Chapel Performance Highlights: STREAM Triad
STREAM Triad is a
memory bandwidth benchmark that multiplies a large 1D array by a
scalar, adds it to a second array, and assigns it to a third. A good
implementation will use no communication during the computation,
making it pleasingly parallel. The performance delta in our results
can be attributed to using hugepages for Chapel but not for
the HPC Challenge reference
version.
Configuration
- System: Cray XC
- Network: Aries
- Processor: 2 x 18-core 2.1 GHz Broadwell processors
- Chapel version: chpl 1.21.0
- C version: gcc 8.3
- MPI version: cray-mpich 7.7.12