20720 | 41499 | 21696 | 18892 | 8188 | 5962 | average latency (in clock cycles) | 7.47246 | 8.29076 | 6.14994 | 7.75 | 7.9006 | 8.58511 | 10.9275 | average latency (in ns) | 7.47246 | 8.29076 | 6.14994 | 7.75 | 7.9006 | 8.58511 | 10.9275 | Static Power | 0.0176561 | 0.0176561 | 0.0176561 | 0.0176561 | 0.0176561 | 0.0176561 | 0.0176561 | Dynamic Energy | 6.8307e-06 | 5.40692e-06 | 8.50509e-06 | 4.03165e-06 | 3.80035e-06 | 4.47535e-06 | 4.02241e-06 | Activity Counters | | | | | | | | Switch Allocator Traversals | 462 | 458 | 674 | 345 | 315 | 251 | 280 | Crossbar Traversals | 4198 | 3322 | 5226 | 2477 | 2335 | 2751 | 2472 | Link Traversals | 4198 | 3322 | 5226 | 2477 | 2335 | 2751 | 2472 | Network model 3 | | | | | | | | num packets received | 0 | 0 | 0 | 0 | 0 | 0 | 0 | num bytes received | 0 | 0 | 0 | 0 | 0 | 0 | 0 | average latency (in clock cycles) | nan | nan | nan | nan | nan | nan | nan | average latency (in ns) | nan | nan | nan | nan | nan | nan | nan | Static Power | 0.0176561 | 0.0176561 | 0.0176561 | 0.0176561 | 0.0176561 | 0.0176561 | 0.0176561 | Dynamic Energy | 0 | 0 | 0 | 0 | 0 | 0 | 0 | Activity Counters | | | | | | | | Switch Allocator Traversals | 0 | 0 | 0 | 0 | 0 | 0 | 0 | Crossbar Traversals | 0 | 0 | 0 | 0 | 0 | 0 | 0 | Link Traversals | 0 | 0 | 0 | 0 | 0 | 0 | 0 | Network model 4 | | | | | | | | num packets received | 1 | 1 | 1 | 1 | 1 | 0 | 0 | num bytes received | 40 | 40 | 40 | 40 | 40 | 0 | 0 | Shmem Perf Model summary | | | | | | | | num memory accesses | 8036 | 2950 | 2950 | 2937 | 2945 | 1487 | 0 | average memory access latency | 4.75075 | 8.31458 | 9.35627 | 8.93667 | 8.39728 | 7.86685 | nan | Cache Summary | | | | | | | | Cache L1-I | | | | | | | | num cache accesses | 0 | 0 | 0 | 0 | 0 | 0 | 0 | miss rate | nan | nan | nan | nan | nan | nan | nan | num cache misses | 0 | 0 | 0 | 0 | 0 | 0 | 0 | Cache L1-D | | | | | | | | num cache accesses | 8066 | 2954 | 2954 | 2941 | 2949 | 1568 | 0 | miss rate | 1.46293 | 3.65606 | 4.40081 | 3.94424 | 3.62835 | 8.09949 | nan | num cache misses | 118 | 108 | 130 | 116 | 107 | 127 | 0 | Cache L2 | | | | | | | | num cache accesses | 118 | 108 | 130 | 116 | 107 | 127 | 0 | miss rate | 74.5763 | 100 | 100 | 100 | 100 | 100 | nan | num cache misses | 88 | 108 | 130 | 116 | 107 | 127 | 0 | Dram Perf Model summary | | | | | | | | num dram accesses | 80 | 68 | 183 | 57 | 46 | 87 | 58 | average dram access latency | 113 | 113.074 | 113.076 | 113 | 113 | 113 | 113.276 | average dram queueing delay | 0 | 0.0735294 | 0.0765027 | 0 | 0 | 0 | 0.275862 | Queue Model | | | | | | | | Queue Utilization(%) | 0.920492 | 0.728351 | 2.14126 | 0.654045 | 0.528698 | 0.996801 | 0.665631 | Analytical Model Used(%) | 0 | 0 | 0 | 0 | 0 | 0 | 0 | Dram Directory Cache | | | | | | | | Total Addresses | 1170 | 1169 | 1166 | 1169 | 1170 | 1174 | 1173 | Average set size | 1.14258 | 1.1416 | 1.13867 | 1.1416 | 1.14258 | 1.14648 | 1.14551 | Set index with max size | 1022 | 1023 | 1017 | 1022 | 995 | 994 | 1014 | Max set size | 6 | 6 | 5 | 5 | 5 | 5 | 5 | Set index with min size | 543 | 543 | 543 | 544 | 543 | 543 | 543 | Min set size | 0 | 0 | 0 | 0 | 0 | 0 | 0 | Average evictions per set | 0 | 0 | 0 | 0 | 0 | 0 | 0 | Set index with max evictions | -1 | -1 | -1 | -1 | -1 | -1 | -1 | Max set evictions | 0 | 0 | 0 | 0 | 0 | 0 | 0 | Address with max evictions | 0xffffffff | 0xffffffff | 0xffffffff | 0xffffffff | 0xffffffff | 0xffffffff | 0xffffffff | Max address evictions | 0 | 0 | 0 | 0 | 0 | 0 | 0 | balewski@debian5:~/graphite-2$ {code} |