This repository has no description
0

Configure Feed

Select the types of activity you want to include in your feed.

C++ 43.0%
C 24.6%
HTML 19.8%
Python 7.9%
Tcl 2.7%
Shell 0.8%
Makefile 0.8%
Other 0.4%
59 2 0

Clone this repository

https://tangled.org/storm.tngl.sh/acce_fpga https://tangled.org/did:plc:aiv6frvo7jipnbyufobvolzs
git@tangled.org:storm.tngl.sh/acce_fpga git@tangled.org:did:plc:aiv6frvo7jipnbyufobvolzs

For self-hosted knots, clone URLs may differ based on your setup.



README.md

ACCE FPGA#

This script collect all required data that was used for the analysis:

./scripts/benchmark.py

Each output can be quickly checked by running:

./scripts/check.py

All other scripts were mainly used during testing. These other scripts also have some dependencies that are not automatically installed and also aren't present on DAS-5. So make sure to install these yourself if you want to give them a try!

Testing harness adaptions#

You'll have to make some changes to your testing harness as we've customised a lot of things to ease automation. Most importantly, our *.tcl scripts require the constants to be given as environment variables. See ./scripts/benchmark.py to see which environment variables need to be given.

It's probably easier to customise our testing harness instead. For testing other input files, please add these to ./scripts/benchmark.py. Specifically, you'll want to add the entries to the configurations list. The tuples are x_count, y_count, unroll_count and file_path.

Also note that by default, the benchmark script uses node 017. If this node is offline for any reason, modify line 60 or comment out both lines 59 and 60.

The ./scripts/check.py automatically adapts to more input files being present. Just make sure that the expected output files are in the cpu_impl folder. The easiest way to do so is to add the files to ./cpu_impl/job.sh and running it with sbatch.