NERSCPowering Scientific Discovery Since 1974

Change Log

Crossroads/NERSC-9 Benchmarks Change Log


Updated to latest version 3.0 release of the HPCG distribution. For the APEX benchmark, the -DHPCG_CONTIGUOUS_ARRAYS flag is now used which improves the baseline performance on Edison significantly. Please refer to the updated SSI spreadsheet for the new baseline value.


Updated source distribution. README.APEX now states to use "Grind Time" as the figure of merit. Failed to do this on the 5/10/2016 update. There are no other distribution changes.

Updated the web site SSI Reference Table to Grind Time for SNAP.


Updated SNAP: New source distribution, LANL release version 1.07
-- See snap_release_v1.07.txt in the distribution for a list of changes.
-- Now using Grind Time as the figure of merit. New Edison baseline performance values.

Updated SSI Spreadsheet: Changed SNAP metric to Grind Time and updated Edison reference value.


Updated GTC-P: Edison SSI reference time was updated. With the source changes made on 3/28/2016 the reference time was not updated to correspond to the new code, and the increase in iteration count for the Edison reference problem. NO source or job scripts are updated, just the reference time in the SSI spreadsheet.

Updated MILC: It was discovered that MILC's random number generator was not thread safe, and hence it was possible for threads to get an improper random number sequence. The source distribution has been updated to fix this problem. To see the updated files, untar the new distribution and use the following commands:
$ cd MILC-apex
$ find . -newer README.APEX -name *.[ch]
This update does not significantly change the overall execution time, but it does change the resultant values that are used to validate the final result. Therefore, the Edison reference result has also been updated in the source distribution in addition to the SSI spreadsheet.
In addition, a license file has been added to the distribution.

Updated MiniPIC: README.APEX was updated. The language on how to construct a weak scaling study was improved. NO other source or directory changes were made.


Updated GTC-P: Source distribution updated; See GTC-P change log for details.


Updated GTC-P: Source distribution updated; See GTC-P change log for details.

Updated MiniDFT: Source distribution updated; Improved OpenMP performance; Updated the figure of merit used for the Edison baseline; SSI spreadsheet has been updated.

Changed PENNANT figure of merit from seconds to seconds/cycle; SSI spreadsheet has been updated.


Updated PENNANT benchmark.
Updated SSI spreadsheet to reflect new PENNANT reference timing.


Updated MPI Benchmarks section. Added link for SMB multi-threaded MPI benchmarks, which are to be used for message rate and latency measurements called out in the Application Performance section of the RFP.


New miniPIC distribution:
1) Updated source and README.APEX, 2) FOM is now updates/second, 3) Edison reference problem has been redefined and new FOM is being used, 4) also updated small and grand challenge problem definitions, 5) code modifications to improve reliability.

Update SSI spreadsheet to reflect changes to miniPIC FOM.


Updated SSI benchmark spreadsheet to address issues with HPCG and UMT FOM calculations.

UMT distribution updated:
Updated source to address missing .F90 files
Updated README.APEX file
Updated benchmarks directory to include example small problem


Updated example results in the PENNANT distribution. Source and problem definitions are unchanged.


Added source distribution for UMT2015.
Added UMT2015 reference values and updated the SSI spreadsheet.


Updated source distribution for mt-dgemm with support for CBLAS, NVBLAS and MKL.


MILC source distribution updated. See MILC page and README.APEX in the distribution.
SSI spreadsheet updated with new reference timing for MILC.


Added source distribution for the mdtest benchmark.
Updated IOR's README.APEX in the source distribution.


Added source distribution and updated web sites for PENNANT, SNAP and IOR.
Provided clearer definition of the reference and target problem definitions to be used in SSI.
Added FOM values for PENNANT and SNAP.
Updated the SSI spreadsheet to include reference FOM values, "c" factors, and weights.
Added MPI benchmarks description.


Added source distribution and SSI FOM reference value for MiniPIC.
Added MILC FOM reference value.


Added source distribution and SSI FOM reference value for HPCG.


Updated to include tentative SSI FOM reference values, source distribution for GTC-P, Meraculous, MILC, MiniDFT, DGEMM and STREAM.


Initial release