https://github.com/andih/cuda-fortran-stream

Variant of STREAM Benchmark in CUDA Fortran
https://github.com/andih/cuda-fortran-stream

cuda cuda-fortran gpu stream-benchmarks variants

Last synced: over 1 year ago
JSON representation

Variant of STREAM Benchmark in CUDA Fortran

• Host: GitHub
• URL: https://github.com/andih/cuda-fortran-stream
• Owner: AndiH
• License: mit
• Created: 2017-01-31T13:57:44.000Z (over 9 years ago)
• Default Branch: master
• Last Pushed: 2017-02-09T12:10:17.000Z (over 9 years ago)
• Last Synced: 2025-01-12T18:37:43.104Z (over 1 year ago)
• Topics: cuda, cuda-fortran, gpu, stream-benchmarks, variants
• Language: Fortran
• Size: 22.5 KB
• Stars: 1
• Watchers: 2
• Forks: 1
• Open Issues: 2
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

          
# CUDA Fortran STREAM Benchmark

Variant of the STREAM benchmark written in CUDA Fortran (and, hence, working on the GPU).

The [PGI compiler](https://www.pgroup.com/resources/cudafortran.htm) is required to compile.

The four STREAM benchmarks are run on the GPU, that is (with arrays `a`, `b`, `c`):

* COPY: `c_j = a_j`

* SCALE: `b_j = scalar * c_j`

* ADD: `c_j = a_j + b_j`

* TRIAD: `a_j = b_j + scalar * c_j`

For information is available in `stream.F90`. The `Makefile` should suffice to call `make run` for compiling and running the four micro-benchmarks.

See also [www.streambench.org](http://www.streambench.org), especially for notes on publishing results based on (variants of) the STREAM benchmark.