From cTuning.org
(Difference between revisions)
(New page: Test) |
Current revision (08:14, 20 April 2014) (view source) |
||
| Line 1: | Line 1: | ||
| - | + | '''Panel at ADAPT 2014''' | |
| + | |||
| + | We had many lively discussions. All pdfs are now available [http://adapt-workshop.org/program.htm online]. | ||
| + | |||
| + | We also discussed our experience with reproducing experimental results from 2 publications at ADAPT. The major problems included: | ||
| + | * lack of explicit software dependencies (small change in the second-tier GPU library caused software crashes) | ||
| + | * complex software and hardware setups, for example including specific Windows version with various installed services (that could be shared as a very large virtual machine which is another problem) and specific GPUs | ||
| + | * very large and not easily shareable/reusable virtual machines or packages | ||
| + | |||
| + | We will continue these discussions at [http://c-mind.org/events/trust2014 ACM SIGPLAN TRUST workshop]. | ||
Current revision
Panel at ADAPT 2014
We had many lively discussions. All pdfs are now available online.
We also discussed our experience with reproducing experimental results from 2 publications at ADAPT. The major problems included:
- lack of explicit software dependencies (small change in the second-tier GPU library caused software crashes)
- complex software and hardware setups, for example including specific Windows version with various installed services (that could be shared as a very large virtual machine which is another problem) and specific GPUs
- very large and not easily shareable/reusable virtual machines or packages
We will continue these discussions at ACM SIGPLAN TRUST workshop.