Metis

examples/metis/README.md

@@ -0,0 +1,100 @@
+# Example: Metis
+
+## Links
+
+- [Metis](http://pdos.csail.mit.edu/metis/)
+- [Github repo](https://github.com/ydmao/Metis.git)
+
+## Setup
+
+### Clone Metis repo
+
+```bash
+cd examples/metis/
+mkdir deps
+cd deps/
+git clone https://github.com/ydmao/Metis.git
+cd Metis
+git apply ../../patch.diff  # apply patch for missing architectures
+cd ../..
+```
+
+### Generate venv & configure it
+
+In the following we assume you are in the `metis` directory.
+
+```bash
+./configure.sh
+. ./venv/bin/activate
+```
+
+### Configure project
+
+In the following we assume you are in the `metis` directory.
+
+To configure and build the project
+
+```bash
+cd deps/Metis/
+./configure
+make
+```
+
+To link with a specific memory allocator
+
+```bash
+cd deps/Metis/
+./configure --with-malloc=<jemalloc|flow>
+make clean
+make
+```
+
+## Available benchmarks
+
+After setup, the following benchmarks work out of the box:
+
+- `wrmem`
+- `kmeans`
+- `pca`
+- `matrix_mult`
+
+The following benchmarks require extra data files:
+
+- `hist`
+- `linear_regression`
+	- sometimes fails with "Segmentation fault" (mistake in the Metis benchmark code)
+- `string_match`
+- `wc`
+- `wr`
+
+Assuming we are in the `metis` directory, download the data files:
+
+```bash
+cd deps/Metis/
+wget http://pdos.csail.mit.edu/metis/data2.tar.gz
+```
+
+Unpack the test data files:
+
+```bash
+tar -xvzf data2.tar.gz
+```
+
+This will create a `data` folder.
+
+## Run campaign with all benchmarks
+
+In the following we assume you are in the `metis` directory.
+
+```bash
+./campaign_metis.py
+```
+
+## TODO
+
+- [ ] Fix `linear_regression` benchmark
+- [ ] Additional benchkit support for
+  - [ ] Different memory allocators
+  - [ ] different map datastructures
+  - [ ] different sorting algorithms
+  - [ ] different modes