This will contain benchmarks for my master's thesis on segmented redomap.
Inspired by the benchmarks for Futhark papers (like https://github.com/HIPERFIT/futhark-pldi17/).
This repo slowly got filled with all kinds of other files that I used in my evaluation. To run the actual benchmarks from my thesis, do the following steps:
-
Install
stack
if it is not already present on your system: https://www.haskellstack.org/ -
Make a clone of the Futhark compiler repo from https://github.com/HIPERFIT/futhark
-
cd
into the futhark directory -
run
git reset --hard dd2d6651fd9ae9de6fcfe408ed02d54e4976b07e
to get the version of the compiler I used. -
run
stack setup
-
run
<path/to/thesis-benchmarks>/build-futhark.sh
-- this will build all four versions of the Futhark compiler I used in my thesis. -
cd into the top level of this repo
-
run
make futhark-benchmarks
. This will get only the newest version of the futhark-benchmarks repo. If significant changes have been made, thereby making the compiler not work, you should manually get the benchmarks from the commit idc5ae760d60749f08968be32c339e49dde3e94b1b
-
run
make bench-res
This will run all the benchmarks used in my thesis, once for every version of the compiler (it will complain about the N-body benchmark not being able to be compiled, and for the versioned-code it will not be able to compute MRI-Q small.) -
After waiting for some time, you should have , but will produce 4 json files in the
results
directory. These contain the runtime for all the benchmarks used in my thesis.