Itasca Benchmark Test (v1.0)

• The Benchmark Speed Test currently allows using 3 Itasca software products:
• FLAC3D (v6.0 and up)
• 3DEC (v7.0)
• PFC3D (v6.0 and up)


• How does the Benchmark Speed Test work?
• The benchmark tests work by generating a relatively large model that is run for 50 calculation steps (cycles) in both small- and large- strain deformations.
• The model specifications used in each software product are as follows:
1. For FLAC3D, the model is discretized into 3,128,000 zones.
2. For 3DEC, the model consists of 34,000 blocks; 1 million zones; and 500,000 sub-contacts.
   Note: The model previously adopted in 3DEC in the benchmark v1.0 release consisted of 18,326 blocks; 212 ,054 zones; and 131,428 sub-contacts.
   Note: The current implementation will be under v2.0
3. For PFC3D, the model is made up of a hexagon packing of 1,430,396 balls.
• The test runs on the software product of choice using two constitutive models:
  1. Mohr-Coulomb and SUBI are used in the FLAC3D and 3DEC software products.
  2. Linear and Soft-Bond are used in the PFC3D software product.
• If the speed test is successful, the results including the speed of each case, and the software product, CPU and memory specifications, are uploaded automatically, and can be viewed and compared to other workstation's performance on http://benchmark.itascacloud.com/ .
• If any error message is encountered, the results won't be uploaded, and please feel free to email the error file to jfurtney@itascacg.com and r.legoc@itasca.fr.


• How is the Benchmark Speed Test Run?
• To initiate the speed test for your workstation, copy and paste the command line corresponding to the software product of interest in the IPython console (Menu Bar->Panes->IPython Console) and run the program.
• The commands according to the software product of interest are as follows:
  1. FLAC3D :
         import six; six.exec_(six.moves.urllib.request.urlopen('http://benchmark.itascacloud.com/_FLAC3D_speed_test.py').read())
  2. 3DEC :
         import six; six.exec_(six.moves.urllib.request.urlopen('http://benchmark.itascacloud.com/_3DEC_speed_test.py').read())
  3. PFC3D :
         import six; six.exec_(six.moves.urllib.request.urlopen('http://benchmark.itascacloud.com/_PFC3D_speed_test.py').read())


• Benchmark speed test website features:
• The website allows to view the Benchmark Speed Test results from different workstations.
• It has a drop-down list at the top that allows for the choice of the software product you prefer to view its results, as well as a button that refers to the documentation.
• For each software product, a table encapsulates the benchmark results as follows:
1. Hostname: the hostname of the computer used to undergo the test.
2. Operating System: the operating system of the computer used to undergo the test (Windows or Linux).
3. Processor: the processor type of the computer used to undergo the test.
4. vCPUs: the max number of threads of the computer used to undergo the test.
5. L3 Cache Size (KB): the size of the L3 cache memory of the computer used to undergo the test, in kilobytes.
6. Memory (MHz): the RAM memory of the computer used to undergo the test, in Megahertz. * This will not show for Linux machines as it requires root access.
7. Software: the software product name and version used to undergo the test.
8. Benchmark Version: the release version of the benchmark website.
9. Max Speed Test: the speed of the test at the max number of threads presented for the different constitutive/contact models used and at both small- and large-strain deformations.
10. File: a downloadable .json file including the corresponding run results, and software product, CPU and memory specifications.
11. Created: the date and time when the run was conducted and the results were uploaded.
• The table can be navigated using:
1. The search box at the top-right of the table that allows filtering the results.
2. The ascending and descending sorting allowed for each column.